0748689605

The Semantics of
Word Formation and

Lexicalization
Edited by
Pius ten Hacken and Claire Thomas
The Semantics of Word Formation and Lexicalization
The Semantics of Word Formation
and Lexicalization
Edited by Pius ten Hacken and Claire Thomas
© editorial matter and organization Pius ten Hacken and Claire Thomas, 2013
© the chapters their several authors, 2013
Edinburgh University Press Ltd

22 George Square, Edinburgh EH8 9LF
www.euppublishing.com
Typeset in Ehrhardt by
Servis Filmsetting Ltd, Stockport, Cheshire,
and printed and bound in Great Britain by
CPI Group (UK) Ltd, Croydon CR0 4YY
A CIP record for this book is available from the British Library
ISBN 978 0 7486 8960 6 (hardback)

ISBN 978 0 7486 8961 3 (webready PDF)
The right of the contributors

to be identified as authors of this work
has been asserted in accordance with
the Copyright, Designs and Patents Act 1988.
Semantics of Word Formation and Lexicalization
Contents
List of figures vii

List of tables ix
1. Word formation, meaning and lexicalization 1

2. Semiproductivity and the place of word formation in
grammar28
Pius ten Hacken
3. Lexicalization in Generative Morphology and Conceptual
Structure45
Claire Thomas
4. Term formation in a special language: how do words specify
scientific concepts? 66
Kaarina Pitkänen-Heikkilä
5. Nominal compounds as naming devices: a comparison of
English and Polish land surveying terminology 83
Pius ten Hacken and Ewelina Kwiatek
6. Semantic and formal structure: a corpus-based study of
Swedish NN compounds and their French counterparts 102
Maria Rosenberg
7. The semantics of lexical modification: meaning and meaning
relations in German A+N compounds 121
Barbara Schlücker
8. Semantic transparency and anaphoric islands 140
Martin Schäfer
9. Semantic coindexation: evidence from Portuguese derivation
and compounding 161
Alexandra Soares Rodrigues and Graça Rio-Torto
vi semantics of word formation and lexicalization
10. Deverbal nominalizations in English: an LMBM approach 180

Maria Bloch-Trojnar
11. Degrees of lexicalization in Ancient Greek deverbal nouns 203
Germana Olga Civilleri
12. How many factors influence the meaning of denominal and
deadjectival verbs? The case of Modern Greek verbs
in -(ι)άζω225
Angeliki Efthymiou
13. Analysing en-and its Romance equivalents in Jackendoff’s
Conceptual Structure 247
Jessica Forse
14. Semantics of diminutivization: evidence from Russian 266
Renáta Panocová
Notes on contributors 286

Bibliography 290
Author index 309
Subject index 315
List of figures
1.1 Latin facio and conficio2

1.2 Model of grammar adopted in LMBM 13
1.3 Distributed Morphology 16
1.4 Parallel Architecture 20
2.1 Male blackbird Turdus merula38
3.1 Parallel Architecture 52
3.2 Representation of settlement57
3.3 Lcp lexicalized as a single lexical item 60
3.4 The logical polysemy of settlement60
3.5 Split lexicalization of an lcp 61
3.6 Lexicalization of motion and manner of motion in French
and English 62
3.7 The projective conclusion space of settlement64
4.1 Emi, hede, sepivä and sulkasuoninen69
4.2 Part of the derivative table by Elias Lönnrot 70
4.3 Purje and kannus72
4.4 Kärhi, sepivä, terälehti and verholehti73
4.5 Lanttopäinen and silposuoninen75
4.6 Sahalaitainen, toissahainen and vastosahainen77
4.7 Puikea vs. vastopuikea and parilehtinen vs. toisparinen78
4.8 Suffix -mAinen in the names of the various forms of a
flower’s corolla: kellomainen, perhomainen, ruusumainen,
ristimäinen79
4.9 Päätöparinen, tasaparinen and vuoroparinen80
5.1 Compounds in the English termbase 93
5.2 Compounds in the Polish termbase 94
5.3 Distribution of compounding types in the English termbase 95
5.4 Distribution of compounding types in the Polish termbase 95
viii semantics of word formation and lexicalization
5.5 Compounds in each subfield of the English termbase 97

5.6 Compounds in each subfield in the Polish termbase 98
5.7 Distribution of compounding types per subfield in English 99
5.8 Distribution of compounding types per subfield in Polish 100
10.1 Derivational and spelling operations in LMBM 189
10.2 LMBM representation of deverbal nominalizations in
English191
11.1 Continuum of the semantic values of AG suffixes 204
11.2 Semantic chain of θυμός [thumós] 216
11.3 Continuum between transparent and lexicalized forms 219
12.1 The distribution of -άζω and -ιάζω verbs in RDMG 229
12.2 The distribution of [+/–learned] and [–learned] -(ι)άζω
verbs in RDMG 229
13.1 The Tripartite Parallel Architecture 249
14.1 Complete semantic definition of deadjectival verbs in Slovak 268
14.2 Denominal diminutive nouns in Russian 273
14.3 Deverbal diminutive nouns in Russian 277
14.4 Deadjectival diminutive nouns in Russian 278
List of tables
2.1 Adjective formation with -al and -ary36

6.1 Total number of tokens in the parallel corpus 106
6.2 Swedish NN compounds and French counterparts in the
parallel corpus 106
6.3 Formal structure of the French counterparts in the parallel
corpus107
6.4 Semantic relations within the Swedish NN compounds and
their French counterparts 113
8.1 Absolute occurrences of the anaphoric reference search
pattern154
8.2 Absolute frequencies of compounds in the corpus 155
8.3 Asymmetry in the absolute frequency of phrasal and
compound A N constructions 157
9.1 Contextualized comparison between the argument and the
lexical-conceptual structures of verbs and their deverbal
nouns163
9.2 Verbal bases and their deverbal nouns with affixes -dura, -ção
and -mento168
10.1 Countability distinctions in lexicalized deverbal nominals 185
10.2 +/− Singular and +/− Plural as operators 189
10.3 An LMBM analysis of deverbal nominalizations in English 197
12.1 Verb-forming suffixes: token frequency in printed school
Modern Greek 229
12.2 The meanings of -ίζω, -(ι)άζω, -ώνω, -εύω, -αίνω, -άρω
derivatives and -ποιώ formations 241
12.3 Verb forming processes 242
12.4 Token frequency in printed school MG 243
12.5 Type frequency in printed school MG 243
x semantics of word formation and lexicalization
13.1 Variables by semantic type 259

14.1 Denominal diminutive nouns in Russian 281
14.2 Deverbal diminutive nouns in Russian 284
14.3 Deadjectival diminutive nouns in Russian 285
chapter 1
Word formation, meaning and

lexicalization
W ord formation is a rule-based process for producing new words.

Often, the study of word formation has been undertaken from a
purely formal perspective. This book looks at how the meaning of the
resulting words is determined and how they are lexicalized. It brings
together new work undertaken from a variety of theoretical perspectives
in order to shed new light on a subject of growing interest in linguistics,
computational science, semantics and lexicography.
This introductory chapter provides a broad historical overview of the
field (section 1) and describes the main current approaches (section 2).
Against this background, section 3 introduces each of the contributions and
explains its position in the resulting theoretical map.
1 HISTORICAL BACKGROUND
In understanding the historical development of linguistic thought in the

twentieth century, it is convenient to distinguish the European tradition
from the American tradition, even though Joseph (2002: 47–70) shows
quite convincingly that it is hard to characterize either in a rigorous way.
The main cause of the difference is that the European tradition of linguistic
research developed out of philology, whereas the American tradition devel-
oped out of anthropology. As we will show, this difference has interesting
consequences for the study of word formation.
1.1 The European tradition

Saussure’s (1916) Cours de Linguistique Générale can be seen as the seminal
work of the European tradition of linguistics. Even though modern
2 pius ten hacken and claire thomas
research has shown that certain aspects of the book may not represent the
actual views held by Ferdinand de Saussure (1857–1913), the posthumous
text compiled by his colleagues and students is the version that inspired
several generations of linguists. The study of word formation hardly
constitutes an issue on its own in the work, but a number of Saussure’s
central tenets are important because they have influenced work on word
formation by later scholars. In order to understand these tenets fully, we
have to consider them in their original context, which was determined by
nineteenth-century comparative-historical linguistics.
The main interest of nineteenth-century linguists was the study of
historical processes. In the early nineteenth century, it was discovered
that many languages could be analysed as related to each other. This led to
efforts to represent these relations in genealogical trees and to reconstruct
earlier stages of the languages than the ones for which direct evidence could
be gathered. August Schleicher (1821–68) was an important representative
of this work. Later, the so-called Junggrammatiker or Neogrammarians,
aimed to formalize and improve the procedures that were used in these
reconstructions. Word formation did not play a significant role in this
type of linguistics. The main objects of attention were phonology and
inflectional morphology. Semantics was important only in the sense that it
helped establish that two word forms from different languages or periods
actually corresponded to each other, as a preliminary to explaining the
underlying phonological or morphological processes.
In an apparent reaction to this general orientation of the field, Saussure
(1916: 114–40) emphasizes the importance of distinguishing synchronic
and diachronic linguistics. Of the two, synchronic linguistics is more basic.
For Saussure, the study of the history of a language is only possible as the
study of a number of successive states. This view is inspired by observed
shortcomings of the earlier practice of comparative-historical linguistics,
but it also has consequences for the study of word formation, not all of
which are made explicit in Saussure (1916). Figure 1.1 represents an
example Saussure (1916: 137) gives to illustrate diachronic and synchronic
approaches.
Saussure objects to the common statement that the short a in facio
(‘make’) becomes an i in conficio (‘produce’). Instead, he argues that we
facio confacio Époque A
facio conficio Époque B

Figure 1.1 Latin facio and conficio
word formation, meaning and lexicalization 3
should describe such a relationship in terms of two successive stages. The

diachronic opposition is then rather between confacio and conficio than
between facio and conficio. This account is typical in its focus on phono-
logical change. The word formation relationship between facio and conficio
is not represented diachronically but only synchronically, i.e. there is no
account of the formation process, but only of its result.
Another opposition that is central to Saussure’s view of language is the
one between langue and parole. For Saussure (1916: 30), langue encom-
passes the social as opposed to individual aspects of language and the essen-
tial as opposed to accidental aspects. It is a system of signs (signes). These
signs are implemented in the speaker’s mind. The formation of sentences
is assigned to the parole, as reflecting the choices of an individual speaker
in expressing thoughts. Word formation can then be seen as the formation
of new signs.
A sign is composed of a signifiant (‘signifier’, i.e. acoustic image) and a
signifié (‘signified’, i.e. concept). As a first principle, Saussure (1916: 100)
proposes ‘l’arbitraire du signe’, i.e. the arbitrariness of the relation between
the signifié and the signifiant. At one level, this principle is straightforward.
If it did not apply, we could not explain that the same animal is called dog
in English, Hund in German, chien in French, and pies in Polish. However,
its validity depends on the strict separation of synchronic and diachronic
perspectives. Historical linguistics explains French chien as derived from
Latin canis by a sequence of regular phonological changes. This means that
the form chien in its relation to the concept it stands for is not (entirely)
arbitrary in this diachronic perspective.
Word formation is for Saussure above all a source of partially motivated
words. Saussure (1916: 181) gives the example of poirier (‘pear tree’), which
is partially motivated by poire (‘pear’) and by the existence of other pairs
such as cérise (‘cherry’) and cérisier (‘cherry tree’). However, Saussure does
not distinguish between these cases and the relation between the various
English plural nouns in -s. In neither case is there an explicit rule under
lying the relationship between the signs. Such a rule would not have any
natural position in Saussure’s theory. Saussure’s system of langue does not
include a rule component for syntax either. As mentioned above, Saussure
assigns the formation of sentences to parole.
For the study of word formation, the most influential continuation of
the general themes of Saussure’s system is the so-called Prague School.
The Cercle Linguistique de Prague was founded in 1926 (Simpson 1995:
247). Its most enduring achievements are in the domains of phonology
and syntax. In phonology, the notion of phoneme was developed by
Trubetzkoy (1939) in a theory that involved minimal pairs, neutralization
and archiphonemes. In syntax, the major contributions were the study of
valency, topic-focus articulation and the opposition between theme and

rheme (see Hajičová 1995). The general view of the nature of language and
the organization of linguistics is formulated in Thèses (1929), a discussion
document presented at the Premier Congrès des philologues slaves.
For the study of word formation, the Prague School emphasized three
aspects that, compared to Saussure’s view, mark more of a shift of emphasis
than a sharp distinction. The first of these is the functionalist approach
to language. Thèses (1929: 33) starts with a section about methodological
problems and the first heading is ‘Conception de la langue comme système
fonctionnel’.1 The idea that the system of the langue should be studied in
its communicative use is also found in Saussure (1916: 27), but it is more
prominent in the work of the Prague School, leading, for instance, to such
analyses as Jakobson’s (1960) model of communication with six functions
linked to each of the factors involved.
A second change of emphasis can be observed in the discussion of
synchronic and diachronic perspectives. According to Thèses (1929:
34), ‘la description synchronique ne peut pas [. . .] exclure absolument
la notion d’évolution’.2 Whereas Saussure kept these two perspectives
apart rather radically, no doubt as a reaction to the preceding school of the
Neogrammarians, the Prague School restored a certain degree of interac-
tion between the two. Thus Thèses (1929: 34) states that ‘Les éléments sty-
listiques sentis comme archaïsmes, en second lieu la distinction de formes
productives et non productives sont des faits de diachronie, que l’on ne
saurait éliminer de la linguistique synchronique.’3 For the study of word
formation, we noted some of the problems involved in a rigorous separation
of synchronic and diachronic views in the discussion of conficio. It is inter-
esting to observe in this respect that Thèses (1929) refers to productive
forms, rather than productive rules.
The third point in which the Prague School diverges from Saussure
is the one most directly relevant to the study of word formation. Thèses
(1929: 38) observes that ‘Le mot, considéré du point de vue de la fonction,
est le résultat de l’activité linguistique dénominatrice.’4 As the quotation
indicates, this observation is directly related to the functionalist perspec-
tive chosen. To the extent that an instance of naming changes the langue,
it is also a case where diachronic aspects interact directly with synchronic
description.
The Prague School view of language created a good background for
the systematic study of terminology. In individual domains, the study of
how concepts could be meaningfully named had been studied before. In
biology, the work of Carl von Linné (1707–78) is the basis of the system of
designating species that is still in use today. In chemistry, the publication
of the Méthode de Nomenclature Chimique in 1787 marked the replace-
ment of a confusing collection of names for chemical substances, based on

various principles, by a set of names indicating their origin (see Bensaude-
Vincent and Stengers 1995: 116–18). However, attempts like these were by
their nature limited to individual disciplines.
In the 1930s, a more systematic approach to terminology emerged. A
landmark was Wüster’s (1931) doctoral dissertation, which is often taken
to be the starting point of modern terminology, e.g. Pearson (1998: 9),
Cabré (1999: 7). Wüster’s approach focused on the systems behind the
naming of concepts and on standardization. It is interesting to note that
the Prague School was also interested in the standardization of national
languages. Thus Thèses (1929: 56–8) discusses the issue of establishing
standard literary languages in the Slavic language family. It is possible, but
by no means obvious, that Wüster was aware of some of the ideas of the
Prague School. Nikolaj Trubetzkoy was professor of Slavic linguistics at
Vienna University from 1922 until his death in 1938 (Honeybone 2005).
According to Lang’s (1998) biographical note, after completing his PhD in
Berlin Eugen Wüster returned to Austria in 1931 to take over his father’s
tool manufacturing business in Wieselburg, a good 100 kilometres from
Vienna. The Technical Committee for the standardization of terminology
of the International Standards Association, the forerunner of ISO/TC37,
was founded in Vienna in 1936, and Lang (1998: 15) states that much of the
work done was carried out in Wieselburg, by Eugen Wüster.
1.2 The American tradition

In North America, at the start of the twentieth century, there was no philo-
logical tradition rooted in indigenous languages as there was in Europe.
Matthews (1993: 5) lets his history of American linguistics begin ‘around
1910’. This excludes William Dwight Whitney (1827–94), but his work
on linguistics, including for instance his 1879 Sanskrit Grammar, was in
many respects closer to the European than to the American tradition. An
important impetus for a specifically American approach to linguistics was
the anthropologically inspired work by Franz Boas (1858–1942).
One of the main differences between the American and the European
approaches to linguistic analysis is that the American approach does not
assume the same level of knowledge of the object language at the start of the
linguistic investigation as does the European approach. The linguist starts
by collecting data and the methods of data collection and analysis should
work equally well for previously unrecorded languages, unknown to the
linguist. As a consequence, word boundaries are not assumed to be known
and the set of contrastive phonemes is to be established without recourse to
traditional analyses of the language.
The first comprehensive overview of linguistic analysis in the American

tradition is Bloomfield (1933).5 Many of Bloomfield’s ideas on word forma-
tion as formulated there still play a role in determining the attitude towards
particular linguistics problems, so it is worth presenting them here as a
background for further developments.
Central to Bloomfield’s (1933) view of linguistics is the assumption
that the study of language should only concentrate on the utterance as
the physical, observable part of communication, to the exclusion of the
mental processes of the speaker and hearer. As described in more detail
by Matthews (1993: 52–75), the stimulus- response model introduced
by Bloomfield (1933: 22–7) constitutes an innovation with respect to his
earlier writings. Bloomfield is reluctant to refer to meaning in crucial
points in linguistic analysis because ‘[i]n order to give a scientifically accu-
rate definition of meaning for every form of a language, we should have
to have a scientifically accurate knowledge of everything in the speakers’
world’ (1933: 139).
Against this background, it is no surprise that the study of morphology
concentrates on forms rather than meaning. Bloomfield’s definition of
morpheme appeals to meaning, ‘[a] linguistic form which bears no partial
phonetic-semantic resemblance to any other form’ (1933: 161), but meaning
is used primarily to recognize homonyms as different morphemes (e.g. pear
and pair) and to exclude arbitrary formal resemblances such as bird and burr
to affect linguistic analysis. Bloomfield (1933: 209) gives re-ceive, de-ceive,
re-tain, de-tain as examples of derived words without an underlying free
form, implying that -ceive and -tain are morphemes, although it is hard to
assign them any coherent semantic value.
In the discussion of the distinction between word formation and inflec-
tion, Bloomfield (1933: 222–6) gives a range of properties that can be used
as criteria to distinguish them, but also states that ‘[t]his distinction cannot
always be carried out’ (1933: 223). Developing Bloomfield’s approach,
Bloch and Trager (1942: 54) describe the nature of the distinction as in (1).
(1) For some languages, it is useful to divide the morphological constructions of

complex words into two kinds according to the grammatical function of the
resulting form: derivational and inflectional.
In a number of ways, (1) plays down the importance of the distinction,

and in particular of the semantic aspect of it. First, the distinction is not
universal but only applies to ‘some languages’. Second, the division is said
to be ‘useful’ rather than natural or compelling. Finally, the basis of the
distinction is the ‘grammatical function’ rather than the meaning of the
resulting form in relation to the base.
As a final component of this overview of how Bloomfield’s theory

approaches the topic of this volume, it is worth outlining his idea of the
lexicon. According to Bloomfield (1933: 274), ‘[t]he lexicon is really an
appendix to the grammar, a list of basic irregularities.’ The distinction
between regular and irregular is that ‘any form which a speaker can utter
without having heard it, is regular’, whereas ‘any form which a speaker
can utter only after he has heard it from other speakers, is irregular’ (1933:
274). It should be noted here that for Bloomfield, ‘the terms regular and
irregular are used only of features that appear in the grammar’ (1933: 275).
The overall picture emerging for Bloomfield’s approach to the meaning
and lexicalization of word formation is then that there is no general class
of word formation (because it is a matter of language-specific convenience
whether it should be distinguished from inflection), no basis for a specific
interest in meaning (because it reaches too far beyond the aspects of lan-
guage we can study) and lexicalization is merely a matter of irregularity.
Given this situation it is not surprising that Bloomfield and his followers
concentrated on other aspects of language. It was a widely shared assump-
tion that the starting point for linguistic research must be a collection of
utterances (see Harris 1951: 12), and the first steps of analysis focused on
the recognition of a set of phonemes, followed by the identification of a set
of morphemes. Thus Bloch and Trager (1942) devote more than half of
their overview of linguistic analysis to phonology and Harris (1951) divides
his overview into methodological preliminaries (ch. 2), phonology (chs
3–11), and morphology (chs 12–19), where the latter is restricted to the
recognition of morphemes and allomorphs.
In what is known as the Chomskyan revolution, a number of foundational
assumptions of Bloomfield’s approach to linguistics were replaced by a new
set. As described by ten Hacken (2007: 156–79), this did not affect every
aspect of linguistic analysis, but it affected above all the stimulus-response
model and its consequences. Whereas Bloomfield insisted on keeping the
scientific study of language clear of mentalism, Chomsky argued for the
systematic study of individual speakers’ competence.
Chomsky’s (1965: 3) distinction between competence and performance
has often been compared to Saussure’s langue and parole. Competence and
langue are both realized in the mind, but whereas competence is individual,
Saussure places the individual aspect of language in parole. For Saussure,
langue is not fully realized in an individual, but only in a speech commu-
nity. This distinction has far-reaching implications for lexicalization, see
ten Hacken and Panocová (2011). The lexicon in Saussure’s theory is part
of the socially shared langue so that a new entry is an expression that has
been accepted by the speech community. In Chomsky’s model, the speech
community does not play an essential theoretical role. Chomsky (1980)
argued that named languages such as English or French do not exist as real
entities in the world. What exists are the speakers of these languages and
their individual competence. An entry in the lexicon is for Chomsky a piece
of knowledge in the individual speaker’s mind.
Apart from competence, another crucial focus of Chomsky’s theories
is syntax. This was quite innovative as older theories had studied syntax
much less than other aspects of language. In the Bloomfieldian framework,
analysis had to start from the phonetic data and, as formulated by Hockett
(1942: 21), ‘[t]here must be no circularity’, i.e. morphological analysis
cannot start until phonological analysis has been completed. This require-
ment made it very hard to reach the level of syntactic analysis. Saussure’s
theory of the signe allocated syntax to the parole, as a component of the
individual speaker’s choice of words in an utterance. This new focus on
syntax meant that Chomskyan linguistics covered new ground in this area,
which therefore attracted most attention.
The earliest study of word formation in the new framework was Lees’
(1960) work on compounding. As described by ten Hacken (2009a), Lees’
basic idea was to generate compounds in much the same way as Chomsky
proposed to generate sentences. Rewrite rules would generate a sentence-
like structure, from which transformations would produce a compound.
The obvious appeal of this theory was that it accounted for both the
meaning and the form of compounds in a way that assimilated them to
sentences. The equally obvious drawback was the lack of constraints on the
power of the rule system.
At the time Lees (1960) proposed his theory of compounding, Chomsky
did not yet assume the existence of a lexicon as part of his linguistic theory.
The grammar would generate a tree structure by means of rewrite rules and
the leaves of the tree would be morphemes. The introduction of the lexicon
was a response to theoretical problems with this model. Chomsky (1965:
142) presents the lexicon as consisting of lexical entries and redundancy
rules. The lexical entries are a combination of phonological, syntactic, and
semantic features that cannot be predicted by rules. The redundancy rules
express generalizations so that they do not have to be specified in the indi-
vidual lexical entries. Therefore ‘the lexical entries constitute the full set of
irregularities of the language’ (1965: 142), an image that corresponds very
closely to Bloomfield’s conception of the lexicon.
The introduction of the lexicon created an alternative approach to the
coverage of word formation. Instead of using syntactic rules to combine
morphemes, word formation could also be accounted for by means of
redundancy rules in the lexicon. This was one of the battlefields in what
Newmeyer (1986) calls the ‘Linguistic Wars’ between generative seman-
tics and Chomsky’s interpretive semantics. The main bone of contention
was the status of Deep Structure. Chomsky (1957) had introduced Deep
Structure as the interface between rewrite rules and transformations. In
generative semantics, Deep Structure was taken to be more and more
abstract, leading to its eventual merger with semantic representation.
In interpretive semantics, Deep Structure was taken to be less and less
abstract, resulting ultimately in its abolition as a level at which constraints
are stated. Traces of movement made it possible to state the constraints on
the structure after movement.
When we consider the study of word formation, it is striking that the
most influential publications reflecting the different sides of the opposition
are of a very different nature. Chomsky’s (1970) contribution to the debate
is a highly programmatic overview that can be seen as the start of lexicalist
morphology. Levi’s (1978) presentation of a fully elaborated system for the
treatment of compounds is rather the endpoint of the generative semantic
approach to morphology.
Chomsky’s (1970) main concern is how the different components of
his grammatical framework are related to each other, in particular the
base component, which contains the rewrite rules, the transformational
component and the lexicon. He discusses nominalizations and argues that
the degree of irregularity they display requires that they be treated in
the lexicon. Jackendoff (1975) elaborates this idea and develops a system
of redundancy rules that encode generalizations about semantic and
form-based regularities. Instead of generating strings of morphemes and
transforming them into a structure that can be interpreted semantically,
Jackendoff proposes a full-entry theory, where all words have an entry in
the lexicon, but regular aspects covered by redundancy rules do not cost as
much to store as idiosyncratic pieces of information. This view of redun-
dancy rules differs somewhat from the one proposed by Chomsky (1965)
and leads to a conception of the lexicon that does not coincide at all with
Bloomfield’s (1933) idea of it as a collection of irregularities.
Levi (1978) can be seen as the culmination of the transformational
approach to word formation initiated by Lees (1960). In line with the
assumptions of generative semantics, Levi assumes that the derivation of
a complex word should account for its meaning. Lees’ (1960) system had
been found too powerful to be explanatory. Chomsky (1964: 41) proposed,
therefore, that ‘[a] deleted element [. . .] is always recoverable.’ In order
to comply with this constraint, Levi (1978) proposes a restricted number
of recoverably deletable predicates (RDPs) that characterize the relationship
between components of a compound. Although this is an ingenious way of
making the transformational approach compatible with Chomsky’s recov-
erability condition, it did not lead to much further research along similar
lines. One of the main reasons was the collapse of the support for generative
semantics in the late 1970s. This not only removed the theoretical founda-
tion of Levi’s work, but, as described by ten Hacken (2009a), also directed
attention away from the research questions she tried to answer.
2 CURRENT APPROACHES
Current approaches to the meaning and lexicalization of word formation

build up in various ways from the historical background described in
the previous section. We will describe here the development of the ono-
masiological approach in terminology and in lexical semantics, Lexeme-
Morpheme Base Morphology (LMBM), Langacker’s Cognitive Grammar
and the most prominent offshoots of generative grammar.
2.1 The onomasiological approach in terminology

In a purely onomasiological approach to terminology, the starting point
is the study of a domain, resulting in the identification of the relevant
concepts and the relations between them. These concepts are then given a
name and a definition. In the 1930s, when terminology emerged as a field
of study, the decision to adopt an onomasiological approach was arguably
motivated as much by practical as by theoretical reasons. Before the advent
of large electronic corpora, the use of text as a basis required a large reading
programme of the type adopted by James Murray for the Oxford English
Dictionary (see Craigie and Onions 1933). Sager (1990) strongly advocates
the use of corpora in terminology. A well-documented early example of
terminological research based on electronic corpora is Pearson (1998).
In principle, one could imagine a clash between terminologists who
persist in using a purely onomasiological approach and those who insist on a
corpus-based approach instead. In practice, a handbook such as Wright and
Budin (1997) shows hardly a trace of such an opposition. Corpus-based work
has been integrated seamlessly into the mainstream of terminology. The
reason that this did not cause any major problems is that the two approaches
complement each other in their strengths and weaknesses. As Wright (2006:
19–20) notes, ‘standards serve their best function when we use them all the
time, but remain oblivious to their presence.’ Therefore standardization
should be based on the knowledge of existing practice. A corpus-based
approach to terminology gives first of all access to the actual forms in use in
a field. An approach starting from the analysis of the relevant domain yields
the concepts to be named. By reconciling these two approaches, a standard
that wherever possible coincides with common usage while at the same time
covering all relevant concepts can be established.
A more pertinent challenge to this mainstream is posed by the socio-

cognitive approach of Temmerman (2000). Following work such as Labov
(1973) and Rosch (1978), it is now widely accepted that the meaning of
general language words has a prototype structure. Temmerman claims
that the meaning of terms also has such a structure. It is certainly true
that in many fields the standardization of terminology is more a matter
of choosing standard names than of establishing exact boundaries of the
concept. As Arntz et al. (2009: 125) state, an important task of terminol-
ogy is the elimination of synonymy, polysemy and homonymy. In order
to incorporate these insights without giving up the core of the classi-
cal approach to terminology, ten Hacken (2008) proposes a distinction
between specialized vocabulary and terms (in the narrow sense), where
the former may involve prototype concepts but the latter not. Precise
definitions are necessary (only) when scientific claims or law enforcement
depend on them. Temmerman’s approach applies to specialized vocabu-
lary. In the case of terms in the narrow sense, terminologists attempt to
counteract the vagueness of the boundaries of the concept they name,
but the interaction between terms and corresponding general language
words complicates this process, as exemplified for some legal terms by ten
Hacken (2010c).
2.2 The onomasiological approach to lexical semantics

As noted in section 1.1, the Prague School developed an interest in the
word as what Thèses (1929: 38) calls ‘le résultat de l’activité linguistique
dénominatrice’, the result of a linguistic naming act. Štekauer (2005b) and
Grzega (2009) give overviews of the work on word formation that emerged
from the elaboration of this perspective of the nature of a word. Word
formation is in competition with semantic change of existing words and
with borrowing from other languages when it comes to naming concepts.
An important milestone in the onomasiological study of word formation
was Dokulil (1962), which Štekauer (2005b: 207) calls ‘the first compre-
hensive onomasiological theory of word-formation’. As Dokulil’s work was
published in Czech, his influence was mainly felt in countries with Slavic
languages.
Dokulil (1962) introduced the term onomasiological category for the
central strategy that language users use to structure a concept in order to
come up with a name for it. He distinguishes mutational, transpositional
and modificational types as different onomasiological categories. In each
of these strategies, the concept is first classified by means of the choice of
an onomasiological base. Then it is further specified by an onomasiological
mark. Thus, in textbook, book is the onomasiological base and determines
the general class, whereas text is the onomasiological mark, which distin-
guishes textbooks from other books. This approach is reminiscent of the
Aristotelian approach to definition, based on genus and differentiae, but
instead of resulting in a definition, it produces a name.
The different onomasiological categories reflect differences in the rela-
tionship between the onomasiological base and the onomasiological mark.
The base is classified according to its conceptual category, e.g. substance,
action, quality. The relationship to the mark constitutes the naming motive.
In the mutational type, the conceptual category of the base is changed by
the mark (e.g. speciality is a substance related to special, a quality). In the
transpositional type, no such change takes place (e.g. curiousness is a quality,
like curious). In the modificational type, a modifying feature is added, e.g.
collectivity in mankind.
Based on Dokulil’s ideas, Horecký (1983) developed a model involving a
sequence of levels of representation. The sequence models the steps in the
process a speaker goes through in determining the name for a particular
new concept. Starting with a description of the concept, in several steps
the speaker comes up with and organizes semantic features, and gradually
specifies properties of the resulting name. Štekauer (1998) developed the
cognitive basis of the model and proposed five onomasiological types, cor-
responding to different patterns for the realization of the onomasiological
base and mark.
Another line of research based on Dokulil’s ideas is the investigation of
the types of meaning that can be expressed by derivation. Szymanek (1988:
93) proposes the Cognitive Grounding Condition, which states that ‘[t]he
basic set of lexical derivational categories is rooted in the fundamental con-
cepts of cognition.’ He then presents a methodology for identifying these
and proposes a set of categories that can serve as a basis for explaining some
of the cross-linguistic similarities and differences in available derivational
types.
2.3 Lexeme-Morpheme Base Morphology

Robert Beard’s system of Lexeme-Morpheme Base Morphology (LMBM)6
can be seen as an attempt to reconcile insights from the onomasiological
approach to lexical semantics that emerged from the Prague School tradi-
tion with some of the basic assumptions of generative grammar. It is inter-
esting to observe in this context that Beard has a background in American
linguistics as well as a specialization in Slavic languages. Beard (1995) gives
the most comprehensive overview of the theory. Concise introductions can
be found in Carstairs-McCarthy (1992: 181–6) and Bloch-Trojnar (2006:
45–56).
The starting point for Beard’s theory is the observation that it is very
difficult to represent morphemes as Saussurean signs, because within a par-
ticular language, the same meaning is often expressed by different forms,
whereas one form can often stand for more than one meaning. When we
consider -al in English refusal as a sign, it shares the same form with -al in
developmental and the same meaning with -ance in acceptance. What Beard
proposes is to solve this problem by radically separating the operation on
the meaning from the operation on the form. He calls the operation on the
meaning derivation and the operation on the form affixation. Derivation
is further divided into L-derivation, corresponding to word formation
(‘lexical’), and I-derivation, corresponding to inflection. This leads to
quite complex mismatches between Beard’s use of the terms derivation and
affixation and the traditional use of these terms. As Carstairs-McCarthy
(1992: 182) observes, Beard’s idiosyncratic use of terminology ‘has almost
certainly hindered discussion of his ideas.’
As a consequence of the separation of derivation and affixation, there is
only a single L-derivation process for refusal and acceptance and a single
affixation process for refusal and developmental. In describing the formation
of refusal, we have to refer separately to the action nominalization process
and to -al suffixation.
Beard (1995: 45) matches the different components of his morphological
theory to the model in Figure 1.2, which corresponds closely to the one
adopted by Chomsky in the 1980s.
In the model in Figure 1.2, the earlier notions of Deep Structure and
Surface Structure have been replaced by D-structure and S-structure.
Whereas in earlier models, phonological interpretation was based on
Surface Structure and semantic interpretation on Deep Structure, now
Lexicon
Base rules
L-derivation
Lexical insertion
D-structure
I-derivation Transformations
S-structure
Affixation Interpretive rules
PF LF
Figure 1.2 Model of grammar adopted in LMBM
phonological form (PF) and logical form (LF) are both derived from
S-structure. To the right of the components of the model, the rule types
assumed by Chomsky are indicated. Chomsky (1981) replaces base rules
with X-bar theory and transformations and interpretive rules with move α
without changing the architecture. The labels to the left of the model indi-
cate the positions of Beard’s L-derivation, I-derivation and affixation. It is
interesting to note that Figure 1.2 implies the rejection of the view about
the distinction between inflection and word formation formulated by Bloch
and Trager (1942) in (1). Beard’s model requires them to be separate in all
languages, because they are in different positions.
2.4 Cognitive Grammar

Cognitive Grammar is a framework developed by Langacker (1987a,
1991). Although it emerged in the US, its assumptions diverge from the
traditional ones in the American linguistic tradition quite significantly.
Thus Langacker (1987a: 5) states that ‘[t]he most fundamental issue in
linguistic theory is the nature of meaning and how to deal with it.’ He
presents it as ‘an alternative to the generative tradition’ and claims that it ‘is
not in any significant sense an outgrowth of generative semantics’ (1987a:
4). Langacker’s idea is that ‘language is shaped and constrained by the
functions it serves’ (2008: 7).
Cognitive Grammar shares with generative semantics the idea that
syntax is not an autonomous component. Langacker (1987a: 57–8) recog-
nizes three types of unit: phonological units, semantic units and so-called
symbolic units, associating phonological and semantic structures. Syntactic
categories are reduced to units of these types, so that a noun is defined as
‘[a] symbolic structure whose semantic pole designates a thing’ (1987a:
183) and a verb as ‘a symbolic expression whose semantic pole designates
a process’ (1987a: 244). The idea of phonological, semantic and symbolic
units constituting the system of language is strongly reminiscent of
Saussure’s theory of the signe. Another point in which Cognitive Grammar
converges with Saussure is Langacker’s (1987a: 62) assumption that
‘[t]he grammar of a language is a characterization of established linguistic
convention. Conventionality implies that something is shared [. . .] by a
substantial number of individuals.’
Heyvaert (2009) gives an overview of the consequences of these assump-
tions for word formation, although her scope is both wider (cognitive
linguistics also beyond Langacker’s theory) and narrower (compounding).
Word formation is the creation of a new, automatized symbolic relationship
in the mind of a speaker and its spread through the speech community.
The main difference from sentence formation is that the formulation of a
sentence is less commonly automatized. However, the difference is one of

degree. As such, it is not substantially different from other oppositions in
linguistics, including the one between linguistic and non-linguistic units
(see Langacker 1987a: 60). Langacker (2008: 346) notes that establishing a
demarcation line between word formation and inflection is ‘both gratuitous
and empirically problematic’. This is in line with Bloch and Trager’s (1),
but the approach is extended to other contrasts. Thus the first example of
a grammatical construction he discusses in detail is in fact a compound,
jar lid factory (2008: 161–74). In cognitive grammar, morphology (and
language in general) is considered a domain with prototypes rather than
precise boundaries.
2.5 Distributed Morphology

Distributed Morphology (DM) is one of the main current approaches to
morphology in generative grammar. It was originally developed by Halle
and Marantz (1993) and has developed significantly since then.
Mainstream generative grammar has always been oriented towards
syntax. For morphology, the consequences are that most of the work
focuses on inflection rather than word formation and on form rather than
meaning. An example of the bias towards inflection is Stump’s (2011)
assessment that the two dominant approaches to morphology are the one
based on morphemes and the one based on paradigms. Paradigms are pri-
marily discussed in the context of inflection, as Stump’s (2001) own study
illustrates. DM is a theory based on morphemes.
The tendency to concentrate on form rather than meaning is illustrated
by Selkirk’s (1982) presentation of her theory of word formation. She
describes her purpose as the investigation of ‘the structure of words and the
system of rules for generating that structure’ (1982: 1), which leaves very
little room for the study of meaning and lexicalization. For Selkirk, the
account of root compounds such as apron string is restricted to the rewrite
rule N → NN (1982: 16) and the right-headedness (1982: 22). In genera-
tive semantics, such a compound was considered the result of transforma-
tions operating on a sentence-like structure, so that the meaning can be
explained as based on the sentence. The lexicalist alternative adopts the
position that the question of where the meaning comes from is not a well-
formed question and should not be studied in morphology. Compared to
Selkirk (1982), DM takes a much more radical approach to introducing
syntactic structure in the domain of morphology. The architecture of the
grammar assumed in DM is represented in Figure 1.3.
The framed items in Figure 1.3 correspond to different components of
what in other approaches is called the lexicon. Harley and Noyer (2003:
feature bundles
syntax
vocabulary
MS
LF
encyclopaedia
PF
meaning
Figure 1.3 Distributed morphology
465) call them ‘List A’, ‘List B’ and ‘List C’. The representation of syntax
has developed considerably over the period since DM was first proposed
and in Figure 1.3 it is therefore simply labeled as ‘syntax’. Halle and
Marantz (1993: 384) adopt a model with D-structure and S-structure as
in government-binding theory (see also Figure 1.2). Embick and Noyer
(2007: 292) assume a ‘syntactic derivation’ resulting in ‘(Spell Out)’, in
accordance with Chomsky’s (2000) phase-based syntactic derivation. ‘MS’
in Figure 1.3 stands for ‘morphological structure’. It is an innovation
introduced by Halle and Marantz (1993: 384) in order to account for the
mismatches between morphosyntactic and phonological ‘pieces’.
One of the central ideas of DM is that morphology does not take place
in the lexicon (see Marantz 1997). There is no lexical component, but its
function is distributed among the three framed components in Figure 1.3.
Syntax operates on feature bundles that are not marked phonologically
or semantically. Lexical insertion takes place at MS (Halle and Marantz
1993: 390). Lexical meaning is encoded in the encyclopaedia. Introduced
by Marantz (1997), this is a list consulted in the semantic interpretation
of LF. LF represents the effects of operations that are not visible in PF,
but it is still a syntactic (as opposed to semantic) representation. Clearly,
the effects of lexical insertion at MS have to be visible in order to consult
the right entries of the encyclopaedia. In their diagram, Harley and Noyer
(2003: 465) bring this about by linking PF to meaning. Embick and Noyer
(2007: 301) propose instead that the encyclopaedia is consulted in a step
they call ‘(Interpretation)’, following the generation of PF and LF. They
leave unspecified how PF and LF are linked to Interpretation.
The unit of operation in DM is the morpheme. Harley and Noyer

(2003: 468) distinguish between l-morphemes and f-morphemes, where
the former are lexical and the latter functional. This distinction has an
impact on the way lexical insertion operates. There is no distinction
between affixes and function words. Marantz (1997) introduces the idea
that Roots do not have a syntactic category but have to be associated with
a functional category in order to get one. Whereas the Root attack can be
associated with empty noun (n0) or verb (v0) morphemes, the Root grow
has an n0 morpheme realized as -th when it becomes a noun. Barner and
Bale (2002) argue that such an analysis is supported by various types of
psycholinguistic data. However, these arguments have been contested (see
Panagiotidis 2005).
As Halle and Marantz (1993: 432) state, in DM ‘word formation is syn-
tactic or postsyntactic, not lexical’. The distinction between inflection and
derivation ‘has no explicit status in DM’ (Harley and Noyer 2003: 474).
Every Root has to pick up a syntactic category from an f-morpheme, which
may or may not induce phonological changes. In the case of developmental,
the Root develop is associated first with a nominal f-morpheme, then with
an adjectival one. These associations are represented in the syntactic tree
and realized through vocabulary insertion at MS. For compounds, Harley
(2009: 135–40) presents a similar analysis. The main difference from
affixation is that for compounds more than one l-morpheme appears in
the tree structure. In the case of phrasal non-heads, e.g. repetitive strain
injury, an XP, here an AP, is recategorized as an N by association with a
phonologically empty n0 head (2009: 143).
Let us now consider how these considerations affect the study of lexi-
calization and meaning in relation to word formation. A first observation is
that the categories in which we state the topic of this volume do not match
the categories adopted in DM. Whereas lexicalization establishes a relation
to the lexicon, DM does not recognize a lexicon. As Harley (2009: 129)
states it, DM ‘attempts to present a fully explicit, completely syntactic
theory of word formation.’ Harley and Noyer (2003: 466–7) present various
interpretations of lexical and show that they are either rejected or have no
significance in DM. Lexical cannot be opposed to syntactic in DM, because
for every expression, the generation is driven by syntax. Idiosyncratic
meanings are treated as idioms in the encyclopaedia.
When we turn to the notion of word formation, an immediate issue is that
DM does not recognize a level of word in between morphemes and syntac-
tic constructs. This is the basis for Williams’s (2007) criticism of DM. Of
course, DM claims to be able to account for all instances of word formation,
but it does not recognize them as a separate category. As mentioned above,
there is no distinction between inflection and derivation. In addition, there
is no distinction between periphrastic and synthetic expressions of the

same information.
Perhaps even more significant than the difference in categorization,
however, is a difference in orientation. The DM framework attracts atten-
tion away from the questions that are central in this volume. Halle and
Marantz (1993) devote their entire article to inflection. The only mention
of derivation is when they point out the parallel between the vowel change
in keep – kept and deep – depth (1993: 397). In her contribution to the
Handbook of Compounding, Harley (2009: 130) admits that ‘there have been
very few Distributed Morphology proposals concerning compounding’. In
fact, her overview of the treatment of different types (2009: 135–40) does
not contain a single reference to DM sources.
In the treatment of meaning, we also find this orientation to other ques-
tions. Meaning is relegated to the encyclopaedia. Halle and Marantz (1993)
do not mention this component. Embick and Noyer (2007: 292) present a
diagram with ‘morphology’ in a box at the point where in Figure 1.3 the
arrow from ‘vocabulary’ arrives at MS. Harley (2009) only discusses the
generation of the correct forms of compounds. Therefore we can conclude
that the general orientation of the research carried out in DM does not
primarily address questions of meaning, but rather the question how the
form of an expression can be accounted for.
In Harley and Noyer (2003: 470), the encyclopaedia is identified as the
place where idioms are stored. In the same way as Hockett (1958), they
assume that any expression whose meaning cannot be derived composition-
ally is an idiom. This includes all unstructured words. Harley and Noyer
correctly point out that questions about the relation between vocabulary
and encyclopaedia are important, but although they mention that it is ‘the
topic of much current debate’ (2003: 471), not much of it is published. This
again indicates that most of the work in DM concentrates on generating the
forms without any concern for the semantic aspect.
In sum, in DM the study of lexicalization and meaning in relation to
word formation is not a cutting-edge research area but rather a backwa-
ter. There is no reason to assume that nothing could be said about it, but
researchers working in DM focus on other questions. The framework
directs attention away from the issues that are central in this volume. This
explains why none of the contributions in this volume takes DM as its point
of departure.
2.6 Other generative approaches

Whereas DM does not concentrate on questions of meaning and lexi-
calization, they are more central in a number of alternative approaches that
developed out of generative grammar. There are at least three such theories
that are of particular relevance here, each starting from a different perspec-
tive. Rochelle Lieber developed a theory starting from morphology while
Ray Jackendoff took semantics as his starting point and James Pustejovsky
the lexicon.
When Lieber (2004) addressed the semantics of word formation, she
observed that ‘[t]o [her] knowledge, there is no comprehensive treatment
of the semantics of word formation in the tradition of generative morphol-
ogy’ (2004: 1–2). In her own earlier work, e.g. Lieber (1983), she had
studied morphology with a set of questions similar to Selkirk’s (1982) in
mind, although arriving at different answers. Lieber (2004) takes the issue
of why there is a many-to-many mapping between form and meaning as the
central question about the meaning of word formation. She identifies four
different shapes of this mismatch (2004: 2): affixes can have more than one
meaning, meanings can be realized by more than one affix, meanings can
be realized in the absence of a correlating form (zero derivation) and affixes
can be required without contributing to the meaning.
The theory she develops takes as its starting point the assumption that
affixes are lexical items and that their meaning is encoded in the lexicon as
two parts, which she calls the skeleton and the body. The skeleton contains
‘all and only those aspects of meaning which have consequences for the
syntax’, whereas the body is ‘encyclopedic, holistic, nondecompositional,
not composed of primitives, and perhaps only partially formalizable’ (2004:
10). In word formation, the skeleton of an affix and a base, or of two bases
in the case of compounds, are combined to create a new lexical item.
Lieber (2004) uses a system of semantic categories and semantic features
to describe the skeleton. Her main semantic categories are substance/
thing/essence and situation. Although they are presented as semantic
categories, they correspond directly to syntactic categories. In fact, sub-
stance/thing/essence is ‘the notional correspondent of the syntactic
category Noun’ (2004: 24) and the disjunctive label is caused by the inho-
mogeneous meanings expressed by nouns. A situation may correspond to
a verb or an adjective. Semantic features are, for instance, [±material] and
[±dynamic]. They are used both as binary-valued and as privative features.
These features take arguments, as in (2), taken from Lieber (2004: 25):
(2) a.
chair [+material ([ ])]
b. leg [+material ([ ], [ ])]
The difference between chair and leg is that something cannot be a leg
without being a leg of something. Therefore (2b) has one more argument
than (2a). The argument in (2a) is the one corresponding to ‘x is a chair’,
formation
rules
phonological syntactic conceptual

structure structure structure
interface
rules
Figure 1.4 Parallel Architecture (after Jackendoff 2002: 125, by permission of

Oxford University Press)
used in model-theoretical semantics as a basis for the expression of quan-

tification. Lieber (2004) uses these variables in her account of word forma-
tion as the basis for combining the affix and the base in affixation and the
two bases in compounding. Thus in table leg, the second argument of leg is
coindexed with the first argument of table, which has the same structure as
chair in (2a).
Whereas Lieber is a morphologist who turned to semantics, Jackendoff
is a semanticist who turned to morphology. Jackendoff was an important
representative of the Chomskyan position in the so-called Linguistic Wars
(see section 1.2). Whereas the mainstream of generative linguistics concen-
trated on syntax after the end of this period, Jackendoff (1983, 1990) devel-
oped a theory of semantics. Jackendoff (2002) presents a fuller account
of the Parallel Architecture (PA), which, as argued by ten Hacken (2007:
245–69), constitutes a research programme separate from Chomsky’s, even
though a number of assumptions are shared. The PA is represented in
Figure 1.4.
The idea in PA is that the phonological, syntactic and conceptual rep-
resentations of an expression are generated by formation rules for each
type of structure and linked by interface rules. Compared to Saussure’s
system described in section 1.1, the PA in Figure 1.4 makes it possible to
include syntax in the language system instead of leaving it to the domain
of language use. Whereas Chomsky in his various architectures goes to the
opposite extreme and instead of excluding syntax takes it to be the only
component with a set of generative rules, Jackendoff proposes to make
phonological and conceptual structure equal to syntax in this respect. This
means that they are not derived from syntax by interpretive rules, but
generated by their own rule sets. The interface rules can further specify
a representation on the basis of information in one of the others, but their
main function is to link the three structures, i.e. establish them as represen-
tations of the same expression by coindexing corresponding components.
The position of the lexicon is not explicitly marked in Figure 1.4 but
Jackendoff (2002: 131) states that ‘the lexicon as a whole is to be regarded
as part of the interface components’. From this perspective, Jackendoff
then considers morphology as the combination of morphemes. Rather than
distinguishing inflection and derivation, Jackendoff (2002: 155) makes a
distinction between regular and irregular morphology. For irregular mor-
phology, the redundancy rules proposed by Jackendoff (1975) take on an
important role. Jackendoff (2009) elaborates this system for a number of
word formation rules and extends its scope to compounding. In Jackendoff
(2010), this system is developed further.
A third example of a generative approach that has a direct rel-
evance to our domain of meaning and lexicalization of word formation
is Pustejovsky’s (1995a) generative lexicon. Although it has generative in
its name, Pustejovsky’s theory makes much less contact with mainstream
generative linguistics than Lieber’s or Jackendoff’s. The reason is that
Pustejovsky presents a theory of lexical meaning with a scope not reaching
beyond the structure of the lexicon. Therefore it is in principle compatible
with any architecture that assumes an autonomous lexical component.
Pustejovsky’s central argument is that the lexicon should not be con-
ceived of as what he calls a Sense Enumerative Lexicon (SEL). A good
example of the problems an SEL raises is the contrast in (3).
(3) a. Last night, Anna started a new book.

b. Reluctantly, Barbara started the exam.
The verb start in (3) requires as its object a process. Neither a book nor an
exam is a process. Nevertheless, (3a) and (3b) can be readily interpreted. The
process in (3a) is probably ‘reading’, but if we know that Anna is a writer, we
can also interpret it as ‘writing’. In (3b), the process depends on Barbara’s
role. If she is a student it is ‘writing’, whereas if she is a teacher it is ‘marking’.
The question is how we know about these processes. If we assume that they
are part of the meaning of start, an SEL will have to list each of these mean-
ings as a separate entry for this verb. This is not attractive, because it would
lead to a very large number of entries for verbs such as start. Moreover, we
cannot be sure that any particular set of entries covers the full meaning of
the verb. Intuitively, the contrast in (3) depends on the object. Therefore
Pustejovsky proposes a theory in which the entries for start and book contain
information that in their interaction have the potential to generate the
meaning of the predicate in (3a), while the same entry for start can interact
with the entry for exam to produce the meaning of the predicate in (3b).
Although Pustejovsky (1995a) does not directly address word forma-

tion, problems similar to (3) occur both in derivation, as the polysemy
of affixes noted by Lieber (2004), and in compounding, as the discussion
in Jackendoff (2009) shows. As an example of a polysemous affix, Lieber
(2004: 2) gives -ize in randomize (‘making . . .’), containerize (‘putting in
. . .’), and anthropologize (‘performing . . .’). In compounding, contrasts
such as the one between water mill (‘. . . powered by . . .’) and paper mill
(‘. . . producing . . .’) are well known.
The solution Pustejovsky (1995a) proposes is that lexical entries
have a much more articulated structure than can be represented in an
SEL and that there are generative mechanisms using elements of this
structure to produce an appropriate meaning in context without having
to list all possible meanings exhaustively. In the case of (3), the so-called
qualia structure is central to the solution. The entries for book and exam
specify in their qualia structure what is typically done with these items.
In the case of book, the agentive quale, which specifies ‘[t]he manner in
which something is created’ (1995a: 97), has the information that it is
written and the telic quale, which ‘defines what the purpose or function
of a concept is’ (1995a: 99), specifies that it is meant to be read. In the
case of exam, the perspectives of writing and marking are encoded in
different qualia as well. The mechanism activating these qualia in (3)
is what Pustejovsky (1995a: 111) calls type coercion. As start requires a
process, book and exam have to become processes by using appropriate
components of their lexical entries. While Pustejovsky does not offer a
full theory of word formation, the qualia structure and type coercion
are particularly useful in accounting for the range of meanings found in
word formation.
3 OVERVIEW OF THE VOLUME
The contributions to the present volume give a multifaceted overview of

current research on the meaning and lexicalization of word formation. The
differences between the approaches appear in a number of dimensions.
First, the type of morphological phenomenon discussed varies. We have
taken this dimension as the leading principle for the organization of the
volume. Although we have not placed the contributions into sections, their
sequence is based on a gradual shift from one chapter to the next. A sec-
ondary criterion in this respect was the relative importance of data analysis
and theory, with the more theoretically oriented contributions at the start
of the volume. Apart from these dimensions, contributions also differ in
the theoretical framework adopted and in the languages their data are taken
from. In this section, we will briefly introduce each chapter, indicating its
theoretical framework and some of its main arguments.
The second chapter, by Pius ten Hacken, addresses the question of the
place of word formation in the model of grammar. He takes Jackendoff’s
(2002) PA as a starting point and concentrates on the question of what
Jackendoff calls ‘semiproductivity’, which he considers ‘one of the central
issues of linguistic theory’ (2010: 34). In English, denominal adjectives for
nouns in -ion and -ment can be formed by means of the suffixes -al (e.g.
national) and -ary (e.g. complimentary). Both processes are productive in
the sense that new items can be formed, but it cannot be predicted which
one of them is used in a particular case. They are productive in the sense
of what Corbin (1987: 176) calls ‘disponible’ (‘available’), but not fully
productive in the sense that syntactic rules are. This raises the problem of
how to account for this type of productivity. Ten Hacken argues that the
solution is to consider such processes from an onomasiological perspec-
tive in the sense that they provide names for a given concept. He claims
that semiproductivity provides evidence for a separate word formation
component, as opposed to Jackendoff’s (2010) approach.
In the next chapter, Claire Thomas discusses the notion of lexicalization.
In generative morphology, two senses of lexicalization occur. On one hand,
a lexicalized output of a word formation rule is listed in the lexicon; on the
other, lexicalization is the process of getting idiosyncratic, unpredictable
senses. Clearly, these two senses are related, but one does not imply the
other. Thomas discusses the interaction of these senses in Lieber’s (2004)
system of morphology, using settlement, in its process and its result senses,
as an example. A third sense of lexicalization is the one used by Jackendoff
in the exposition of his PA. The lexicalization of a Conceptual Structure
is the choice of lexical items to represent particular portions of it. Thomas
argues that some concepts from Pustejovsky’s (1995a) theory of the genera-
tive lexicon can be used to reconcile the different perspectives and express
the meanings of settlement as well as the relation between them in a way that
leads to a more convincing, integrated approach.
The fourth chapter, by Kaarina Pitkanen-Heikkilä, is devoted to termi-
nology. The question addressed is how names are chosen for concepts in
a specialist field, more specifically botanical terms in Finnish. In the nine-
teenth century, inspired by nationalist and romanticist currents, borrowing
from Latin or neighbouring languages was rejected in many countries as a
source of naming. In the case of Finnish botany, Elias Lönnrot (1802–84)
introduced a large number of terms that are still in use and exploit the
language-internal sources for new names available in Finnish. This study
of the choice of names is a good example of the onomasiological approach
to terminology. It also connects to the question of lexicalization in the sense
that Pitkanen-Heikkilä discusses to what extent the new terms are trans-
parent and have been accepted by the language community and specialists.
Staying within the field of terminology, Pius ten Hacken and Ewelina
Kwiatek consider terms from a different perspective. Instead of looking at
the naming devices used in one language, they consider the use of a single
naming device, nominal compounding, comparing two languages, English
and Polish. In order to compare compounding in the two languages,
they first propose a language- independent definition of compounding
and discuss its application to the systems of English and Polish. The
terminological domain they study is land surveying, a domain in which
conceptualization is in large part rooted in national traditions rather than
determined by international communication. In order to exploit the con-
trast between the parts with and without a strong international influence,
they also consider the domain of the global positioning system (GPS),
where technological innovations are central. Ten Hacken and Kwiatek’s
focus is on the question of how the different systems of compounding in
English and Polish affect the use of compounds of different types in naming
terminological concepts. This chapter therefore constitutes a transition
from terminology to compounding, to which the next chapters are devoted.
Maria Rosenberg also compares nominal compounding in two lan-
guages, Swedish and French, but her study has a somewhat different focus.
She uses a parallel corpus from the European Parliament and analyses the
translation equivalents of Swedish compounds in French. Again, this first
requires a procedure to identify compounds. As a next step, Rosenberg
classifies the compounds according to the semantic relation between the
two components of the compound. She discusses a number of alterna-
tive classification schemes and uses one based on the system proposed by
Jackendoff (2009). Her findings suggest that compounding in French is
much less flexible than in Swedish, so that translators often have to use
alternative expressions. Many of these alternatives require a more explicit
description of the relationship between the components than the com-
pounds in Swedish. In some cases, however, the French texts are less spe-
cific than the Swedish, for instance when Swedish bekämpningsmedelsrester
(‘pesticide residues’) is rendered in French as pesticides.
In the next two chapters, we turn away from the contrastive study of
nominal compounding to focus on German A+N compounds. In the first
of these, Barbara Schlücker explores the differences that mark the bound-
ary between compounds and phrasal combinations of adjectives and nouns,
e.g. the phrase alte Stadt (‘old city’) and the compound Altstadt (‘historic
city centre’). The central issue is the semantics of compounding, or, more
precisely, the question whether lexical modification differs systematically
in meaning from phrasal modification. Schlücker argues that in fact there is
a basic meaning difference, such that compounds generally have a classify-

ing meaning whereas phrases do not. Without strongly committing herself
to any particular framework, she uses a variety of predicate logic with a λ
operator as a formalism to express meanings.
The other chapter of this pair, by Martin Schäfer, considers the contri-
bution of the individual components of German A+N compounds to the
sentence they occur in. More specifically, he studies in which contexts and
under which conditions the component Stadt (‘town/city’) in Großstadt
(‘big city’) can serve as an antecedent to an empty pronoun, as in eine
kleine PRO (‘a small one’). The question of anaphoric islands of this type
was originally introduced in the domain of linguistic discussion from the
perspective of generative semantics, see Postal (1969), but later framed
mainly in a pragmatic context, e.g. by Ward et al. (1991). Schäfer applies
these insights to a set of German A+N compounds and corresponding
phrases in a large corpus and argues that the possibility of anaphoric refer-
ence depends in part on what Langacker (2000a: 93) calls entrenchment.
An important concept in lexical acquisition, entrenchment refers to the
formation of ‘a well-rehearsed routine’ through repetition.
The chapter by Alexandra Soares Rodrigues and Graça Rio-Torto
constitutes a transition between the chapters on compounding and those
on derivation, because it compares the processes by which derivations and
compounds get their meanings. They study Portuguese data in a frame-
work that is based on Jackendoff’s PA, but also incorporates insights from
Langacker’s Cognitive Grammar and Pustejovsky’s Generative Lexicon.
A central component of the framework is the notion of coindexation,
which according to Soares Rodrigues and Rio-Torto is a purely semantic
mechanism, operating on semantic features of the morphemes involved in
the formation of a word. They argue that syntax does not directly influence
the assignment of meaning and that the difference between the complex-
ity of the meaning of derivations and compounds can be explained by the
richer semantics of bases. Whereas derivations combine a single base with a
(semantically less complex) affix, the two bases in a compound can combine
their richer semantic specification in more elaborate ways.
The next two chapters are studies of deverbal nouns. Maria Bloch-
Trojnar studies English deverbal nouns in Beard’s (1995) framework of
Lexeme Morpheme Base Morphology. As in this framework the semantics
and the form of the result are produced by different sets of rules, a central
question is how meanings and forms are paired up. Although it is possible
to identify a number of factors involved in this mapping, e.g. ± Latinate,
these factors cannot fully predict the form of a nominalization derived from
a particular verb. Often meanings are much more specialized than can be
predicted on the basis of the verb and the derivation rule. Pustejovsky’s
(1995a) theory of the generative lexicon can be invoked to explain some of

these cases.
The second contribution on deverbal nouns is by Germana Olga
Civilleri. She takes Langacker’s Cognitive Grammar as her framework,
which means that the analysis concentrates on somewhat different proper-
ties. Her study is corpus-based, taking the Ancient Greek of the Homeric
poems as its basis, but she also draws parallels with English and Italian.
Here it is not so much the pairing of form and meaning, but the continuum
between compositional and lexicalized (in the sense of idiosyncratic)
meanings that constitutes the main focus. Civilleri develops an analysis in
which individual derivations are placed on this continuum. When moving
towards the lexicalized pole, we can also observe a gradual loss of typically
verbal features and a parallel increase in typically nominal features.
Following these two chapters on deverbal nouns, there are two chapters
on verb formation. The first of these, by Angeliki Efthymiou, concentrates
on the suffix -(ι)άζω [-(i)ázo] in Modern Greek. One of the problems with
this suffix is that it triggers a wide range of different meanings. The first
question Efthymiou addresses is whether we are dealing here with a single
suffix or with two separate suffixes. She concludes that there are two dif-
ferent suffixes, one of which is [–learned] and still productive. This suffix
also involves a phonological sequence that has a general role as a marker of
pejorative evaluation in Modern Greek. In order to assess the productiv-
ity of the suffix, Efthymiou compares it to the competing verb-forming
suffixes of Modern Greek. She concludes that the [–learned] nature of
-(ι)άζω as well as its pejorative connotation reduce its productivity,
although it is still among the more productive competitors.
In the second chapter on verb formation, Jessica Forse presents an
analysis of the English prefix en- and its counterparts en- and a- in French,
Spanish and Portuguese. She takes Jackendoff’s PA framework as a back-
ground, concentrating on the Conceptual Structure. In analysing the set of
relevant verbs in English, she arrives at a single basic Conceptual Structure
with a small number of variables. The instantiation of these variables leads
to seven more specific classes that can be grouped into three clusters. For
each of these seven classes, examples can also be found for en- in French,
Spanish and Portuguese. These three languages have a competing prefix
a-, which does not occur in English. In each of the three languages, a- has a
more limited distribution in the sense that examples can only be found for
three of the semantic classes distinguished for en-. Interestingly, these are
not the same three in French, Spanish and Portuguese.
The final chapter, by Renáta Panocová, is, like the preceding one, a
demonstration of how a particular framework can be applied to a par-
ticular word formation process. In this case, the framework is Horecký’s
onomasiological model of word formation and the word formation process

is Russian diminutivization. The contrast with the preceding chapter is a
good illustration of how the framework shapes the questions to be addressed
in word formation research. Horecký’s model requires the description of
the so-called complete semantic definition of a word formation process such
as diminutivization, irrespective of the formal realization. The complete
semantic definition is a classification of the base and the output in terms of
a set of semantic features. The result is a tree structure with, as its leaves,
classes of semantically similar words. Whereas most of the contributions
to this volume are broadly based on generative ideas, this chapter gives an
impression of the very different type of word formation research that is
carried out in the onomasiological tradition.
Together, the contributions to this volume give a wide-ranging exem-
plification of the type of research that is currently carried out in the study
of the meaning and lexicalization of word formation throughout Europe.
Some of the approaches represent widely held views, others are only
pursued in more specialized areas, but all have their role to play in further-
ing the study of word formation. We hope that collecting these examples of
different currents and traditions in one volume will provide inspiration to
researchers in this domain.
NOTES
1. ‘Language conceived of as a functional system.’ Our translation,

PtH&CT.
2. ‘synchronic description cannot categorically exclude the notion of evo-
lution’. Our translation, PtH&CT.
3. ‘Stylistic elements felt to be archaic, as well as the distinction between
productive and non- productive forms, are diachronic facts which
cannot be eliminated from the domain of linguistics even in a syn-
chronic approach.’ Our translation, PtH&CT.
4. ‘The word, considered from the point of view of its function, is the
result of linguistic naming activity.’ Our translation, PtH&CT.
5. Bloomfield intended his book as a general overview of the entire field, so
he also included chapters on areas such as historical linguistics and dia-
lectology, which do not have a strong position in the type of linguistics
typically associated with the American tradition.
6. In earlier work, Beard refers to his framework as Lexeme-Morpheme-
Based Morphology, e.g. Beard (1988).
chapter 2
Semiproductivity and the place of

word formation in grammar
Pius ten Hacken
I n this chapter I will address the question of how to account for what
appear to be different degrees of productivity of word formation
processes in the framework of Jackendoff’s (2002) Parallel Architecture
(PA). Section 1 explains Jackendoff’s approach to the notion of lexical
entry, which is quite different from the traditional one. Section 2 turns
to the analysis of productivity in PA and demonstrates why accounting
for degrees of productivity that are neither maximal (full productivity)
nor minimal (no new cases) is problematic in Jackendoff’s (2002, 2009)
approach. Section 3 proposes an alternative approach that requires a
separate word formation component but incorporates the semiproduc-
tivity of word formation rules in a more natural way. The conclusion
in section 4 is that it is not so much certain classes of word forma-
tion rules, but rather semiproductivity that should be regarded as an
epiphenomenon.
1 LEXICAL ENTRIES IN JACKENDOFF’S PARALLEL

ARCHITECTURE
Jackendoff’s Parallel Architecture is a model of language that is intended

to account for its integration into human cognition. Jackendoff (1983)
develops conceptual structure as a representation of meaning in the mind
that is compatible with the interpretation of vision and other sensual input,
motor instructions, inferences and language. Conceptual structure itself is
not part of language, but interacts with it. Jackendoff (1990) elaborates the
formalism of conceptual structure by analysing a variety of complex lin-
guistic phenomena. The interaction of conceptual structure with syntactic
and phonological structures is indicated in early sources, e.g. Jackendoff
semiproductivity and the place of word formation 29
(1983: 21), but the position of the lexicon is only fully developed later, in
particular in Jackendoff (1997). Jackendoff (2002) develops the argument
that linking rules and formation rules should all be considered as lexical
entries.
Central in PA is the idea that a linguistic expression such as (1) has
three correlated mental representations: phonological, syntactic and
conceptual.
(1) Anna has a new car.
The phonological structure of (1) indicates how the sentence can be pro-
nounced. The syntactic structure gives the constituents. These two struc-
tures are purely linguistic. The conceptual structure of (1) indicates the
meaning. If we are interested in the linguistic expression of this meaning,
only those aspects that can be mapped to syntactic and/or phonological
structures need to be encoded. The three representations of (1) are given
in (2). As the phonological representation is not central to the point to
be made here, I will generally use orthographic representations instead.
Jackendoff (2002: 6) gives a more detailed representation of a different
example.1
(2) a. Annai hasj ak newl carm

b. [S NPi [VP Vj [NP Detk [AP Al]p Nm]q ] ]r
c. [State BEPoss ([Thing CAR; [Property NEW]p ]q, [Place ATPoss [Thing ANNA]i
]) ]r
The details of the analysis in (2) are less important than the distribution of
information among the three representations and the links between them.
Each word in (2a) is linked to a syntactic category in (2b) by means of an
index. (2b) does not have the words themselves in it. Without the indices,
it is indistinguishable from the syntactic representations of the sentences
in (3).
(3) a. Swansea is a sunny city.

b. Bernard won a major prize.
The representation of meaning in (2c) is underspecified. Only the concepts

that correspond to linguistic expressions and the relations between them
are represented. In the syntactic structure (2b), the higher nodes tend to be
coindexed with the conceptual structure (2c) and the lower with the pho-
nological structure (2a). This is typical of the way the three r epresentations
are linked.
30 pius ten hacken
The information necessary for generating (2) is in the lexicon. This is

most obvious for words such as car. The lexical entry for car can be seen as
a triple of linked phonological (or orthographic), syntactic, and conceptual
structures, indicated in (4).
(4) a. car
b. N[+Count]
c. [Thing CAR]
It is essential to see how (2) and (4) are of a different nature. Whereas
(2) represents a sentence as written and understood, (4) represents infor-
mation that is stored in a speaker’s mind. Whereas (4) is part of linguistic
competence, (2) is an example of performance. The lexicon in PA consists
of items such as (4).
Jackendoff (1997) develops the point that the lexicon contains a lot more
than entries of the type in (4), extending it to various types of multi-word
units. Jackendoff (2002: 153–82) generalizes this argument to an even
wider range of items. The type of reasoning can be illustrated with the
examples in (5).
(5) a. Carol made a genuine attempt to put Swansea on the map.

b. It has finally stopped raining.
The example in (5a) is typical of the cases considered by Jackendoff (1997).

In order to understand the meaning of put Swansea on the map, the reader
will need a lexical entry along the lines of (6).2
(6) a. putp onq ther maps

b. [VP Vp NPn [PP Pq [NP Dr Ns] ] ]m
c. [CAUSE ([X]i, [INCH [BE ([Y]n, [AT [WELL-KNOWN] ]) ]) ]m
The lexical entry in (6) only specifies for each representation what infor-
mation the idiom contributes to a sentence that contains it and how this
information is linked across the representations. The fact that there has to
be an NP in the position of Swansea but that the choice of this NP is free
is expressed in (6a) by not specifying any word form, in (6b) by having
the NPn in the structure, and in (6c) by the variable [Y]n which indicates
the position in the conceptual interpretation and the link with the syntac-
tic structure. What (6) indicates is that lexical entries can have a rather
intricate structure in their representations.
In (5b), the expletive it is an example of the type of additional cases
considered by Jackendoff (2002). The lexical entry for it is (7).
(7) a. it
b. Det[3 sing neuter]
c. Ø
While it as used in (5b) has a form and syntactic features, it does not have a
meaning. The contribution to conceptual structure, as indicated in (7c), is
empty. If we can have a lexical entry with structure at each representation,
as indicated by (6), and a lexical entry need not contain information for
each representation, as indicated by (7), we can model formation rules for
each representation in the same way as (other) lexical entries. This point is
made by Jackendoff (2002: 178–82) and can be illustrated with the rule for
the object NPs in (1) and (3), given in (8).
(8) a. Ø
b. [NP Det AP N]
c. Ø
The entry in (8) does not specify any phonological or conceptual informa-
tion, because the only contribution it makes to the specification of the
representation of an expression is that it combines a Det, an AP, and an N
into an NP.3 This means that formation rules and linking rules are of the
same formal type and that they are both part of the lexicon.
On the basis of this reasoning, the lexicon in PA becomes a much more
encompassing component than in more traditional approaches. It contains
all the information a speaker needs in order to build up representations
of linguistic expressions such as (2). The starting point for the process of
building such representations can be a thought or it can be a visual (e.g.
orthographic) or acoustic input. Individual lexical entries encode content
words, function words, multi-word expressions and formation rules. They
constitute the core of linguistic competence.
2 REPRESENTING MORPHOLOGICAL PRODUCTIVITY
In a sense, the need to account for productivity can be seen as one of the
main forces driving the generative enterprise throughout its history. As
Chomsky formulates it, ‘the central fact to which any significant linguistic
theory must address itself is this: a mature speaker can produce a new sen-
tence of his language on the appropriate occasion, and other speakers can
understand it immediately’ (1964: 7). As this quotation indicates, however,
the main emphasis has not been on morphological productivity. When
discussing nominalizations, Chomsky (1970) argues for the ‘lexicalist
32 pius ten hacken
hypothesis’, which implies ‘that derived nominals will correspond to

base structures rather than transforms’ (1970: 193), i.e. are in the lexicon
rather than the result of syntactic rule application. In Chomsky’s (1970)
framework, there are two equivalent ways of formulating the lexicalist
hypothesis, given in (9).
(9) a. Morphology is in the lexicon.

b. Morphology and syntax are in separate components.
Given the interpretation of the lexicon in PA, (9a) and (9b) are no longer
equivalent. The reason is that in PA, syntactic rules have become lexical
entries. Therefore (9b) implies that morphology is not in the lexicon.
It is not surprising that (9a) should be the starting point for Jackendoff,
because it is the simpler hypothesis. Adopting (9b) requires the introduc-
tion of a new component, which by Occam’s razor should only be consid-
ered if it can be used to solve problems arising from the absence of such a
component. Jackendoff’s (1975) Full Entry Theory, as discussed in section
2.1, was proposed as an elaboration of Chomsky’s Lexicalist Hypothesis.
Jackendoff (2010: 35–9) presents this as the first publication that fore-
shadows PA. Against this background, section 2.2 presents the approach
Jackendoff (2002) develops for productive morphological rules and section
2.3 his approach to less than fully productive rules.
2.1 Full Entry Theory

The immediate question when we adopt (9a) is how much of morphology
is stored in the lexicon and how it is represented there. Jackendoff (1975)
argues for a Full Entry Theory with redundancy rules as a framework for
the lexicon. One of his arguments is based on pairs and triplets such as (10).
(10) a. aggression – aggressive – aggressor

b. aviation – aviator
c. retribution – retributive
In (10), the nouns in -ion may have associated adjectives in -ive and agent
nouns in -or, but the underlying root does not exist or is only a back forma-
tion. The idea of the Full Entry Theory is that each of the nouns and adjec-
tives in (10) has an entry, but that the cost of specifying these entries is less
than the cost of seven otherwise unrelated words. There is a redundancy
rule for -ion, which specifies common properties of nouns with this ending.
The entry for aggression will then refer to this redundancy rule so that the
information specified in the rule for -ion will not be counted in calculating
the cost of the entry aggression. As Jackendoff states it, the cost of storing
a word such as aggression is ‘the information that there is a word, plus the
cost of the root, plus the cost of referring to [the redundancy] rule’ (1975:
648). The cost of the root is basically the specification of what is common
to the triplet in (10a). For the cost of the reference to the redundancy rule,
Jackendoff (1975: 666) proposes (11) as a measure.
(11) The cost of referring to redundancy rule R in evaluating a lexical entry W is

IR,W × PR,W, where IR,W is the amount of information in W predicted by R and
PR,W is a number between 0 and 1 measuring the regularity of R in applying to
the derivation of W.
In calculating PR,W, Jackendoff (1975: 667) proposes to take as a basis the

potential uses of R as the sum of the actual uses (i.e. lexical pairs related
by R) and the non-uses (i.e. cases where only the input or only the output
of R exists). This means that the cost of referring to a redundancy rule is
related to the degree of its productivity. Fully productive rules will have a
cost of 0.
In this context, it is interesting to consider Corbin’s (1987: 176–8)
analysis of productivity into three different properties, régularité (‘regular-
ity’), disponibilité (‘availability’) and rentabilité (‘profitability’). Regularity
refers to the predictability of the form and meaning of the outcome of a
word formation process. Availability is the property of a rule that means
it can be applied to form new words. Profitability refers to the degree to
which a rule is actually used to form new words. Availability is a binary
property, whereas regularity and profitability can be expressed on a scale
from 0 to 1. The measures (11) refers to are regularity and profitability.
However, as Corbin (1987: 177) observes, availability is the more basic
property, underlying the other two. Only available rules are a part of the
speaker’s competence. Profitability can only be talked about for avail-
able rules. Regularity is a property of individual rule applications, not
of the rule as such, but a rule has to be available before it can be applied.
Jackendoff’s (11) fails to put availability in its proper place at the centre of
the concept of productivity.
2.2 Productive morphology in PA

Jackendoff (2002) does not distinguish inflection and word formation as dif-
ferent processes. He makes two distinctions that cut across both categories.
On one hand, he distinguishes affixation from compounding. He suggests
that compounding, with its potential to produce long strings of nouns,
is a relic of protolanguage, a version of language that preceded modern
34 pius ten hacken
language in evolutionary history and did not include a syntactic compo-

nent (2002: 249–50). This motivates a separate treatment, elaborated in
Jackendoff (2009; 2010: 413–51). Affixation includes both inflection and
derivation. On the other hand, Jackendoff distinguishes productive and
non-productive morphology, where ‘[p]roductive morphology is totally
regular’ (2002: 155).
Jackendoff’s treatment of fully productive rules can be illustrated with
the English nominal plural. The fact that this is generally considered a case
of inflection does not make it unsuitable to illustrate the procedure, because
inflection and derivation are treated in the same way. The entry for the
regular plural affix can be represented as in (12).4
(12) a. [Wd Wdn [Cl s]q ]m

b. [N Nn [Num Pl]q ]m
c. [PLURAL [X]n ]m
In (12a), s is represented as a clitic which attaches to a word so that the two

form a word together. In (12b), Number is represented as a constituent
specified as plural and attaching to a noun. In (12c), what is referred to by
the noun is the argument of the function PLURAL. Whereas (12) accounts
for cars and buses, it does not cover feet. For buses, we would need to assume
a rule inserting -e- in the relevant contexts.5 For feet there is a separate
entry (13).
(13) a. feetm
b. [N Nn [Num Pl]q ]m
c. [PLURAL [FOOT]n ]m
Both (12) and (13) are lexical entries. As such, they are in the same
category as the lexical entries for car, bus and foot. This means that in
some cases plural nouns are produced by combining two lexical entries,
whereas in other cases they are retrieved from the lexicon directly.
Clearly, the latter route is more efficient in terms of processing time.
This is probably at least a partial explanation for the non-occurrence
of *foots. Whenever we try to lexicalize a conceptual structure includ-
ing the part in (13c), the lexical entry in (13) is triggered immediately.
However, Jackendoff (2002: 50) also refers to morphological blocking as
a meta-constraint.
As an example of a productive derivational process, Jackendoff (2002:
155) mentions pre- as in pre-season. In his analysis, the affix pre- attaches to
nouns that denote time periods or events to form adjectives. Therefore a
lexical entry could be (14).
(14) a. [Wd [Cl pre]n Wdq ]m

b. [A Nq ]m
c. [BEFORE [Time/EventX]q ]m
An interesting observation if we compare (12) and (14) is that the affix in

(12) is represented in syntax, whereas the affix in (14) is not. This cor-
relates more or less with the common distinction between inflection and
derivation.
Whether the analysis in (14) is the best one is something that may be
doubted. The alleged adjective resulting from the application of (14) has
a very limited distribution, basically limited to the attributive modifier
position of nouns. In ten Hacken (1994), I propose a model of compound-
ing in which the non-head of a compound is not categorially specified. As
shown by Wiese (1996), compounds can have foreign-language expres-
sions or non-linguistic items (characters, gestures) in this position. In ten
Hacken (2003a, 2003b), I discuss the case of phrases and other linguistic
expressions that would not otherwise occur in the language. An example is
copulative compounds of the true dvandva type, as in attribute-value pair.
In English, structures of the type attribute-value cannot occur other than
as the non-head component of a compound. This distribution is similar to
that of pre-season. A detailed discussion of the two options would lead us
too far afield.
2.3 Less productive morphology in PA

Let us now turn to less productive morphological processes. Some of these
are entirely non-productive. An example of the prototypical behaviour of
such processes is English -th as illustrated in (15).
(15) a. warm warmth

b. long length
c. deep depth
The suffix -th makes nouns out of adjectives, but there are very few
adjective-noun pairs illustrating it. Moreover, most of them involve a vowel
change in the stem, as illustrated in (15b and c). It is not necessary to set
up a lexicon entry for -th, because all nouns resulting from this process are
in the lexicon as nouns. Nevertheless, speakers can optimize their mental
lexicon by noting the correlation between the pairs in (15) and a few other,
similar ones. The result of this optimization is a redundancy rule, a gener-
alization that reduces the cost of the information specified for each of the
individual entries. Jackendoff (2002: 165–7) discusses a similar situation
36 pius ten hacken
Table 2.1 Adjective formation with -al and -ary

Base noun Adjective in -al Adjective in -ary
function functional (functionary)
nation national
foundation foundational (foundationary)
caution cautionary
evolution (evolutional) evolutionary
development developmental
department departmental
compliment complimentary
rudiment (rudimental) rudimentary
fragment fragmental fragmentary
for patterns of irregular past tense formation involving the same vowel
changes. He proposes that ‘no rule [for such cases] is stored explicitly in
the speaker’s head’ (2002: 166). The regularity is only an epiphenomenon,
emerging implicitly from the way the brain stores information.
While such a solution may be attractive for (15), it is less so for the
patterns illustrated in Table 2.1. For nouns ending in -tion and in -ment,
there are two regular adjective forming processes, one with the suffix -al
and one with the suffix -ary. Table 2.1 gives examples of five nouns each
for -tion and -ment and indicates the corresponding adjectives. Forms in
brackets are variants found in a dictionary that do not belong to my active
vocabulary. In the case of fragment, the two adjectives have different mean-
ings. Fragmental can be said of geological deposits put together from frag-
mented rocks, whereas fragmentary can be said of a manuscript of which
only fragments survive.
The details of the data in Table 2.1 reflect a single speaker’s mental
lexicon, supplemented by dictionary information from Collins (1986).
However, for the purpose of the discussion here, there is no need to
‘verify’ the data, for instance by means of a large-scale questionnaire.
There is no higher authority to be found in order to determine whether
a particular adjective ‘really’ exists or is used in a particular sense. As stated
by Uriagereka (1998: 27), ‘English does not really exist.’ What exists as a
natural entity is each speaker’s competence and performance, but not a
named language such as English (see ten Hacken 2007: 274–81 for more
discussion).
The situation illustrated in Table 2.1 can be summarized as follows.
• There are open-ended sets of nouns ending in -tion and -ment.

• There are two competing processes of adjective formation, suffixation
with -al and with -ary.
• Both processes are productive in the sense that new words can be formed
(i.e. available).
• Not all nouns can combine with both suffixes.
• No obvious generalizations can be made on which of the suffixes a par-
ticular noun combines with.
• For some nouns, both adjectives exist with different meanings.
For Jackendoff, it is not immediately obvious how to account for the

data illustrated in Table 2.1. If -al and -ary are lexical entries of the type
illustrated in (12) and (14), many non-existing words are predicted. This
includes not only the words in brackets, but also, for instance, *develop-
mentary and *cautional. Basically, all slots in Table 2.1 are predicted to be
filled.
It may then seem better to treat the data in Table 2.1 in the same way
as (15). In that case, there are no lexical entries for -al and -ary, so that no
unexpected new words are predicted. However, this also means that only
the words already existing in the lexicon are covered. The suffixes -al and
-ary are not available for the formation of new entries.
Jackendoff (2010: 28–34) returns to the issue of semiproductivity and
comes to a different conclusion than in his 2002 book. He concludes:
Hence, apparently the only difference between a productive and a

semiproductive rule is that productive rules license one to go beyond
the listed instances without any special effort. This suggests that the
formal distinction between the two sorts of rule is specifically local-
ized in a diacritic on the variable: those in productive rules are marked
[+productive] and those in semiproductive rules are not. (Jackendoff
2010: 32)
In interpreting the scope of Jackendoff’s proposal, we should keep in mind

two points of his discussion. First, according to Jackendoff (2010: 29),
‘Semiproductivity is not confined to morphology.’ Jackendoff does not
want to separate the cases in Table 2.1 from syntactic constructions such
as the NPN construction discussed by Jackendoff (2008) and illustrated by
face to face. He also gives past tense vowel change in English as an example.
This means that instead of emergent generalizations, all cases of less than
full productivity are treated as rules encoded as lexical entries.
The introduction of the diacritic feature [+productive] implies that
without such a feature a lexical entry cannot be used to construct an expres-
sion that is not also a lexical entry. This means that regular lexical entries
for words, such as car, and formation rules, such as the rule combining
a determiner and a noun into a noun phrase, should have this feature.
38 pius ten hacken
One may wonder why Jackendoff does not propose a diacritic feature
[–productive] instead, which would have a more limited distribution. The
reason is probably that Jackendoff uses the feature as a privative one, i.e.
either it is there or it is not there, rather than a binary one with two values.
This means that adding the feature in the course of acquisition must be
based on positive evidence. A feature [–productive] would not be learnable,
however, without explicit correction.
Jackendoff’s solution with a diacritic feature distinguishing fully pro-
ductive and semiproductive rules is not particularly elegant. One of its
disadvantages is that it loses the distinction between unavailable processes
as in (15) and available but not fully productive processes as in Table 2.1. It
is therefore worth considering alternatives.
3 WORD FORMATION AS THE FORMATION OF LEXICAL

ENTRIES
In English, blackbird is used to refer to the species Turdus merula.

Accounting for the meaning of blackbird is less straightforward than it may
seem. Male blackbirds are black except for the yellow bill (see Figure 2.1)
whereas female blackbirds are brown.
There are other birds that are black, e.g. crows. In fact, crows (genus
Corvus) are more black than blackbirds, because specimens of both sexes
are entirely black. Yet it is only Turdus merula that is referred to as black-
Figure 2.1 Male blackbird Turdus merula

(© Mike Langman, rspb-images.com)
bird. We can explain this if we assume that motivation is not directed from
the name to the object but from the concept to the name. When speakers
of English wanted a name for the bird in Figure 2.1, they came up with
blackbird. At that point, the existence of crows did not play a role. Once
assigned to the species in Figure 2.1, the name was no longer available for
other species.
The observation that the motivation for new words comes from naming
needs is of course not new. Downing (1977) discusses it in the context of
compounding and proposes a distinction between compounds used as the
name of a concept and deictic compounds, which are used as ad hoc names
for individual objects in a particular context. Downing’s example for the
latter is apple-juice chair, used to identify a particular chair in a particular
context. Regular compounds are not deictic; they are used as a name for a
concept. This insight was also important in the onomasiological approach
to compounding, following on from Dokulil’s (1962) seminal work. Grzega
(2009) gives an overview of work in this tradition.
The general idea that the motivation of the meaning of a word as the
name of a concept is rather from the meaning to the form than the other
way round can also be illustrated with examples from derivation. The suffix
-ism is attached to a proper name to indicate a set of ideas associated with a
person by that name. However, whereas Jansen is one of the most common
Dutch surnames, Jansenism only refers to the set of ideas proposed by
Cornelis Jansen (1585–1638) about salvation and determinism. This can
hardly be seen as an example where the meaning gradually specialized,
finally to settle on this particular doctrine. Rather, speakers were looking
for a suitable name for this doctrine and found Jansenism appropriate
because Cornelis Jansen had defended it in his posthumously published
work Augustinus (1640).
It is worth considering what happens in communication when a new
word gets into use. The speaker or writer will have as their overriding
concern to be understood. Therefore a name should be related to what the
hearers or readers have in their mental lexicon. This is why new names are
normally related to existing words, as metaphoric or metonymic meaning
extensions, as input to word formation rules or as borrowings from another
language. Often, more than one form will be possible. The choice will be
governed by the intended impression on and the expected knowledge of
the hearers or readers, the intended relation to other lexicon entries, exist-
ing regularities, etc. These considerations are of a fundamentally different
nature to the ones that play a role in lexical retrieval, as used in the compo-
sitional formation of an expression. Once a name is established, the entry
will be activated immediately when the concept is triggered, as described
by Jackendoff (2002: 200–5).
40 pius ten hacken
At this point it is interesting to consider the data in Table 2.1 again. In

English, national is so strongly established that *nationary does not have a
chance to be formed to refer to the same concept. This is of course a matter
of individual speakers. For me, *functionary is as impossible as *nation-
ary, but apparently Collins (1986: 614) found enough evidence to give the
adjectival sense and describe it as ‘a less common word for functional
or official’.6 Of particular interest is the contrast between fragmental and
fragmentary. According to the OED (2011), fragmentary was attested in
1621, whereas fragmental did not appear until 1763 and was generally rarer.
When the specialized geological sense emerged in the nineteenth century,
fragmentary had a longer history of use. Therefore fragmental was a better
choice for this new sense. However, the distinction is not clear-cut for all
speakers. Collins (1986: 600) gives ‘another word for fragmentary’ as the
second sense of fragmental and ‘Also: fragmental’ as information about
fragmentary.
An essential consideration in the choice of a name is the opposite per-
spective, processing by the hearer or reader. In ten Hacken (2010b: 246–9)
I discuss the case of image converter. A reader coming across this word while
unfamiliar with it will have the information in (16) to work from.
(16) a.
Image converter is a word. If it is not part of my mental lexicon, it probably
designates a concept I am not familiar with.
b.
Image converter is a complex word. Its morphological analysis suggests a
range of possible meanings.
c. Image converter is used in a particular linguistic context.
Jackendoff (p.c.)7 objects to this analysis, suggesting the analysis in (17).
(17) a. ‘The morphological analysis, given the usual meaning of the -er morpheme,
suggests that image is to be understood as the patient of convert. Thus an
image converter is something that converts images.
b. So whatever readers can construe from context about the VP convert an
image, such as what the image is converted into, they can also ascribe to
image converter.
(17a) corresponds quite closely to (16b), although it is somewhat more

specific about the nature of the inferences in this particular case. However,
I think (16a) is an essential step. It makes the reader hypothesize a concept,
which (16b) suggests is a particular device or profession. The phrase convert
an image only evokes the process. This explains the difference between the
way the context is used in (16c) and in (17b).
In fact, an image converter is a device for producing an image on the
basis of invisible radiation, e.g. X- rays. Clearly, the precise meaning
cannot be determined on the basis of the word formation analysis alone.
The reader can anticipate, however, that the meaning will have this degree
of precision. Both writer and reader are intuitively aware of the nature of
the naming process. Therefore, depending on the context in (16c), there
may well be a sufficient degree of specification for communicative success.
Without an equivalent to (16a), however, no such precision is expected in
(17b).
The distinction between the word image converter and a corresponding
phrase can be seen by comparing their use in two scientific articles in (18).
(18) a. Before beginning the operation, it was verified that the entire upper or
lower leg was accessible with the image converter in both planes. (Müller et
al. 1998: 462)
b. The screen is used to convert X-ray energies into light, which is absorbed
by photodiodes integrated into the active-matrix flat-panel array and stored
as charge on the capacitance of the photodiodes. (Mail et al. 2007: 138)
Even with reduced context, (18a) clearly evokes the idea of a particular
concept corresponding to image converter. It is not just anything converting
from or into images, but a particular kind of thing with a specific function.
(18b) gives more context, but does not invoke any specific concept. Here,
only the conversion of X-ray energies to light is referred to, not a particular
device. The difference between the word image converter in (18a) and the
paraphrase in (18b) is that the former is associated with a lexical concep-
tual structure whereas the latter is not. The process in (16) modifies the
mental lexicon, i.e. competence. In interpreting (18b), information from
the mental lexicon is used to create a conceptual structure, but this remains
purely at the level of performance. Competence is not affected. The crucial
distinguishing step is (16a), which is immediately invoked when image
converter is encountered as a new word in (18a), but for which there is no
correlate in the interpretation of (18b).
The discussion in this section highlighted the difference between lexical
and compositional processing. The examples of blackbird, Jansenism and
image converter all illustrate how the naming function is implemented in
relation to grammar and the lexicon.
4 WORD FORMATION AND PRODUCTIVITY
Let us now consider how the preceding discussion affects Jackendoff’s

(2010) model of the lexicon in PA. A first observation is that this model
42 pius ten hacken
does not incorporate the analysis of productivity by Corbin (1987) suf-

ficiently. Of the three components of productivity, availability should be
central. This is a binary property of rules as encoded in the competence of
individual speakers. However, as the discussion of (11) showed, Jackendoff
puts more weight on profitability and regularity. These are derived con-
cepts because they address the application of a rule rather than its presence
or absence.
If we want to put availability back into the centre of productivity we have
to find a place in the architecture of grammar where rules can be present or
absent in the relevant sense. In ten Hacken (2010b), I propose to introduce
a separate word formation component in PA. Admittedly, this goes against
the programmatic tendency to reduce the entire account of linguistic
competence to a single lexical component. However, as the discussion in
section 3 showed, there is ample evidence for a distinction between on the
one hand a lexical component used to build up expressions and thoughts in
performance and on the other a concept naming component used to extend
the lexical component and change a speaker’s competence. Whereas syn-
tactic rules belong to the former, word formation rules belong to the latter.
The central difference is the concept naming nature of word formation, as
formulated in (16a). This maintains the lexicalist hypothesis in the sense
of (9b).
When we adopt a separate word formation component as part of the
architecture of grammar, we can account straightforwardly for the avail-
ability aspect of productivity. Available word formation rules exist in the
word formation component, whereas non-available rules do not.8 This
also means that we can restore the distinction between the data in (15)
and in Table 2.1, which is lost in Jackendoff’s (2010) revised proposal.
Non-productive rules such as -th suffixation are not in the word formation
component. For these cases, any generalization is of the emergent type, as
proposed by Jackendoff (2002, 2009). Semiproductive rules such as -al and
-ary suffixation exist in the word formation component.
This brings us to the question of profitability and regularity, i.e. how to
explain the semiproductive nature of such rules as -al and -ary suffixation.
In the model I propose here, the starting point for word formation is the
need to name a new concept. If a word formation rule is found that pro-
duces a form that is likely to be understood, it may be chosen. Regularity
depends therefore at least as much on the nature of concepts to be named
as on the word formation rule that is applied. Profitability is even more
of a derived phenomenon. It depends not only on the rule for which the
profitability is to be calculated, but also on the nature of concepts to be
named and the alternative naming procedures that are available. These
factors are usually not taken into account in measuring productivity, but
they are at least as important as communicative factors, normative pressure

and personal preferences. The interaction of all these factors yields profit-
ability. The relation between profitability and availability can be compared
to the one between performance and competence. Therefore I propose that
semiproductivity is an epiphenomenon.
NOTES
1. In different publications, Jackendoff uses various, slightly different

notational conventions. Here I will use a variety that I think is transpar-
ent and explicit enough for the expository needs in this chapter.
2. In order to increase readability, labels of functions and constituents as
illustrated in (2b) have not been specified in (6c). These labels are not
essential for the discussion, but should be included in a full Lexical
Conceptual Structure.
3. Following Culicover and Jackendoff (2005), I adopt here a flat struc-
ture of the noun phrase with N as the head, rather than Abney’s
(1987) DP hypothesis, which holds that the determiner is the head
of the noun phrase. It should be noted, however, that Culicover and
Jackendoff (2005: 78) argue against the DP analysis only by attacking
the Uniform Theta Assignment Hypothesis (UTAH). Abney (1987)
supports his hypothesis by a range of observations, such as that that
car can be replace by that but not by car. As far as I know, these
points have not been addressed systematically in the Simpler Syntax
Hypothesis.
4. The representation of each level in (12) follows the one for the regular
past in Jackendoff (2002: 160), but replaces tree structures with labelled
bracketing.
5. This could treated be along the lines of Two- Level Morphology,
proposed by Koskenniemi (1983). Sproat (1992: 145–70) gives a brief
overview of this formalism. Given its tier-based nature, it would match
the general approach of Jackendoff’s PA very well.
6. On the interpretation of the information in dictionaries in relation to the
mental lexicon and the text corpora used as a basis in their compilation,
see ten Hacken (2009b).
7. Jackendoff originally reacted to a pre-publication of ten Hacken (2010b)
in an email dated 19 August 2009, but provided the quotation in an
email of 7 August 2011. I added the division into a and b for conveni-
ence of reference.
8. As argued in ten Hacken (1994), I assume that inflection and word
formation are different. There is no space here to make this argument in
44 pius ten hacken
any detail, but one aspect that Jackendoff’s (2002) account of the English
past tense does not account for sufficiently in my view is the paradig-
matic nature of inflection as opposed to the case-by-case application of
word formation rules.
chapter 3
Lexicalization in Generative
Morphology and Conceptual
Structure
Claire Thomas
L exicalization has an important place in theories of word formation.

From a very general point of view, it refers to the integration of an
item into the lexicon, but what this actually implies is open to interpreta-
tion. The term has been used somewhat ambiguously, as Brinton and
Traugott (2005) note in their effort to integrate the various different per-
spectives. They identify two divergent conceptualizations, one diachronic
and the other synchronic. The diachronic one, found largely within
Generative Morphology, characterizes lexicalization as both integration
into the lexicon and meaning change. The synchronic perspective is found
in theories of lexical semantics, notably Jackendoff’s theory of Conceptual
Structure, and examines the way that systematic relations between seman-
tic constituents are expressed by lexical items. The view that Brinton and
Traugott arrive at is that
Lexicalization is the change whereby in certain linguistic contexts

speakers use a syntactic construction or word formation as a new
contentful form with formal and semantic properties that are not
completely derivable or predictable from the constituents of the con-
struction or word formation pattern. Over time there may be further
loss of internal constituency and the item may become more lexical.
(Brinton and Traugott 2005: 91–2)
While uniting divergent diachronic perspectives, this definition does not

incorporate the synchronic view of lexicalization. The aim of this chapter
is to show how these two seemingly divergent conceptualizations of lexi-
calization can be integrated. In section 1, I begin by outlining the way that
it is dealt with within the field of Generative Morphology, in particular in
Lieber’s (2004, 2009) theory. I then turn in section 2 to Jackendoff’s theory
46 claire thomas
of Conceptual Structure, showing how this can broaden our understanding

of lexicalization. However, the integration of Jackendoff’s theory does not
complete the picture, so in section 3 I turn to Pustejovsky’s Generative
Lexicon to show how this theory can fill in the missing link. Finally, in
section 4 the three perspectives are integrated to form a more complete
view of lexicalization.
1 LEXICALIZATION IN GENERATIVE MORPHOLOGY
Lexicalization within the field of Generative Morphology can be viewed as

a largely diachronic process, examining changes to a formation over time.
Within this, there are two broad and related conceptualizations. The first
of these characterizes lexicalization as permanent incorporation into the
lexicon (e.g. Plag 2003; Booij 2005). What this entails is a shift away from
online production towards storage in the lexicon as a lexical item in its own
right. In this sense, lexicalization is an inevitable step if a new formation
is to become established: if the process does not take place the derivation
remains a nonce word.
Often associated with this permanent incorporation is the change or
specialization of meaning that many (e.g. Anderson 1992; Kastovsky 1982;
Lieber 2004) refer to in discussing lexicalization. This can usually be seen
as a shift away from compositional, transparent meaning and towards a
more idiosyncratic meaning (Plag 2003). This can in turn be taken as an
indication that a new word has been incorporated into the lexicon:
Having idiosyncratic properties thus implies for a word that it has to

be listed, but the inverse is not necessarily true: a complex word that
is listed may have fully predictable properties, and may be listed only
because it is an established word. (Booij 2005: 17–18)
How ‘idiosyncratic’ the meaning of a word has to be in order for it to count

as lexicalized is not obvious. Bauer (2003: 272) states that ‘lexicalization
ignores morphological structure. Typically, in lexicalization, the internal
structure of a word is lost sight of’, which seems to imply that lexicalization
is the endpoint of a process of shifting away from compositionality. For
others, however, the entire cline is seen as part of lexicalization. Kastovsky,
for example, takes the view that lexicalization is ‘a gradual phenomenon,
both diachronically and synchronically’ (1982: 205). The diachronic scale is
in line with the explanations pursued above, where a particular derivation
or compound progresses from entire transparency to complete idiosyn-
crasy, as seen in the case of compounds such as hogwash, whose meaning
lexicalization 47
has shifted from ‘slops fed to pigs’ to ‘nonsense’. From a synchronic per-
spective, at any particular point in time, there will be different lexical items
at different points on the lexicalization scale. So, for example, there are
derived words in Modern English whose meanings are entirely predictable
according to their form (e.g. unhappy), derived words whose meanings are
entirely non-compositional (e.g. considerable) and derived words which lie
at some point along this scale. Here we might include deverbal nouns such
as those in (1).
(1) a. Access to attorneys might compromise Padilla’s ongoing interrogation by the

military.
b. Remember it’s an interview, not an interrogation.
c. Condensation of water vapour on cold surfaces is a daily experience to all of
us.
d. Within two hours, condensation on the walls had disappeared.
e. Political involvement in the appointment of judges would mean a more
representative judiciary.
f. This new appointment will complement and significantly strengthen the
Partners’ highly regarded service.
g. Indian resistance, sectionalism and racism forced some pauses in the process
of westward settlement.
h. The remains of an Iron Age settlement have been unearthed by
archaeologists.
In (1a, c, e and g), the noun refers to the process of the action denoted by
the base verb. In contrast (1b, d, f and h), denote the result of this action.
The group of suffixes which includes -ment and -ation is often classified as
deriving abstract nouns from verbs (Lieber 2004), and the process-result
alternation is well-attested, but (1d and h) are examples of concrete usage.
The fact that this is neither entirely compositional nor entirely unpredict-
able seems to indicate the position of such a derivation somewhere in the
middle of the more extreme cases mentioned above.
The perspectives outlined here seem to imply that lexicalization in
terms of meaning change is equivalent to idiosyncrasy. This would appear
to be what Lieber has in mind when she discusses lexicalization in the
context of her theory of morphology and lexical semantics. As this theory
is of some significance in the discussion of the lexicalization of derivations,
it will be outlined here. According to Lieber (2004, 2009), the semantics
of a lexical item (which includes affixes) is made up of two parts, which
she calls the skeleton and the body. The skeleton contains ‘all and only
those aspects of meaning which have consequences for the syntax’ (Lieber
2004: 10). The skeleton is made up of a set of cross-categorial, e quipollent
48 claire thomas
and privative features which are used in the characterisation of nine

semantic categories. For example, the four semantic categories which
make up the ontological class of substances/things/essences, which cor-
respond to the syntactic category of nouns, are composed of the features
shown in (2).
(2) a. simple, concrete substances/things/essences: [+material] (man)

b. simple, abstract substances/things/essences: [-material] (morning)
c. concrete, processual substances/things/essences: [+material, dynamic]
(author)
d. abstract, processual substances/things/essences: [-material, dynamic]
(war)
For Lieber, word formation is a mechanism for extending the simplex

lexicon and she assumes that usually the addition of an affix adds a func-
tion to the lexical item corresponding to the semantic category of that affix
(2004: 37). For example, the skeleton of the nominalising suffixes -ation
and -ment, among others, is illustrated in (3).
(3) [-material, dynamic ([ ], <base>)]
The feature [-material] is associated with abstract nouns, and the feature
[dynamic] indicates that this is a noun whose meaning involves the unfold-
ing of some kind of process. The way that this interacts with the semantics
of the base verb is illustrated in (4) for the verb settle.
(4) a. [+dynamic, +IEPS] (settle)

b. [-material, dynamic ([ ], [+dynamic, +IEPS])] (settlement)
For verbal meanings, Lieber uses the feature [dynamic] in an equipollent

way. Its positive value in (4a) shows that this is an event or a Process1
rather than a state. Meanwhile, the positive value for [IEPS] (which
stands for Inferable Eventual Position or State) indicates that there is a
progression from one position or state to another. In the case of settle, this
would be the transition from the state of not being settled to the state of
being settled. The features added by the affix -ment in (4b) then show that
settlement is an abstract noun denoting the process of settling. As discussed
above, however, one of the features of nominalizations of this kind is that
they tend to be polysemous between a process and a result reading, which
presents a problem for Lieber’s framework given that (4b) encodes the
process reading only. In order to avoid this problem, she suggests that the
binary nature of the feature [dynamic] might be relevant here, although it
lexicalization 49
is not for other nouns. The interpretation of the noun is often dependent
on the context, as shown in (5).
(5) a. The frequent expression of one’s feelings [+dynamic]

b. An old expression [-dynamic]
Lieber argues that the complex event reading of (5a) and the result
reading of (5b) must be encoded at a higher level. This may be permissible
given that in nouns, the feature [dynamic] does not already have a value.
Less easy to explain is the fact that in a number of cases the result reading
is a concrete noun, as shown by (1h), repeated in (6).
(6) The remains of an Iron Age settlement have been unearthed by archaeologists.
(6) refers to the concrete, rather than abstract result, of the process of
settling. This requires the feature [material] to have a positive value
rather than a negative one, so in line with Lieber’s (2004) argument about
[dynamic], the binary nature of [material] would have to be determined
at a higher level. This is problematic, first because it does not work for
nouns whose meaning in this respect is not context-dependent, i.e. where
the lexical meaning of the noun is either [+material] or [-material], and
second because allowing the binary nature of the feature to be determined
at a higher level in these cases would mean that derived nouns had a differ-
ent status to simplex nouns. As mentioned above, word formation extends
the simplex lexicon: derived nouns and simplex nouns should therefore
have the same status. Finally, Lieber claims that each affix has a unified
skeleton and therefore cannot belong to more than one semantic category.
Allowing the binary nature of [dynamic] to be encoded at a higher level
does not change the fact that the affix belongs to the semantic category in
(2d), but the binary nature of [material] represents the difference between
the semantic categories in (2c) and (2d). This means that the affix would
have to belong to both categories, which is incompatible with Lieber’s
approach.
This seems to indicate that the encoding of the alternation between
abstract and concrete deverbal nouns of this kind cannot be carried out by
the skeleton. One solution is to call this ‘lexicalization’ in the sense of the
development of an idiosyncratic meaning. This is how Lieber characterizes
lexicalization, as shown in (7).
(7) Both derived words and compounds may, however, over time, develop
substantial and distinctive bodies as a function of their lexicalization.
Lexicalization [. . .] proceeds on an item by item basis, thus allowing a wide
50 claire thomas
range of meanings to exist in items formed by the same process of derivation

and compounding. (Lieber 2004: 10–11)
For Lieber, lexicalization is a function of the body, which as noted above is

the second aspect of the meaning of a lexical item. Lieber (2009) explains
that the body is in turn made up of two separate components. The first
consists of ‘universal semantic features that are not syntactically active
in the language in question’ (Lieber 2009: 83). The second consists of
encyclopaedic ‘assorted bits of information’ and varies between speakers.
If we are to treat the problem of concrete result readings in deverbal nouns
as lexicalization, we should be able to encode it here. However, the first
component does not permit this, as it has already been established that the
alternation is between the two values of [material], a feature that Lieber
claims is encoded in the skeleton (in English at least), meaning that it
cannot also be encoded in the body. The second component is also not suit-
able, as this is meant to encode the more idiosyncratic aspects of meaning,
and the alternation seen here is found in many different cases. The problem
posed by these nominalizations for Lieber’s account then is that they are
neither the kind of systematic polysemy that her skeleton is able to encode,
nor the more ‘idiosyncratic’ form of lexicalization that is attributed to the
body. It would seem that phenomena that fall somewhere in the middle of
Kastovsky’s scale, ‘where some relation to the rule by which it was formed
remains present, and it is at least partly interpreted on the basis of this
rule’ (1982: 205), cannot easily be encoded in Lieber’s framework. If the
specialization of meaning associated with lexicalization in cases like this is
not entirely idiosyncratic, then how should it be accounted for?
One solution lies in what Brinton and Traugott (2005) characterize as a
synchronic view of lexicalization, which differs from the synchronic per-
spective discussed by Kastovsky (1982). They treat this as unrelated to the
integrated view of lexicalization that they build up, but I will argue that it
should not be disregarded for two reasons. The first comes from a practical
standpoint. Theories of lexical semantics such as Jackendoff’s theory of
Conceptual Structure have been used increasingly in the characterization
of the semantics of word formation, as will be discussed in the next section.
Given the usefulness of this approach, it is worthwhile seeing what else it
can contribute to our understanding of lexicalization.
The second reason concerns the lexicon. Theories of lexical semantics
can tell us a great deal about the nature of the lexicon, and given that
lexicalization from a very general point of view involves incorporation into
the lexicon, it is worth examining the insights contributed by the theory.
The following section will focus on Jackendoff’s view of lexicalization and
examine how it might inform the perspective from Generative Morphology.
lexicalization 51
2 LEXICALIZATION IN THE PARALLEL ARCHITECTURE
Jackendoff’s theory of Conceptual Structure has been used by several dif-

ferent theorists as a basis for investigating the semantics of word formation.
Jackendoff himself employed the framework to characterize English N+N
compounds (Jackendoff 2009). While this has limited applications to the
rest of morphology if we assume with Jackendoff that compounding differs
from other morphology in many respects (2009: 114), it nevertheless dem-
onstrates how the framework can be extended to cover it, which is what
others have done. Plag (1998), for example, uses Conceptual Structure
to characterize the semantics of -ize derivation. Still others, in particular
Lieber (2004), have used Jackendoff’s framework along with other theories
of decompositional semantics as a starting point from which to develop
their own theories.
In the current context, the clear compatibility of the framework
with Generative Morphology leads to the question of how it might be
relevant to the discussion of lexicalization. There is very little overt
discussion of the process within Jackendoff’s work, but an investigation
of the assumptions that underlie it allows the elaboration of a picture of
lexicalization within the Parallel Architecture. This reveals that there are
two strands relating to lexicalization, one of which corresponds closely
to the diachronic view of lexicalization within Generative Morphology,
but gives a somewhat more complete picture. The other adds a new
dimension to the discussion by focusing on an area that is not normally
dealt with in Generative Morphology as part of the phenomenon of
lexicalization: the perspective that Brinton and Traugott (2005) identify
as synchronic.
In order to investigate these two strands, it is necessary first to take a step
back from conceptual structure and examine Jackendoff’s view of language
as a whole and the nature of the lexicon. Figure 3.1 illustrates the Parallel
Architecture, a tripartite system consisting of three independent genera-
tive systems which are linked by interface rules: Phonological Structure,
Syntactic Structure and Conceptual Structure.
Within this framework, a lexical item should be viewed as a linkage of
elements of Conceptual Structure, Syntactic Structure and Phonological
Structure which is stored in the long-term memory (Jackendoff 2002,
2009, 2010). As Jackendoff explains, ‘each lexical entry can be thought of
as a small-scale interface rule’ (2010: 15), meaning that the lexicon itself is
made up of a collection of interface rules. The example in (8) shows how a
typical lexical item can be broken down into pieces of phonology, syntax
and semantics, and how these elements are linked.
52 claire thomas
Phonological Syntactic Conceptual

formation rules formation rules formation rules

structures structures structures
Interfaces to PS-SS SS-CS Interfaces to

hearing and interface rules interface rules perception and
vocalization action
PS-CS
interface rules
Figure 3.1 Parallel Architecture (Jackendoff 2002: 125, by permission of Oxford

University Press)
Wdi
(8)
N Thing CAT
-plural i
i
k a t
From this arises the question of how far we can extend this picture of a
typical lexical item. Jackendoff’s view of the lexicon is a broad one, and
the status of ‘lexical item’ is not confined to words. Rather ‘words, idioms,
rules of grammar and regular affixes are all stated in a common format,
namely as pieces of stored structure’ (2010: 19). Particularly important for
the purpose here is the status of affixes, as this allows us to encode an affix
in exactly the same way as a word. Jackendoff (2009: 118) gives the follow-
ing as his analysis of -er.
(9) a. V1-er2 = PERSON2α; [F1, (α, . . .)] (agentive -er)

b. V1-er2 = OBJECT2α; [F1, (INDEF, . . . WITH α)] (instrument -er)
(9) shows the formalism used in Jackendoff (2009), where = is now used
to show the interface between syntax and semantics and F stands for ‘a
variable function of some unspecified number of variables’ (2009: 118)
The two different structures indicate the two different meanings that the
suffix -er can contribute. Note, however, that this does not show a link
between these two different meanings, a point that will be important
later on.
Turning back to the idea of each lexical item comprising several linked
structures, this is where we can begin to talk about lexicalization in one
lexicalization 53
of its guises. Suppose we take a word and an affix and combine them to
make a new word. When this word is stored permanently in the lexicon,
it is possible to speak of it as having been lexicalized. The view thus far
is identical to that within Generative Morphology. What Jackendoff’s
framework allows us to do is be more specific about what this actually
means. Lexicalization here refers to the process by which a long-term
memory linkage between the phonology, semantics and syntax of the
newly derived word is formed. This, along with the associated meaning
change, is the process Jackendoff refers to when he discusses the princi-
ples for forming new compounds and claims that ‘[l]exicalized compounds
are for the most part specialized instantiations of these principles’ (2009:
114). In this way, one view of lexicalization in the Parallel Architecture
gives a more detailed description of the process described in Generative
Morphology.
However, this is not the whole story: the framework of the Parallel
Architecture and in particular Conceptual Structure adds a further per-
spective to the discussion of lexicalization. To see this, it is necessary to
focus first on Conceptual Structure as independent of language. Conceptual
Structure is universal and language-independent; where differences in lan-
guages occur this can be attributed to ‘different strategies in how they typi-
cally bundle up conceptual elements into lexical items’ (Jackendoff 2010:
24), or in other words how they lexicalize them.
An oft cited example of these cross-linguistic differences is that discussed
by Talmy (1985). This is the tendency displayed in Romance languages to
lexicalize a path and motion together, in contrast to English where there is
a tendency to lexicalize manner and motion together, as can be seen in the
way that the conceptual structure in (10) is expressed.
(10) GO ([ThingBOY], [PathTO [PlaceIN [ThingHOUSE]]])
[BY [MOVE ([ThingBOY])]]

Event
(10) shows a universal Conceptual Structure, where the top line shows that
a Thing traversed a Path into a Place and the second shows the manner
of motion. (11) shows how (10) is lexicalized differently in English and
French.
(11) a. The boy runs into the house.

b. Le garçon entre dans la maison en courant.
‘The boy enters the house running.’
54 claire thomas
In (11a) the motion and the manner of motion, i.e. GO and BY + MOVE,
are lexicalized by the verb run, while in (11b) the motion, i.e. GO + TO,
is lexicalized by entre and the manner of motion, i.e. BY + MOVE, by en
courant. Jackendoff (2010: 24) also discusses the fact that while some lan-
guages require certain elements of Conceptual Structure to be lexicalized,
others do not. Jackendoff calls these ‘patterns of lexicalization’ (2002: 292),
and they demonstrate an earlier stage in the lexicalization process: which
elements of Conceptual Structure are selected to become a part of the long-
term memory linkage that makes up a lexical item.
Again, Jackendoff’s framework allows this to be extended to cover
morphology. It makes it possible to explore which elements of Conceptual
Structure are expressed by which affixes. For example, Lieber (2004) iden-
tifies the various elements of meaning lexicalized by English affixes, and
shows that some are not expressed by affixes at all but rather by other types
of word formation such as conversion. If elements of Conceptual Structure
are not lexicalized in this sense, i.e. ‘bundled up’ by a particular affix, they
can never be lexicalized in the other sense, i.e. stored permanently in the
lexicon as part of the tripartite structure of that particular lexical item.
To sum up so far, three closely related strands in the conceptualiza-
tion of lexicalization have been identified in Jackendoff’s framework and
Generative Morphology. Firstly, lexicalization is the selection of elements
of CS expressed by a lexical item. Secondly, the term can be used to refer
to the permanent storage of these elements as part of a small-scale interface
rule between Conceptual Structure, Syntactic Structure and Phonological
Structure (a lexical item). Finally, lexicalization can refer to the specializa-
tion in meaning that often occurs when this permanent storage takes place.
However, it could be argued that there is still a missing link. Taken
together, Generative Morphology and the Parallel Architecture undoubt-
edly build up a fairly comprehensive picture of lexicalization, but they do
not really account for the link between meanings of a lexicalized item. This
was illustrated by (1), partially repeated here as (12).
(12) a. Political involvement in the appointment of judges would mean a more

representative judiciary.
b. This new appointment will complement and significantly strengthen the
Partners’ highly regarded service.
c. Indian resistance, sectionalism and racism forced some pauses in the process
of westward settlement.
d. The remains of an Iron Age settlement have been unearthed by archaeologists.
In Generative Morphology, any shifts in meaning are treated as idiosyncra-

sies, while Jackendoff’s framework requires these two meanings of -ment to
lexicalization 55
be treated as separate lexical entries, as -er was in (9), whether or not the
result meaning is abstract as in (12b) or concrete as in (12d). Neither of
these adequately captures the systematic nature of the way that alternations
like these are lexicalized. To do this it is necessary to look elsewhere.
3 PUSTEJOVSKY’S GENERATIVE LEXICON
The problem we are facing at this point concerns the question of how to
represent the relationship between the compositional meaning and the lexi-
calized meaning of a derived word which essentially coexist. Jackendoff’s
(2009) theory allows the description of the semantics of a derived word, but
is not ideally designed to account for the more flexible picture of meaning
that emerges. (13) illustrates a regular alternation between two different
senses of a derived word.
(13) a. The settlement of the first immigrants took place in the 1600s.
b. There was a large settlement on top of the hill.
(13a) uses the derived noun settlement in its compositional sense, meaning
‘act of settling’, while in (13b) it refers to ‘result of settling’, further
complicated (at least for Lieber’s framework) by the fact that the result
meaning is concrete. It is important to note that the existence of the lexi-
calized meaning does not preclude the continued use of the compositional
meaning. Obvious though this point may seem, it is crucial: both meanings
must be available in the lexicon, and there is clearly a relationship between
them. How should this relationship be accounted for?
What we are dealing with here is a case of polysemy. Polysemy has been
discussed in relation to word-formation before, as in the case of Lieber’s
exploration of what she calls the ‘polysemy question’ (2004: 2). However,
Lieber (2004) dealt with the polysemy of certain affixes, and the question
here concerns the polysemy of the derivation itself. (14) is an example
of the process-result alternation which is common across many deverbal
nominalizations independent of the affix; another example is given in (14).
(14) a. The construction of the Taj Mahal was entrusted to a board of architects
under imperial supervision.
b. The Millennium Dome is a huge construction being built on derelict land.
It is not appropriate to treat these alternations as separate lexical entries

because of the link between the meanings: The process referred to in
(14a) is what yields the result referred to in (14b). A framework is needed
56 claire thomas
therefore that can account for the polysemy demonstrated here. This is one
of the problems that Pustejovsky (1995a) takes as his starting point for his
account of the lexicon, so this is where I turn to complete this exploration
of lexicalization.
The alternation illustrated above is a case of what Pustejovsky calls
logical polysemy, ‘where there is no change in lexical category, and the
multiple senses of the word have overlapping, dependent, or shared mean-
ings’ (1995: 28). Traditionally, in cases like this, each of the different senses
would have been listed as an individual entry in the lexicon, as in (15).
(15) Settlement1: act of colonising an area

Settlement2: place where settlers live
Pustejovsky calls a lexicon which takes this approach a Sense Enumerative

Lexicon (SEL) and identifies several problems with it, the most relevant
here being that word definitions are not atomic and distinct, but rather
overlap and refer to one another. In (13a) settlement conveys a process; in
(13b) it conveys a concrete result. An SEL would require these different
senses to be encoded as distinct lexical entries, just like Jackendoff’s entries
for -er in (9). This means that the systematic relationship between these
two senses cannot be expressed. While one conveys a process and the other
a result, both meanings are intrinsically related to the meaning of the verb
settle and to list them as distinct senses is to miss this relationship.
Pustejovsky’s solution to this problem as well as to several others con-
cerning SELs is to propose an entirely different system, where each lexical
entry has four levels of representation. These are as follows:
• Event Structure: definition of the event type of a lexical item (STATE,

PROCESS, TRANSITION).
• Argument Structure: specification of the number and type of logical
arguments and how they are realised syntactically.
• Qualia Structure: modes of explanation, composed of FORMAL,
CONSTITUTIVE, TELIC and AGENTIVE roles.
• Lexical Inheritance Structure: identification of how a lexical s tructure
is related to other structures.
The first three of these can be used to build up a semantic representation of

the lexical item. Figure 3.2 shows the semantic representation of settlement.
Each of the levels in Figure 3.2 plays an integral part in the meaning
of the word. The Qualia Structure defines the central meaning, which is
then constrained by the Event Structure and Argument Structure. The
representation in Figure 3.2 gives us the following information. The
lexicalization 57
settlement
E1 = process
EVENTSTR = E2 = result
RESTR = <α
ARG1 = animate_ind
ARGSTR = 1
FORMAL = physobj
ARG2 = 2 physobj
FORMAL = entity
FORMAL = settle_result (e2, 2 )

QUALIA =
AGENTIVE = settle_act (e1, 1 , 2 )
Figure 3.2 Representation of settlement
Event Structure shows that there are two events, a process and a result.
There is also a restriction on the ordering of the events: <α means that the
process takes place before the result. The Argument Structure includes
two arguments, one of which is the animate individual who carries out
the process, and the other the physical object, i.e. the place that is settled.
Both the Event Structure and the Argument Structure are involved in the
characterization of the Qualia Structure. The Formal quale ‘distinguishes
the object within a larger domain’ (Pustejovsky 1995a: 85) in terms of
values such as orientation, magnitude and colour. Here, it shows that a
settlement is the concrete result of settling, incorporating event E2 and
argument ARG2. The integration of this aspect allows us to encode the
concrete meaning of some deverbal nouns that proved difficult in Lieber’s
framework. The Agentive Quale shows how an entity comes about, in this
case through the process of settling which is carried out by Argument 1 on
Argument 2.
Pustejovsky’s account is thus very different from an SEL. Rather than a
list of atomic definitions, word meanings are complex, with several levels of
representation that interact with each other. These levels of representation
work in conjunction with a set of three Generative Mechanisms.
• Type Coercion: a semantic operation that converts an argument to the

type which is expected by a function, where it would otherwise result in
a type error.
• Co-composition: allows a qualia structure for a phrase which reflects
aspects of both constituents.
58 claire thomas
• Selective Binding: a semantic device which applies an adjective to a

particular quale of the noun it is modifying.
Together, the semantic levels of representation and the generative mecha-

nisms build up a much more complex picture of meaning.
In Figure 3.2, it was shown that the lexical item settlement comprises two
events: a process and a result. Pustejovsky (1995a) claims that where there
are two simple types like these, it is possible for them to combine to form a
complex type, or dotted type. These are relevant to non-complex nouns as
well as complex ones.
(16) a. The door slammed.

b. The man came through the door.
c. The man came through the door that was hanging off its hinges.
In (16a) door refers to the physical object; in (16b) it refers to the aperture.
It is possible for both these senses to unite as a dotted type in (16c), where
door makes reference to both the physical object and the aperture. These
types cluster together in a meta-entry called a lexical conceptual paradigm
(lcp).
(17) physobj.aperture = {physobj.aperture, physobj, aperture}
What this means is that in order for a lexical item to refer to a complex or
dotted type (physobj.aperture), it must also be able to refer to both of the
simple types that make up the dotted type (physobj and aperture). The
reading in (16c) would not be possible if door did not also have the senses
illustrated in (16a) and (16b). The same clustering of types into an lcp can
be observed in the process-result readings of nominalizations discussed
above, as shown in Pustejovsky’s (1995a: 170) example, reproduced here
as (18).
(18) a. The house’s construction was finished in two months. (process.result)

b. The construction was interrupted during the rains. (process)
c. The construction is standing on the next street. (result)
The semantic representation in Figure 3.2 contained three of the four

levels of representation identified by Pustejovsky. The fourth level,
lexical inheritance structure, is concerned with the way that lexical items
are ‘globally related to other concepts in the lexicon’ (Pustejovsky 1991:
419). This is achieved through two inheritance mechanisms, as laid out in
Pustejovsky (1991). The first of these, fixed inheritance, concerns the static
lexicalization 59
relations that exist between lexical items, such as hyponymy. For example,
(19) shows the relationship between the words bird and robin.
(19) a. bird (x)

CONST = {beak, feathers . . .}
FORMAL = animal (x)
TELIC = fly (x)
b. robin (x)
CONST = {red feathers . . .}
FORMAL = bird (x)
Inheritance in (19) takes place through the formal qualia role. While the more
specific constitutive value of robin overrides the value in the constitutive quale
of bird, the values for the other qualia roles are still inherited (Pustejovsky,
1995b), so we know, for example, that a robin flies. Pustejovsky (1995b) dis-
cusses the same hyponymous relationship between car and Honda.
The second mechanism is projective inheritance, which ‘allows us to
dynamically create arbitrary concepts through the application of certain
transformations to arbitrary meanings’ (Pustejovsky, 1991: 434). These
transformations include negation (¬), temporal precedence (≤), temporal
succession (≥), temporal equivalence (=) and agency (act) which operate
over the various values of the qualia roles. The application of these trans-
formations generates the projective expansion of a predicate, and the set of
all projective expansions generated on all the roles of a qualia structure
forms the projective conclusion space. This allows us to explain the more ad
hoc relationships between concepts. Pustejovsky (1991) demonstrates this
through the examples in (20).
(20) a. The prisoner escaped last night.

b. The prisoner ate dinner last night.
The fact that (20a) sounds intuitively more prototypical than (20b) is
explained by the projective conclusion space of the lexical item prisoner.
The telic qualia role of prisoner means roughly ‘be confined in a prison’.
If the projective transformation ¬ (negation) is applied, the resulting
predicate is not-confined. The application of the temporal operators ≤ (prec-
edence) and ≥ (succession) generates two states, free before capture and free
after capture. Finally, the application of act (agency) and the selection of the
relevant agent (the prisoner) generates the concept escape. Escape thus lies
within the projective conclusion space of prisoner, which is why it sounds
60 claire thomas
more prototypical than eat. The effects of collocation must perhaps be

allowed for in cases such as these, but the idea of the projective conclusion
space nevertheless provides a useful way of linking concepts that only seem
intuitively close and therefore cannot be so easily explained through fixed
inheritance structures such as hyponomy.
Having examined the essentials of Pustejovsky’s (1991, 1995a, 1995b)
theory, we turn to Pustejovsky’s (1995a: 177–80) account of lexicalization.
The central concept here is the lcp (lexical conceptual paradigm), which
is distinct from the italicised lcp (type constructor). The type constructor
(lcp) creates complex, or dotted, types from simple types, as shown in (18).
These types are clustered together in a lexical conceptual paradigm (lcp).
There are two possible ways for the lcp to be lexicalized, or expressed as
lexical items. The first possibility is that the lcp is lexicalized as a single
lexical item, as Figure 3.3 illustrates.
Figure 3.3 shows that an lcp consisting of the basic types t1 and t2 and
the dotted type t1.t2 is expressed by the lexical item wi. This results in a case
of logical polysemy, as discussed above. This is the lexicalization pattern
that results in the process-result alternation we have been examining,
shown in Figure 3.4.
t1 t2
t1.t2
Wi
Figure 3.3 Lcp lexicalized as a single lexical item
process result
process.result
settlement
Figure 3.4 The logical polysemy of settlement
lexicalization 61
give take
give.take
sale transaction purchase

Figure 3.5 Split lexicalization of an lcp
The other option is for both the simple types and the dotted type that
make up the lcp to be lexicalized separately. In this case, t1 is expressed
by one lexical item, t2 by another and the dotted type t1.t2 by a third.
Pustejovsky points out that this is the case for the lcp which is partially
expressed by the lexical item transaction; the lexicalization pattern is shown
in Figure 3.5.
The lexical item transaction can only refer to the dotted type and not
to the simple types, which must be expressed by the nominals sale and
purchase. This is a case of split lexicalization of an lcp.
The picture of lexicalization emerging here corresponds very closely
to that expounded by Jackendoff, where the focus is on the mapping of
meaning onto certain lexical items. The differences lie in how they see the
process of linking certain elements of conceptual structure (Jackendoff)
or semantic types (Pustejovsky). For Pustejovsky, types are already built
into an lcp by the type constructor, which may either be lexicalized as
one lexical item or several. In the case of Jackendoff, however, different
languages ‘bundle up’ conceptual elements into lexical items in different
ways: it is effectively the process of lexicalization which is concerned with
the linking together of conceptual elements. The difference can be dem-
onstrated if we examine how the difference between Romance languages
and English identified by Talmy (1980) would be characterized using
Pustejovsky’s framework.
Figure 3.6 shows not that motion and manner of motion cannot be lexi-
calized separately in English, but that there is a way of lexicalizing them
together. In French the verb has to lexicalize motion alone. Furthermore,
the patterns of lexicalization are different but, crucially, the lcp is the
same for both. This is a small point, but this area where Pustejovsky and
Jackendoff diverge will be shown to be significant in the discussion of
lexicalization within morphology.
62 claire thomas
motion manner motion manner
motion.manner motion.manner
swim traverser traverser nager

across en
nageant
Figure 3.6 Lexicalization of motion and manner of motion in French and English
4 LEXICALIZATION: AN INTEGRATED PERSPECTIVE
We can now begin to build a more complete picture of lexicaliza-

tion informed by the synchronic perspectives offered by Jackendoff and
Pustejovsky. It has already been established that the different conceptual-
izations of lexicalization are not as divergent as they first appear. The aim
now is to integrate them.
In section 2, it was pointed out that within Generative Morphology,
the term lexicalization is often used to refer to completely idiomatized
meaning, even though it may be more useful to conceive of there being a
scale of lexicalization (Kastovsky 1982). It was suggested that by equating
lexicalization with idiosyncrasy, it is possible to miss some of the sys-
tematicity that can be seen within items that have not completely lost their
compositional meaning. This systematicity can now be explained through
the dotted types in the Generative Lexicon.
One of the systematic alternations that was identified concerned the
process and result readings of nominalizations, illustrated again in (21).
(21) a. The settlement of the area took several years.

b. The settlement was on top of the hill.
It can be argued that the existence of the process.result lcp makes it more
likely that other, newer, derivations with -ment will have the same reading.
This can be explained through Jackendoff’s account of priming. Jackendoff
(2002) describes how both lexical items and syntactic structures prime
other lexical items or syntactic structures. It is not much of a stretch to
suppose that conceptual structures, or indeed lexical conceptual para-
digms, might also be primed. Such a process can be demonstrated through
lexicalization 63
the nominalization of the verb adjectify in the sense ‘turn [someone’s name]
into an adjective’, which for many speakers is not lexicalized.
(22) a. The newspapers frequently adjectify Bill Clinton.

b. The media is employing increasing adjectification of politicians’ names.
The noun adjectification in (22b) has a process reading, but it can be argued
that a result reading is also activated. Suppose that on encountering an
unfamiliar word, adjectification, a speaker’s interpretation is aided in part
by the activation of other words ending in -ation. The interpretation
of the new lexical item may be helped by redundancy rules which have
been created from generalizations made about existing items (Jackendoff,
1975). The meaning of these words is encoded by the lcp process.result =
{process.result, process, result}. Thus not only is the process reading
understood as a result of the lcp and interpretation of the context, but the
result reading is also brought onto the ‘blackboard’ (Jackendoff, 2002).
This makes it more likely that the other meanings encoded in the lcp will
also be lexicalized as part of the meaning of the nominalization, so the noun
may begin to occur with a result reading, e.g. in (23).
(23) The adjectification the newspaper used was insulting.
Thus by applying the idea of semantic and syntactic priming to the lcp, we
are able to account for the more systematic side of lexicalization, where a
particular pattern can be identified.
We move now to examining how we can account for the relationship
of lexicalized derivations to other items in the lexicon. Lieber expresses
clearly how we should view the products of word formation processes:
Noninflectional word formation – derivation, compounding, and

conversion – serves to create lexemes to extend the simplex lexicon;
for that reason, I believe that the meanings it expresses ought to reflect
the semantic distinctions that are salient in the simplex lexicon. That
is, to the extent that we find semantic classes that are significant in
distinguishing the behaviour of underived lexemes, we might expect
derivation, compounding, and conversion to extend those classes.
And to the extent that we find polysemy in complex words, it ought to
be like the polysemy we see in simplex lexical items. (Lieber 2004: 9)
Lexicalized complex items, then, have the same status in the lexicon as
simplex items. I argue here that Pustejovsky’s framework allows us to
characterize the relationship between derived and simplex words in the
64 claire thomas
lexicon. The assumption to be taken from Lieber’s argument is that the

derived word is filling some kind of ‘gap’ in the lexicon, as the processes
exist to extend it. Pustejovsky’s lexical inheritance structure can be used to
characterize the semantic space of a lexicalized derivation in order to show
how it relates to other already existing items within the lexicon. This can
be demonstrated by returning to the example of settlement. We know that
settlement has some relationship to the simplex lexical item town, in that
they seem to be at least partially synonymous, as (24) shows.
(24) a. They built a town on the hill.

b. They built a settlement on the hill.
However, it is also clear that their meanings differ enough to stop them
from being interchangeable in most contexts, as illustrated in (25).
(25) a. The town of Guildford in Surrey is largely populated by commuters.

b. ?The settlement of Guildford in Surrey is largely populated by commuters.
c. The archaeologists uncovered an iron-age settlement.
d. ?The archaeologists uncovered an iron-age town.
Neither (25b) nor (25d) is semantically or grammatically wrong, it is just

that we intuitively feel that (25a) and (25c) are pragmatically more plau-
sible. This arguably has something to do with the projective conclusion
space of the lexical items, as Figure 3.7 shows. Town and settlement are
related through the formal qualia role, so we know that the entities they
denote may be similar in terms of form. Where they differ is in terms of the
formal qualia role, i.e. how they came about. Encoded within the meaning
of settlement is the idea that it was brought about by the act of settling, and
it evokes all the connotations of colonising and historical events that this
entails. Pustejovsky’s projective conclusion space allows us therefore to
build_act
settle_act
A
town
A
F
A = Agentive qualia role
settlement F = Formal qualia role
Figure 3.7 The projective conclusion space of settlement
lexicalization 65
characterize and explain the relationships of newly lexicalized words with

those that already exist in the lexicon.
5 CONCLUSION
Section 4 showed how Pustejovsky’s theory can be used to extend the account
of lexicalization provided by Generative Morphology, in particular Lieber
(2004, 2009) and Jackendoff’s Conceptual Structure. The existence of the
lcp means that certain possible meanings are already primed and therefore
more likely to be lexicalized, so while Generative Morphology provides an
account of idiosyncratic lexicalizations and Conceptual Structure can be
used to characterize these meanings, only the integration of Pustejovsky’s
framework can explain the more systematic patterns found in lexicaliza-
tion. Both Jackendoff’s theory of Conceptual Structure and Pustejovsky’s
Generative Lexicon can contribute greatly to our understanding of how a
lexicalized word gets its meaning. Jackendoff’s framework demonstrates
exactly what lexicalization is in terms of the formation of a long-term
memory linkage between a piece of syntax, a piece of phonology and a piece
of meaning; Pustejovsky’s framework shows how the relationship of this
newly lexicalized item to other items can be characterized.
I started by noting that Brinton and Traugott (2005) did not see syn-
chronic approaches to lexicalization as relevant to the various diachronic
perspectives, including that in Generative Morphology. It has been
demonstrated here that by examining two synchronic approaches, the
Generative Morphology account of lexicalization can be better understood
and the process itself better described.
NOTE
1. Lieber does not take Process to be a third category along with event and
state, but claims that ‘Process readings arise from the interaction of
events with unbounded arguments’ (2004: 24, fn. 6).
chapter 4
Term formation in a special

language: how do words specify
scientific concepts?
Kaarina Pitkänen-Heikkilä
T he topic of this chapter is the formation of scientific vocabulary in

nineteenth-century Finland, a process which my thesis (Pitkänen
2008) examined for the field of botany. This study examined how the
Finnish language was intentionally developed to meet the demands of civi-
lized society and Finnish-language science. It dealt with the methods used
to form scientific terms, and considers why a certain word (formed by word
formation) was chosen to represent a particular concept. The questions
discussed in this chapter are: how do words specify scientific concepts and
how do terms acquire their meaning?
Term formation in specialised language usually requires a significant
contribution from experts in different disciplines. One must not only
understand the scientific concepts thoroughly, but also be familiar with the
tradition of term formation in the discipline and the methods and norms of
word formation in the language.
Firstly, in section 1, I will describe some crucial principles in the theory
of terminology and present my research questions in detail. In section 2,
I will describe the background and materials of the study, as well as the
three methods used in the nineteenth century when forming new Finnish
terms for new concepts. I will go on to demonstrate the influence of the
indigenous language, foreign languages and the history of botany in the
process of term formation in section 3. Finally, in section 4, I will compare
the influence of these various factors on the naming of botanical concepts
in nineteenth-century Finnish.
term formation in a special language 67
1 MEANING AND MOTIVATION OF SCIENTIFIC TERMS
A term is the name of a special purpose concept. In the traditional theory

of terminology (e.g. Wüster 1979; Felber 1984), terminology is seen as
starting from a concept that has a place in a concept system. A term is the
established name of a concept, and it has been given through an agreement
and wide acceptance in the language society of a certain subject field.
Words in general language, particularly simplexes, can be arbitrary, as
argued by Saussure, but the terms that name concepts in special languages
typically have motivation. The motivation of naming can be examined
through form and meaning; a word can be motivated phonologically,
morphologically and semantically (Ullmann 1962: 82–93; Arnzt et al.
2002: 123–4). Phonologically motivated words are, for example, imita-
tions of animal cries (such as miaow), or words formed with the help
of sound-symbolic elements. A motivation of this type is not typical in
scientific terminology. Semantic motivation is typically related to meta-
phoric and metonymic naming (such as mouse in computer technology),
and often concerns semantic transfer of simplexes used also in terminol-
ogy. Morphological motivation concerns complex words and typically
influences the formation of new terms: if one knows the meaning of the
constituents, one can predict the meaning of the entire combination (this
includes all transparent compounds and derivatives).
Scientific terms can also be morphologically motivated in view of the term
system: such motivation can transpire through the utilization of various but
repetitive derivational and compound types. Terms that belong to the same
conceptual system, for instance, can have similar structures. Names of super-,
sub-and co-ordinate concepts have possibly been formed so logically that the
term systems reflect the concept systems. For example, in German chemical
vocabulary, the suffix -ium is a marker of metals; -um is a marker of non-
metals; -on is a marker of noble gases; -id, -it and -at are markers of various
salts; and prefixes an-, en- and in-are markers of organic bonds (Drodz and
Seibicke 1974: 91). This kind of structural congruence of term systems could
be an objective, particularly in naming the concepts of special languages,
because it would increase transparency of term systems and at the same time
make new terms more predictable. According to Picht and Draskau (1985:
114), ideal terms are highly motivated, logical and self-explanatory.
This chapter integrates lexicological and terminological methods. In lex-
icology, the word and its various meanings serve as the focus, whereas the
theory of terminology focuses on the concept and concept systems (Sager
1990: 55–6). A new, consciously developed terminology can be understood
through the old, familiar vocabulary and structures as well as through the
new, logical term system.
68 kaarina pitkänen-heikkilä
This chapter investigates what factors build the meaning of a term, what
the motivations of the chosen name of a concept are, and what the relation-
ship is between the term and the concept. Is the new term motivated in
the concept system of the subject field (e.g. certain repeated structures
describing certain conceptual relationships)? Is it understandable in rela-
tion to other terms in the term system, or in other words, from the term’s
logical systems? Or is the term formation motivated or influenced by earlier
indigenous vocabulary or by vocabularies of some foreign languages (e.g.
metaphoric terms and established morphologic elements based on classical
languages)? What is the relationship between a new term and its models?
Such questions are also possible when we study terminology in national
languages. How are concepts named? With the help of old indigenous
naming resources or with the help of well-known terms in other languages?
Can we understand a new term on the basis of knowledge of the term systems
in other languages? Because Latin was the scientific lingua franca of Europe
until the eighteenth century, many terms in natural sciences come from
Latin or Greek. Because of this, we recognize many naming resources, such
as classical words and structures that have been used in Latin for scientific
concepts, and this makes foreign terms and term systems more transparent
to us. For example, Felber (1984: 169–77) calls repeating components such
as infra-and -graphy ‘term elements’, while Bauer (1983: 213–16) calls new
English words that include combining forms based on Latin such as astro-,
electro-, -crat and -phile ‘neo-classical compounds’.
When forming new terminologies, the interesting question is what
we should say about the concept. Do we need to describe the concept by
naming it, or is it possible to simply choose any name without a motivation?
A good term should be neutral and unambiguous (Sager 1990: 89–90).
However, when we name a concept, we cannot express everything about
its content. In metaphorical vocabulary, for instance, metaphors emphasize
some aspects of a phenomenon and at the same time exclude other aspects
if they are not shared with the source of the metaphor (Lakoff and Johnson
1980); thus they direct our attention to certain aspects of a phenomenon.
In special languages, too, naming always hides some aspects, but highlights
others. That is why it is very important to find the essential features when
naming scientific concepts, because language not only reflects reality, it also
builds it.
2 EARLY BOTANICAL FINNISH
Pitkänen (2008) examines the development of Finnish botanical terminol-

ogy in the 1860s, when the first flora publication and its second edition
emi
hede
Figure 4.1 Emi, hede, sepivä and sulkasuoninen
were published in Finnish. The material consists of 1,500 botanical terms

from plant morphology, which denote and describe the parts of plants
and the relationships between those parts. They are simplex, derivations
and compounds, and they are typically formed in a very productive way
from the old vocabulary. New simplex terms, however, are quite rare.
Such terms, for example, are emi ‘the female organs of a flower’ (Lat.
pistil, En. carpel, pistil), hede ‘male fertilizing organ of a flower’ (Lat./En.
stamen), sepivä ‘stem-clasping’ (Lat. amplexicaulis, En. amplexicaul) and
sulkasuoninen ‘pinnately nerved’ (Lat. pinnatinervis), which are illustrated
in Figure 4.1.
The formation of botanical terminology in Finnish differs considerably
from that of botanical terminology in English and the modern Romance
languages. Colloquial words such as those for root, leaf and flower also
differ widely in these languages because they have such a long history
behind them, but technical words introduced during the past three centu-
ries are cognates in all these languages. For example, petal and pollen are in
French pétale and pollen, in Italian petalo and polline, in Spanish pétalo and
polen, in Romanian petale and polenul (Stearn 2004: 44). In Finnish, they
are terälehti and siitepöly, both compound terms formed by word formation
in the nineteenth century.
2.1 First Finnish flora

In the spirit of nationalism and national romanticism it was considered
desirable to preserve the unique features of the language. Thus the trans-
lators and writers of early Finnish textbooks aimed to form new words
on the basis of the old vocabulary by derivation and compounding. Elias
Lönnrot (1802–84), the creator of Finnish botanical terminology, believed
that loanwords from Latin, Greek or Swedish were not suitable because
these languages were not related to Finnish. Nevertheless, even though
linguistic purism was emergent in the nineteenth century and loans were
mainly avoided, the majority of the terms in Finnish textbooks of that time
Figure 4.2 Part of the derivative table by Elias Lönnrot (from the archives of the
Finnish Literary Society)
are different loan translations. Generally they were formed either loosely or
precisely according to the models of Latin or Swedish terms.
New terms were used for the first time in the first book of Finnish flora,
Flora Fennica (Lönnrot 1860). The terminology proved to be very useful,
and its use in botany was soon established. The basic terminology is still
in use. The author, Elias Lönnrot, was famous for collecting folklore and
writing the Finnish national epos, Kalevala. He was also a professor of
Finnish at the University of Helsinki as well as the writer and editor of a
large Finnish–Swedish dictionary (1866–80). He was familiar with plants
because of his background as a doctor of medicine. Lönnrot published a
test list of new vocabulary in 1858, and waited for comments on the termi-
nology before he used and expanded it in the book.
Lönnrot researched and lectured on Finnish word formation at the uni-
versity in the 1850s. In addition, he developed a particular formula from the
Finnish derivation system for the editors of his dictionary (see Figure 4.2).
This derivative table describes and names Finnish nominal suffixes used
with verb stems. Such tables and notes in the personal archives of Lönnrot
(at the Finnish Literary Society) indicate that he knew the Finnish deriva-
tion system well and was able to use various, also very rare, suffixes in his
word formation. His notes include a broad collection (c.3,700 words) of
Finnish derivatives from various dialects, which helped him to build a good
general view of the Finnish derivation system of that time (Pitkänen 2005).
Lönnrot believed that the rich vocabulary of Finnish dialects offered
good material for new terminologies, and that the free methods of word
formation offered opportunities for many potential words. For example,
emi is the diminutive of emä ‘mother, dam’, sepivä is the present participle
form of the verb sepiä ‘to bind’ (SSA 1992–2000). In the preface of Flora
Fennica (1860), Lönnrot writes that new botanical terms in Finnish are easy
for Finns to understand, as opposed to the corresponding Swedish terms,
which are more often borrowed from Latin (e.g. Lat. pistillum>Swe. pistill).
2.2 Various naming strategies

Finnish botanical terminology can be divided into three groups depend-
ing on the origin: the terms have been (a) accepted, (b) chosen from the
existing vocabulary, or (c) created on the basis of the existing vocabulary.
Accepted terms are old words used in their original meanings. In total,
accepted terms (c.290) account for 19 per cent of all the terms (c.1,500).
They are such terms as those in (1) and belong to the basic vocabulary of
the subject field.
(1) a. juuri ‘root; usually underground part of a seed plant body’

Lat. radix
b. varsi ‘stem; one of the three main organs of vascular plants’
Lat. caulis
c. lehti ‘leaf; a lateral organ for photosynthesis and gas exchange’
Lat. folium
In the examples, after the definition1 of a term, I express the equivalents in

Latin (which are often originally Greek) because it is the language of scien-
tific botanical nomenclature in which many morphological terms are used.
It is also the language from which many languages have borrowed their
own terms. In addition, I express equivalents in Swedish when it is useful
for understanding the background of Finnish terms. We must remember
that Latin and Swedish terms were Lönnrot’s models when he formed the
Finnish terminology.
Chosen terms are used for new, specific botanical meanings. They are
old words used with new meanings. Such terms are domestic meaning
shifts or semantic loan translations. These terms total 11 per cent (c.160) of
all the terms. This is typically metaphoric and metonymic naming, as in (2).
(2) a. hede ‘a rye’s flower’ → ‘a male fertilizing organ of a flower’

Lat. stamen, En. stamen
b. purje ‘sail’ → ‘the large upright petal of a sweet pea or related flower’
Lat. vexillum, Swe. segel, En. vexillum
c. kannus ‘spur’ → ‘hollow nectar-producing appendage of calyx or corolla’
Lat. calcar, Swe. sporre, En. spur
Figure 4.3 Purje and kannus
In (2a), hede was chosen from the old Finnish vocabulary where it meant ‘a
rye’s flower’ (SSA 1992–2000, henceforth SSA). In botany this word has
been used metonymically, and the new, botanical meaning is ‘a male ferti-
lizing organ of a flower’. In the Latin equivalent, the word stamen is used
metaphorically, because the original meaning of stamen in Latin (according
to the Concise Oxford English Dictionary, henceforth OED 2002) is ‘warp in
an upright loom; thread’.
The examples in (2b–c) are both metaphorical terms. In (2b), purje is a
strict meaning loan from Swedish segel ‘sail’. The Latin term vexillum is
also metaphorical; its original meaning is ‘martial flag, banner’ (OED 2002).
This kind of upright petal is illustrated in Figure 4.3. In (2c) kannus (lit.
‘spur’) was chosen through the models in Latin (calcar ‘spur’) and Swedish
(sporre), where the concept has been named metaphorically. A spur-like
‘hollow nectar-producing appendage of calyx or corolla’ is illustrated in
Figure 4.3.
Created terms have been formed on the basis of the existing vocabu-
lary and used for new, specific botanical meanings. They are new words
used in new, botanical meanings. They are either loan translations from
Latin, Greek or Swedish, or motivated by old domestic words. They
amount to c.1,050 terms, a total of 70 per cent of all terms. They are
new simplex, derivatives and compounds formed by word formation,
such as (3).
(3) a. kärhi ‘a slender thread-like appendage of a climbing plant’

Lat. cirrhus, Swe. klänge, En. tendril
b. sepivä ‘a leaf with its base clasping the stem’
Lat. folium amplectens, En. amplexicaul
c. terälehti ‘each of the segments of the corolla of a flower’
Lat. petalum, Swe. kronblad, En. petal
terälehti
verholehti
Figure 4.4 Kärhi, sepivä, terälehti and verholehti
d. verholehti ‘each of the parts of the calyx of a flower; enclosing the petals and
typically green and leaf-like’
Lat. sepalum, Swe. foderblad, En. sepal
The created terms are discussed further in this chapter because they are the
only true terms formed by word formation. Two-thirds of the created new
words were formed either loosely or precisely according to either Latin or
Swedish terms. For example, sepivä (3b) is based on sepä, an old Finnish
name for ‘neck’; it comes from the verb sepiä ‘bind, intertwine’ (SSA) and
the present participle marker -vA ‘-ing’. In Latin, amplectens ‘clasping’ is
a verbal derivative of amplexus ‘embrace’ (OED 2002). Similarly, the term
verholehti (verho ‘cover’, lehti ‘leaf’) (3d) has been translated morpheme for
morpheme from the Swedish foderblad. The Latin term sepalum ‘sepal’,
however, is not a compound; it comes originally from the Greek skepe ‘cov-
ering’ (OED 2002). The Latin petalum ‘petal’ in (3c) (<Gk petalon ‘leaf’,
OED 2002) is kronblad (‘corona leaf’) in Swedish and terälehti (terä ‘spike,
ear of corn; corolla’, lehti ‘leaf’) in Finnish.
One-third of the created terms were formed completely differently
from their equivalents in these foreign languages (alkio [alkaa ‘to begin’]
Lat. embryo [<Gk bryein ‘to grow’, OED 2002]). It is worth noting that
many loan translations contain rare vocabulary from Finnish dialects (for
example, sepä is an obsolete name for ‘neck’) used as equivalents for foreign
components of terms. Lönnrot’s idea was that the scientific terminol-
ogy should be inspired by the indigenous language, so that the Finnish
agricultural population would also be able to understand scientific texts.
In most cases, when new terms for new concepts were formed on the
basis of the existing vocabulary, new terms were made very transparent.
In addition to this, many metaphoric terms, which commonly exist in
botanical language, were changed to domestic metaphors (such as in (3a),
the Finnish equivalent of the Latin term cirrhus ‘curl’ is kärhi, a retrogres-
sive derivation from kärhys ‘big, branchy tree used for the hay pole’, SSA).
Nevertheless, the influence of foreign languages was still strong in the

process of forming Finnish equivalents for Latin and Swedish terms. Even
if linguistic purism was emergent (e.g. Thomas 1991: 204–5) and loans
mainly avoided, calques (where a word from another language is translated
morpheme by morpheme) and loan shifts (where an existing word acquires
a new meaning under the influence of another language) were extremely
common.
Characteristic of Lönnrot’s botanical terms is the selection of vocabulary
from various Finnish dialects, using compounds reflecting the generic
and partitive relations of concepts and exploiting particular, repetitive
structures for the co-ordinate concepts (for instance, certain affixes). These
kinds of structure form term systems that reflect the scientific concept
systems.
3 HOW UNDERSTANDABLE ARE NEW TERMS FORMED BY

WORD FORMATION?
This section deals only with created terms, because with accepted and
chosen terms there is not actually a question of word formation.
There are three ways of understanding the meanings of the terms.
Terms can be understood through the knowledge of old indigenous vocab-
ulary, or through the knowledge of basic botanical Latin or Swedish. They
can also be understood on the basis of special-purpose expertise as encoded
in concept systems and term systems in botany.
Although terms might seem very odd and unfamiliar, they usually have
a very clear motivation. Nowadays, however, many of them are transpar-
ent for only those few who are familiar with the old vocabulary and rare
structures used in the nineteenth century.
3.1 Old indigenous words

Firstly, motivation comes from old indigenous words. New terms have
many rare stem words and structures from various dialects. Lönnrot used
very rare vocabulary from the Finnish dialects for the equivalents of Latin
or Swedish words, and it is not always easy to ascertain whether, in fact, it
is a question of loan translation.
In addition to kärhi and sepivä, in (3a–b), good examples of the motiva-
tion of old Finnish words are illustrated in (4).
(4) a. emi ‘the female organs of a flower’

Lat. pistillum, En. pistil, carpel
Figure 4.5 Lanttopäinen and silposuoninen
b. itiö ‘a minute, typically single-celled, reproductive unit characteristic

of lower plants, fungi and protozoans, capable of giving rise to a new
individual without sexual fusion’
Lat. spora, En. spore
c. lanttopäinen ‘(of a leaf) with rounded shallowly notched end’
Lat. folium retusum
d. silposuoninen ‘(of a leaf) with undivided veins’
Lat. folium simplicinerve
Emi in (4a) is the diminutive form of emä ‘mother, dam’. Emi means ‘the
female organs of a flower’, the equivalent of which in botanical Latin is pistil-
lum, which means ‘pestle’. The Latin term is based on a different metaphor
from the Finnish term. The Finnish term references the function while the
Latin term is based on the physical shape. Itiö ‘spore’ (itää ‘to sprout’ +
noun suffix -iO), in (4b), is also a new derivation with a domestic motivation.
It has been formed without a foreign model: for example, the Latin term
comes from the Greek spora, which means ‘sowing’ or ‘seed’ (OED 2002).
Lanttopäinen (lantto in dialects ‘dell’ (SSA), pää ‘end’ + suffix –inen), in
(4c), and silposuoninen (silpa ‘bald, branchless’ (SSA), suoni ‘vein’ + (i)nen),
in (4d), are new compounds with an indigenous naming idea in their deter-
minative parts. The meanings of these terms are illustrated in Figure 4.5.
3.2 Foreign language terms

Secondly, motivation comes from the terms in foreign languages. Even
though Lönnrot’s terminology does not have loan words, approximately
two-thirds of all terms have some kind of foreign model for naming. All
terms in (5) are literal loan translations of foreign terms, although they have
very domestic material, being based, for example, on rare dialect words.
(5) a. hattumainen ‘hat-shaped’

Lat. corolla hypocrateriformis, Swe. hättelik
b. kalpamainen ‘scimitar-shaped’
Lat. folium acinaciforme, Swe. sabelformig
c. munuamainen ‘kidney-shaped’
Lat. folium reniforme, Swe. njurformig
d. haaralehti ‘leaf belonging to a branch’
Lat. folium ramea, Swe. grenblad
e. neuvoton ‘neuter; (of a plant or flower) having neither functional pistils nor
stamens’
Lat. flos neuter, Swe. könlös
The derivative term hattumainen (shape of the corolla: literal ‘hat-shaped’)

(5a) consists of the stem hattu ‘hat’ and suffix -mAinen ‘-ish, like’, and is
translated morpheme for morpheme from the Swedish term hättelik ‘hat-
like’. Kalpamainen (5b) and munuamainen (5c) have been formed similarly;
both name the shape of a leaf’s blades. Haaralehti (haara ‘branch’, lehti
‘leaf’) (5d) is a new compound that is translated word for word from
Swedish and Latin terms. Neuvoton (5e) is a privative (-tOn ‘less’) deriva-
tion from neuvo, an outdated dialect word for ‘men’s genitals’ or ‘tool’
(SSA). Nowadays, however, neuvo typically means ‘advice’, which is why
it is not easy to identify (5e) as a loan translation. Nevertheless, it is a loan
translation from the Swedish term könlös (kön ‘sex, gender’ + suffix -lös
‘less’). The Finnish word, however, refers only to male genitals whereas
the Latin and Swedish words refer to both sexes.
In addition to literal loan translations, my material also contains many
free loan translations and meaning loans. Meaning loans are indigenous
words that have acquired a new meaning because of foreign words. Good
examples of meaning loans were presented previously in (2b–c). Purje
originally means ‘sail’, but in botany it means ‘a large upright petal of
flower’, and this ‘loan shift’ is a Swedish influence on botanical Finnish.
Kannus originally means ‘spur’, but in botany it means ‘a slender
projection from the base of a flower’, such as calcar ‘spur’ in botanical
Latin.
3.3 History of botany

Thirdly, the history of botanical science has markedly influenced the
Finnish term system. The concept systems are based on studies by bota-
nists, and the logical systems have been reflected in the term systems of
different languages.
For example, certain repetitive term elements reflect the relations of
concepts such as tois-‘double’ in (6b) and (8b) and vasto-‘reverse’ in (6c)

and (7b)2 where the same prefixes or combining forms indicate relations of
concepts in the different systems.
(6) a. sahalaitainen ‘(of a leaf) having a jagged edge; saw-like’

Lat. serratum, Swe. sågad, En. serrated
b. toissahainen ‘serrate with smaller teeth on the larger ones’
Lat. duplicato-serratum, Swe. dubbelsågad, En. doubly serrate
c. vastosahainen ‘reversed serrate’
Lat. retrorsum serratum, Swe. omvändt sågad
(7) a. puikea ‘flat object egg-shaped in outline; oval, egg-shaped’

Lat. folium ovatum, Swe. äggrund, En. ovate
b. vastopuikea ‘eggshaped in outline, broadest above the middle; reversed
ovate’
Lat. folium obovatum, Swe. omvändt äggrund, En. obovate
(8) a. parilehtinen ‘having leaflets arranged on either side of a stem’

Lat. folium pinnatum, Swe. parbladig, En. pinnated
b. toisparinen ‘of a pinnate leaf in which the primary divisions are themselves
pinnate’
Lat. folium bipinnatum, Swe. dubbelt parpladig, En. bipinnate
The examples in (6a), (7a) and (8a) illustrate names of basic-level concepts
in the concept systems. Within each numbered example (6–9), the terms in
a, b, c and d are the names of co-ordinate concepts.
The examples in (6) name various leaf margins that are also illustrated
in Figure 4.6. Sahalaitainen (saha ‘saw’, laita ‘margin’) is a basic term,
and toissahainen and vastosahainen contain the translated classical e lements
Figure 4.6 Sahalaitainen, toissahainen and vastosahainen

Figure 4.7 Puikea vs. vastopuikea and parilehtinen vs. toisparinen
tois- (duplicato) and vasto- (retrorsum). Latin elements duplicato- and

bi-have systematically been translated to tois-‘double’ in Finnish, and
retrorsum and ob- to vasto-‘reverse’. In examples (7b) and (8b), the Latin
terms have these prefixes. Puikea is an old Finnish adjective suitable for
naming this egg-shaped form of a leaf blade. Vastopuikea is the same form
but reversed (see Figure 4.7).
Tois-‘double’ and vasto-‘reverse’ are not generally established first parts
of compounds in Finnish special languages, but they are firmly established
in botany. It is brought about by Lönnrot’s translation of the repetitive
structures of Latin and Swedish terms. In English, Bauer (1983: 213–16)
calls those terms that have such elements borrowed from Latin neoclassical
compounds. Such ultimately classical structures form term systems that
reflect scientific concept systems.
In botanical Latin, pinna is one of the term elements, as is the Swedish
prefix par-, from which comes the Finnish element pari- ‘pair’. Latin
pinnatum means ‘feathered’, and pinna ‘wing, fin’ (OED 2002). This
arrangement of leaflets3 (see Figure 4.7) has been named in Latin through
metaphor, but not in Swedish or Finnish. These terms will also be illus-
trated in (10) and (11) with another precept of compound formation.
In addition, the same suffix -mAinen can indicate that the terms are co-
ordinate concepts for each other. This is illustrated in the examples in (9).
(9) a. kellomainen ‘bell shaped’

Lat. corolla campanulatus
b. perhomainen ‘butterfly like’
Lat. corolla papilionaceus
c. ruusumainen ‘rose shaped’
Lat. corolla rosaceus
d. ristimäinen ‘cross shaped’
Lat. corolla cruciatus
Figure 4.8 Suffix -mAinen in the names of the various forms of a flower’s corolla:
kellomainen, perhomainen, ruusumainen, ristimäinen
The examples in (9) are all literal loan translations from Latin and name
different types of corolla or the flower’s shape (see Figure 4.8).
Botanical Finnish also has an interesting compound formation method
between super-and subordinate concepts in generic concept systems. By
this I mean a method in which the name of a subordinate concept is formed
by using the first part of a superordinate concept as the second part of the
subordinate concept. In (10–11), (a) illustrates the name of superordinate
concepts and (b–d) illustrate the names of various subordinate concepts.
(10) a. parilehtinen ‘having leaflets arranged on either side of a stem; feathered’

Lat. folium pinnatum
b. päätöparinen ‘unequally feathered, has a single terminal leaflet’
Lat. folium impari-pinnatum
c. tasaparinen ‘pinnate and ending with a pair of leaflets, not a single terminal
leaflet’
Lat. folium paripinnatum
d. vuoroparinen ‘of a pinnate leaf where the leaflets are not opposite’
Lat. folium alterne pinnatum
(11) a. sahalaitainen ‘serrated; (of a leaf) having a jagged edge, saw-like’

Lat. folium serratum
b. hienosahainen ‘finely serrated’
Lat f. serrulatum
c. toissahainen ‘double serrated’
Lat. f. duplicato-serratum
d. vastosahainen ‘reverse serrated’
Lat. f. retrorsum-serratum
In (10), parilehtinen (pari ‘pair’, lehti ‘leaf’) is the name of the superordi-
nate concept, and the names of its subordinate concepts are päätöparinen,
tasaparinen and vuoroparinen (see Figure 4.9).
This naming method makes it possible to predict the superordinate
concept on the basis of a new name of a subordinate concept. It has been
Figure 4.9 Päätöparinen, tasaparinen and vuoroparinen
used in Finnish botany since Lönnrot’s time. When a term system has
been established (and the terms lexicalized), new terms can be formed on
the basis of it. For example, a new term from 1995 is hienonirhainen (‘finely
dentate’), which obviously names a subordinate concept of the old term
nirhalaitainen (‘of a toothed leaf-margin which has symmetrically triangu-
lar teeth, rather than rounded or oblique, saw-like teeth’ En. dentate). The
same is true for the term herttasepoinen (‘cordated clasping’, En. cordate
‘heart-shaped leaf blade: broad at the notched base, narrow at the tip’)
from 1903, which names a subordinate concept of Lönnrot’s term sepokan-
tainen (‘stem clasping’). This seems to be a productive method in botanical
Finnish and has not come from Latin or Swedish models. It is, however, an
analogical rather than a productive method, because it has never been used
in standard Finnish (for a discussion of analogy, see Bauer 2001: 75–96).
Another interesting, exceptional method for forming new compounds
has been used in forming names for co-ordinate concepts in partitive
concept systems. Lönnrot tended to form subordinate compounds in
which the name of the partitive superordinate concept is the first part and
the name of the subordinate concept is the second part. This is illustrated
in (12).
(12) lehti ‘leaf’ (superordinate concept in a partitive concept system)

a. lehtilapa ‘the broad flat part of a leaf’
Lat. lamina, Swe. bladskifva
b. lehtiruoti ‘the stalk that joins a leaf to a stem’
Lat. petiolus, Swe. bladskaft, En. petiole
Such compounds have their head as the first part and the qualifying sub-
ordinate element as the last component, as in lehtilapa (lehti ‘leaf’, lapa
‘blade’) and lehtiruoti (ruoti ‘bone; rib’), where the first part is in the nomi-
native case, not the genitive. Similar compounds are possible in botanical
Latin, as in phyllophorus (‘leaf-bearing’<Gk. fyllon ‘leaf’, ferô ‘bear’),
Cheilolepton (sometimes used instead of the generic name Leptocheilus
‘narrow lip’) and phyllomegus (used instead of megalophyllus ‘large-leaved’)
(Stearn 2004: 259).
4 CONCLUSION
In conclusion, I would like to briefly consider the capacity of words speci-

fying scientific concepts and the balance between various factors behind
the naming of scientific concepts. It is important to realize that naming in
botanical Finnish has typically been influenced by both Finnish and foreign
languages. New terms have been formed by derivation and compounding
from indigenous material, but at the same time, the idea for naming might
come from a foreign language. Botanical Finnish, however, did not have
any loan words in the nineteenth century, and they are still fairly rare
in plant morphology. Rather than borrowing or simply forming a loan
translation it was typical of Lönnrot’s method to choose certain rare words
from dialects as the material for the equivalents of the Latin or Swedish
terms – even if Finnish had more common equivalents. The central terms
of Lönnrot’s terminology have been used for over 150 years, and they have
shown that new, transparent, domestic terms, or in other words new words
formed by word formation, can be operational in scientific terminology.
Domestic structures, understandable metaphors and new analogical forms
that have become established for a special language can build clear, logical
and self-explanatory terminology.
The salient question in naming is, of course, the requirements of science
and its concept systems. It is important to be familiar with the history of
the science in question when forming new terms; for example, when a
system has been established for naming concepts, new terms follow the
model and structures of the lexicalized terms, as well as terms that have
been formed analogically. In addition, it is important to remember that
when our understanding of a concept changes, we may also need a new
term that describes the concept better. Consequently, in special-purpose
term formation, the history and concept systems of the field are possibly as
important as the language we use in the naming of the concepts.
According to Sager (1990: 89–90) the meaning of a term should be inde-
pendent of the context. Thus a term should always mean the same thing
in different contexts. The meaning of a term is explained in a permanent
definition. In the present study, the meanings of new terms are predictable
on the basis of the old, indigenous word and structures from which they
have been formed and the logical system of terms that comes from botani-
cal Latin. Štekauer (2005a) points out that there is something systematic
about the interpretation of context-free, novel words. The same can be said
of special languages: when the language and terminology become familiar,
it is possible to predict the meanings of new terms with the help of intui-
tions that can arise gradually by using a special language, even though new,
analogical methods have also been used in addition to familiar, productive
methods. In the interpretation of a term it is essential to know the meaning
(or etymology) of the word and structures used, the history of the special
field and its vocabulary, and the history of the scientific concept and
concept system. Thus a new, consciously developed terminology can be
understood through the old, familiar vocabulary and structures as well as
through the new, logical system of terms.
NOTES
1. English definitions of the terms are typically presented according to

Stearn (2004) or the OED (2002).
2. The Finnish language has no prefixes; tois- and vasto- are the repeated
first parts of compounds. The equivalents in Latin and Swedish also
have prefixes.
3. leaflet ‘a small leaf; a component of a compound leaf’
chapter 5
Nominal compounds as naming

devices: a comparison of English
and Polish land surveying
terminology
Pius ten Hacken and Ewelina Kwiatek
C ompounding is a frequently used word formation process in many

languages, but different languages tend to have slightly different
systems. In terminology, there is a strong urge to use corresponding terms
in translation. Many terms are compounds. Therefore, when compound-
ing systems diverge, this leads to translation problems for terms. In this
chapter, we consider the nature of these problems in the context of English
and Polish terminology in the domain of land surveying. We start by
presenting our definition of compounding (section 1). Then we describe
how this definition applies to English and Polish (sections 2 and 3). Next,
we describe how we collected a set of terms for both languages (section 4).
Finally, we analyse the set of terms for each language and compare the use
of compounding in them (section 5).
1 A DEFINITION OF COMPOUNDING
In order to analyse compounding in a language, we need a working defini-

tion of what constitutes compounding. It is generally agreed that defining
compounding is not straightforward. After an overview of criteria used
for definitions, Lieber and Štekauer (2009: 14) conclude that ‘there are
(almost) no reliable criteria for distinguishing compounds from phrases
or from other sorts of derived words’. Similarly, Dressler (2006: 24)
claims that ‘[m]ore explicit universal definitions of the intensional type
are not only theory-dependent [. . .] but also cross-linguistically never
watertight – in many languages there are exceptions or fuzzy transitions to
non-compounding.’
84 pius ten hacken and ewelina kwiatek
In our view, this approach to the definition of compounding, although

widespread, is not warranted. As argued in ten Hacken (2010a), the exist-
ence of different definitions for a concept such as compounding should not
be approached as an indication that there is a single natural concept that
each definition tries to approximate with a larger or smaller degree of
success. Instead, each definition sets up a different concept and the discus-
sion about which definition is the best actually concerns the question which
of these concepts serves us best. When Dressler (2006: 30) states that ‘an
intensional definition of compounds must be preferential, by referring to
prototypes, rather than discrete’, he presupposes that compounding is a
pre-theoretical concept that can be used as a standard for the evaluation of
definitions. Prototypes may be the best way to approach natural concepts in
the mental lexicon, but if compounding is a theoretical concept, it must have
a precise definition and the definition takes priority over any individual
person’s intuitions.
In this light, the absence of ‘reliable criteria’ for Lieber and Štekauer or
the ‘exceptions or fuzzy transitions’ for Dressler can only be interpreted in
two ways. Either they consider compounding as a pre-theoretical category
which can be used as a point of departure for research but not as the object
of theoretical claims, or they claim that the definitions under consideration
are not adequate because they are underspecified. The latter occurs, for
instance, when we consider a definition such as (1).
(1) An A+N combination is a compound if the A is not inflected although it

should be inflected if it were a syntactic unit.
The problem with (1) is that it is only a sufficient, not a necessary condi-
tion. It implies that German Rotwein (‘red_wine’) is a compound, because
rot would have to be inflected in a syntactic context, e.g. der rote Wein (‘the
redINFL wine’). (For more discussion of these cases, see Schlücker (this
volume) and Schäfer (this volume).) What (1) does not do is to decide
whether English A+N combinations are compounds, because in English
adjectives are never inflected.
The choice of a definition of compounding is inevitably at least to some
extent theory-dependent. The motivation for a definition is that it sets up
a useful category. It is not a theoretical claim, because it cannot be refuted,
but if the category that results from the definition is interesting, it fulfils
its purpose. The definition we will use is the one presented in ten Hacken
(1994, 1999) and given in (2).
(2) A compound is a structure [X Y]Z or [Y X]Z, such that:

a. The denotation of Z is a subset of the denotation of Y;
nominal compounds as naming devices 85
b. If S is a possible way of specifying Y, the denotation of Z is determined by

the range of Ss that are compatible with the semantics of X;
c. X does not have independent access to the discourse.
The definition in (2) sets up compound as a binary structure. (2a) states that
it is semantically headed, but it does not specify whether it is left-headed or
right-headed. (2b) is a paraphrase of Allen’s (1978) Variable R condition.
In particular, it implies that there is no limited set of construction-specific
relations that characterize the relationship between head and non-head, as
proposed, for instance, by Levi (1978).1 (2c) specifies that the non-head of
a compound is not introduced as a new entity. This means that unless it
works as a proper noun and identifies a single entity in the outside world
on its own, it cannot be referred to in an unmarked way by means of a
pronoun.
As emphasized in ten Hacken (1994, 1999), (2) should be used to iden-
tify compounding constructions. For some individual compounds, it may
be difficult to apply the definition. However, by varying the components
of the compound without changing the construction, it is possible to
determine whether the construction belongs to compounding. This will be
illustrated in sections 2 and 3.
2 NOMINAL COMPOUNDING IN ENGLISH
Although (2) is a definition of compounding in general, we will here

concentrate only on nominal compounding. A nominal compound is a
compound with a noun as its head, i.e. Y = N in (2). The reasons for this
restriction are mainly practical. As we are interested in the terminology of
a specific domain and in this domain almost all terms are nouns or nominal
phrases, it would be difficult to collect enough instances of adjectival and
verbal compounds. Moreover, the issues they raise are not the same as for
nominal compounds.
In discussing the way compounding works in English, we should not
start from our termbase, but from general language examples. There are at
least three constructions we should consider, exemplified in (3).
(3) a. garden party

b. gentleman’s agreement
c. solar panel
All examples in (3) are entries in the Collins (1986) dictionary, which
provides an indication that they are used as names for specific concepts.
The construction in (3a) is the prototypical compounding construction,

discussed in much of the literature. It is therefore a good example to
explain the application of the definition in (2). All instances of this con-
struction are right-headed. In this specific example, the interpretation
‘party in a garden’ can be derived from the components because party
refers to an event that takes place at a specific location and garden can fill
this slot. The final condition in (2) can be illustrated with the sentence
in (4).
(4) Anna went to a gardeni partyj. It*i/j was bigger than she expected.
As the coindexation indicates, it can refer only to party or garden party,

not to garden, although semantically there is nothing abnormal about
the latter interpretation. The reason is that garden does not introduce
a particular entity in the discourse if it occurs in the expression garden
party.
The construction illustrated in (3b) is in many respects similar to the one
in (3a), but it is important to distinguish it from the syntactic construction
exemplified by Ben’s agreement. The contrast is illustrated in (5).
(5) a. This went counter to a gentlemani’s agreement. *Hei objected quite

strongly.
b. We need Beni’s agreement to proceed. However, hei is unlikely to give it.
c. a gentleman’s full agreement
d. these gentleman’s agreements
Whereas in (5a), he cannot refer to gentleman, in (5b) he can very well refer
to Ben. This is because the compound gentleman’s agreement does not intro-
duce gentleman as a discourse element. Inserting an adjective as in (5c) is
not ungrammatical, but the expression no longer involves the compound,
as the meaning demonstrates. (5d) shows that the determiner agrees with
the compound as a whole, not with gentleman. Therefore, we can conclude
that the construction illustrated in (3b) can be distinguished from the
syntactic construction as in Ben’s agreement. Whereas (3b) is a compound,
Ben’s agreement is not.
The status of (3c) is much less clear. The problem is that solar is an
adjective and seems to denote a property that is predicated of panel. This
suggests that we are dealing with a syntactic construction similar to, for
instance, big panel. However, there is a difference in the way the meanings
of solar and big are determined in these contexts. As a starting point, let
us compare the way the adjectives are described by Collins (1986), quoted
in (6).2
(6) a.
big: of great or considerable size, height, weight, number, power or capacity
b. solar: of or relating to the sun
The meaning of big depends in part on the noun it modifies. Not only
the dimension varies, as (6a) illustrates, but also the scale. A big spider is
much smaller than a big dog. Nevertheless, there is an inherent meaning
component in the sense of big. It refers to the top end of a scale that is prob-
ably most centrally determined by size. Underspecified as this may be, it
is much more specific than (6b). The only meaning that can be assigned to
solar is a relationship to an entity designated by a noun, sun. As a result, big
panel is interpreted as panel of a particular size, whereas solar panel can only
be assigned a meaning if we can establish a relationship between panel and
sun. The way this relationship is determined is exactly what is described in
(2b) for compounding.
The formal relationship between sun and solar is not based on a mor-
phological rule. Solar was not formed in English by word formation but
borrowed from Latin. Levi (1978) was one of the first to make the argu-
ment that expressions such as (3c) should be treated in a way parallel
to expressions such as (3a). She worked in the framework of generative
semantics in which sun would appear in the deep structure of (3c). It is not
necessary, however, to make such an assumption if we want to analyse (3c)
as a compound. The same can be achieved if we assume that solar is linked
to sun in the lexicon in a way suggested by (6b).
Therefore we conclude that all of the examples in (3) are compounds
under our definition (2). For the Saxon genitive construction in (3b) and
the relational adjective construction in (3c), we should be careful to dis-
tinguish individual occurrences from examples of syntactic constructions
of the same form. In cases such as children’s film or legal action, only the
context and the interpretation can decide whether we are dealing with a
compound or a syntactic construction. Without sufficient context, they are
ambiguous. Some minimal pairs are given in (7).
(7) a. this children’s film

b. these children’s film
c. non-legal action
d. illegal action
In (7a–b), the agreement of the determiner demonstrates that (7a) includes

a compound but (7b) does not. In (7c–d) the ambiguity is resolved by the
negation, which makes (7c) a compound and (7d) a syntactic adjective-
noun combination.
3 NOMINAL COMPOUNDING IN POLISH
The traditional concept of compounding in Polish is quite different from

the concept defined in (2). Szymanek (2010: 225) gives the contrast in (8)
as an example.
(8) a. dobranoc (‘Good night!’)

b. dobra noc (‘a/the good night’)
In (8), stressed syllables are underlined. Polish has a very regular stress
assignment rule. With very few exceptions, stress is on the penultimate
syllable of a word. Formally, the difference between (8a) and (8b) is that
the former is a word, the latter a phrase. As indicated by the glosses, Polish
does not have articles. (8a) can be considered as a compound if that concept
is defined along the lines of (9).
(9) A compound is a combination of two stems that constitutes a single prosodic

word.
The definition in (9) does not refer to semantic differences. In the case of
(8), the difference in status correlates with a difference in meaning. Whereas
(8b) is a noun phrase with compositional meaning, (8a) is, in Wray’s (2002)
terms, a formulaic expression, i.e. an expression that is used in specific,
pragmatically determined contexts. It is not possible, however, to general-
ize over the semantic difference in any meaningful way. Moreover, the
definition in (9) is not immediately applicable in many languages other than
Polish. Its applicability depends on such specific rules as stress assignment
in Polish that clearly identify the word as a prosodic domain. According to
our definition in (2), (8a) is not a compound.
Following Grzegorczykowa and Puzynina (1999: 455–68), Szymanek
(2010) includes the types in (10) in his concept of compound.
(10) a. gwiazd-o-zbiór (‘star-Ø-collection’, i.e. constellation)

b. łam-i-strajk (‘break-Ø-strike’, i.e. strike breaker)
c. dług-o-dystans-owiec (‘long-Ø-distance-er’, i.e. long-distance runner)
d. prac-o-daw-ca (‘job-Ø-giv-er’, i.e. employer)
In (10), the Polish words are divided into formatives by hyphens that do
not appear in normal orthography. The formatives glossed Ø are linking
elements. (10a) and (10d) are compounds of the type corresponding to (3a)
in English. In (10d), dawca (‘giver’) is the head. (10b–c) are excluded by
our definition in (2) because they are not headed. In (10b), the stem łam
of the verb łamać (‘break’) cannot be the head, because the whole word is
a noun, not a verb. The construction is similar to English pickpocket. In
(10c), the suffix -owiec functions in much the same way as English -er in
four-poster. It determines the basic semantic category of a concept that is
specified by an A+N combination. Ten Hacken (2010b) gives an overview
of such constructions and proposes an analysis which does not involve
compounding.
Szymanek (2010: 218–19) discusses a number of Polish translations of
English compounds of the type in (3a), listed in (11).
(11) a. numer telefonu (‘number telephonegen’, i.e. telephone number)

b. papier komputerowy (‘paper computeradj’, i.e. computer paper)
c. papier do komputera (‘paper for computergen’, i.e. computer paper)
d. pasta do zębów (‘paste for teethgen’, i.e. toothpaste)
Szymanek (2010: 219) states that ‘[w]hat is important is the fact that the
Polish expressions just cited are syntactic objects, and that they may involve
both inflection and derivation, but not compounding.’ It is interesting that
this statement is presented as a ‘fact’, even though no explicit definition of
compounding is given. Under a definition such as (9), the observation is
correct, but under (2) this is much less obvious.
The genitive construction in (11a) is similar to (3b). In the discussion of
(3b), we noted how important it is to distinguish the compounding variant
and the syntactic variant, see (7a–b). The analogous contrast for (11a) is
(12).
(12) a. ten numer telefonu

(‘this number telephonegen’, i.e. this telephone number)
b. numer tego telefonu
(‘number thisgen telephonegen’, i.e. the number of this telephone)
As the glosses indicate, when the demonstrative modifies telefonu, the

meaning is no longer that of the concept of ‘telephone number’, but
refers to some other type of number assigned to a particular telephone
set. Therefore numer telefonu has the same ambiguity as analogous con-
structions in English. In the most common sense it is a compound under
the definition in (2), but with appropriate coercion, as in (12b), it can be
re-analysed as a syntactic construction. As with the English counterparts,
the preference for the compound sense in (11a) depends on the individual
nouns combined.
The relational adjective construction in (11b) corresponds to (3c).3 The
formation of relational adjectives by means of the suffix -owy is a very
roductive process in Polish. Kallas (1999: 496) gives a table of adjective

p
formation processes and their meanings in which -owy is given as triggering
seventeen of the eighteen meaning relations between the adjective and the
underlying noun. Szymanek (2010: 87) calls the combinations of relational
adjectives and nouns ‘Noun Phrase’. This is understandable because Polish
does not have articles and a condition such as (1) is traditionally assumed
to distinguish compounds and syntactic constituents. However, for the
same reason as the one that made us conclude that (3c) is a compound if
we adopt the definition in (2), we should analyse (11b) as a compound.
This is supported by the nature of the discussions of relational adjectives
in Kallas (1999: 485–94) and Szymanek (2010: 85–97). Both go through a
range of thematic roles and meanings that relational adjectives may have
which corresponds largely to the range of meanings found in English N+N
compounds. Therefore we will assume that constructions such as (11b) are
compounds.
Finally, (11c–d) illustrate a prepositional construction. Szymanek (2010:
219) highlights the alternative structures (11b) and (11c) for the same
concept. This should not be taken as an argument that if (11b) is a com-
pound, (11c) should be as well. We have to distinguish the concept named
and the range of meanings a particular construction may have. Whereas
violinist and violin player are synonymous, the former is a derivation and
the latter a compound.
A first observation about (11c–d) as opposed to (11a–b) is that the former
have an explicit marker of the relationship, the preposition do, whereas the
latter do not. Bielec (1998: 216–18) describes the meaning of do when fol-
lowed by the genitive as ‘by [+ time]’, ‘to [+ place]’, or ‘for’. As the glosses
indicate, we take the last of these to be the core meaning. The question is
then whether N+do+N constructs such as (11c–d) meet condition (2b) on
the range of relationships between the two nouns.
It is interesting to compare the Polish N+do+N construction with
the French N+de+N and N+à+N constructions discussed by Nicoladis
(2002). Examples are given in (13).
(13) a. sac de toilette (‘bag of toilet’, i.e. toilet bag)

b. tasse à café (‘cup with/for coffee’, i.e. coffee cup)
Nicoladis (2002: 49) notes that (13b) refers to a kind of cup, not to any cup
filled with coffee. On the basis of language acquisition data from bilingual
and monolingual children in Canada, she concludes that ‘prepositions
are becoming linking items in French’ (2002: 58). Linking elements are
elements that have no or very little meaning of their own but contribute
to the form of a compound. Such elements are traditionally recognized as
Fugenzeichen in German morphology, as in Engel (1996: 520). A German

example is Schönheit-s-königin (‘beauty-Ø-queen’). The vowels glossed as
Ø in (10) can be seen as examples in Polish. Linking elements in German
are strongly connected to the genitive and Nicoladis (2002: 49) notes that
both à and de can be used to express possession. Given the absence of a
morphological genitive marker in French, we can see the constructions in
(13) as the French correlates of the genitive compounds in (3b) and (11a).
When we now return to Polish do in (11c–d), we note that it has a much
more specific meaning than de and à in (13). In Levi’s (1978) system of
Recoverably Deletable Predicates (RDPs), do corresponds to the RDP for.
For French de and à no such characterization in terms of a single RDP is
possible. Clause (2b) of the definition requires that the range of possible
relationships is determined by the meaning of the components, as opposed
to the construction. While admitting that more research into N+Prep+N
constructions across languages remains necessary, we conclude that Polish
N+do+N constructions are not compounds according to the definition
in (2).
Overall, then, Polish compounds are of three types. First, there are
N+N compounds such as (10a) and (10d). Second, there are genitive
compounds such as (11a). Finally, there are compounds with relational
adjectives as in (11b). The first type is right-headed, whereas the other two
are left-headed.
4 ENGLISH AND POLISH TERMBASES
In order to investigate how compounding is used in English and Polish

terminology, we used two termbases for the domain of land surveying that
were created for the research in Kwiatek (2013). As one of the purposes of
that research was to investigate conceptual mismatches between English
and Polish terms, the termbases for each language were created separately.
The English termbase has 490 term records, the Polish one 459.
In the selection of the terms, a combination of top-down and bottom-up
approaches was adopted. The top-down approach is the classical, onoma-
siological approach to terminology. The bottom-up approach starts from
the collection of terms from corpora of specialized text. As Arntz et al.
(2009: 220) indicate, the starting point for a terminological project is the
delimitation of the field and its division into subfields. In order to do this,
we used the Universal Decimal Classification (UDC, 2008), the system
of Encyclopaedia Britannica (1986), and a standard university textbook,
Bannister et al. (1998). Combining the information from these three
sources, we found the ten subfields of land surveying listed in (14).
(14) a. Analysis and adjustment of errors

b. Satellite positioning system (GPS)
c. Geodetic surveying
d. Cadastral surveying
e. Topographic surveying
f. Engineering surveying
g. Hydrographic surveying
h. Photogrammetry
i. Geographic information systems
j. Cartography and mapping
Rather than attempting to cover all of these subfields, we concentrated on

three of them which show interesting terminological differences, namely
(14b), (14c) and (14j). Whereas (14c) and (14j) are fields that have estab-
lished terms in the course of a more nationally oriented history, (14b) is a
new field with many recently added terms. For the English termbase, we
took the British terminology as used in England and Wales as the standard.
For the purpose of term extraction, we collected a corpus consisting of
textbook chapters, articles from scientific journals and technical magazines,
professional and educational websites, and manuals for specialized soft-
ware. In order to cover the terminology of a field such as geodetic survey-
ing, it is necessary to include terms that occur in related fields. While the
corpus texts were all from the three selected subfields, we also recorded
terms from other subfields that were found in these texts. In addition, we
recorded terms from adjacent domains where necessary. In the case of
geodetic surveying, we included, for instance, measuring units from math-
ematics, aberration from optics, and latitude from geography.
For each term, a term record was completed which specifies the standard
information types in a way similar to the examples given by Cabré (1999:
124–5) and Arntz et al. (2009: 223–5). They include citation form, abbre-
viation, grammatical information, subject field, definition, three examples,
synonyms, status (in relation to standardization), author and date. In
addition, we included lexical relations (hyperonym and/or holonym, as
appropriate), and entity type (based on Jackendoff’s (1983) conceptual
constituents).
As mentioned above, the termbases were intended to be used also
for research into mismatches and gaps. In addition to the two monolin-
gual termbases, we also produced a set of correspondence records (see
Cabré 1999: 127). In these records, not only translational equivalents are
recorded, but also conditions on how a term in one language is restricted
as a translation of a corresponding term in the other. In order to produce
correspondence records, the monolingual termbases were extended where
necessary. Thus, if the translation of a Polish term was not in our English
corpus, we added it after verifying that it is used in the relevant type of
English texts. While we included terms from subfields other than the three
we focused on in our monolingual termbases, we did not produce corre-
spondence records for these terms. This explains the different number of
term records for each language.
5 COMPOUNDS IN THE TERMBASES
In this section, we give an overview of the use of compounding as a naming

device in the English and Polish termbases. As a starting point, we consider
the proportion of compounds in the termbases. Figure 5.1 shows the distri-
bution for the English termbase.
Four classes are distinguished in Figure 5.1. The first two are com-
pounds. They constitute almost half of the termbase, 238 of 490 terms.
Most of these have two components, e.g. carrier frequency, but a significant
minority is the result of two or more applications of compounding, e.g
deeds registration system. Non-compounds are also divided into two classes.
A small minority have a compound as part of the term, e.g. mean sea level.
In this case, the full term combines an adjective and a compound. The
compound is not a term on its own. There were twenty-five such cases.
The remaining large category includes terms of various forms, e.g. contour,
public right of way, prime meridian. Figure 5.2 shows the corresponding
distribution in the Polish database.
When we consider the diagrams, the most striking differences are that
compounds are more frequent in the Polish termbase and that there is a
new category of borrowings. Compounds are a clear majority of the Polish
9%
Compounds (recursive)
Compounds (2 components)
Non-compounds (compound components)
46% Non-compounds (no compound components)
40%
5%
Figure 5.1 Compounds in the English termbase

8%
Compounds (recursive)
34%
Compounds (2 components)
Non-compounds (compound components)
Borrowing
Other non-compounds
51%
5%
2%
Figure 5.2 Compounds in the Polish termbase
terms, 268 of 459. All twenty-four borrowings in our Polish termbase are
from English. They include terms such as replica code and abbreviations
such as EGNOS (European Geostationary Navigation Overlay Service).
It is remarkable that there are no fewer than thirty-nine recursive
compounds, because Szymanek (2010: 223) suggests that compounding is
not recursive in Polish. However, as we saw in section 3, Szymanek uses
a rather different concept of compounding than we do. An example from
this class is (15).
(15) gleboznawcza klasyfikacja gruntów

‘pedologicalfem classification landgen-pl’ i.e. soil-based land classification
In (15), we see a combination of a relational adjective, a noun and a genitive

noun. We will come back to the analysis of this class after considering the
relative frequency of compound types in the two languages.
So far, we have only considered compounding as a single category. It
is also interesting to look at the distribution of compounds among com-
pounding constructions. Figure 5.3 gives the distribution in English.
N+N compounds such as ground antenna, corresponding to the type
(3a) in section 2, constitute more than half of all compounds. Most of the
remaining compounds combine a relational adjective with a noun as in
(3c), e.g. autumnal equinox. Perhaps surprisingly, there were only two com-
pounds with a Saxon genitive, Peter’s projection and Tissot’s indicatrix. Both
of these include a proper name, but six other compounds with a proper
name do not have the genitive marker, e.g. Robinson projection. The cat-
egory of compounds with phrasal non-heads is illustrated by dual frequency
receiver. Here, dual frequency is not a compound but an adjective-noun
combination, see ten Hacken (2003a, 2003b).
30% N+N
Proper name non-head
Neoclassical
Phrasal non-head
52% Possessive non-head
RA+N
1%
7%
7%
3%
Figure 5.3 Distribution of compounding types in the English termbase
The final category in Figure 5.3 is that of neoclassical compounds. In

neoclassical word formation, stems based on Ancient Greek and Latin
words are used to form new words in a modern language. As many
European languages have a similar system of neoclassical word forma-
tion, it is not always possible to distinguish borrowing and actual word
formation in this domain. Following Petropoulou (2009), we assume that
neoclassical word formation is not a homogeneous category. Only those
neoclassical words whose components can be analysed in accordance with
our definition in (2) are considered compounds here. This includes, for
instance, tacheometry but not perimeter. Whereas tacheo-, based on Ancient
Greek ταχύς [tachýs] (‘fast’), is used in the sense of speed in the former, the
latter includes peri-, based on the preposition περί [perí] (‘around’).
In Polish, the distribution among compounding constructions is given
in Figure 5.4.
7%
8%
N+N
Neoclassical
21% N + GN
N + RA or RA + N
64%
Figure 5.4 Distribution of compounding types in the Polish termbase

A comparison of Figure 5.4 with the English distribution in Figure 5.3

shows some striking differences. The only type that is equally represented
in both languages is the neoclassical one. Relational adjectives and genitive
nouns, as exemplified in (16), are much more frequent in Polish.
(16) a. południk zerowy

‘meridian zeroadj’, i.e. prime meridian
b. efekt Dopplera
‘effect Dopplergen’, i.e. Doppler effect
The examples in (16) correspond to (11b) and (11a), respectively. Both

relational adjectives and genitive nouns generally follow the head, but rela-
tional adjectives precede the head if the head is itself a compound with a
genitive noun. This is illustrated in (15). In both Figure 5.3 and Figure 5.4,
only the topmost operation is taken into account, so that (15) is classified
as RA+N.
The category of N+N compounds in Figure 5.4 does not correspond to
the type discussed in section 3 and illustrated in (10a) and (10d). Whereas
(10a) and (10b) are right-headed, the N+N compounds in our termbase are
all left-headed. Some examples are given in (17).
(17) a. sygnał PRN

‘signal PRN’, i.e. PRN signal
b. technika cross-correlation
‘technique cross-correlation’, i.e. cross-correlation
In all of the nineteen N+N compounds in our termbase, the non-head is

a code, abbreviation or borrowing. Most cases are of the type illustrated
in (17a). They can be seen as loan translations in which the order of the
constituents is adapted to match the one found with genitive nouns and
relational adjectives. The abbreviation PRN in (16a) stands for pseudoran-
dom noise, which is a binary signal similar to noise but which can be exactly
reproduced. In the case of (17b), the borrowing cross-correlation is made
more explicit by classifying it as a technika. As codes, abbreviations and bor-
rowings such as in (17) are not inflected in Polish, it is impossible to deter-
mine which case they appear in. Szymanek (2010: 217) gives the example
of film wideo (‘film video’, i.e. video film) and calls it ‘a somewhat irregular
syntactic phrase of the Noun + Relational Adjective type’. However, they
could also be analysed as genitives without overt inflection. This analysis is
supported by their left-headedness, aligning them with cases such as (16b).
One of the ideas underlying the decision to concentrate on three sub-
fields of land surveying was the prospect of identifying different trends that
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Total Cartography Geod. Surv. GPS Other
Non-compounds Non-compounds
(no compound components) (compound components)
Compounds (2 components) Compounds (recursive)
Figure 5.5 Compounds in each subfield of the English termbase
are characteristic of the individual fields. Figure 5.5 gives an overview of

the three subfields in English.
As explained in section 4, the selection of terms was based on texts
from three subfields, cartography, geodetic surveying and GPS. The first
column gives the distribution in the overall database and therefore corre-
sponds exactly to Figure 5.1. The last column gives the distribution of land
surveying terms that do not belong primarily to any of the three subfields
we concentrated on, but nevertheless occurred in the corpus of texts we
collected for these subfields. The variation from one field to the next is not
very big. It is interesting to compare Figure 5.5 to the corresponding data
for Polish in Figure 5.6.
The most striking observation about Figure 5.6 is the concentration of
borrowings in the subfield of GPS. This is not surprising and indeed the
choice of GPS as one of the fields was motivated by the expectation that
there would be more borrowing from English here. It is interesting to see
how this affects the distribution between the other naming mechanisms.
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Other non-compounds Compounds (recursive)

Compounds (2 components) Non-compounds (compound components)
Borrowings
Figure 5.6 Compounds in each subfield in the Polish termbase
In general, Polish terminology has more compounds, but the relative

distribution in the subfields is remarkably similar to the one in English.
Cartography has somewhat more compounds than average and geodetic
surveying somewhat fewer. For recursive compounds this is the same and
GPS is more or less at the average. Compared to English, borrowings in
Polish encroach somewhat on the territory of compounding, pushing its
proportion below the one in geodetic surveying, but much more on that of
the ‘other non-compounds’.
These observations suggest that certain concepts are more likely to be
named by a compound than others. After all, a field such as cartography
will have largely the same concepts in English and in Polish. Another con-
clusion these observations suggest is that borrowing is more likely to occur
if no compound can be formed.
We can also compare the distribution of different compounding con-
structions among the subfields. Figure 5.7 shows the distribution in
English.
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
RA+N Possessive non-head Phrasal non-head

Neoclassical Proper name non-head N+N
Figure 5.7 Distribution of compounding types per subfield in English
The differences between the three subfields are not very striking.
Cartography has relatively many RA+N and geodetic surveying relatively
many N+N compounds. The ‘other’ subfields have a rather different
distribution, but the total for this category is only thirty-one. We can now
compare this to the corresponding Polish data in Figure 5.8.
An interesting observation in Figure 5.8 is the concentration of N+N
compounds in the field of GPS. As illustrated in (17), these compounds are
loan translations. As such, we expect to find them in the field of GPS for
the same reason that borrowings are more frequent in this field.
When we compare the proportion of RA+N compounds in the differ-
ent fields for English and Polish, no particular trend can be discovered.
The N+N and N+GN compounding types are less directly comparable
between English and Polish because the former is marked in Polish and
the latter almost non-existent in English. It is interesting to note that
in Polish, as in English, neoclassical compounds are more frequent in
the ‘other’ subfields. We can think of two possible explanations. First,
neoclassical compounds can be more frequent in the other subfields of
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
N + RA or RA + N N + GN Neoclassical N+N
Figure 5.8 Distribution of compounding types per subfield in Polish
land surveying. Alternatively, neoclassical compounds designate more

general concepts, which are more likely to occur in texts in other subfields
than the one the term belongs to. As the neoclassical compounds in other
subfields do not belong to the same subfield, the latter seems to us more
plausible.
6 CONCLUSION
Adopting the definition of compounding in (2), we found that both in

English and in Polish a large proportion of terms in land surveying are com-
pounds. Although both languages have noun-noun compounds, genitive
compounds and relational adjective compounds, noun-noun compounds
are marginal in Polish and genitive compounds are marginal in English in
our termbases. The distribution of compounding types over the individual
subfields cannot be correlated between the two languages. If a subfield has
more relational adjective compounds than the average in one language,
this does not predict that it has more relational adjective compounds in the
other language.
A much stronger correlation can be observed for the use of compound-

ing as such. When a subfield has many compounds as terms in one lan-
guage, it tends to have many in the other language as well. This suggests
that concepts in certain domains are particularly prone to be named by a
compound. More research is needed to explore whether this hypothesis can
be substantiated.
As expected, Polish terminology in the subfield of GPS is marked by
borrowings from English. However, they mainly affect the number of
non-compounds (simple or derived words, phrasal expressions). Among
compounds, a special type of N+N compound based on loan translation
from English is relatively frequent in this subfield.
Finally, it would be worth investigating the distribution of neoclassical
compounds in more detail. As our termbase was based on the extraction
of terms from three subfields and neoclassical compounds were relatively
frequent among other land surveying subfields, we suggest that they are
used to designate concepts that are more general and therefore more likely
to occur in texts of other subfields.
NOTES
1. Jackendoff’s (2010: 413–51) system for characterizing the relationships

is compatible with (2b) because the system is generative. Although it
specifies fourteen primitive relations, it also specifies mechanisms for
combining them in an in principle unrestricted way.
2. Only the first sense is given in (6). Other senses are specialized for indi-
vidual nouns the adjectives modify. The function of these further senses
is to illustrate how the general first sense is used in particular contexts.
In (6b) the example, which serves the same purpose, is also left out in
order to make it more directly comparable to (6a).
3. Both (11b) and (11c) sound somewhat marked in Polish, but not more
so than computer paper in English. More common is papier do drukarki
(‘paper for printergen’, i.e. printer paper). A frequency check in Google,
carried out on 19 July 2011, suggests that this Polish expression is more
frequent than (11b) and (11c) combined. In English, printing paper and
printer paper are each 3.5–4.5 times as frequent as computer paper.
chapter 6
Semantic and formal structure:

a corpus-based study of Swedish
NN compounds and their French
counterparts
Maria Rosenberg
T his chapter addresses word formation and lexical representation.

It examines the relation between semantic and formal structure in
Swedish NN compounds and their French counterparts. The underlying
assumption is that semantic structures are lexicalized, and thus can be more
or less productive. The study adopts a primarily onomasiological perspec-
tive, which goes from meaning to form (see Marchand 1969; Štekauer
2005b), although the semasiological perspective will also be present. I
borrow Downing’s (1977: 838) quotation of Bolinger here:
Words are not coined in order to extract the meanings of their ele-
ments and compile a new meaning for them. The new meaning is
there FIRST, and the coiner is looking for the best way to express it
without going to too much trouble. (Bolinger 1975: 109)
Swedish NN compounding is highly productive, and NN compounds

are easily formed for new concepts (see Josefsson 2005). French seems to
prefer lexicalized NP constructions or free syntactic phrases, although NN
compounding is productive (see Fradin 2009). Hence one might predict
that, for the most part, Swedish NN compounds do not correspond to
NN compounds in French. Another important difference between the two
languages is that the semantic relation within Swedish NN compounds is
implicit, whereas it can be partly specified by prepositions in their French
counterparts.
The main objective of this study is twofold: to examine the formal
structure of the French counterparts of the Swedish NN compounds, as
well as the semantic relations expressed by the Swedish NN compounds
semantic and formal structure 103
and the corresponding French constructions. One further objective is to

examine to what extent Swedish NN compounds correspond to French
NN constructions and in what way the latter are more restricted compared
to the former. Given these objectives, it seems appropriate to use a method
based on empirical data. Hence I have compiled a small parallel corpus. In
that regard, what is being conducted here is a pilot study. Still, the analy-
sis aims to find systematic generalizations, and is both quantitative and
qualitative with respect to formal and semantic structure. Despite formal
differences between the two languages, a unified semantic account, based
on Jackendoff (2009), is adopted.
The chapter will be organized as follows. Section 1 accounts for the
theoretical context for this study. The parallel corpus and the data are
presented in section 2. The classification of the formal structure of
the data follows in section 3. Section 4 deals with previous accounts
of the meaning of compounds in different frameworks, and section 5 with
the semantic analysis of the data. Section 6 contains some concluding
remarks.
1 THEORETICAL CONTEXT
This section accounts for the theoretical context and presents some mor-
phological notions relevant for the present study. The delimitation between
compounds and lexicalized phrases in French is also briefly addressed.
1.1 Lexeme-based morphology

In this chapter, a compound is defined as a complex lexeme, formed by
morphological rules which associate lexemes (Amiot 2005: 190). The term
lexeme is taken in the sense, for example, of Lyons (1963), Matthews (1972)
or Aronoff (1992). According to the latter, a lexeme is ‘a (potential or
actual) decontextualized vocabulary word’ (Aronoff 1992: 13), underspeci-
fied for inflection and member of one of the open grammatical categories,
N, V or A. In French, a compound is either a noun or an adjective (see
Namer 2005: 133).
In general, the morphological head is said to correspond to the con-
stituent that determines the grammatical category and/or the major part
of the meaning of a complex word (see, for example, Carstairs-McCarthy
1992: 20). Distributional criteria are relevant for gender assignment to NN
compounds, but otherwise the head of an NN compound can be based on
semantic grounds. Haspelmath (2002: 87) and Rainer and Varela (1992:
22) among others define the head constituent of a compound as being its
104 maria rosenberg
hyperonym. Generally, Romance compounds are said to be left-headed,

as opposed to right-headed Germanic compounds: salade-santé vs. health
salad (see Williams 1981; Selkirk 1982; Scalise 1986; Booij 2009b).1 The
head notion in morphology has been debated (see, for example, Bauer
1990; Haspelmath 1992; Beard 1998; Fradin 2003); I return to this issue in
section 5.3.
This chapter adopts Jackendoff’s (e.g. 2002, 2009) architecture of
grammar as tripartite and parallel: phonology, syntax and semantics
are independent generative components, connected by interface rules.
Furthermore, the lexicon is assumed to have a hierarchical structure with
general abstract schemas at the highest level and single instantiations at the
lowest level. Redundancy is allowed, so that outputs of a productive rule
can also be listed (Booij 2009b). In addition to simplex words or lexemes,
the lexicon also contains larger units, such as compounds, meaningful con-
structions and idioms. These more complex lexical items are constructed
by either morphological or syntactical means, and can be referred to as syn-
tactic atoms (see Cruse 1986: 35–6; Di Sciullo and Williams 1987; Fradin
2003: 192–5).
1.2 Interpretation, lexicalization and productivity of compounds

As to the role of productivity in the semantics and lexicalization of word
formation, Ryder (1994) claims that new compounds are created and inter-
preted through knowledge about productive semantic patterns. According
to Jackendoff (2009) as well, the productivity of compounding involves a
set of principles, a productive rule system, for the interpretation of new
compounds. Lexicalized compounds conform mostly to these principles,
although compounds exhibit idiosyncrasies, and children’s evidence for
the general principles must come from generalizing over the learned
compounds. Jackendoff’s (2009) view resembles Booij’s constructional
approach to morphology, where abstract word formation patterns are gen-
eralizations from existing complex words, correlating form and meaning
(see Booij 2009b: 201).
The psycholinguistic findings of Libben (2006) indicate that the human
mind seems to both store and compute as much as possible. A compound
which has been encountered often enough can be lexicalized. Its repre-
sentation is then stored as a whole, with morphological structure, but can
still be decomposed into its constituents (Libben 2006: 6). Semenza and
Mondini (2006) note that aphasia research, experimental psychology and
formal linguistics reach the same conclusions regarding the representation
and processing of compounds. They claim that during lexical retrieval,
there are strong indications for the parsing (composition or decomposi-
tion) of compounds, transparent as well as opaque,2 where the head has no

privileged role.
1.3 Compounds vs. lexicalized phrases in French

As to the delimitation between compounds and lexicalized phrases in
French, I refer to Corbin (1992: 51), who claims that only those complex
sequences which cannot be generated otherwise than by morphological
mechanisms, i.e. lexical rules of composition, are compounds (e.g. ouvre-
boîtes ‘can opener’). In contrast, sequences that have regular syntax, such
as bas bleu ‘bluestocking’, are not compounds but syntactic construc-
tions. However, with regard to French NN constructions, the division
between syntax and morphology is not clear-cut. According to different
criteria (e.g. that the first noun is a hyperonym of the compound, no
preposition can be inserted between the two nouns, the two nouns are
not coordinated, proper nouns are not included, and depending on the
nature of the classifying relation) some of them are NN compounds (e.g.
attentat-suicide ‘suicide attack’), deriving from morphology, whereas
others are NN phrases (e.g. sortie piétons ‘pedestrian exit’), formed by
syntax (see Fradin 2003: 199, 202–6).3 Lesselingue (2003) deals with
what she refers to as [N1N2]N3 compounds in French as a continuum. At
one end, N3 is a hyponym of the class denoted by N1, e.g. poisson-chat
‘catfish’. At the other end, there are hybrids, e.g. porte-fenêtre ‘French
door’, corresponding to endocentric co-ordinate compounds. NN con-
structions, such as physique-chimie ‘physics-chemistry’, with two distinct
referents, are syntactic rather than compounds according to Lesselingue
(2003). In sum, not every French NN construction is an NN compound.
Note finally that Swedish NN compounds are written as one word
(blomkål ‘cauliflower’), whereas French NN constructions either contain
a hyphen (chou-fleur ‘cauliflower’) or are written as two words (chapeau
cloche ‘cloche hat’).
2 DATA
The data, consisting of Swedish NN compounds and their French coun-

terparts, draws from a parallel corpus. Cysouw and Wälchli (2007) signal
the relevance of parallel corpora for morphological and typological studies.
Furthermore, Tegelberg (2000) claims that, at the word level, the risk of
interference between source language and target language is low. The par-
allel corpus that I have compiled for this study is limited to written texts,
drawing from debates of the European Parliament held in the European
106 maria rosenberg
Table 6.1 Total number of tokens in the parallel corpus

Swedish French Total
Token 20,286 23,640 43,926
Table 6.2 Swedish NN compounds and French counterparts in the parallel corpus
Swedish NN compounds French counterparts
Token 983 940
Type 398 475
Union’s languages and rendered in both Swedish and French versions

(hence, neither French nor Swedish is necessarily the source or the target
language). In this respect, and as shown in Table 6.1, the corpus is rather
limited. As mentioned, this study is part of ongoing research.4
Nevertheless, the corpus seems to confirm the conclusion drawn by
Eriksson (1997), namely that French uses more words than Swedish does
in order to express the same content (partly dependent on the frequent use
of compounding in Swedish).
As shown by Table 6.2, I managed to find attestations for almost a
thousand Swedish NN compounds along with their French counterparts,
despite the limited size of the corpus.
We see that nearly 5 per cent of the Swedish part of the corpus con-
sists of NN compounds. These figures are consistent with findings of the
quantity of German nominal compounds in much larger corpora (see, for
example, Baroni et al. 2002; Schiller 2005; Junczys-Dowmunt 2008). In a
corpus of 1 million Swedish tokens, one would expect to find about 50,000
NN compounds. The lower number of tokens of the French counterparts
is due to the fact that some of them are abbreviations that correspond to
two Swedish compounds, such as LMR (limite maximale de résidus (des
pesticides)) ‘MRL (maximum residue level (for pesticides))’ vs. gränsvärde
för bekämpningsmedelsrester ‘threshold value for pesticide residues’. In con-
trast, Table 6.2 shows that the French counterparts are more diverse and
correspond to more different types.
3 FORMAL STRUCTURE OF THE FRENCH COUNTERPARTS
The French counterparts were classified according to their formal struc-

ture. Table 6.3 shows the frequency of the different formal structures
found in the corpus, exemplified with an attestation of a Swedish NN
compound and its French counterpart.
Table 6.3 Formal structure of the French counterparts in the parallel corpus
Swedish NN Token Type French counterparts Token Type
compounds
983 398 940 475
NN: beslutsprocess NA: processus décisionnel 265 140
‘decision process’ ‘decision process’
NN: dagcentrum N de N: centre de jour 165 114
‘day centre’ ‘day centre’
NN: människohandel N: traite 236 84
‘human trafficking’ ‘trafficking’
NN: våldsvirus N de DET N: virus de la violence 118 58
‘violence virus’ ‘virus of violence’
NN: medlemsparti NN: parti membre 58 12
‘member party’ ‘member party’
NN: könsbudgetering N (A) à DET N: 17 10
‘gender budgeting’ budgétisation sensible au genre
‘budgeting sensitive to gender’
NN: sysselsättningsstrategi N pour DET N: 9 6
‘employment strategy’ stratégie pour l’emploi
‘strategy for employment’
NN: småjordbrukare AN: petits agriculteurs 6 6
‘small farmers’ ‘small farmers’
NN: hälsopåverkan N (A) sur DET N: 6 5
‘health effects’ impact sur la santé ‘impact on health’
NN: (högsta prioritet på) Omission: une priorité absolue ø – 5
dagordningen ‘(highest priority on) ‘an absolute priority’
the agenda’
NN: barnsjukhus N pour N: hôpital pour enfants 14 4
‘children’s hospital’ ‘hospital for children’
NN: ansträngningar på gräsrotsnivå A: les efforts populaires 4 4
‘efforts at the grass-root level’ ‘popular efforts’
NN: hemarbete N à N: travail à domicile 4 4
‘housework’ ‘work at home’
NN: samhällsklasser N dans DET N: horizons dans la 4 3
‘social classes’ société ‘horizons in society’
NN: jämställdhetsperspektiv English loanword: 5 2
‘gender perspective’ gender mainstreaming
NN: (vara ett viktigt) önskemål Other: devrait tous nous préoccuper 24 23
‘(be an important) wish’ ‘ought to concern us all’
In my data, NA and N de N constructions, as well as simple Ns, are

the most frequent French counterparts of Swedish NN compounds. In
contrast, French NN constructions are rarely attested. Only 0.25 per cent
of the total number of tokens in the French half of the corpus are NN
constructions, not all of which are NN compounds (compared to nearly 5
per cent of the total tokens in the Swedish half). Given the low numbers of
attestations of French NN compounds in the parallel corpus, I consulted
108 maria rosenberg
a set of compounds used in a previous study (Rosenberg 2007), which

aimed to examine the pluralization of French compounds. That study
lends support to the result of the present study, namely that compounding
is a marginal phenomenon in French word formation.5 The French N de
N and N à N constructions are highly productive syntactic phrase types,
called ‘synapsies’ in French (Benveniste 1966), corresponding to stable
lexical units (see pomme de terre ‘potato’ or assiette à soupe ‘soup plate’).
Nicoladis (2002) suggests that the prepositions à and de in these kinds
of constructions are in an intermediate state to become linking elements.
Note, however, that N à N constructions are remarkably rare in the data.
4 SEMANTIC RELATIONS WITHIN NN COMPOUNDS
In the absence of any formal clues, the particular meaning of an NN com-

pound must be examined through its internal semantic relation and the
meaning of each constituent. Hence my study builds upon works such as
Downing (1977), Levi (1978), Gagné and Shoben (1997) and Jackendoff
(2009), which all examine the relation holding between the constituents of
a compound.6 However, I do not agree with Lieber (2004: 49) that a general
characterization of the semantic relations within root compounds is prob-
ably impossible (see also Selkirk 1982). According to Isabelle (1984), odd
interpretations of compounds are in fact rare. Clark and Berman (1987)
report that from the age of four children can interpret various relations
(e.g. possession, location, containment, material and purpose) between
head and modifier in Hebrew compounds (see Jarema 2006: 58).
4.1 Downing’s (1977) list

Although Downing (1977: 828) emphasizes that it is impossible to come
up with a finite list of the semantic relationships within NN compounds,
she compiles an inventory of semantic relations. Her inventory draws from
earlier classifications proposed by scholars such as Jespersen (1922), Li
(1971) and Levi (1975). These show considerable overlap and therefore seem
to indicate that some relations are perceived as more typical than others:
• Whole-part (duck-foot)
• Half-half (giraffe-cow)
• Part-whole (pendulum clock)
• Composition (stone furniture)
• Comparison (pumpkin bus)
• Time (summer dust)
• Place (Eastern Oregon meal)

• Source (vulture shit)
• Product (honey glands)
• User (flea wheelbarrow)
• Purpose (hedge hatchet)
• Occupation (coffee man)
We will see that Downing’s (1977) list with its general scope is not as elabo-
rated and detailed as the semantic relations proposed by Jackendoff (2009)
(see section 4.3).
4.2 Gagné and Shoben’s (1997) list

Although Gagné and Shoben (1997: 72–3) agree with, for example,
Downing (1977) that a list of the semantic relations within NN compounds
can never be exhaustive, they firmly believe that a relatively short list
can account for the majority of the conceptual combinations. Gagné and
Shoben’s (1997: 74) list follows in large parts the taxonomy of Shoben
(1991), which is in turn largely based on Levi (1978). In their description,
the head is labelled as ‘noun’ and the non-head as ‘modifier’.
• noun causes modifier (flu virus)

• modifier causes noun (college headache)
• noun has modifier (picture book)
• modifier has noun (lemon peel)
• noun makes modifier (milk cow)
• noun made of modifier (chocolate bird)
• noun for modifier (cooking toy)
• modifier is noun (dessert food)
• noun uses modifier (gas antiques)
• noun about modifier (mountain magazine)
• noun located (at/in/on) modifier (mountain cloud)
• noun used by modifier (servant language)
• modifier located (at/in/on) noun (murder town)
• noun derived from modifier (oil money)
• noun during modifier (summer cloud)
Gagné and Spalding (2006: 148) claim that it is important to examine

how complex words are formed and interpreted in order to understand
their processing and mental representation. ‘Conceptual combination’,
i.e. the process of combining two concepts to form a new concept (in the
conceptual system), is in their opinion highly relevant to the study of
110 maria rosenberg
compounds (in the language system), and a new combined concept is not
just an additive process, it is ‘more than a hybrid of its parts’ (Gagné and
Spalding 2006: 149). Gagné and Spalding (2006: 154) specify that nothing
prevents, for example, a relation involving ‘noun has modifier’ to be further
discriminated into has-part (picture book) or has-ownership (student book).
However, they emphasize that a more general relation does still provide a
lot of information and is important in the course of processing.
Gagné and Spalding (2006: 154–5, 162) underline that, during inter-
pretation, several relations compete for selection, thereby allowing com-
pounds to be potentially ambiguous. In this way, the active processing of
both stored and novel compounds is similar: they are all first decomposed
and then re-composed by use of the relation that links the constituents.
However, the interpretation tends to converge on a fixed meaning of the
compound that fits the context (Gagné and Spalding 2006: 163). In addi-
tion, it is easier to interpret and use a highly frequent relation than a less
frequent one (Gagné and Shoben 1997: 81) (see also section 5.4).
4.3 Jackendoff’s (2009) semantic relations

In Jackendoff’s (2009) Conceptual Semantics, a compound’s meaning is
a function of the meaning of its constituents: an N1N2 compound has the
meaning F (X1, X2). There are of course exceptions, such as compounds
with cranberry or strawberry constituents, which play no role in the com-
pound’s meaning, and so called ‘promiscuous’ compounds, which can have
several meanings simultaneously, such as boxcar (‘car that carries a box’,
‘that resembles a box’ or ‘that serves as a box’).7 According to Jackendoff
(2009: 122), in order to retrieve the semantic structure of an N1N2 com-
pound, one has to designate a head and determine the semantic relation
between the constituents. Jackendoff (2009: 123–4) enumerates several
basic functions for the semantic relation within NN compounds (based on
the analysis of 2,500 English compounds). Without claiming to be exhaus-
tive, he gives the following list:
• CLASSIFY (X1, Y2), ‘N1 classifies N2’ (beta-cell)

• Y2(X1), ‘(a/the) N2 of/by N1’ (sea level)
• BOTH (X1, Y2), ‘both N1 and N2’ (politician-tycoon)
• SAME/SIMILAR (X1, Y2), ‘N1 and N2 are the same/similar’ (zebrafish)
• KIND (X1, Y2), ‘N1 is a kind of N2’ (puppy dog)
• SERVES-AS (Y2, X1), ‘N2 that serves as N1’ (extension cord)
• LOC (X1, Y2), ‘N2 is located at/in/on N1’ (sunspot), also ‘N1 is located
at/in/on N2’ or ‘N2 with N1 at/in/on it’ (icewater)
• LOCtemp (X1, Y2), ‘N2 takes place at time N1’ (spring rain)
• CAUSE (X1, Y2), ‘N2 caused by N1’ (knife wound)

• COMP (Y2, X1), ‘N2 is composed of N1’ (rubber band), also ‘N1 is com-
posed of N2’ or ‘N2 that N1 is composed of ’ (bathwater)
• PART (X1, Y2), ‘N2 is part of N1’(apple core), also ‘N2 with N1 as a part’
(wheelchair)
• MAKE (X, Y, FROM Z), ‘X makes Y from Z’: a. ‘N2 made by N1’ (horse
shit) vs. ‘N2 that makes N1’ (honeybee); b. ‘N2 made from N1’ (olive oil)
vs. ‘N2 that N1 is made from’ (sugar beet)
• PROTECT (X, Y, FROM Z), ‘X protects Y from Z’: a. ‘N2 protects N1’
(lifeboat); b. ‘N2 protects from N1’ (mosquito net)
In addition to the semantic relations listed above, Jackendoff (2009: 120)

borrows the term ‘Proper Function’ from Millikan (1984) and treats it as
an action modality,8 applying to entities that are supposed or designed to
perform specific functions. Proper Function is true irrespective of actual
situations and can be further divided into (a) artefacts (e.g. can opener);
(b) parts of artefacts (e.g. back of a chair) or of organisms (e.g. leaves of a
plant); (c) objects destined to become something (e.g. seed>plant). In an
N1N2 compound, N1 can be an argument of the Proper Function of N2, e.g.
in coffee cup, cup has a Proper Function to contain coffee (Jackendoff 2009:
125). This notion corresponds more or less to PURPOSE in Downing’s
(1977) list and ‘noun for modifier’ in Gagné and Shoben’s (1997) list.
To sum up, we see that the three lists of semantic relations within NN
compounds, although not totally identical, have a lot in common. Note,
however, that the first two lists (Downing 1977; Gagné and Shoben 1997)
intend to capture the most general relations, while the last one (Jackendoff
2009), although not exhaustive, aims to account for an unlimited amount
of relations found within compounds. Consequently, I have chosen to use
Jackendoff’s list for the classification of the data.
5 SEMANTIC ANALYSIS OF THE DATA
This section deals with the semantic analysis of the data by focusing on
different aspects, namely semantic classification, variants, head and mor-
phological family, which will be accounted for in separate sections.
5.1 Semantic classification

In what follows, the collected data is classified according to the semantic
relations listed by Jackendoff (2009), with the addition of the last version
of CAUSE, ‘N2 causes N1’, drawn from Gagné and Shoben’s (1997) list, as
112 maria rosenberg
well as PURPOSE, described as ‘N2 is intended for N1’. Hence, I prefer to

use PURPOSE instead of Proper Function, an action modality according to
Jackendoff (2009), and therefore, in my opinion, situated at a different level
than the other semantic relations. Note that Jackendoff’s relations, elaborated
for right-headed English compounds, apply directly to Swedish NN com-
pounds. With respect to French, the ordering between N1 and N2 has to be
reversed. Let me also signal that my classification is tentative in some cases.
I will not go into details of the classification, just point to some ten-
dencies which merit further investigation. First, four semantic relations
dominate in the data, namely PURPOSE, the argument schema Y2(X1),
LOC (with its three variants) and CLASSIFY. Apart from the LOC rela-
tions, the most frequent relations are rather general and occasionally vague;
therefore it is not surprising that they dominate in the classification.
With regard to the relation between semantic and formal structure, an
analysis of the data I collected shows that PURPOSE and CLASSIFY can
be rendered by nearly all of the formal structures in French. Inversely,
NA and N de N constructions, the most frequent French counterparts of
Swedish NN compounds, are open to nearly all of the semantic relations
which are expressed by Swedish NN compounds. Recall the claim made by
Nicoladis (2002) that the French prepositions à and de in N à N and N de
N constructions are in an intermediate state of becoming linking elements.
Given the flexible semantics of the N de N constructions, my data actually
seems to give independent evidence to support her claim, although only with
respect to de ‘of/about/from’. In contrast, the few attestations of French
constructions involving the preposition à ‘to/in/at/with/of’ appear to be
limited to a few relations, such as PURPOSE, LOC and PART.9 Likewise,
French constructions with pour ‘for’ render, quite obviously, a PURPOSE
relation. Only two out of the four MAKE relations in Jackendoff (2009)
were attested in the data, corresponding to ‘noun makes modifier’ and ‘noun
made of modifier’ in Gagné and Shoben’s (1997) list, or, more or less, to
Product and Source in Downing’s (1977) list. They are rendered by NA
and N de N constructions in French. French NN constructions are only
attested for two relations, namely CLASSIFY and BOTH. Hence the
narrow semantic scope of French NN constructions can provide a clue to
their marginality. The BOTH relation can be considered as identical to
coordination, and coordinate constructions, overall, are rare in the data. In
my opinion, coordinate NN compounds are also rare in Swedish. Note also
that the PROTECT and KIND relations were not attested at all in my data.
In sum, Table 6.4 shows that several of the semantic relations have few
attestations in my data. They can thus be assumed to be less productive.
However, this fact might also be dependent on the selected text genre in
the corpus.
Table 6.4 Semantic relations within the Swedish NN compounds and their French
counterparts
Semantic relation Swedish NN compound French counterpart Type
PURPOSE sysselsättningsstrategi stratégie pour l’emploi 112
‘N2 is intended for N1’ ‘employment strategy’ ‘strategy for employment’
Y2(X1) kvinnoslaveri esclavage des femmes 80
‘(a/the) N2 of/by N1’ ‘woman slavery’ ‘slavery of women’
LOC hemarbete travail à domicile 30
‘N2 is located at/in/on N1’ ‘housework’ ‘work at home’
LOC köttdisk rayon boucherie 25
‘N1 is located at/in/on N2’ ‘meat counter’ ‘meat counter’
LOCtemp vårtoppmöte sommet de printemps 16
‘N2 takes place at time N1’ ‘spring summit’ ‘spring summit’
CLASSIFY rambeslut décision-cadre 70
‘N1 classifies N2’ ‘framework decision’ ‘framework decision’
PART familjemedlem membre de la famille 13
‘N2 is part of N1’ ‘family member’ ‘member of the family’
PART databaser bases de données 8
‘N2 with N1 as a part’ ‘databases’ ‘bases of data’
COMP glastak plafond de verre 13
‘N2 is composed of N1’ ‘glass ceiling’ ‘ceiling of glass’
COMP foderärter pois fourragers 2
‘N1 is composed of N2’ ‘feed peas’ ‘feed peas’
MAKE mjölkko vache laitière 4
‘N2 makes N1’ ‘dairy cow’ ‘dairy cow’
MAKE köttmjöl farine de viande 4
‘N2 made from N1’ ‘meat meal’ ‘meal of/from meat’
CAUSE bekämpningsmedelsrest résidu de pesticide 3
‘N2 caused by N1’ ‘pesticide residue’ ‘residue of pesticide’
CAUSE våldsvirus virus de la violence 4
‘N2 causes N1’ ‘violence virus’ ‘virus of violence’
BOTH invandrarkvinna femme migrante 6
‘both N1 and N2’ ‘immigrant woman’ ‘immigrant woman’
SERVES-AS gränsvärde limite maximale 5
‘N2 that serves as N1’ ‘threshold value’ ‘maximal limit’
SAME/SIMILAR lagstiftningsdjungeln législation, laquelle est 3
‘N1 and N1 are the same/ ‘legislation jungle’ extrêmement confuse
similar’ ‘legislation which is
extremely confusing’
Types of Swedish NN compounds 398
5.2 Variants
For the most part, Swedish NN compounds and their French counter-
parts attested in the data express the same semantic relation. For example,
Swedish hemmafru ‘housewife’ corresponds to French femme au foyer
‘woman at/of the house’ ’housewife’ where the use of the preposition makes
114 maria rosenberg
the LOC relation explicit. In some cases, the Swedish NN compound and
its French counterpart are not semantically equivalent: djurbestånd ‘animal
stock’ manifests a COMP relation (‘N2 is composed of N1’) whereas
animaux d’élevage ‘breeding-livestock’ manifests a PURPOSE relation (‘N1
is intended for N2’). In other cases, such as fodertillsats ‘feed additive’ in
example (1), the use of the prepositions dans ‘in’ or pour ‘for’ in the French
counterparts affects the semantic relation, PART or PURPOSE:
(1) a. fodertillsats (PART) additif destiné à l’alimentation des animaux

‘feed additive’ ‘additive destined for animal nutrition’
(PURPOSE)
b. additif dans l’alimentation animale (PART)
‘additive in animal nutrition’
c. additif pour l’alimentation animale (PURPOSE)
‘additive for animal nutrition’
Habert and Jacquemin (1993: 38) point to the fact that the same concept
can be rendered by variant nominal constructions, and that morphology
often intervenes in such cases. Thus variants are important because they
can permit us to account for different, competing linguistic forms used to
describe the same concept. A parallel corpus is highly useful in order to
shed light upon this phenomenon, since there are cases where the source
language linguistically codes an established concept which does not exist
in the target language but must be rendered somehow. Some Swedish NN
compounds for particular concepts correspond consistently to one and
the same French construction (and concept), such as arbetsplats ‘work-
place’ vs. lieu de travail ‘workplace’, or folkhälsa ‘public health’ vs. santé
publique ‘public health’, whereas others, such as those in examples (2–5),
correspond to different constructions in French, which thus seems to lack
established linguistic forms.
(2) a. födelsetal taux de natalité (4 attest.)

‘birth rate’ ‘birth rate’
b. taux de croissance démographique (1 attest.)
‘rate of demographic growth’
(3) a. arbetsmarknad marché du travail (11 attest.)

‘labour market’ ‘labour market’
b. marché de l’emploi (3 attest.)
‘employment market’
c. monde du travail (1 attest.)
‘labour world’
(4) a. arbetsliv vie professionnelle (4 attest.)

‘working life’ ‘professional life’
b. monde du travail (1 attest.)
‘labour world’
(5) a. barnomsorg structures d’accueil pour enfants (4 attest.)

‘child care’ ‘reception facilities for children’
b. crèches (2 attest.)
‘kindergartens’
c. garde-éducation (1 attest.)
‘care-education’
d. infrastructures d’accueil des enfants (1 attest.)
‘reception facilities for children’
e. modes de garde (1 attest.)
‘minding facilities’
f. prendre soin des enfants (1 attest.)
‘take care of children’
g. garde des enfants (1 attest.)
‘care of children’
h. prise en charge pour les enfants (1 attest.)
‘care of children’
The most striking example is barnomsorg ‘child care’. A well-established

concept in Swedish, it has several French counterparts in the data, as indi-
cated in (5).10 The different competing forms are partly explained by the
fact that French prefers freely generated or lexicalized syntactic construc-
tions instead of compounding.
5.3 Head
Gagné and Spalding (2006), who assume that NN compounds are inter-
preted through the semantic relation between the constituents, claim that
the modifier (non-head) is more relevant here than the head N: the head
N provides the category name, whereas the modifier implies a contrast
among members, indicating the subcategory (see honey bee vs. chocolate
bee). Thus they question the predominance attributed to the head noun for
a compound’s interpretation. Without going too far into debate about the
complexity of the morphological head (see section 1.1), it should be noted
that there are examples in the data, such as those in (6–11), where Swedish
NN compounds correspond to simplex Ns in French, and at least in (6a,
7–10) the head of the Swedish NN compound is not represented in the
French counterpart.
116 maria rosenberg
(6) a. bekämpningsmedelsrester pesticides

‘pesticide residues’ ‘pesticides’
b. bekämpningsmedelsrester résidus
‘pesticide residues’ ‘residues’
(7) aromämne arôme11

‘aroma compound’ ‘aroma’
(8) minoritetsgrupper minorités12

‘minority groups’ ‘minorities’
(9) lobbygrupper lobbies

‘lobby groups’ ‘lobbies’
(10) analysmetoder analyses

‘analysis methods’ ‘analyses’
(11) våldsoffer victimes

‘violence victims’ ‘victims’
Arguably, most of the Swedish NN compounds above draw most of their

semantic content from the first constituent. This generalization is not valid
for all Swedish NN compounds, such as shown by the last example (11)
våldsoffer ‘(violence) victim’. It seems likely that whether the head or the
non-head of the Swedish NN compound is expressed by a simple N in
French has to do with the particular semantic relation within the compound.
5.4 Morphological family and relation frequency

Schreuder and Baayen use the term morphological family ‘to denote the
set of words derived from a given stem by means of either compounding
(tablespoon, timetable) or derivation (tablet, tabular)’ (1997: 121). Jarema
notes that ‘[m]orphological family size, the type count of a morphological
family, has been found to influence word recognition, while morphological
family frequency ([. . .] token count of a morphological family) has not’
(2006: 52), referring to Schreuder and Baayen (1997). This evidence
relates, with regard to the modifier (non-head) constituent of the com-
pound, to Gagné and Shoben’s (1997: 73–4) claim that, during conceptual
combination, people make use of distributional knowledge about how
frequently a particular relation combines with a particular modifier. They
argue that since the locative relation is highly frequent for the modifier
mountain, for example, it should also be the easiest to interpret. In my data,
there were some instances of what might be called morphological families,

such as (12).
(12) a. huvudorsak (CLASSIFY) cause principale

‘main cause’ ‘main cause’
b. huvudmål (CLASSIFY) but ultime
‘main objective’ ‘ultimate goal’
c. huvudpunkt (CLASSIFY) point essentiel/point principal
‘main point’ ‘essential point/main point’
d. huvudskäl (CLASSIFY) raison principale
‘main reason’ ‘main reason’
Swedish NN compounds with huvud ‘head’ in initial position, as in (12),

manifested only one of its available relations, CLASSIFY, mostly cor-
responding to the French adjective principal ‘main’. On the other hand, we
also find cases such as (13–14).
(13) a. livsmedelskedja (Y2(X1)) chaîne alimentaire

‘food chain’ ‘food chain’
b. livsmedelssäkerhet (PURPOSE) sécurité des aliments/alimentaire
‘food safety’ ‘food safety’
(14) a. låginkomstfamiljer (PART) les familles à bas revenus

‘low income families’ ‘families with low income’
b. familjeliv (CLASSIFY) vie familiale
‘family life’ ‘family life’
c. familjebegrepp (CLASSIFY) concept de la famille
‘family concept’ ‘concept of family’
d. familjepolitik (PURPOSE) politique familiale
‘family policy’ ‘family policy’
Examples (13–14) show Swedish NN compounds containing livsmedel

‘food’ and familj ‘family’ along with their French counterparts. Here, there
does not seem to be any preferred relation for the modifier (or non-head)
constituents. Hence one might wonder to what extent different modifiers
actually have preferred relations in the sense of Gagné and Shoben (1997).
6 CONCLUDING REMARKS
In conclusion, in order to render the semantic relations expressed within

Swedish NN compounds, French makes use of several different formal
118 maria rosenberg
structures. My data indicates that French NA constructions are the most

frequent counterparts, followed by N de N constructions and simple Ns.
NA and N de N constructions are able to express nearly all of the semantic
relations proposed for NN compounds. Hence this result can be taken as
support for Nicoladis’ (2002) claim that the preposition de is on its way
to become a linking element, which in turn could imply that N de N con-
structions should have the same status as compounds in French, namely
deriving from morphology. When the French counterparts of Swedish NN
compounds correspond to simple Ns, they can correspond to either one of
the two constituents in the Swedish compound, thus not necessarily to the
head constituent. One might assume that this has to do with the semantic
relation expressed in the Swedish compound, a hypothesis which needs to
be further investigated. The distribution of semantic relations in combi-
nation with particular constructions in French also merits more detailed
study than is carried out here. New constructions and variations may
be helpful to shed light upon semantic restrictions governing particular
constructions and their productivity. Overall, the semantic classification
of the data indicates that the most productive semantic relations are quite
general.
There were few French NN constructions in the data. However, the
analysis indicates that they are semantically less flexible than Swedish NN
compounds. This might thus partly explain why they are also considerably
less productive. In light of this finding, it would be interesting to examine
Swedish counterparts of French NN constructions. One can predict that
they would correspond to NN compounds to a very high extent.
To conclude, the contrastive perspective is highly profitable and pre-
sents the advantage of having potential application within second-language
acquisition and natural language processing, such as machine translation
and disambiguation analysis (see Johnston and Busa 1999). Moreover,
parallel corpora are most useful for studying the often ignored impact of
contextual and pragmatic factors on the interpretation of compounds (see
Lapata 2002). According to Pustejovsky (1995a) and Jackendoff (2009),
among others, the lexicon is a generative system allowing meaning to be
flexible. In this context, further important issues are to examine to what
extent compounds are open to different interpretations or have fixed mean-
ings, as well as to continue to explore the semantic structure of nouns.
NOTES
1. In French, neoclassical compounds, e.g. océanographe ‘oceanographer’,

are traditionally seen as right-headed (Namer 2005: 22).
2. However, Gagné and Spalding’s (2006: 147) state that ‘whether a com-
pound is represented as a whole word or in terms of its constituents
appears to be related to semantic transparency’. Dressler (2006: 41)
distinguishes four degrees of morphosemantic transparency within
compounds. Transparency of the head is most important: (a) trans-
parency of both constituents (doorbell); (b) transparency of the head,
opacity of the non-head (strawberry); (c) transparency of the non-head,
opacity of the head (jailbird); (d) opacity of both constituents (humbug).
Following Semenza and Mondini’s (2006) claim, I will not consider
transparency degrees in this study.
3. Fradin (2003: 196, 199) has a restricted view of French NN com-
pounds, in contrast to Noailly (1990: 65–93) who seems to treat all
French NN constructions (including compounds) as syntactic.
4. A bidirectional, extensive and well-balanced Swedish-French parallel
corpus is being elaborated along the lines proposed by McEnery et al.
(2006: 46–51).
5. Only 303 (0.008 per cent) French plural NN compounds were attested
in a French corpus consisting of 3,702,906 tokens compiled by Umeå
University, Sweden and containing text from two Belgian newspapers,
La libre Belgique and Le Soir, and two French newspapers, Le Monde
and Le Monde Économique.
6. Note that Bisetto and Scalise (2005) propose a different classification of
compounds, based on the grammatical relation between the constitu-
ents: subordinate, attributive or coordinate. Subordinate compounds
involve a complement relation, such as chèque-restaurant ‘dinner
cheque’. In attributive NN compounds, the non-head expresses, often
metaphorically, an attribute of the head, such as camion-suicide ‘suicide
lorry’. The constituents of coordinate compounds are linked by the
conjunction ‘and’, such as boulanger-pâtissier ‘baker-confectioner’.
7. It is not ambiguous in the way football is between ‘game’ or ‘ball’, since
it picks out the same object no matter which interpretation you give it
(Jackendoff 2009: 117).
8. That is, ‘variant interpretations [. . .] under which a nominal can be
understood’, which can be part of its lexical meaning (Jackendoff 2009:
119).
9. According to Cadiot (1993), there are two principal types of N1 à N2
constructions: à/AVEC ‘with’ (verre à pied ‘goblet’) and à/POUR ‘for’
(verre à dents ‘tooth mug’), corresponding to PART and PURPOSE in
Table 6.4.
10. In the Swedish data, there are two occurrences of barnkrubba ‘kin-
dergarten’, which is archaic and non-idiomatic, and must be due to a
translation of an expression such as crèche ‘kindergarten’ in French.
120 maria rosenberg
11. The French counterpart substance aromatique ‘aroma compound’ is also

attested.
12. The French counterpart groupes minoritaires ‘minority groups’ is also
attested.
chapter 7
The semantics of lexical

modification: meaning and
meaning relations in German A+N
compounds
Barbara Schlücker
T his chapter explores the semantics of German A+N compounds as

opposed to A+N phrases. Taking into account various data, the chapter
shows that in their semantics, German A+N compounds are far from being
simple or consistent. The main claim is that there are systematic meaning
differences between lexical and syntactic adjectival modification but that
these differences do not arise from differences on the level of semantic form.
That is, the internal modification relations found in A+N compounds and
A+N phrases are basically the same, contrary to what has been proposed
in the literature. Furthermore, the chapter also discusses differences and
similarities between the semantic properties of A+N and N+N compounds.
The structure of the paper is as follows: section 1 introduces the issue;
section 2 gives a brief overview of the formal properties of German A+N
compounds and phrases. The third section discusses the main point of the
chapter, that is the semantics of lexical adjectival modification as opposed
to phrasal adjectival modification. The fourth section examines the differ-
ence between adjectival lexical modification and nominal lexical modifica-
tion and is followed by a summary in section 5.
1 INTRODUCTION
Research on the semantics of nominal compounds concentrates for the

most part on N+N compounds, in English as well as in German and other
languages. The reason for this is obvious: N+N compounding is highly
productive, both in English and German, and it seems reasonable to
assume that one explanation for this productivity lies in the fact that N+N
122 barbara schlücker
compounding combines two concepts but leaves open the exact nature of
the relation between them. From this point of view, the meaning of com-
pounds is inherently underspecified and therefore very flexible.
In comparison to N+N compounds, the semantics of A+N compounds
seem to be rather simple: the property denoted by the adjective is predi-
cated to the noun referent, such that a blackbird is a bird that has the prop-
erty of being black, or, more precisely, a blackbird is an individual that is
black and a bird. Thus, instead of the various meaning relations found with
N+N compounds (e.g. a dollhouse is a house FOR dolls, a garden house is
house LOCALIZED in the garden and a glasshouse is a house MADE-OF
glass), there seems to be just a simple attributive modification relation in
the case of A+N compounds.
Of course, A+N compounding very often involves semantic speciali-
zation, as can be seen from the above example: not every black bird is a
blackbird; rather, blackbird denotes a certain species, Turdus merula. Thus
one of the questions to be answered is whether compounding always and
inherently involves semantic specialization and whether this contributes to
the semantic difference between lexical and syntactic modification.
Another question has to do with the claim that, despite the above
description, the meaning relation between the adjective and the noun is not
generally as simple. Although such an ‘attributive’ relation can be found
in most German A+N compounds, there are also deviant cases, such as
Gelbfieber ‘yellow fever’: this is not a fever that is yellow but rather a fever
that makes the person concerned become yellow (or, more precisely, have a
yellow skin). So, instead of modifying the head noun, the adjective in these
cases modifies an implicit referent (‘skin’).
The main questions to be addressed in this paper are, therefore, first:
What is the effect of word formation? That is, what is the difference
between syntactic and lexical modification, as in (1a) and (1b), respectively.
(1) a. hohes Haus ‘high house’
b. Hochhaus ‘high_house’, i.e. skyscraper
And second: Do compounds have an abstract meaning? How does the modi-
fier contribute to the compound meaning? As can be seen from the examples
in (2), the function of the modifier in a compound can roughly be described
as identifying and selecting a subconcept of the concept denoted by the head
noun (see Zimmer 1971; Downing 1977; Berman and Clark 1989).
(2) a. FertigAhausN, HochAhausN
‘ready house’, i.e. prefabricated house, ‘high house’, i.e. skyscraper
b. KrankenNhausN, EckNhausN
‘invalid house’, i.e. hospital, ‘corner house’
the semantics of lexical modification 123
c. WohnVhausN, BadeVhausN
‘living house’, i.e. residential house, ‘bathing house’, i.e. bathhouse
d. HinterPhausN, OberPhausN
‘back house’, i.e. rear building, ‘above house’, i.e. upper house
The modifier, whether it is an adjective, a noun, a verb or a preposition,

selects one particular meaning aspect (e.g. the intended use, the form or
location) in order to do so, although the subconcepts will usually not only
differ from each other in this particular respect. I will not consider V+N
and P+N compounds in the remainder of the chapter; however, I will
argue that, despite the common function of modifiers as just described,
there may also be functional differences between adjectival and nominal
lexical modifiers.
2 FORMAL PROPERTIES OF GERMAN A+N COMPOUNDS AND

PHRASES
In a way, the above mentioned example blackbird is rather atypical for

English, because it is unquestionably regarded as a compound. This status
is also reflected by the fact that it is consistently written as one word. There
are a number of similar, often quite old compounds, such as poorhouse,
shortbread, hotbed or greenhouse, but in general it seems rather difficult to
establish a clear formal difference between nominal compounds and cor-
responding phrases in English, both for A+N and N+N combinations.
Accordingly, a discussion about the classification of forms like black board,
silk tie and apple pie has been going on at least since Bloomfield (1933) and
has not yet been decided (see Bauer 1998 and Giegerich 2004, 2006, among
many others).1
German, on the contrary, exhibits a clear formal difference between
nominal compounds and corresponding phrases, as they can be distin-
guished unambiguously by means of stress and inflection: the main stress
is on the head of the phrase but on the modifier of the compound and the
phrasal modifier is inflected but not the lexical one. This distinction is also
reflected by spelling, as compounds are consistently written in one word
whereas phrases are written in two words – see (3).
(3) alte Stádt – Áltstadt

‘old city’
Morphologically complex adjectives are excluded from A+N compound-

ing, with the exception of adjectives suffixed by -al, -ar, -är, -iv, -ig (e.g.
Instrumentalmusik ‘instrumental music’) and some participle forms. There

are no restrictions whatsoever on the noun.
The question of the semantic effects of word formation and the semantic
difference between lexical and syntactic modification hinges crucially on
a clear formal difference between lexical and syntactic modification. As
German allows such a clear formal distinction, the German data are espe-
cially well-suited for investigating the semantic properties of lexical and
syntactic modification.
3 SYNTACTIC AND LEXICAL MODIFICATION
A very common assumption about the semantics of compounds as opposed

to phrases is that phrases have compositional meaning whereas compounds
have non-compositional meaning, i.e. they exhibit semantic s pecialization –
see (4).
(4) a. alte Stadt – Altstadt

‘old city’ – ‘historic city centre’
b. süßes Wasser – Süßwasser
‘sweet water’ – ‘freshwater’
c. frischer Käse – Frischkäse
‘fresh cheese’ – ‘cream cheese’
The compositional interpretation of phrases has often led to an analysis

according to which syntactic modification is intersective modification: alte
Stadt denotes an individual that is both old and a city. However, it has often
been mentioned that such an analysis is deficient. First, it does not work
for non-intersective adjectives like ehemalig ‘former’ or angeblich ‘alleged’:
ehemaliger Präsident ‘former president’ does not refer to the intersection
of the set of individuals that are former and the set of individuals that are
a president. Second, intersective modification often also seems to be inap-
propriate (or at least inaccurate) when dealing with intersective adjectives
in cases such as (5).
(5) a. red apple, pink grapefruit

b. roter Apfel, rosa Grapefruit
A red apple is an apple with a red peel whereas a pink grapefruit is a

grapefruit with pink pulp (but a white or yellow peel) – see Lahav (1989,
1993), Sweetser (1999). The same holds for the German counterparts in
(5b). Interestingly, these examples are phrases as can be seen from the
inflected adjective and stress, indicating that these meaning properties

do not originate from the supposed non-compositional interpretation of
compounds.
All in all, these few examples show that the analysis of syntactic
adjectival modification is far from simple. It becomes even more com-
plicated if relational adjectives such as nuklear ‘nuclear’, sozial ‘social’
or medizinisch ‘medical’ are taken into consideration. These adjectives
do not denote a property on their own but rather establish a meaning
relation between the base noun they are derived from and the head
noun. Furthermore, they deviate from other adjectives with regard to
their morphosyntactic behaviour. Therefore I will not consider these
adjectives here.
3.1 Intersective and partial modification

Cases like (4) show that there is a very clear meaning difference between
A+N phrases and A+N compounds. However, one could easily argue
that this meaning difference does not stem from a basic difference
between the semantics of syntactic adjectival modification and lexical
adjectival modification, but that it rather has to do with the fact that
Altstadt, Süßwasser and Frischkäse are lexicalized compounds. For this
reason, I will use newly coined, non- lexicalized combinations of an
adjective and a noun in order to investigate the semantic difference
between syntactic and lexical modification, such as rot ‘red’ and Ball
‘ball’. Of course, the phrase roter Ball is a common expression, but
neither the compound Rotball nor the phrase roter Ball is a lexicalized
sequence in German.
To begin with, the adjective is not accessible for syntactic modification
after compounding. In (6a), the degree particle (or adverb) sehr ‘very’
modifies the adjective rot; in (6b), however, rot, in the modifier position of
the compound, cannot be modified. In this prenominal position, sehr can
only modify the entire compound, which is, however, a noun and for this
reason cannot be modified by an adverb.
(6) a. Das ist ein sehr roter Ball

‘This is a very red ball’
b. *Das ist ein sehr Rotball
‘This is a very red_ball’
The non-accessibility of the adjective in the modifier position of the

compound can also account for semantic differences between phrases and
compounds, as illustrated by the examples (7)–(9).2
(7) a. *Das ist ein blauer roter Ball

‘This is a blue red ball’
b. Das ist ein blauer Rotball
‘This is a blue red_ball’
(8) a. *Das ist kein roter Ball, obwohl er rot ist

‘This isn’t a red ball though it is red’
b. Das ist kein Rotball, obwohl er rot ist
‘This isn’t a red_ball though it is red’
(9) a. ??Das ist ein roter Ball, weil er rot ist

‘This is a red ball because it is red’
b. Das ist ein Rotball, weil er rot ist
‘This is a red_ball because it is red’
Example (7) shows that compounds apparently allow for incompatible

attribution whereas phrases do not. That is, adding an attribute that is
semantically incompatible with the modifier yields a semantic contradic-
tion in the case of the phrase but not with the compound. Similarly, the
negation in (8) causes a contradictory reading in the case of the phrase but
not with the compound. And finally, the causal subordinate clause in (9)
leads to a tautological interpretation with the phrase but it is perfectly
informative with the compound.
All these data can be explained on the assumption that the adjective
embedded in the compound is semantically non-accessible. The non-
accessibility of the internal constituents of morphologically complex words
has first been observed by Postal (1969). Postal claimed that complex (as
well as monomorphemic) words are anaphoric islands, i.e. neither the inter-
nal constituents of a complex word nor the entities that constitute part of
the meaning may serve as antecedents for pronominal reference.3 However,
the data discussed by Postal (1969) are restricted to N+N combinations.
Whereas anaphoric reference can easily be made to a nominal antecedent
by means of pronouns, this is more difficult in the case of adjectival ante-
cedents. A proposal for anaphoric reference to an adjective is given in (10)
and indeed it shows that the adjective embedded in the compound cannot
easily be accessed (although such an interpretation is not totally excluded).4
(10) a. Ich habe nicht nur einen roten Ball, sondern auch einen solchen / so einen
Schläger
b. ??Ich habe nicht nur einen Rotball, sondern auch einen solchen / so einen
Schläger
‘I do not only have a red ball but also such a racket’
The question then is how this non-accessibility can be explained. Bücking

(2009, 2010) proposes an explanation on the level of semantic form. He
adopts the abstract modification template MOD* (see (11a)) which is
accompanied by the structural condition in (11b) (see Bücking 2010: 256).5
(11) a. MOD*: λQ λP λx [P(x) ˄ R (x, v) ˄ Q(v)]

b. Condition: if MOD* is applied at the lexical level, then R is instantiated as
Rintegral, if it is applied on the phrasal level, R corresponds to the identity
function.
According to (11a), a modifier-head structure consists of two predicates

P and Q whose arguments x and v stand in a relation R to each other.
According to (11b), R may be spelled out in two different ways. At the
lexical level, i.e. in the case of compounds, R is instantiated as Rintegral.
Rintegral ‘identifies some entity’s integral constituents’ (Bücking 2010: 257).
At the phrasal level, R is the identity function. Phrasal modification, then,
is intersective modification whereas lexical modification can be described
as partial modification. (12) and (13) illustrate how the application of
the template yields the respective interpretation of syntactic and lexical
modification.
(12) a. Syntactic modification: λQ λP λx [P(x) ˄ identity (x, v) ˄ Q(v)]

= λQ λP λx [P(x) ˄ Q(x)]
b. roter Ball: λx [ball (x) ˄ red (x)]
(13) a. Lexical modification: λQ λP λx [P(x) ˄ Rintegral (x, v) ˄ Q(v)]

b. Rotball: λx [ball (x) ˄ Rintegral (x, v) ˄ red (v)]
Thus, according to this analysis, non-accessibility is the result of a partial

modification relation: the adjectival modifier is inaccessible as it is bound to
a non-specified variable v that represents an integral constituent of the head
noun. This analysis can explain the data in (7)–(9), as illustrated in (14): the
(a)-sentence is semantically contradictory because x cannot be blue and red
at the same time. (14b), however, denotes individuals x that are a ball and
blue and that have an integral part v (not specified) that is red. In this way,
the predication of blue and red does not conflict.
(14) a. *blauer roter Ball: λx [blue (x) ˄ ball (x) ˄ red (x)]
b. blauer Rotball: λx [blue (x) ˄ ball (x) ˄ Rintegral (x, v) ˄ red (v)]
However, there are some objections to be raised to this analysis. First, an

intersective analysis for syntactic modification sometimes seems to fail
(or at least to be inaccurate) for intersective adjectives. For cases like (5)
the partial analysis seems to be much more adequate than the intersective
analysis. Lahav (1993) makes a similar argument regarding cases like (15),
which can be perfectly applied to the German phrasal equivalents.
(15) a. braune Kuh ‘brown cow’

b. brauner Kristall ‘brown crystal’
c. brauner Keks ‘brown cookie’
A brown cow is a cow whose body’s surface (or most of it) is brown, but
not the internal organs or the face, a brown crystal must be brown outside
and inside and for a cookie to count as brown it suffices to be covered with
brown sugar. Lahav (1989) argues that this problem applies to almost all
adjectives and that the part of noun referent that is to be modified by the
adjective varies considerably and non-systematically with the types of noun
referents. Consequently, the interpretation of syntactic modification must
be highly flexible, just like lexical modification, and involves contextual and
conceptual knowledge.
On the other hand, the meaning of many compounds can be captured
more adequately with an intersective interpretation than with a partial
one – see (16):
(16) a. Rotwein ‘red wine’

b. Direktflug ‘non-stop flight’
c. Billigfleisch ‘cheap meat’
d. Blaubeere ‘blueberry’
e. Feingold ‘fine gold’
It is the complete wine that is red and not an integral part thereof, and the
flight as a whole is direct, not a part of it. This is not to say that there may
not be semantic specialization in some of these cases (of course, not every
berry that is blue is a blueberry and only gold with a fineness of 999 counts
as fine gold). But the underlying modification structure in all of those cases
is intersective, not partial.
And finally, there are three subclasses of A+N compounds that can be
captured neither with an intersective nor with a partial analysis as they
exhibit a structurally more complex modification relation. In these cases,
the adjective does not modify the head noun but rather an implicit referent
that is semantically related to the head noun. Note that these modification
relations cannot be captured by the partial analysis as this is a relation
between the modifier and an integral part of the head constituent. In the
cases at hand (see (17)–(19)), however, there is a modification relation
between the adjectival modifier and a noun referent which is distinct from
the head noun. In the first subclass, exemplified in (17), there is generally a
causal relation, bringing about a change of state such that the implicit noun
has the property denoted by the adjective.6
(17) a. Gelbfieber ‘yellow fever’

b. Jungbrunnen ‘young fountain’, i.e. fountain of youth
c. Trockenhaube ‘dry hood’, i.e. dryer hood
d. Magersucht ‘meagre addiction’, i.e. anorexia
The compound Jungbrunnen, for instance, does not denote an entity that
is a fountain and young at the same time, nor is any part of the fountain
young. In fact, the internal semantic structure is much more complex: it
is a fountain that causes the person who takes a bath in the water of this
fountain to become young (see also Gelbfieber – section 1).
The class in (18) is similar, but it lacks the causal relationship: the adjec-
tive modifies an implicit noun referent that is semantically related to the
head noun. So warm in (18a) is related to an implicit noun ‘house’, such
that Warmmiete denotes the rent that has to be paid for a heated house.
(18) a. Warmmiete ‘warm rent’, i.e. rent including heating

b. Leerfahrt ‘empty drive’, i.e. journey without cargo
c. Nacktkultur ‘nude culture’, i.e. nudism
d. Trockenschnitt ‘dry haircut’
In the last subclass, as exemplified in (19), the adjective is interpreted as an

adverb and it modifies an implicit verb that is semantically related to the
expressed noun: a Schnellgericht is a meal that is prepared quickly.7
(19) a. Schnellgericht ‘quick meal’, i.e. instant meal

b. Scharfschütze ‘sharp shooter’
c. Simultandolmetscher ‘simultaneous interpreter’
To summarize, the data in (7)–(9) provide evidence for the assumption

that there is a semantic difference between lexical adjectival modification
and syntactic adjectival modification. As these data are newly coined,
non-lexicalized A+N combinations, this difference cannot originate from
idiosyncratic lexicalized meaning aspects. Taking into account various
types of data, I have argued that an approach that operates on the level
of semantic form, analysing syntactic modification as intersective modi-
fication and lexical modification as partial modification, cannot explain
the semantic properties of A+N compounds and phrases in general, as
there are phrases with a partial meaning relation and compounds with an
intersective meaning relation as well as compounds that have a much more
complex semantic structure.
3.2 Lexical modification as classifying modification

The claim advanced here is that there is no difference between lexical and
syntactic adjectival modification in terms of semantic form. The modifica-
tion relation between the adjective and the noun can be very well captured
by a modification template such as that in (11a), given that R receives a
more flexible interpretation so that it can also account for compounds with
a complex internal semantic structure, such as in (17)–(19). But, impor-
tantly, there is no correlation between the different specifications of R on
the one hand and the lexical/syntactic distinction on the other. In fact,
phrases as well as compounds can realize any of the possible specifications
of R. The only exception in this regard are complex internal structures
where the adjective modifies an implicit noun referent as such structures
are only rarely found with phrases.
Instead, lexical modification should be regarded as classifying modi-
fication, which can best be characterized from two perspectives. First,
compounds refer to a complex concept which is a subconcept of the concept
denoted by the head. According to Gunkel and Zifonun (2009, 2011), clas-
sifying modification means that the complex expression refers to a concept
that is a subconcept of the concept denoted by the head alone (i.e. the
hyperconcept) and that this subconcept is created by conceptual restriction
(see also Rijkhoff 2004, 2008). This sets compounds apart from derivations
and exocentric compounds and also from some (though not all) lexicalized
A+N phrases such as grüner Daumen (‘green thumb’), because the subcon-
cept has an inherent systematic relationship to the hyperconcept and also to
potential co-subordinate concepts. This gives rise to contrasts such as (20).
(20) a. Junghase ‘young rabbit’, i.e. leveret

b. alter Hase ‘old rabbit’, i.e. old stager
Junghase has an implicit meaning relation to the hyperconcept Hase as well

as to potential co-hyponym concepts such as Althase lit. old rabbit, ‘leading
rabbit’, but the lexicalized phrase alter Hase does not.
Classifying modifiers, according to Gunkel and Zifonun (2009), need
to meet two conditions: they must be non-referential and restrictive.
Adjectives are not referential in principle (for further discussion of this
point see section 4) and they can, normally, be used restrictively (although
they also may be used non-restrictively), i.e. they are used in order to
determine the reference of the NP, whereas a non-restrictive modifier

adds further information about an independently established referent – see
Partee (1997: 319).
However, an adjective such as rot may be used restrictively both in a
compound and a phrase, so this does not yet explain the difference between
syntactic and lexical modification. The second claim therefore is that
compounding creates a new, self-contained concept. This sets compounds
apart from phrases. Syntactic modification may be restrictive and it adds
information to one particular instance of the concept denoted by the head,
such as in (21).
(21) Ich möchte den roten Ball kaufen (nicht den blauen)
‘I want to buy the red ball (not the blue one)’
Importantly, however, no new concept is created. The phrasal modifier

specifies one particular instance of the concept denoted by the head but
leaves the concept as it is. Lexical modification, on the other hand, yields
a new, individual subconcept. Similar distinctions have been made by
Bolinger (1967; reference vs. referent modification) and Rosenbach (2006,
2007; type vs. token restriction).
Gunkel and Zifonun (2011) introduce several structural types of clas-
sifying modification among which are N+N compounds and A+N phrases
with relational adjectives, such as medizinische Untersuchung ‘medical
examination’ or häusliche Gewalt ‘domestic violence’.8 Obviously, A+N
compounds are also generally classifying.
If lexical modification creates a new, individual subconcept, it can easily
be explained why compounds are well-suited for use as common nouns and
why they often (though not in general) become lexicalized (see also Dahl
2004). Lexicalization, in turn, often comes along with semantic specializa-
tion. More precisely, if it is the inherent function of lexical modification to
identify an individual subconcept then this subconcept may be narrowly
delimited. This is exactly what semantic specialization means – see (22).
(22) a. dicker Hase ‘fat rabbit’

b. Junghase ‘young rabbit’, i.e. leveret
Junghase is not only a rabbit that is young (such as a dicker Hase is a rabbit
that is fat) but it also has an additional meaning component ‘offspring’, that
is it is the term used in order to denote the next generation. On this basis
we might conclude that compounds are inherently apt for semantic spe-
cialization, but semantic specialization is bound to lexicalization. Rotwein
‘red wine’ and Direktflug ‘non-stop flight’, on the other hand, are examples
of lexicalized compounds that refer to established subconcepts but do not

exhibit semantic specialization: every red wine is a Rotwein, and every
direct flight a Direktflug, without further meaning restrictions.
If compounds are inherently classifying, speakers will use compounds
rather than phrases if they want to express that an adjective and a noun fuse
into a new single complex concept.9 On the other hand, hearers use this
conceptual knowledge when interpreting new, unknown A+N combina-
tions: when encountering an unknown A+N compound they will accom-
modate the existence of such an individual, self-contained subconcept.
In the light of this view on compounds, it is interesting to consider again
the formal properties of compounds and phrases which, as stated above,
can be clearly distinguished from each other, among other things, by stress.
First, as pointed out by Eisenberg (2002), the initial stress on the
modifier constituent of compounds can be interpreted as morphologized
contrastive stress. That is, the stress on the modifier is used to refer to
alternative subconcepts of the head concept in the case of compounds,
but not in the case of phrases, as exemplified in (20). Second, the initial
stress on the modifier constituent in compounds can be interpreted as
an indication for a holistic interpretation of these complex expressions.
Jacobs (1993, 1999) suggests a connection between the stress patterns of
heads and their sister constituents and their semantic properties. The
basic idea is that there is an ‘integrational’ relation between two sister
constituents that conflate them to one single semantic entity in such a way
that their meaning is ‘semantically processed at one fell swoop’ (Jacobs
1999: 57). That is, those entities are referred to holistically, there is ‘only
one property, although its quality is determined by combining the mean-
ings of the two constituents.’ The relation of integration is asymmetric
as there is always one constituent that is integrated into the other, so that
it does not correspond to a separate processing step. The stress pattern
indicates whether there exists such a relation of integration between two
constituents or not: in case of integration the integrated constituent bears
the stress, otherwise the stress is on the head. This is exactly what is found
in the case of A+N compounds and phrases: the compounds have main
stress on the modifier. Thus the meaning of the adjective is not processed
separately but forms a complex semantic entity with the head noun. In
the case of A+N phrases, however, the main stress is on the head noun,
indicating that the adjective and the noun are to be processed separately
and that they do not form a semantic unity. That is, the initial stress on
the adjective in the case of a compound works as an indication for the
hearer that the processing of the adjective has to be postponed and that it
rather has to be interpreted together with the noun as part of the complex
meaning unit.
3.3 The identification of the subconcept

In this section, I will discuss the way in which lexical modifiers identify new
subconcepts and provide an alternative explanation for the data in (7)–(9).
As stated above, there are several kinds of adjectival modification relations,
including intersective and partial modification as well as more complex
modification structures involving an implicit referent. The lexical modifier
singles out a salient feature of the subconcept but this is not (necessarily)
the only one that discriminates this particular subconcept from other co-
subconcepts. It is just one that seems to be suitable to identify this particu-
lar subconcept. Often, therefore, these are properties that can be observed
in one way or another, such as outer shape, localization, taste, sound, etc.
However, less observable properties such as abstract properties and prop-
erties referring to the function of the subconcept are also found, as in (23).
(23) a. Großhandel ‘big trade’, i.e. wholesale trade

b. Vollmilch ‘full milk’, i.e. whole milk
c. Magermilch ‘meagre milk’, i.e. low-fat milk
d. Intensivstation ‘intensive ward’, i.e. intensive care unit
e. Rundschreiben ‘round letter’, circular letter
Even if the property denoted by the adjectival modifier is not at all predi-
cated to the head noun, as it is the case with compounds with an implicit
referent, this modifier can be used in order to identify a subconcept of the
head noun. This can be seen from the examples in (17)–(19) as well as from
the recently attested, non-established examples in (24).
(24) a. Schlankfrucht ‘slim fruit’

b. Nacktgebot10 ‘naked instruction’
The first example refers to fruit that makes one stay (or become) slim, namely
melons (which certainly do not have a slim shape themselves), the second to
usage instruction(s) concerning nudity for sauna visitors. It is, however,
impossible to receive the same interpretation for the corresponding phrases
schlanke Frucht and nacktes Gebot. It seems, then, that any kind of modifica-
tion relation can be used for identifying a subconcept of the head noun.
We can now get back to the data in (7)–(9). Comparing the (a)-sentences,
repeated here as (25), reveals that the reader has to accommodate different
interpretations of the (non-existent) concept ‘Rotball’.
(25) a. Das ist ein blauer Rotball

‘This is a blue red_ball’
b. Das ist kein Rotball, obwohl er rot ist

‘This isn’t a red_ball though it is red’
c. Das ist ein Rotball, weil er rot ist
‘This is a red_ball because it is red’
The surface of ‘Rotball’ as referred to in (25a) is obviously not red. Rot

may instead be used in order to indicate the function of this concept, for
instance marking the road surface with red signs. It may also be the case that
the concept ‘Rotball’, linked to a particular subconcept, originally meant
‘ball with a red surface’ but that in the course of time the colour changed
while the concept remained stable.11 The surface of ‘Rotball’ as referred
to in (25b) may be red or not; in any case this is not the reason for it being
called a ‘Rotball’. Only in the concept of ‘Rotball’ as referred to in (26c) is
the red surface the salient feature that indicates this particular subconcept.
Note, however, that there must be other properties, too, that distinguish the
concept ‘Rotball’ from the concept ‘Ball’, because otherwise (25c) would be
not informative but rather tautological, as its phrasal counterpart in (9a).
The acceptability of the sentences in (25) as opposed to their phrasal
counterparts (7b), (8b) and (9b) can be explained on two assumptions.
First, lexical modification is classifying modification. The function of the
lexical modifier is to identify a new, self-contained concept. The modifica-
tion relation is very flexible and it can modify the noun referent as well as
any other, semantically related referent. Second, speakers generally try to
give a meaningful interpretation to any sentence they hear. It is not at all
problematic to infer three different interpretations of the newly coined
compound Rotball as in (26) in order to yield three meaningful, acceptable
sentences, given the classifying function of compounds as described.
4 NOMINAL LEXICAL MODIFICATION
In this last section I will compare adjectival and nominal lexical modifica-
tion. The preceding sections have presented both shared and different
properties of A+N and N+N compounds.
To start with the differences, nominal lexical modification gener-
ally includes an additional relational meaning predicate, such as
INSTRUMENT, LOC, PART-OF etc., as illustrated in (26).
(26) a. Fingersprache INSTRUMENT (x, y)

‘finger language’‚ i.e. sign language
b. Fingerring LOC (x, y)
‘finger ring’
c. Fingerspitze PART-OF (x, y)

‘fingertip’
Adjectival lexical modification, on the other hand, lacks such a relational

predicate, although the meaning relation between the adjective and the
noun is not always straightforward, as discussed above. So, the modifica-
tion template introduced in (11a) might also be used for nominal lexical
modification, given an additional condition that accounts for the fact
that R in N+N compounds is to be specified as a lexically filled relational
meaning predicate. In fact, such analyses have frequently been proposed in
the literature – see Motsch (2004), Guevara and Scalise (2009), Jackendoff
(2010), among many others.
On the other hand, adjectival as well as nominal lexical modification
seems to be classifying modification. According to Gunkel and Zifonun
(2009), non-referentiality is a precondition for classifying modification.
They argue that adjectives are inherently non-referential, in contrast to
nouns.12 However, from a semantic point of view, nouns are referential
(or non-referential) in the same way as adjectives are. Whether a noun is
referential depends on the particular use of the noun phrase it is part of.
So, to be more precise, a modifier may not be used referentially in order
to serve as a classifying modifier. This explains very well ‘[. . .] the well-
known fact that referential expressions do not fit well into the internal
structure of lexical items, which has its counterpart in the apparently uni-
versal reluctance to incorporate definite noun-phrases’ (Dahl 2004: 180),
as illustrated in (28):
(28) a. *Die-Birnenblume
‘the pear flower’
b. Birnenblume
‘pear flower’
Referential modification, i.e. modification by means of a modifier that refers

to an individual non-linguistic object, seems to be restricted to phrasal
modification, such as in (29), and it does not have a classifying function.
(29) a. das Zimmer mit den bunten Bildern

‘the room with the colourful pictures’
b. der Tisch meines Vaters
‘the table of my father’
However, as has often been observed in the literature, there are also
non-
classifying N+N compounds with a modifier that is interpreted
r eferentially. These compounds do not denote self-contained subconcepts.

Their function has rather been characterized as descriptive, deictic or
discourse-structuring. They are used to compress information, e.g. in a
newspaper headline (see, among others, Downing 1977; Kastovsky 1982;
Booij 2009a). Downing’s (1977) well-known apple juice seat forms an illus-
trative example: this compound (as well as its German counterpart) can be
used, in a particular situation, in order to refer to a seat in front of which
stands a glass of apple juice. Certainly, this use does not imply the existence
of a subconcept ‘apple juice seat’. Interestingly, however, such a reading is
dependent on an appropriate context. In context-free use the hearer will
always accommodate an adequate subconcept.
Other examples are taken from newspaper headlines. The compounds in
(30) do not denote particular subconcepts of debates or videos respectively
but they are condensed expressions, referring to recent events.
(30) a. Grundschul-Debatte: Was ist wirklich gut für unsere Kinder?

‘Primary school debate: what is really good for our children?’
b. Tierquäler-Video – Esel an Fallschirm gebunden
‘Animal torturer video – donkey tied to parachute’
For instance, (30a) refers to a recent debate on the referendum on the dura-
tion of primary schooling in Hamburg. In particular, the modifier nouns
are interpreted referentially, referring to a specific primary school system
and a specific animal torturer respectively.
Similarly, if the modifier is a proper name, as in (31), this modifier refers
to an individual non-linguistic object and is therefore often interpreted
referentially, such that the compound does not receive a classifying inter-
pretation (and is therefore normally not lexicalized – see Gaeta and Ricca
2009).13
(31) a. Rumänien-Diktator
‘Romania dictator’
b. Berlusconi-Prozess
‘Berlusconi process’
The data from nominal lexical modification thus confirm the idea that the
(default) meaning of lexical modification is classification. They also support
the claim that classifying modifiers must be interpreted non-referentially.
Accordingly, A+N compounds are always classificatory whereas N+N
compounds may or may not be classificatory.
5 CONCLUSION
In this chapter, I have argued that the semantic structure of German A+N
compounds is far from being clear or uniform across the data. In the litera-
ture, relatively little attention has been paid to the internal semantic struc-
ture of A+N compounds. Quite often, the semantics of A+N compounds
are described as a simple ‘attributive’ structure, i.e. the property denoted
by the adjectival modifier is predicated to the nominal referent. However, I
have presented various data supporting the claim that there is much more
variation and argued that, contrary to N+N compounds, this variation does
not stem from an underspecified mediating variable, but rather from the
fact that the referent the adjectival modifier is predicated to may be either
the head noun referent or an implicit, semantically related referent.
The second claim is that A+N compounds and A+N phrases do not
differ with regard to their internal semantic structure but that in prin-
ciple the same range of semantic relations can be found both with A+N
compounds and phrases. The only difference in this regard are complex
modification relations with an implicit noun referent to be modified as
such relations are almost exclusively found with compounds. However,
the different semantic behaviour A+N compounds and phrases exhibit
is not related to the underlying modification relation but to the fact that
compounds denote new, self-contained, stable subconcepts of the concept
denoted by the head noun whereas phrases add information to one par-
ticular instance of the concept denoted by the head, leaving, however, this
concept unchanged.
ACKNOWLEDGEMENTS
This research was supported by a grant from the Deutsche

Forschungsgemeinschaft (HU 1635/1-1). I would like to thank the audi-
ence of the workshop ‘Meaning and Lexicalization of Word Formation’
at the 14th International Morphology Meeting 2010 in Budapest for the
stimulating discussion. Special thanks go to Martin Schäfer.
NOTES
1. In view of the problems of differentiating between English compounds

and phrases, Sweetser (1999) does not make a difference between
A+N compounds and A+N phrases with regard to the semantics
of English A+N modification. Other approaches, such as Jespersen
(1942: 137), claim that English compounds can only be identified on

the basis of their deviant semantic properties. Bloomfield (1933: 227),
on the other hand, argues that semantic specialization (as exemplified
in blackbird as opposed to black bird) is not an appropriate criterion for
compoundhood.
2. These are similar to examples discussed in ten Hacken (1994: 99–100)
and Bücking (2009: 185).
3. See also Schäfer (this volume).
4. Ten Hacken (1994: 100) argues that there is no test that can identify
the syntactic category of the modifier constituent of a compound.
He suggests that an adjective such as rot ‘red’ in Rotbus may also be
analysed as a noun ((das) Rot ‘(the) red’) as adjectives can be converted
into nouns on a regular basis, such that the pronominal reference
test can be applied. The test shows that pronominal reference to the
alleged noun Rot is excluded. However, this argument does not work
for all adjectives. Importantly, the nominalized forms of most adjec-
tives do not coincide with the base form as they end with a schwa, see
hart – das Harte ‘hard’, alt – das Alte ‘old’. Also, the fact that adverbial
modification of the modifier constituent is excluded (see (6b)) does not
provide evidence against the adjectival status of this constituent but
rather indicates that the modifier constituent is semantically unacces-
sible. While I am not convinced that the adjective can be analysed as a
converted noun, from a broader, cognitive view it can be argued that
the word class is of minor importance as the modifier serves to identify
a subconcept indepent of the word class it belongs to (see the examples
in (2)), or, as Jackendoff (2010: 424) puts it: ‘the syntax of English
compounding is to some extent even blind to syntactic category’. From
this point of view, the modifier constituent seems to be constrained by
semantic and pragmatic restrictions rather than by syntactic category
restrictions (see ten Hacken 2003b).
5. This template is based on a proposal by Maienborn (2003).
6. Related classifications can be found in Ortner and Müller-Bollhagen
(1991), Simoska (1999) and Motsch (2004).
7. Yet another semantic subclass of A+N compounds are the so-called
Bahuvrihi or possessive compounds (exocentric compounds), such as
Rothaut ‘redskin’, Dummkopf ‘dumb head’ or Rotkehlchen ‘red throat’,
i.e. robin redbreast. However, their internal semantic structure can be
regarded as an ‘ordinary’ intersective relation. Crucially, those com-
pounds receive a metonymic interpretation – see Booij (2007).
8. Note the semantic difference between A+N phrases with relational
adjectives and A+N phrases with non-relational adjectives: while the
former denote new, self-contained subconcepts, this is not true for
the latter as they do not receive a classifying interpretation – see teure

Untersuchung ‘expensive examination’, große Gewalt ‘heavy violence’.
Admittedly, there are also cases of A+N phrases with non-relational
adjectives that nevertheless have a classifying meaning, e.g. saure
Sahne ‘sour cream’. However, within the limits of this paper, they
cannot be further discussed.
9. This is of course reminiscent of the view expressed in morphological/
typological works that word formation entities provide names whereas
syntactic entities provide descriptions – see Bauer (1988). The claim
that compounds are inherently classifying is closely linked to the idea
that they exhibit a naming function. However, the question of naming
forms an issue on its own, as this notion is far from being clear and
uncontroversial. For this reason, it is not addressed in the present
chapter.
10. (a) BILD-Zeitung, 21-07-2010; (b) Tagesspiegel, 14-02-2010
11. For a similar example involving Rotbus (‘red bus’), see ten Hacken
(1994: 99).
12. A similar, typological view is taken in Croft (2000) who states that
the (unmarked) pragmatic function of nouns is reference to an object,
whereas it is modification by a property in the case of adjectives.
13. There are also cases of lexicalized compounds with a proper name
modifier, such as Molotowcocktail ‘molotov cocktail’ or Röntgenstrahlen
‘X-rays’. In these cases, however, the proper name is not inter-
preted referentially. Rather, it has a commemorative function, i.e. the
subconcept is named after the person denoted by the proper name
modifier – see Koptjevskaja-Tamm (2009).
CHAPTER 8
Semantic transparency and

anaphoric islands
Martin Schäfer
T his chapter investigates the relationship between semantic transpar-

ency of compounds and their status as anaphoric islands. More specif-
ically, I will take a detailed look at the behaviour of German adjective-noun
compounds in this respect. The chapter argues that semantic transparency
plays a crucial role in accessing compound-internal components for ana-
phoric reference and discusses a number of factors that motivate the actual
usage of anaphora.
1 INTRODUCTION
Postal (1969) argued that words, whether monomorphemic or derived, are

anaphoric islands.1 That is, neither internal constituents of morphologi-
cally complex words nor entities contained in the meaning of a word can
serve as antecedents to a following anaphoric element, or, in Postal’s words,
allow outbound anaphora.2 Two pieces of data supporting his claim are
reproduced in (1) and (2), his (3) and (53), respectively.
(1) a. Max’s parentsi are dead and he deeply misses themi.

b. *Max is an orphan and he deeply misses them.
(2) a. Harry was looking for a rack for booksi but he only found racks for very
small onesi.
b. *Harry was looking for a bookrack but he only found racks for very small
ones.
In (1a), the pronoun them refers anaphorically to the referent of the noun
phrase Max’s parents. In contrast, in (1b) them cannot refer to Max’s
semantic transparency and anaphoric islands 141
parents, although the meaning of the word orphan ‘involves reference to the
parents of an individual’ (Postal 1969: 206). Example (2) aims to show that
even complex words are islands. In the first sentence, the noun phrase books
can be picked up by the pro-form one in the second clause, contrasting with
the morpheme book contained in the compound bookrack in the second
sentence, which cannot be picked up by the pro-form one.
In contrast to Postal’s findings, German adjective-noun-compounds do
not act as barriers for anaphoric reference – see (3) and (4).
(3) Ich bin das Grünglas losgeworden, das weiße liegt noch im Auto.
‘I am the green-glass got.rid.off, the white lies still in.the car’
i.e. I got rid of the green glass, the white glass is still in the car.
(4) Ich liebe Großstädte, in kleinen gehe ich ein.

‘I love big-towns, in small go I in’ [cf. ein-gehen ‘to perish’]
i.e. I love big cities, I cannot exist in small cities.
Standardly, both examples are interpreted as involving anaphoric reference

by an empty element to the head of the compound. Thus we can assume
the structure in (5) for (4), where PRO stands for an empty element:
(5) Ich liebe [Groß[städte]i], in kleinen PROi gehe ich ein.
On the other hand, this kind of anaphoric reference does not seem to be
available across the board, as, for example, (6) shows, where # marks
pragmatic deviance.
(6) #Mein Vater hat in seinem Garten schon mal einen Grünspecht gesehen, aber
noch nie einen schwarzen.
‘My father has in his garden already once a green-woodpecker seen but so far
never a black.’
Intended: My father once saw a green woodpecker in his garden, but he has
never seen a black woodpecker.
In (6), we have the same basic configuration, that is an AN construction,

followed by an A pro-form construction, but the strongly preferred inter-
pretation is with anaphoric reference to the whole compound, not to just
its head. Thus, instead of the intended interpretation of the second, elliptic
AN construction as Schwarzspecht ‘black woodpecker’, the preferred inter-
pretation is schwarzer Grünspecht, ‘black green woodpecker’.
These two points, first that there are counter-examples to anaphoric
islandhood, and secondly that anaphoric reference is sometimes more or
142 martin schäfer
less easily available, are not surprising, given that both aspects have been
repeatedly pointed out in the literature on English. Why then this chapter
with its focus on AN compounds in German? Firstly, two recent papers
dealing with AN constructions use the status of compounds as anaphoric
islands as a test for assigning compound or phrasal status to A N construc-
tions for which no other measures are available (see Giegerich 2005 for
English and Paul 2005 for Mandarin Chinese AN constructions), making it
worthwhile to gather together the counter-evidence to the idea of anaphoric
islandhood as a syntactic test once again. Secondly, the pattern found in the
German data cannot be found in English, allowing us to elaborate on the
mechanism behind these types of anaphoric relations.
The chapter is organized as follows: section 2 introduces two rival
accounts of anaphoric islandhood data and reviews the psycholinguistic
evidence discussed in this context. Section 3 discusses the details of
the German data and some findings from an explorative corpus study.
Section 4 concludes the chapter.
2 ACCOUNTING FOR ANAPHORIC ISLANDS
The pattern in Postal’s data has been interpreted in two ways. In the first
view, the pattern is taken as evidence for the existence of an anaphoric
island and the relationship between the anaphoric element and its anteced-
ent is seen as a syntactic relation. Based on this interpretation, the pattern
has also been used as a diagnostic for compoundhood. In the second view,
the key to the patterns lies in the pragmatics involved. I will discuss the two
accounts in turn, dismissing the former in favour of the latter.
2.1 A syntactic interpretation

A syntactic interpretation contains two core ingredients: (a) a classical
anaphora account and (b) a strong version of the lexicalist hypothesis (also
referred to as lexical integrity hypothesis). (7) contains a typical state-
ment of the classical anaphora account, a quote from Hankamer and Sag’s
description of what ‘by now may be called the “classical” position’ (1976:
394).
(7) Classical anaphora account:

‘[A]ll anaphoric processes are transformations that involve deletion (or
conversion to a pro-form) of an underlyingly present, fully lexical segment
under conditions of identity with an antecedent segment; [. . .]’
(Hankamer and Sag 1976: 394)
The lexicalist hypothesis can likewise be given in a maximally strong form–

see (8).
(8) Lexicalist hypothesis [maximally strong version]:

‘The syntax neither manipulates nor has access to the internal form of words’
(Anderson 1992: 84)3
In weaker variants of the lexicalist hypothesis, anaphoric reference to

word-internal constituents is allowed, for instance Selkirk’s (1982: 70)
Word Structure Autonomy Condition.
With (7) and (8) in place, the difference in the pair of example sentences
in (2), repeated here for convenience, is accounted for as follows.
(9) a. Harry was looking for a rack for booksi but he only found racks for very
small onesi.
b. *Harry was looking for a bookrack but he only found racks for very small
ones.
The underlying form of small ones in sentence (9a) is small books, where
books, stripped of its inflectional ending, is a fully lexical segment that is
identical to the preceding segment book in the phrase a rack for books. The
conditions of the classical anaphora account are fulfilled and the second
occurrence of books can be replaced by the pro-form ones. The very same
conditions seem to be met in (9b): we have the underlying small books,
with book again being identical to book in bookrack. However, this time,
it is blocked by the Lexical Integrity Hypothesis: the syntactic process of
anaphora formation is not allowed to look into the compound bookrack.
Therefore, no identity can be established and no pro-form can be intro-
duced. Following the logic of this explanation, this pattern, or rather the
unavailability of this pattern, has been used to establish whether AN con-
structions in languages with few if any morphosyntactic marking of com-
pounds are compounds or not. A simple example from English will clarify
the logic behind this test. English has no clear morphosyntactic criteria for
AN compounds, because the word forms of the adjective and the noun are
unaffected by phrasal or compound status, and secondary criteria like spell-
ing and stress placement are not decisive, that is a compound does not have
to be spelled as one word, nor does it necessarily need to carry main stress
on its left-hand constituent.4
Thus an AN construction like young dogs could, in principle, be either
a phrase or a compound. However, employing lexical integrity and the
classical anaphora account, we can settle this issue. First, we need to find a
sentence containing a second AN construction with identical head, e.g. old
144 martin schäfer
dogs in I like young dogs, but Sue prefers old dogs. Secondly, we need to check
whether it is possible to delete or replace the second head. In this case, it is
possible – see John likes young dogs, but I prefer old ones. Combining the clas-
sical anaphor account and lexical integrity, there is only one possible way in
which this deletion/substitution can take place, schematically given in (10).
(10) Schema for head noun deletion/substitution:

a. A N . . . A N [Two AN constructions]
b. A Ni . . . A Ni [Head-Head identity can be established]
c. A Ni . . . A Ni or A pro-formi
[Head of second AN is deleted or replaced by a pro form]
In order to end up with the configuration in (10c), that is with a deleted

or replaced head of the second AN construction, we need to start with
two AN constructions in (10a) where the head to be deleted or replaced is
still present. Only then can we identify two identical lexical segments, as
required by the classical anaphora account, in this case the two heads, as
indicated in (10b) through the use of the identical subscript. At the same
time, the identity of the two heads can only be established if the first head
can be seen by the syntactic component. According to the maximally strong
version of the Lexical Integrity Hypothesis, this is only possible if the
first AN construction is phrasal. Therefore the availability of the ANi . . .
ANi/A pro-formi pattern shows that the first AN construction is phrasal.
Recently, this argument has been employed in Giegerich (2005) and Paul
(2005) for AN constructions in English and Mandarin Chinese, respec-
tively. Thus, according to the pattern outlined in (10), Giegerich’s data in
(11) – see (4) in Giegerich (2005: 579ff.) – can be taken to show that medical
appointment is a phrasal construction whereas mental disorder is a compound.5
(11) a. Do you have a medical appointment or a dental one?

b. *Is this a mental disorder or a nervous one?
Paul also uses the anaphoric island diagnostic in her discussion of Mandarin
AN compounds, a representative minimal pair being (12), taken from (51a)
in Paul (2008) and (19) in Paul (2005).
(12) a. *Wǒ xǐhuān lǜ-chá, hóng de yě kěyǐ

I like green-tea red SUB also possible
i.e. ‘I like green tea, but black [lit. red] tea is also OK.’
b. Amei bù xǐhuān huáng méigui, hóng de hái kěyi
Amei NEG like yellow rose red SUB still acceptable
i.e. ‘Amei doesn’t like yellow roses, red ones are still OK.’
According to the syntactic reading of the contrast in (12), which is based on

the parallel A N constructions lǜ-chá ‘green tea’ and hóng -chá ‘black [lit.
red] tea’, the ungrammaticality of (12a) shows that lǜ-chá ‘green tea’ is a
compound, whereas the grammaticality of (12b) shows huáng méigui ‘yellow
rose’ (paired with hóng méigui ‘red rose’) to be a phrasal A N construction.
The problem with the argument is that there are counter-examples for
almost every prediction of the syntactic account, which are briefly summa-
rized below, where we will start with some general remarks and then focus
on counter-evidence involving compounds in English.
In general, the counter-evidence shows two important things. Firstly,
it shows that the syntactic explanation as stated by Postal is insufficient.
Secondly, it shows that any explanation for the data must be able to explain
different degrees of acceptability. This cline in acceptability was observed
in one of the first reactions to Postal’s original paper, a paper by Lakoff and
Ross (1972) giving the data and judgements reproduced in (13), their (2b)
and (3a–b), where one in (13a) and it in (13b) and (13c) are intended to refer
to the guitar.
(13) a. *A guitarist bought one yesterday

b. ?*The guitarist thought that it was a beautiful instrument.
c. ?John became a guitarist because he thought that it was a beautiful instrument.
Clearly, this cline cannot be explained by Postal’s original proposal, which

makes a categorical difference between islands and non-islands. Other
authors offering counter-examples to Postal’s strong claim include Tic
Douloureux (1971), Corum (1973), Browne (1974) and Watt (1975), whose
main claims and accounts are discussed in Ward et al. (1991) as well as
Levi (1977). A representative set of counter-examples involving English
compounds is presented below, first with anaphoric references to the first
element of the compound – see (14) – and secondly with anaphoric refer-
ence to the second part of the compound – see (15).
(14) a. Although casual cocainei use is down, the number of people using iti
routinely has increased.
b. Patty is a definite Kal Kani cat. Every day she waits for iti.
c. I was an IRSi-agent for about 24 years . . . I stopped working for themi.
The examples in (14) are from the appendix to Ward et al. (1991). Cocaine
use in (14a) is a synthetic compound which should disallow anaphoric
reference to its two constituents. However, it refers back to cocaine and
not to cocaine use. Similarly, it in (14b) refers back to the denotation of Kal
Kan, that is to a specific brand of catfood, where Kal Kan is embedded in a
146 martin schäfer
standard endocentric compound, and finally, in (14c), it refers back to the

IRS.6
The data discussed by Ward et al. (1991) is restricted to anaphoric refer-
ence to the first element of the compound, whereas Levi (1977) presents
data where we see that reference to the second element is also possible – see
(15), Levi’s (17b), (18b) and (19a).
(15) a. State taxesi were higher than municipal onesi.

b. Steam ironsi need more maintenance than thosei that iron dry.
c. Student poweri is insignificant compared to thati of the Dean.
In (15a), ones refers back to the denotation of taxes and not to that of state
taxes, those in (15b) refers to that of irons and not to the denotation of steam
irons, and finally that in (15c) refers back to the denotation of power and not
to the denotation of student power.
The German data presented in section 1 strengthens Levi’s observation
insofar as the first A N constructions are clear compounds according to
morphosyntactic criteria, whereas for English similarly clear-cut criteria
are not available. In addition, the difference between the data in (3)–(4) and
(6) shows also a cline in acceptability.
Thus, while the patterns observed by Paul and Giegerich still need an
explanation, it seems clear that anaphoric islandhood of the antecedents
is not the correct one. The main alternative to the original proposal is the
pragmatic account outlined below.
2.2 A pragmatic interpretation

Ward et al. (1991) offer a pragmatic account for the anaphoric island data,
summed up in the following quote: ‘. . . the degree to which outbound
anaphora is felicitous is determined by the relative accessibility of the
discourse entities evoked by word-internal lexical elements, and not by any
principle of syntax or morphology’ (1991: 449). The question that emerges
from this account is, naturally, what then determines the relative accessibil-
ity of the discourse entities involved. Ward et al. (1991) distinguish two sets
of factors: (a) morphosyntactic/semantic factors and (b) pragmatic factors,
that is contrastiveness and topicality. They discuss semantic transparency
as the key morphosyntactic and semantic factor. Semantic transparency of
a compound is understood by Ward et al. to involve accessing the meanings
of both parts of the compound. Thus they claim that interpreting cocaine
use and Kal Kan cat in the examples (14a) and (14b) above both require the
hearer to access the meanings of the two constituents of the compounds.
Due to this access to the meaning of the individual constituent, the corre-
sponding discourse entities are evoked and, in turn, available for anaphoric
reference. In contrast, institutionalized, idiosyncratic compounds can
acquire opaque meanings, that is meanings that cannot be ‘straightfor-
wardly’ (1991: 454) interpreted on the basis of the meanings of the con-
stituents of the compound. Once a compound has acquired an idiosyncratic
meaning, ‘a hearer may access the meaning of the compound directly, i.e.
without morphologically decomposing it’ (1991: 454). In consequence, the
potential discourse entities are not evoked. This explains the pragmatic
deviance of (16), Ward et al.’s (23a).
(16) Fritz is a [cow]iboy.

# He says theyi can be difficult to look after.
While this explanation itself rests on a categorical contrast (either the

meanings of the individual constituents are accessed or not), Ward et al. are
careful to point out that opacity is a gradient phenomenon: ‘. . . the distinc-
tion between transparent words and opaque or institutionalized words is
gradient rather than categorical. We would therefore expect word-internal
morphemes to evoke discourse entities with a greater or lesser degree of
accessibility depending, inter alia, upon the relative transparency of the
containing word’ (1991: 455). However, a relative measure of semantic
transparency is not introduced by the authors.
Interestingly, Ward et al. argue that ‘While semantically transparent
compounds do allow felicitous outbound anaphora, it is also true that
anaphora involving antecedents within compounds is, other things being
equal, more difficult to construe than anaphora involving non- word-
internal antecedents’ (1991: 455). To account for this, they speculate that it
may have something to do with the fact that in modifier-head constructions,
which the compounds in their data essentially are, the modifiers are more
backgrounded than the heads, a claim they support by referring to psycho-
linguistic evidence from McKoon et al. (1990).7 Note that this explanation
only targets anaphoric reference to the non-heads of compounds, whereas
the German data introduced in section 1 (examples (3), (4) and (6)) in all
cases involves anaphoric reference to the head of the compound.
As for pragmatic factors, Ward et al. (1991: 456) discuss contrast and
topicality. Discourse entities seem to be more accessible if in salient oppo-
sition to some other discourse entity (see also Watt 1975): ‘topical discourse
entities evoked by word-internal elements facilitate outbound anaphora
. . .’ (Ward et al. 1991: 456). The role of topicality in the facilitation of
outbound anaphora is also discussed by Bosch (1983: 238, note 116), who
notes that this factor might be responsible for the difference in acceptability
in the following example sentences from the original Postal paper:8
148 martin schäfer
(17) a. Prejudice against Jews makes rich ones donate money.

b. *Anti-Semitism makes rich ones donate money.
(18) a. Max hunts for wild animals but Pete only kills domesticated ones.
b. *Max is a wild-animal hunter but Pete only kills domesticated ones.
(19) a. Harry solicits for prostitutes and Pete arrests them.

b. *Harry is a pimp and Pete arrests them.
All three b-sentences contain different examples for anaphoric islands, e.g.
a neoclassical derivation in (17b), an [[AB]C] compound in (18b) and a
monomorphemic word in (19b), yet for all three ‘islands’, the same, simple
explanation can account for the missing availability of anaphoric reference:
‘The (a) sentences can be about Jews, wild animals, or prostitutes respec-
tively, and accordingly anaphoric relations linking up to the correspond-
ing referents are possible. In the (b) sentences there is no possibility for
corresponding aboutness relations and hence no anaphora either’ (Bosch
1983: 238). In the German data, topicality and contrastiveness are inherent
features of the pattern under investigation.
2.3 Psycholinguistic evidence

The pragmatic account is partially supported by findings from psycho-
linguists. On the one hand, Ward et al. (1991) report the results of a
number of psycholinguistic experiments9 that show the importance of
the two factors of transparency and topicality. On the other hand, their
claims can be fruitfully linked to other psycholinguistic research on
compounds.
Ward et al. (1991) report experiments where the accessibility of dis-
course units serving as antecedents for anaphora was manipulated by
(a) varying morphosyntactic structure (the antecedent either occurred in
a nominal compound or in a verb phrase) and (b) varying topicality and
contrast. The two most important results of the experiments are (a) when
the antecedent is topical, there is no significant difference in reading times
regardless of the antecedent occurring in a compound or not and (b) when
the antecedent is non-topical, the reading times for the compound internal
versions and the VP version differ significantly. In McKoon et al. (1993),
these findings are accounted for by assuming that we use a discourse
model, i.e. a continually updated representation of information built up in
comprehension of texts. The model ‘is made up of the entities evoked by
linguistic and contextual information, the relations among the entities, and
their accessibilities relative to potential referential cues’ (McKoon et al.
1993: 72). Importantly, they propose that a number of factors, such as topi-
cality and morphosyntactic as well as extralinguistic context, influence the
accessibility of a given discourse entity, and that the referential cue, i.e. the
anaphoric element itself, also determines relative accessibility. Therefore
what might be, at a specific position in the text, accessible through use of a
personal pronoun might not be accessible by other pro-forms in the same
position.10
Another aspect of compounds, namely their internal semantics, has also
been the subject of psycholinguistic investigations. Zwitserlood’s (1994)
study on Dutch compounds addresses exactly this point. Zwitserlood clas-
sifies the compounds used in her study into three different groups: (a) fully
transparent compounds; (b) truly opaque compounds; and (c) partially
opaque compounds. In the case of fully transparent compounds, the
meaning is synchronically related to the meaning of the individual con-
stituents, as in, for example, milkman. The truly opaque compounds, in
contrast, bear no semantic relation to any of the constituents. Zwitserlood
uses the English blackguard ‘one of the idle criminal class’ to illustrate
this; a Dutch example she uses is klokhuis, lit. ‘clock house’ but meaning
‘core of an apple’. Finally, partially opaque compounds are linked to the
original meaning of one of the constituents, e.g. jailbird refers to a person
that is often in jail. The most important finding of Zwitserlood’s experi-
ments is that all compounds, including the truly opaque compounds,
are represented as morphologically complex at some level. Transparent
compounds and partially opaque compounds11 facilitate semantic access to
the meanings of their constituents. This is interpreted by Zwitserlood as
evidence that transparent and partially opaque compounds are linked (a) to
their own semantic representation and (b) to the semantic representation
of their constituent words. Truly opaque compounds, in contrast, behave
semantically like monomorphemic words in that they are only linked to
their own semantic representation. Libben et al. (2003), in a study on
English compounds, also found that all compounds show morphological
constituency. In addition, they found that the transparency of the head
played a decisive role for the time it takes to make a lexical decision, i.e.
the time it takes to determine whether or not a string of letters is a word of
one’s language.
3 THE GERMAN DATA
Ward et al.’s (1991) data contains a large number of compounds, but

anaphoric reference is never made to the right-most constituent. Instead,
the anaphoric reference is made to the first constituent, as in the examples
150 martin schäfer
in (14). In addition, in Ward et al.’s examples, the anaphoric element is

always (a) a pro-form on the surface and (b) a form serving as a noun phrase
(or a prepositional phrase) by itself. As a consequence, the anaphoric ele-
ments are mostly personal pronouns or there, and only in a few cases is one
used as an anaphoric pro-form. The anaphoric element thus typically refers
to a concrete discourse entity introduced by the antecedent, whereas the
German pattern as introduced in examples (3) and (4) always involves the
deletion of the head noun. The anaphoric element is a pro-form that is not
visible on the surface. This pro-form does not serve as a noun phrase but
just replaces the head noun. In addition, the whole test pattern already rests
on contrastive topicality, which allows focusing on other factors that lead to
anaphoric reference in these examples.
Below, I will first introduce a standard classification of German AN
compounds and discuss their behaviour with regard to anaphoric refer-
ence. In a second step, I will discuss the results of an explorative corpus
search for the pattern under investigation.
3.1 Subclasses of German A N compounds and anaphoric

islandhood
We get a first feel for the interrelatinship between the semantic trans-
parency of German A N compounds and the availability of our deletion
pattern if we go through examples of the standard subclasses of A N
compounds. An endocentric compound is understood here as allowing the
paraphrase ‘[AB]N is a BN’, that is the compound is a hyponym of its head.
The classes described here are based on the classification schemes for AN
compounds in Fahim (1977), who employs simple paraphrase tests for the
distinction between all in all five classes. Three of the five classes are sub-
categories of endocentric compounds, here referred to as endocentric class
A, endocentric class B and endocentric class C. The other two classes cover
the subcategories of exocentric compounds, here referred to as exocentric
class A and exocentric class B.
Endocentric class A compounds are compounds that correspond in
meaning to the respective phrasal versions – see (20).
(20) Endocentric class A:

[AN]N = [AN]NP
Rotwein = roter Wein ‘red wine’
This class, which includes the compound pair Grünglas ‘green glass’ and
Weißglas ‘white glass’ from example (3), allows for anaphoric reference –
see (21).
(21) Ich hab keinen Rotwein gekriegt, es gab nur noch weißen.
‘I have no red.wine received, there was only still white.’
i.e. I didn’t get any red wine, they only had white wine left.
Other examples for compounds of this class are Grüntee ‘green tea’ and
Schwarztee ‘black tea’.
The second subgroup of endocentric compounds, endocentric class B,
still allows phrasal paraphrases, but the meaning of compound and cor-
responding phrase is not completely equivalent – see (22).
(22) Endocentric class B:

[AB]N ≈ [AB]NP
Großstadt ≈ große Stadt ‘big city’
The meaning of the phrase große Stadt ‘big city’ is less specific than the
meaning of the compound and can be used to refer to any large-sized city.
In contrast, the compound Großstadt has two different idiosyncratic mean-
ings. On the one hand, it is the technical term in Germany for any city with
at least 100,000 inhabitants. On the other hand, it can be used for cities in
the sense of ‘metropolis’, that is a large, bustling city.
Despite the specialized meaning in this class, the corresponding patterns
are acceptable – see (4), repeated here as (23) – where Großstadt contrasts
with Kleinstadt, which exhibits the same two kinds of specialized meaning,
i.e. it either refers to a town with between 5,000 and 20,000 inhabitants or
to a town with a provincial feel to it.12
(23) Ich liebe Großstädte, in kleinen gehe ich ein.

‘I love big-towns, in small go I in’ [cf. ein-gehen ‘to perish’]
i.e. I love big cities, I cannot exist in small cities.
The possibility of pragmatically unmarked anaphoric reference might be

connected to the fact that for endocentric A as well as for endocentric B
compounds the contribution of the adjective fully overlaps with its pre-
dicative usage. For class A compounds, although Rotwein is a specific kind
of wine, it is also wine that is red, a feature perhaps even more obvious for
the three common colour adjective + glas compounds, Weißglas ‘white
glass’, Grünglas ‘green glass’ and Braunglas ‘brown glass’. These three
compounds refer to the three kinds of coloured glass bottles relevant for the
recycling schemes in Germany, and one encounters them when standing in
front of the correspondingly labelled containers in order to get rid of one’s
old bottles, where one typically checks the colour of one’s bottles before
disposing of them in the corresponding containers. For endocentric class B
152 martin schäfer
compounds, the pattern still holds, i.e. a Kleinstadt is small, although the
specific meanings mentioned above are lost.
Endocentric class C compounds, finally, are those that are so opaque that
the phrasal version does not correspond to the compound any more – see (24).
(24) Endocentric class C:

[AB]N ≠ [AB]NP
Grünspecht ≠ grüner Specht ‘Green woodpecker’
That is, the green woodpecker is a member of the woodpecker family Picidae.
Green woodpeckers all have green upper parts, paler yellowish underpants
and a red crown, but the properties expressed by the adjective do not hold
for the whole entity referred to by the head noun and the corresponding
predications are false (i.e. a green woodpecker is NOT green). In contrast,
the intuitive interpretation of the phrasal grüner Specht is intersective, its
referent is a woodpecker and it is green. Anaphoric reference to the head
is pragmatically heavily marked as illustrated in (6), repeated here as (25).
(25) #Mein Vater hat in seinem Garten schon mal einen Grünspecht gesehen, aber
noch nie einen schwarzen.
My father has in his garden already once a green-woodpecker seen, but so_far
never a black.
Intended: ‘My father once saw a green woodpecker in his garden, but he has
never seen a black woodpecker.’
On its preferred reading, (25) is interpreted with anaphoric reference to the

whole compound, that is to a black green woodpecker.
For the two groups of exocentric compounds, Fahim (1977) distin-
guishes between those where the first part is true of the referent of the
whole compound and those where this is not the case. Thus exocentric
class A conforms to the pattern in (26).
(26) Exocentric class A:

[AB]N(x) → [A]N(x)
Ein Dummkopf ist dumm. ‘A stupidhead is stupid’
As in (25), anaphoric reference into exocentric A class compounds is prag-

matically marked, intuitively bordering on the ungrammatical – see (27),
based on the pair Dummkopf ‘stupid head’– Schlaukopf ‘smart head’.13
(27) #In dieser Klasse gibt es reichlich Dummköpfe, aber zum Glück auch schlaue.
In this class gives it abundantly stupidheads, but luckily also smart.
Intended: ‘This class has lots of boneheads, but, luckily, also some smart
students.’
Exocentric class B compounds, finally, are those compounds where neither

the first part nor the second part is a predicate of the referent of the whole –
see (28).
(28) Exocentric class B:

[AB]N(x) NOT→ [A]N(x)
Ein Rotkehlchen ist nicht rot. ‘A robin [red.throat] is not red.’
Thus the referent of the noun phrase ein Rotkehlchen ‘a robin’ in (28) is
a bird with a red throat. It is not a red entity nor is it a throat. It is thus
a prototypical instance of a possessive compound: the head, Kehlchen
‘throat’, is metonymically reinterpreted and stands for the whole bird, and
the property expressed by the adjective is a property of the entity normally
referred to by the reinterpreted head. As was the case for exocentric class A
compounds, anaphoric reference is heavily marked – see (29), with the pair
Braunkehlchen [lit. brown throat] ‘whinchat’– Rotkehlchen [lit. red throat]
‘robin’ .
(29) #Guck mal, ein Braunkehlchen und ein rotes!

Look, a brown.throat and a red
Intended: ‘Look, a whinchat and a robin!’
The paraphrase tests used to establish the five different classes are one way
to operationalize the notion of semantic transparency. On this view, endo-
centric class A compounds are the most semantically transparent subclass
of AN compounds, while exocentric class B compounds represent the least
transparent subclass. The intuitive judgements of the possibility of ana-
phoric reference into these different classes of compounds support the idea
that semantic transparency plays a key role. In the next section, we attempt
to give empirical backing to these intuitions by looking at corpus data.
3.2 The data in the corpus

The data so far was based solely on intuitive judgements. In order to give
some empirical support to these intuitive judgements, I report here the
result of an exploratory corpus study. I used the COSMAS IIweb corpus
tool14 developed and maintained by the Institut für Deutsche Sprache
(Mannheim), accessing all written corpora of the Deutsche Referenz
Korpus (DeReKo)15 available through that system.
154 martin schäfer
The main idea behind this exploratory study is quite simple: if there is
anything to the intuitions reported in the previous sections, then we should
find a reflection in the corpus. First of all, we should find the pattern of
interest as such, i.e. occurrences of an AN compound with a following
inflected adjective and a deleted nominal head. Secondly, there should be
a higher relative number of tokens exhibiting this pattern for the endocen-
tric class A compounds, and a lower relative number for those compound
classes where the pattern was judged to be pragmatically deviant.
Trying to find this kind of data in the corpus proved to be difficult. First,
the corpus is not tagged for the internal structure of compounds, so that
no search for AN compounds as such is possible. Secondly, the second
element of the pattern, the occurrence of an inflected adjective followed by
a zero element, provides no help, since, unsurprisingly, the zero element
is not encoded and inflected adjectives occur in large numbers. The only
reasonable strategy would be to search standard lists of AN compounds
for same-headed pairs with a contrasting first adjectival element and then
using these pairs as a basis to search for the patterns individually. While
this involves a rather large-scale study, I present below the results of an
exploratory corpus search for all the patterns discussed in the examples
so far. The search pattern used was always similar; I illustrate it here for
the pair Großstadt/Kleinstadt ‘big city/small city’. In order to capture
both possible linearizations, two searches were carried out, one with the
search pattern ‘Großst?dt+ /+s0 kleine*’ and one with the search pattern
‘Kleinst?dt+ /+s0 große*’.16
The results can easily be summarized: there were no hits for the
pattern involving any of Weißglas/Grünglas ‘white glass/green glass’,
Grüntee/Schwarztee ‘green tea/black tea’, Schwarzspecht/Grünspecht
‘black woodpecker/green woodpecker’, Dummkopf/Schlaukopf ‘stupid-
head/smarthead’ and Rotkehlchen/Braunkehlchen ‘robin/whinchat’ as the
first element. For the four remaining compounds, I obtained the results
presented in Table 8.1. These results can partly be explained by the abso-
lute frequencies of the corresponding compounds in the corpus (2.3 billion
words) – see Table 8.2.
The two members of the pair Weißglas/Grünglas occur eighty-six and
Table 8.1 Absolute occurrences of the anaphoric reference search pattern

Pattern Number of occurrences
Rotwein(e) . . . weiße(n/r) [Wein] 25
Weißwein(e) . . . rote(n/r) [Wein] 55
Großstadt/städte . . . kleine(n/re) [Stadt/Städte] 1
Kleinstadt/städte . . . große(n)/größere [Stadt/Städte] 2
Table 8.2 Absolute frequencies of compounds in the corpus

Rotwein ‘red wine’ 12,465 Weißwein ‘white wine’ 5,209
Weißglas ‘white glass’ 86 Grünglas ‘green glass’ 65
Grüntee ‘green tea’ 336 Schwarztee ‘black tea’ 328
Großstadt ‘big city’ 17,638 Kleinstadt ‘small city’ 12,276
Schwarzspecht ‘black woodpecker’ 304 Grünspecht ‘green woodpecker’ 319
Dummkopf ‘stupidhead’ 948 Schlaukopf ‘smarthead’ 76
Rotkehlchen ‘robin’ 873 Braunkehlchen ‘whinchat’ 281
sixty-five times, compared to Rotwein/Weißwein with 12,465 and 5,209,

respectively. In fact, the absolute frequency for the compounds where the
pattern does not occur are all relatively low, except for the pair Dummkopf/
Schlaukopf, which shows a high asymmetry with 948 to 76 occurrences.
The absolute frequency of the Großstadt/Kleinstadt ‘big city/small city’
pair, 17,638 and 12,276, respectively, is the highest of the sample and
clearly higher than that of white/red wine.
Taking a closer look at the data, a few other aspects are noteworthy.
First, the two instances of the ‘Kleinstadt/städte . . . große(n)/größere’
pattern both involve the comparative form of the second adjective, as in the
example below.
(30) Erst mit der Erschließung des Umlandes durch Eisen-und Straßenbahn
konnten Kleinstädte zu größeren wachsen, so auch Salzburg. (N97/
AUG.33561)17
‘Only with the development of.the surroundings through trains and trams
could small.cities to bigger grow, as also Salzburg.’
i.e. Not until the urban hinterland had been developed through trains and
trams could small cities grow into bigger ones, as did Salzburg.
Here, the corresponding compound could not have been used (the phrase
could, though). Similarly, the single example for the ‘Großstadt/städte . . .
kleine(n/re)’ pattern contains multiple modification, again not realizable
through a compound, see (31).
(31) Nein, nicht in einer Großstadt, in einer kleinen, überschaubaren, ohne Hektik
und Trubel. (O95/JUL.69073)
‘No, not in a big city, in a small, manageable [one], without hustle and bustle. ’
Thus, although in all three cases the compounds in question are topical and
the anaphoric usage is contrastive, that is the general conditions for ana-
phoric reference are fulfilled, we have, in addition, clear grammatical con-
straints that make the usage of the corresponding compound impossible.
156 martin schäfer
This differs radically from the picture that presents itself when looking
through the sentences involving wine as the head of a compound. In all but
one instance, the compound could be used instead of the adjective + pro-
form. A typical example is given in (32).
(32) Die viele Sonne ist gut für den Rotwein, schlecht für den Weißen. (NON07/
JUN.16950)
‘The large amount of sunshine is good for the red wine, bad for the white.’
And while the choice of the construction with a deleted head in these sen-
tences can thus be seen as an instance of linguistic economy, it is the only
available encoding in the sentences involving Stadt ‘city’ as the head of a
compound, where the usage of the corresponding phrasal variant would
lose the specific meaning characteristics of the compound. If it is true that
we can thus distinguish between two different motivations for the usage of
the pro-form anaphora, on the one hand grammatical necessity in the case
of the city examples, on the other hand simple linguistic economy in the
case of the wine examples, the difference between the sentences with com-
pounds headed by Wein ‘wine’ and compounds headed by Tee ‘tea’ comes
as a surprise. Looking at the number of absolute occurrences given above,
one is led to believe that it is simply an effect of the low overall number of
the two types of tea compounds. However, a closer look at the data reveals
a fundamental difference between the two cases. Tea compounds co-occur
frequently with full tea phrases, as in (33).
(33) Wie Schwarztee ist grüner Tee koffeinhaltig, . . . (HMP08/NOV.01318)

‘Like black tea is green tea caffeine-containing, . . .’
i.e. Like black tea, green tea contains caffeine . . .
Out of the seventeen instances that fit the general search pattern used for
the detection of anaphoric reference, sixteen instances show this pattern.
In contrast, this pattern never occurs for the wine examples, where we have
181 instances matching the search pattern, of which 79 show the anaphoric
reference phenomena. Why might this be? The key seems to lie in the rela-
tion between the phrasal and the compound versions. As was mentioned in
section 3.2, the defining characteristic of endocentric class A compounds is
the full equivalence of the phrasal and the compound version. This is true
for all four compounds. However, the absolute number of occurrences of
the compound variants and the phrasal variants differ asymmetrically (see
Table 8.3).
Thus Rotwein is nineteen times more frequent than roter Wein and
Weißwein is forty-nine times more frequent than its phrasal counterpart
Table 8.3 Asymmetry in the absolute frequency of phrasal and compound A N

constructions
Rotwein 12,465 roter Wein 668
Weißwein 5,209 weißer Wein 106
Grüntee 336 grüner Tee 1,082
Schwarztee 328 schwarzer Tee 549
weißer Wein, contrasting with the relation between grüner Tee and Grüntee,
where the former is 3.2 times more frequent than the latter, and between
schwarzer Tee and Schwarztee, where the former is 1.7 times as frequent as
the latter.
The picture that emerges is that the relative collocational strength of the
phrasal tea constructions is far greater than that of the wine constructions.
In other words, for the AN constructions with tea, the phrasal variant is
more entrenched than the compound variant, whereas for the wine A N
constructions, it is the other way around.18 The effect of this difference in
collocational strength is that once we decide to use the phrasal AN con-
struction to refer to black or green tea, we are forced to add the head, even
if this head is easily anaphorically recoverable – see (33).
4 CONCLUSION AND OPEN ENDS
The aim of this chapter was to discuss the relationship between semantic
transparency and anaphoric islands. In particular, it discussed the two
competing analyses of the phenomena, reproduced the most compelling
evidence in favour of the pragmatic account, and introduced German data
through which a deeper understanding of the processes behind anaphora
into compounds can be gained. Thus for English we already have data
which shows that the first, modifying part of compounds as well as the
second part, the head, can serve as antecedents for anaphora. This can easily
be explained by the pragmatic account, whose insistence on the importance
of semantic transparency and contrastive topicality is backed up by findings
from psycholinguistics. The German data shows that while standard clas-
sifications of A N compounds follow the general tendencies predicted by
the pragmatic account, we can identify other relevant factors by looking at
corpus data. Thus, at least for compounds that are fully transparent, i.e. of
the endocentric A class, the entrenchment of the phrasal variant relative
to the compound variant plays a decisive role in opting for a construction
involving anaphoric reference. And for compounds that are not fully trans-
parent, starting with endocentric class B, we apparently need morphosyn-
tactic constraints to force us into the usage of an anaphoric construction.
158 martin schäfer
However, since we did not find any data in the corpus involving the
other compound classes, there is still much work left to do. According
to the pragmatic account, anaphoric reference should be possible in all
cases, something which we do not find reflected in the data. This might
ultimately be due to the relatively low frequency of the base constructions,
and maybe only psycholinguistic tests will allow further insights into the
detailed workings behind anaphoric reference into AN compounds.
NOTES
1. Note that Postal’s anaphoric islands are only terminologically similar

to Ross’s (1967) island phenomena.
2. Besides outbound anaphora, Postal also discusses inbound anaph-
ora. Harris (2006) provides a comprehensive discussion of inbound
anaphora.
3. Note that this version of the Lexicalist Hypothesis is used by Anderson
as an illustration of a maximally strong position.
4. Note that these criteria usually work the other way around, though:
one-word-spellings usually indicate one-word status and stress on the
left-most member of an AB construction usually indicates compound
status.
5. Giegerich (2005: 579ff.) interprets this data pattern in a different way,
in that he believes that it is actually the underlying second AN construc-
tion that is shown to be phrasal or compound-like, that is, for (11a), the
availability of dental one shows us that dental appointment is phrasal. On
this view, replacement by one itself is enough to show phrasal status, a
test criterion also used by Bauer (1998: 76ff.) Technically, this is based
not on the prohibition of outbound anaphoras into anaphoric islands,
but to the prohibition of what Postal calls inbound anaphora, that is
anaphoric elements that occur inside of words. In order to make his
point, though, Giegerich would have to show that the antecedents in
(11) both allow anaphoric access to their head. That is, if mental disor-
der is a compound and therefore an anaphoric island, its head disorder
cannot serve as an antecedent in the first place. Note that Levi (1977:
332) uses A one constructions similar to those used by Giegerich for
arguing that complex nominals do in fact allow inbound anaphora.
6. Note that it does not seem to be accidental that in two of the three
examples in (14) the pronominal element refers back to the referent of
a proper noun. Ten Hacken (1994: 76) points out that this occurs in
two-thirds of the cases involving pronominal reference to non-heads in
the corpus investigated by Ward et al.
7. This manuscript is not available to me.

8. (17) is Postal’s (59a–b) and (18) his (54a–b). (19) corresponds to
(27a–b) in the appendix to Postal’s paper.
9. In reporting these experiments, Ward et al. always refer to the
aforementioned manuscript McKoon et al. (1990). McKoon et al.
(1993) seems to be the publication that this manuscript resulted in, as
it contains discussion of all the experiments mentioned below.
10. A small example from native speaker judgements on English AN
compounds illustrates this sensitivity to the anaphoric element itself:
while the native speakers I asked all judged (a) as out, most thought (b)
was OK.
(a) Peter uses a blackboard, but I prefer the white ones.
(b) Blackboards require more care than those with a white surface.
However, the target of the anaphoric references is in both cases
identical.
11. In Zwitserlood’s study, only partially opaque compounds with a trans-
parent first constituent and an opaque second constituent were used.
12. Not surprisingly, there is also a technical term for towns with between
20,000 and 100,000 inhabitants: Mittelstadt ‘middle city’. However,
this term is not used in everyday language.
13. One additional problem with this pair is that Schlaukopf ‘smart.head’ is
often interpreted with the somewhat negative tinge also present in, for
example, English smart alec.
14. COSMAS II (Corpus Search, Management and Analysis System),
http://www.ids-mannheim.de/cosmas2/, © 1991–2010 Institut für
Deutsche Sprache, Mannheim.
15. Das Deutsche Referenzkorpus DeReKo, http://www.ids-mannheim.
de/kl/projekte/korpora/, am Institut für Deutsche Sprache,
Mannheim.
16. The search syntax and the special characters in these two search pat-
terns have the following functions: /+s0 requires the two other con-
straints to be matched in the same sentence, the question mark stands
for any single character, allowing simultaneous search for the singular
word forms and the plural forms, which require an umlaut. The +
at the end of the compounds stands for zero or one more character,
capturing thus all inflectional endings except the dative plural, the star
stands for any number of further characters.
17. The short signatures after the examples from the DeReKo expand to
the following full references:
N97/AUG.33561 = Salzburger Nachrichten, 19.08.1997, Ressort:
LOKALES; Erreichbarkeit ist der zentrale Punkt Seit Minister
Einems.
160 martin schäfer
O95/JUL.69073 = Neue Kronen-Zeitung, 15.07.1995, S. 7; Zum

Schnuppern nach Baden.
NON07/JUN.16950 = Niederösterreichische Nachrichten,
27.06.2007, S. 3; Mödlinger forscht an Klimazukunft
HMP08/NOV.01318 = Hamburger Morgenpost, 13.11.2008, S. 27;
Der Wundertrank Grüner Tee.
18. For the notion of entrenchment, see Langacker (2000b: 3).
chapter 9
Semantic coindexation: evidence

from Portuguese derivation and
compounding
Alexandra Soares Rodrigues and Graça Rio-Torto1 Soares Rodrigues and Rio-Torto
T he parallel between meaning construction in derivation and com-

pounding has received little discussion. Some works such as Lieber
(2004) and Fradin (2005) focus on it. However, more empirical data is
needed to contribute to the understanding of how meaning construction
works. Our contribution brings more data on the phenomena by comparing
derivation and compounding in Portuguese.
The aim of this article is to analyse the way meaning construction occurs
in derivation and compounding. We try to answer the questions in (1).
(1) a. How do words formed by derivation and compounding get their meaning?
b. What are the factors involved and what is the balance between them?
c. Are compounding and derivation rules sensitive to the semantics of their
bases in the same manner?
The analysis is focused on Portuguese deverbal nouns and adjectives

(section 1) as well as on nominal compounds (section 2) formed by noun-
noun [NN]N and noun-adjective [NA]N.
We assume that the formation of meaning in the word is independent
of syntax (see Lieber and Scalise 2007 regarding compounds), since there
is a discrepancy between the meanings provided by syntactic arrangement
and argument structure and the developed meanings of the coined word.
Semantic coindexation is responsible for the construction of meaning in
word formation. In our proposal, coindexation operates between seman-
tic features of the constituents (affix and base or compound bases) and
those of the ‘maximal semantic frame’ (Fillmore 1977a, 1977b; Langacker
1987a; Jackendoff 1997, 2002) associated with them. Coindexation is ruled
by what we will call the maximal compatibility principle. According to
our proposal, semantic coindexation between the involved features is
162 soares rodrigues and rio-torto
ependent on the degree of semantic similarity between them.2 This prin-

d
ciple prevents chaotic linking between features, because it only allows the
linkage of those that best fit semantically with each other.
The semantic framework adopted here corresponds to a conceptualist
version of semantic processing (Jackendoff 2002, 2007), according to which
the meanings of words must conform to human categorization, to mental
representations connected to perception and action and to the speaker’s
experience with language and the world. Thus the meaning of complex
words comes from the lexical units involved, as well as from other informa-
tion sources, such as the referential and/or pragmatic (Jackendoff 1997,
2002, 2007).
1 DERIVATION
Underlying our perspective in recent research (e.g. Rodrigues 2008), we

assume that meaning construction of derivation products, at the level of
the lexicon, is not dependent on syntax. It depends on a purely semantic
mechanism which is called ‘coindexation’.
1.1 Absence of syntactic factors in meaning construction in word

formation
Traditionally, deverbal derivations are seen as the result of either the pro-
jection of the argument structure of the verbal base, in the case of deverbal
event nouns (e.g. Grimshaw 1990), or of one of the arguments of the base
verb, in the case of deverbal agent nouns and adjectives (Rappaport Hovav
and Levin 1992). This perspective has already been questioned by authors
such as Hoekstra and van der Putten (1988) and Rodrigues (2008).
As an example, let us consider agent nouns/adjectives such as coloni-
zador (‘colonizer’). In this case, apparently, the argument structure of the
verbal base creates the agent meaning of the noun. Without giving many
details, the external argument of the verb colonizar (‘to colonize’) is topical-
ized in the meaning of colonizador. The internal argument of colonizar is
still available outside the noun – see (2).
(2) O colonizador do Peru

The colonizer of.the Peru
‘The colonizer of Peru’
The problem arises when we apply this syntactic explanation to nouns and
adjectives such as (3).
semantic coindexation 163
(3) a.
lambedorN (from lamber ‘to lick’), which, apart from the meaning of ‘licker’,
also has the meaning of ‘syrup’
b. chovedorA ‘that makes rain’
c.
suadorA/N ‘that makes sweat’, besides the more prototypical meaning of ‘that
sweats’
The syntactic explanation does not fit with these deverbal nouns. The
syntactic, the argument and the lexical-conceptual structures of the verbal
bases of the nouns in (3) are given in (4).
(4) a. chover ‘to rain’: intransitive verb, expletive subject that corresponds to a
syntactic function that is argumentally empty (no theta-role)
b. suar ‘to sweat’: intransitive verb (unergative), external argument (internal
cause [– volitional])
c.
lamber ‘to lick’: transitive verb, external argument (external cause [+
volitional]) and internal argument (theme)
Table 9.1 gives an overview of these data with the structures of the nouns
and adjectives.
From the data shown in Table 9.1, we can see that chovedor, suador and
lambedor display an agent meaning ‘external cause’. However, their verbal
bases lack this external argument that would correspond to the one that
appears in the argument structure of the nouns. Where does this ‘external
cause’ meaning come from?
In these cases, the ‘external cause’ meaning must come from a maximal
semantic frame, which is not particular to any lexeme but available in con-
ceptual structure in general.3 This maximal semantic frame also explains
Table 9.1 Contextualized comparison between the argument and the lexical-

conceptual structures of verbs and their deverbal nouns
Verb Noun/adjective
Chover: Chovedor:
Chove Substância chovedora
Rains Substance that.makes.rain.fem
(It rains)
Suar: Suador:
O João suou Exercício suador
The John sweated exercise that.makes.sweat
‘John sweated’
Lamber: Lambedor:
O João lambeu o xarope O João comprou um lambedor
The John licked the syrup The John bought a that.makes.lick
‘John licked the syrup’ ‘John bought a syrup’
deverbal nouns with ‘locative’ meanings that do not correspond to any

argument of the argument structure of the verb base. This is the case of
bramadeiro ‘place where deer join together when in rut’, from bramar ‘to
bellow (deer)’; miradouro ‘viewpoint’, from the verb mirar ‘to watch’; and
matadouro ‘slaughterhouse’, from the verb matar ‘to kill’. None of these
verbs displays a locative argument in their argument structure.
These are not marginal examples of deverbal nouns. In fact, many of
them do not correspond to the argument structure of the base in what
concerns meaning and in what concerns their proper argument structure
capacity (Rodrigues 2008: 80–93). These examples show that the relation
between the meaning of the deverbal nouns and the verbal base must be
founded on fine-grained semantic structures.
1.2 Semantic coindexation

We propose that meaning construction in word formation is sustained by
the combination of semantic features of the base, the affix (if there is an affix
involved, as is the case for the lexemes under analysis) and the maximal
semantic frame. The mechanism that is responsible for the activation of
those connections is coindexation.
In contrast to Lieber (2004), we propose that coindexation is a purely
semantic mechanism (Rodrigues 2008: 60–9). The reason we reject any
kind of syntactic intervention in coindexation is based on the fact that there
may be no relation between the arguments of the verb and the meaning of
the deverbal noun, as we have seen in section 2.1.
To understand coindexation, we must conceive of semantics as a domain
structured in tiers (Jackendoff 2002). The components of these tiers of a
lexeme are able to be dynamically linked to components of other tiers or of
the same tiers of other lexemes (Rodrigues 2008: 60). It may be objected
that a process of coindexation totally based on semantic structures would
lead to an overgeneration of derivations. However, semantic coindexation
is based on the degree of compatibility between the features of the base,
the features of the affix and the features of the maximal semantic frame
(Rodrigues 2008: 227–74).
As an example, consider the affix -dor, which prototypically generates
agent nouns. As Rodrigues (2008: 340–53) explains, the semantic feature
of -dor [that has the function of] is very close to the meaning of the feature
[agent] of the lexical-conceptual structure of a verb, but is not so close, for
instance, to the feature [place]. Indeed, [that has the function of] alludes
to something or someone that will bring about a given event. This is a
meaning quite similar to [agent]. Semantically, due to the presence of an
[active] feature, the degree of compatibility between [that has the function
of] and [agent] is higher than the one between [that has the function of]
and [place].
As observed in Rodrigues (2008), the most prototypical meanings and
derivatives result from the coindexation of features that are semantically
closer to each other. If the feature [that has the function of] of -dor coin-
dexes with the agent feature of the lexical-semantic structure of the verbal
base, the obtained meaning is ‘agent’, which is a prototypical meaning of
-dor derivatives. This situation illustrates a process of maximal compat-
ibility between the features of the suffix and the base. If the same feature
coindexes with [place], the obtained meaning is ‘place’, which is far from
being a prototypical meaning of -dor nouns.
Due to this need for compatibility between the features, the overgen-
eration of derivations is avoided. The maximal compatibility between the
features of the affix and those of the base represent the most prototypical
derivatives/meanings of that paradigm (e.g. -dor nouns meaning ‘agent’,
such as conquistador ‘conqueror’). A minimal compatibility between the
features leads to the least representative derivatives of that paradigm
(e.g. -dor nouns meaning ‘place’, such as toucador ‘dressing table’). The
boundaries provided by minimal and maximal are dependent on the kinds
of meanings (least and most prototypical) that the derivatives of each suffix
display. In the interior of this scale there are many derivations, such as
assador ‘dish where to put the food to be roasted’ and apontador ‘notebook’,
which designate instruments that are used to accomplish an event, but that
by themselves do not accomplish it.
Following Plag (1999, 2003), we assume that affixes are provided with
semantic features. An affix is not simply a formal operator of a word forma-
tion rule. This explains some kinds of constraint between affix and base, for
instance why an affix occurs with one kind of verbal base (e.g. comiseraçãoN
from comiserarV ‘to move to pity’) and not with other kinds of verbal base
(e.g. *envelheceçãoN from envelhecerV ‘to become old’) (see envelhecimento vs.
*comiseramento), although both affixes generate deverbal event nouns.
Semantic features of the affix are not accessible when the affix is on its
own. As a non-autonomous morpheme, semantic structures of the affix are
only observable when the affix is integrated into the derived word. This
is to say that the semantic contribution of the affix is placed in an implicit
structure. To determine the semantic contribution of the affix we need
to compare the derivatives of that affix both to each other and with the
derivatives of other affixes that operate by the same rule.
As an example, consider deverbal event nouns. Although they share the
same verbal base and a general meaning ‘event of V’, the nouns reveal differ-
ent semantics according to the affix. Let us compare event nouns from the
verb andar ‘to walk’. The noun with the suffix -nça, [[anda]Vnça]N, means
‘adventure, journey’; the noun with -mento, [[anda]Vmento]N, means ‘speed

or way of something going’; and the noun with -dura, [[anda]Vdura]N, ‘physi-
cal way of moving’. The differences between their meanings come from the
coindexation of semantic features of each affix with semantic features of the
base.
The semantic features of the base belong to the event structure and to
the lexical-conceptual structure. From the analysis of 8,414 deverbal nouns
constructed with 23 affixes, Rodrigues (2008: 227–74) has determined the
following event features available in the verbal bases:
• [±punctual] The event occurs at a particular point on the temporal line

and not along that line (e.g. estalar ‘to click’).
• [±durative] The event occurs along the temporal line (viver ‘to live’).
• [±composed of individuals] This feature was presented by Lieber
(2004: 136). It refers to an event that is symmetrically divided into parts
repeated along the temporal line (saltitar ‘to hop’).
• [±composed of different operations] The event (e.g. conduzir ‘to drive (a
car)’) is composed of different subevents (such as ‘clutching’, ‘braking’,
‘changing direction with the steering wheel’, etc.).
• [±point of departure] The event has a beginning. This feature permits
us to distinguish durative verbs such as distar ‘to be distant from’, which
is a state verb, from durative verbs such as caminhar ‘to walk’. Distar
does not indicate an event with a clear temporal beginning. On the
contrary, caminhar has an implicit beginning. This difference shows that
event structure contains subcomponents and does not behave like an
indivisible whole.
• [±point of arrival] The event has an end point. This allows us to distin-
guish a verb such as construir ‘to build’ from a verb such as trabalhar ‘to
work’. The former has a point of arrival while the latter does not.
• [±telic] A telic verb presupposes that a change of state occurs, for
instance estar ‘to be’ vs. cozinhar ‘to cook’. The latter is [+ telic] while
the former is not.
• [±perfect] The event is irreversible and not prolongable (e.g. matar ‘to
kill’).
These features do not correspond to classes in the sense that each feature
on its own does not characterize the event type of the verb. Each verb
may have a set of features. Semantic coindexation does not operate with
semantic boxes, but with subcomponents of those boxes. This assump-
tion is based on the observation that there is no relation between the event
class of the verbal base (e.g. Vendler’s classes: accomplishment, achievement,
activity, state) and the selected suffix. For instance, some verbs of accom-
plishment such as enrolar (‘to wrap’) select the affix -mento (enrolamento
‘event of V’) and so do some verbs of achievement such as salvar (‘to
save’) (salvamento ‘event of V’), some verbs of activity such as respirar (‘to
breathe’) (respiramento ‘event of V’) and even some verbs of state such as
preceder (‘to precede’) (precedimento ‘event of V’). However, not all verbs of
those classes select this suffix. We need to observe the features that charac-
terize each event structure of each verb. Those features, and not the entire
event structure as a whole, reveal themselves to be important not only to
the selection of the suffix, but also to the determination of the meaning of
the deverbal noun (Rodrigues 2008: 201–2).
The verb relaxar ‘to relax’ contains features such as [durative], [telic]
and [point of arrival], since they behave as accomplishment verbs. The
event deverbal nouns from this verb are relaxamento and relaxação, respec-
tively. Although both are deverbal event nouns, relaxamento presents
different semantic shades in comparison with relaxação. Nouns with the
affix -mento display a meaning of ‘state’ that co-occurs with the course of
the event. Nouns with the affix -ção have a meaning of ‘state’ that does not
co-occur with the course of the event but occurs after or as a consequence
of the point of arrival of the event.
This can be explained if we assume that affixes have a semantic structure.
The affix -mento contains the feature [process], whilst -ção is characterized
by the feature [effectuation]. Note that these features are not included in
the above list, since that list only shows semantic features of the verbal
bases, not semantic features of the affixes delimited in Rodrigues (2008:
227–74). We will limit ourselves to the affixes shown here.
The feature [effectuation] refers to an event that is presented as realized
and completed. In contrast, the feature [process] refers to the course of the
event and not its ending. It stresses the unfolding of the event and not its
conclusion. Once again, we need to emphasize that semantic features of
affixes become available to our explicit knowledge when we compare dever-
bal nouns with different affixes and the same verbal bases and deverbal
nouns with the same affix and different verbal bases.
Producing relaxamento uses the following procedures: the affix -mento
has the feature [process]. This feature is maximally compatible with the
feature [durative] and minimally compatible with the feature [point of
arrival]. Thus -mento, or, specifically, its feature, will coindex with [dura-
tive] and not with [point of arrival]. In consequence, relaxamento means
a state that co-occurs with the course of the event. In contrast, -ção has
the feature [effectuation]. This feature is maximally compatible with the
feature [point of arrival] and minimally compatible with the feature [dura-
tive]. Thus relaxação means a state that occurs after the end of the process,
that is at the point of arrival (Rodrigues 2008: 291–315).
Table 9.2 Verbal bases and their deverbal nouns with affixes -dura, -ção and -mento
Verbal bases Affix -dura Affix -ção Affix -mento
serrar ‘to saw up’ serradura ‘sawdust’ serração ‘event of
sawing; sawmill’
amolgar ‘to dent’ amolgadura ‘dent, amolgamento ‘event of
depression’ denting’
pisar ‘to bruise’ pisadura ‘bruise’ pisamento ‘event of
bruising’
abotoar ‘to button’ abotoadura ‘set of abotoação ‘event of
buttons’ buttoning’
This difference between the semantic features of - ção and - mento

explains why -mento attaches to durative verbs such as balancear ‘to swing’
and espigar ‘to ear (cereal)’. The feature [process] of this affix prefers the
feature [durative] of these verbs. We find the nouns balanceamento, espiga-
mento, but not *balanceação, *espigação.
From this perspective, the maximal semantic compatibility principle
explains why -mento attaches to -ec-and -esc-verbs and not to -iz- and
-ific-verbs. Nouns such as envelhecimento ‘growing old’ from envelhecer
‘to grow old’ and amarelecimento ‘yellowing’ from amarelecer ‘to yellow’
are common. Nouns derived from these verbs with -ção are ungram-
matical (*envelheceção; *amareleceção). This is because -mento selects verbs
where the feature [durative] is foregrounded. The same factor explains
why -mento is the affix that occurs with verbs such as abairrar ‘to divide into
wards’ (abairramento) and arruar ‘to divide into streets’, which indicate the
unbounded division of an object in infinite parts.
Apart from these differences in the meaning of ‘event’ and the corre-
lated consequences in the constraints between bases and affixes, semantic
coindexation also reveals itself in the concrete meanings that the deriva-
tives of each affix exhibit. This can be observed for the deverbal nouns in
Table 9.2.
All of the deverbal nouns presented in Table 9.2 have an event meaning
related to their verbal bases. However, -dura nouns also have concrete
meanings of ‘portion’, ‘residue’, ‘amounts’, ‘concrete result’, which are
absent from the other deverbal nouns. Where do those concrete meanings
come from?
In the theory we propose here, the suffix -dura has the feature [refer-
entiation]. Once again, this feature only comes to light indirectly when
we observe the meanings of -dura nouns in comparison to the other event
affixes derivatives. Relevant data are provided if we compare deverbal
nouns of the same verb with different event affixes. What we observe in
-dura nouns is that, besides the meaning of ‘event’, many of them manifest
a concrete meaning that can be subsumed as ‘something concrete that

results from the event’, ‘a physical result of the event’. This is specified
as ‘residue’ (serradura ‘sawdust’, cevadura ‘remains of the bird that a bird
of prey has fed on’) and/or ‘physical wound’ (pisadura ‘bruise’). This affix
only attaches to verbs that possess a concrete meaning. This emphasizes
the semantic compatibility between affix and base.
Nevertheless, what explains the construction of these particular mean-
ings is coindexation. All the verbal bases of these nouns possess the feature
[telic], among others. The feature of the affix, which is [referentiation],
indicates a segmentation and an identification of a certain event, detaching
it from the continuum of realia. The semantic effect of this affix is not to
provide an event shade of the event, i.e. [durative], [actualized], etc., but
simply to identify it as a referent (Rodrigues 2008: 315–20). Therefore the
feature [referentiation] is maximally compatible with the feature [telic] of
the verbs. The segmentation and identification of the referent of [telic],
operated by the feature of the affix, result in these peculiar concrete
meanings.
1.3 Conclusions on derivation

In relation to meaning construction in derivation, we have come to the fol-
lowing conclusions. First, meaning construction in derivation takes place
in a phase where argument structure does not play a role. In this phase,
lexical-conceptual structure and the ‘maximal semantic frame’ associated
to each lexeme are considered. Their role consists of the coindexation of
semantic features belonging to the lexical-conceptual structure of the base,
or to a maximal semantic frame, which is subsumed in mental conceptual
structures (Jackendoff 2002; Pustejovsky 1995a).
Meaning construction in word formation does not occur at the level of
argument structure. If the deverbal noun has an argument structure, it is
the result of the combination of a series of factors (among them the kind of
meaning of the noun and the affix that formed it).4 It is not inherited from
the verbal base.
Our second conclusion is that the mechanism that is responsible for
meaning construction is coindexation. Coindexation operates between
features of the affix, of the base and of the maximal semantic frame associ-
ated with it. Although strictly based on semantic parameters, coindexation
is ruled by the maximal compatibility principle. This principle prevents
chaotic linking between features, because it only allows the linkage of those
that semantically best fit with each other.
2 COMPOUNDING
A compound is a plurilexematic structure that is used as a holistic denomi-

nation, that resists any internal alteration and that is characterized by
denotational unity. This conceptualization is based on research carried
out by Lieber and Scalise (2007), Rio-Torto and Ribeiro (2009, 2010) and
Scalise and Bisetto (2009), and is supported by Scalise and Vogel’s (2010)
cross-disciplinary approach to compounding applied to several languages
studied by the Morbocom team (http://morbo.lingue.unibo.it/).
Portuguese [NN]N and [NA]N compounds are conceived as morphologi-
cal objects,5 whose components are linked by a modification relationship
(Rio-Torto and Ribeiro 2009). As the internal structure of the compounds
is fixed (determination, quantification, intensification are not allowed),
the diversity of semantic relations they express cannot be explained
syntactically, but semantically.
The understanding of compounds is anchored on the assumption that
there exists a semantic relation between the referents of the two concepts
being combined (NN) or that one or more properties of the modifier
constituent are attributed in some way to the head concept (NA).6 We
will address here the questions of what the principles and the devices that
govern this semantic coindexation are and how the emergence of idiomatic
meanings should be explained.
On the basis of the discussion in this section, we will claim that a theory
of enriched meaning and processing (Jackendoff 1997) is necessary in cases
where simple composition does not suffice.7 Semantic procedures, like coer-
cion, reference transfer and figurative shifts, conceived in an enriched com-
position frame, are responsible for the conventionalized meaning of a word.
2.1 Some assumptions

Compounding is associated with idiomaticity.8 There are indeed com-
pounds whose meanings are compositional and transparent, but many
compounds have an idiomatic meaning: a garbage man is a ‘man who
handles garbage’, but a snow man is a ‘simulated man made of snow’.
We assume that the meaning of a compound incorporates, even in a
sophisticated, unpredictable and idiomatic fashion, the meanings of its
components. Nevertheless, the meaning of a compound is not necessar-
ily confined to the meanings of its parts and the rules by which they are
combined. The maximal semantic frame of a compound includes all the
features associated with it in a specific cultural universe, namely the fea-
tures associated with the profiles, the roles and/or the proper functions of
denotata, as well as the pragmatic purposes words can serve.
We claim that idiomaticity is the result of forced meaning shifts

with respect to the compositional one (Rio-Torto and Ribeiro 2010).
Compositionality and idiomaticity are inversely related. However, as
empirical data illustrate (see (6) below), sometimes earlier stages of seman-
tic construction may display the underlying compositionality dissipated by
idiomaticity (Rio-Torto and Ribeiro 2010).
2.2 Trends of idiomaticity

Within a compound, the head and the modifier must be coindexed semanti-
cally in accordance with (a) the (maximal) information they convey and
(b) the possible grammatical and LCS relationships linking them (Bisetto and
Scalise 2005). When the final meaning does not include the compositional
meanings of the constituents, other semantic devices must be managed. In
order to guarantee the internal compatibility and plausibility of the whole,
coercion procedures are activated. Metaphor, metonymy, referential shift
and/or specialization are often activated for this task, and they are mainly
responsible for the idiomaticity of the compounds, as (5) illustrates.
(5) a. [NN]N visita-relâmpago (lit. ‘visit-lightning’)

b. [NN]N pontapé (lit. ‘fronting foot’)
c. [NA]N saco azul (lit. ‘bag blue’)
(5a) denotes not a visit of lightning, but ‘a flying visit, an unexpected and
brief one’: a metaphor is activated as the features [brief and unexpected] are
transferred from the lightning to a visit. (5b) denotes a kick; a metonymy
is activated, as the act is denominated by the ‘actor [foot]’.Usually, when
figurative tools are activated, a reference transfer occurs, e.g. in (5c). Saco
azul (lit. blue bag) denotes not a specific blue bag, but illicit funding. The
older motivation – an (ancient) bag containing money from an unofficial
source and lined with blue fabric – is lost. The metonymic meaning is not
semantically compositional because the meaning of the whole is not com-
putable from the meaning of the constituents. (5b–c) display a reference
transfer from the denotation and ontological class of the head to those of the
compound (foot > physical aggression; bag > funding). Their interpreta-
tion is opaque to native speakers if they have not previously encountered it.
Semantic specialization between coindexed compound members can be
sustained by polysemy. A polysemic adjective like civil adjusts its meaning
in accordance with the LCS of the noun whose intension it circumscribes.
The specific meaning of the adjective is delimited by the lexical-conceptual
relation between N and A. (6) illustrates four different types of meaning
that Portuguese [NA] nominal compounds with the adjective civil can have:
(6) a. event: [[guerra]N [civil]A]N ‘civil war’, [[casamento]N [civil]A]N ‘civil

marriage’
b. state: [[estado]N[civil]A]N ‘marital status’
c. human institution: [[polícia]N [civil]A]N ‘civil police’
d. specialized professional domain: [[engenharia]N [civil]A]N ‘civil engineering’
The meaning of the adjective varies according to the meaning of the noun
it modifies. The history of the entities highlights the semantic features
focused in each case.
A guerra civil ‘civil war’ is a war between organized groups within a single
nation. The adjective means, in this context, ‘intra-national’, by opposition
to international. A casamento civil is a secular marriage as opposed to a reli-
gious one. Since the Middle Ages, civil has been used in contrast to ecclesias-
tic. In both cases the adjective modifies an eventive noun; nevertheless, the
meanings displayed by the adjective are independent of the eventive class:
they are correlated with the specific lexical meaning and profile of each noun.
The Brazilian polícia civil ‘civil police’ denotes the investigative state
police forces. In this case civil is opposite to ‘military’ (cf. polícia militar
‘military police’).
The estado civil denotes the ‘marital status’, the legal standing of a
person in regard to his/her marriage state. In Portuguese, the adjective
civil covers, as a hyperonym, all the types of marital status: single, (un)
married, divorced, widow(er).
Engenharia civil ‘civil engineering’ is a hyponym of engenharia. The
adjective presents a technical meaning, referring not only, as in the past,
to non-military and/or non-ecclesiastic engineering, but also to a wide
variety of subdomains, including all the classes of construction engineering
and construction materials. This hyponym term opposes civil to electrical,
biomedical, geological engineering.
The semantic diversity and specialization of the adjective is correlated
with the semantics of the noun it modifies. The history of the culture and
of the society highlights the motivations of this variation.
2.3 [NN]N: semantic frames and world knowledge

When two nouns form a compound, the meanings of both must be analysed
in order to construct a plausible meaning. As a large variety of semantic
relationships is possible between the nouns, a large frame of conceptualiza-
tion and of reference is necessary to explain NN semantic profusion and
diversity, namely when unexpected meanings emerge. But if a gap between
the plausible meanings and the conventionalized one remains, then only a
specialized source of information can provide the idiomatic meaning.
2.3.1 Proposals
A speaker uses compounds as memorized constructions whose holistic
meaning is understandable, despite their degree of semantic idiomaticity.
The speaker is able to use the word, without knowing how the idiomatic
meaning has been built. However, the speaker must understand the idio-
matic meaning of the word if he intends to use it. When the comprehension
of the meaning is not straightforward, which happens when the meaning
of the whole is not literal and compositional regarding the meaning of the
parts, what are the means applied to understand the word?
First of all, the speaker tends to perceive the meaning of each compound
component. The semantics of each component is mentally constructed
in accordance with the conceptual and the denotational representations
associated with it.
According to its ontological nature, each N is characterized by a cluster
of semantic or thematic roles (Dowty 1989, 1991) that are connected with
the LCS schemata associated with it and by the network of possible seman-
tic relations the N can establish. A constellation of conceptual functions,
like BE, DO, HAVE, SEEM, BEHAVE, CAUSE, which underlie LCS
structures, profiles the semantics of a lexical item and, in consequence, the
thematic relations it supplies.
In order to bring plausibility and transparency to the semantic relation-
ship built by the compound, the speaker takes into account all the features
and scenarios – the more and the less prototypical – associated with each
word and its denotational frame: if necessary, even the possible semantic
features of each word are mapped for this demanding computation task.
World knowledge and/or referential coercion can also be used as ways of
reconciling the constraints from various sources. As Jackendoff (2002: 250)
says, in order to determine the meaning of a newly encountered compound
‘one uses the Head Principle, plus the repertoire of possible semantic rela-
tions, plus a dose of pragmatics, to put together a meaning that makes sense
in context’.9
2.3.2 Semantic roles and the ‘maximal semantic compatibility principle’

Let us consider some Portuguese [NN] compounds in order to describe
how coindexation underlies their semantic processing. We begin with four
NN compounds lexicalized in the 1990s and then we analyse two novel or
possible compounds.
Recent psycholinguistic research (Gagné and Shoben 1997; Gagné
and Spalding 2006) emphasizes the relevance of compound interpreta-
tion based on thematic relations. Several sets of thematic roles have been
proposed.10 We adopt Rio-Torto and Ribeiro’s (2013) framework, with

the thematic relations that are suitable for [[N[PN]]N, [NN]N, [NA]N
Portuguese compounds. Here, four thematic relations are taken into
account: container/recipient, goal, similarity and source. We claim that a
dynamic framework of thematic roles is needed to explain their frequent
intersections.
The meaning of the nominal modifier N2 and that of the nominal head
N1 must coarticulate and converge into an output compatible with human
categorization, with human world knowledge and with human experience.
Coindexation involves the literal, denotational, figurative meanings and
uses coercion so that a plausible meaning emerges. This is illustrated in (7).
(7) [[bébé]N-[proveta]N]N ‘test-tube baby’
How should the typical characteristics associated with a baby be reconciled

with a test-tube, in the interpretation of (7)? This container represents
the place where fertilization takes place. In fact, (7) designates a baby con-
ceived by in vitro fertilization. In comparison with other properties, con-
tainers do not represent salient features of a human being. But for specific
types of human beings – for some foetuses – a container such as a test-tube
denotes something fundamental in their conception: an artificial womb
where fertilization of the ovum took place and where the human embryo
developed before being transferred to the mother’s body. So, in this case
the (artificial) place of fertilization replaces the biological container (the
uterus) where conception typically occurs. World knowledge is here crucial
for the recognition of the specific meaning of the word.
This example shows to a higher degree the denotation coercion imposed
on the semantic relationship between N1 and N2, as [container] is not an
expected prototypical feature of N1 when it denotes a baby. As with several
containers, two semantic roles are also involved here, because the artificial
fertilization occurs by means of a test-tube and in its interior. Another
example is (8).
(8) [[criança]N-[soldado]N]N ‘child soldier’
A child has no professional activity, and a soldier is a military professional.

However, (8) refers to children that act as and become combatants. It is the
goal, the telic function that is focused. Once again, coindexation is forced
to select a non-typical and non-expected feature of N1. Otherwise, this
compound could be understood as denominating a child pretending to be
a combatant, which is not in accordance with the extralinguistic reality.11
A rather different type of example is (9).
(9) [[homem]N-[rã]N]N lit. ‘man-frog’, i.e. frogman
In (9), prominence is given to the similarity to a frog in appearance and

function.12 A homem-rã is a diver, a person who explores under water,
especially equipped with breathing apparatus and weighted clothing. As
the similarity involves the appearance and the function, two thematic
dimensions are also merged. As a final example, consider (10).
(10) [[Retrato]N-[robot]N]N lit. ‘portrait-robot’, i.e. photofit picture
(10) refers to the product of a specialized method of combining photo-

graphs of facial features, hair, etc., into a composite picture of a face. This
method is used by the police to trace suspects from witnesses’ descriptions.
Like a robot, that is an electro-mechanical machine that is supposed to
function as a human being, the photofit picture is obtained by computer
means in order to create a schematic and artificial sketch of a real being.
Regarding its semantic role, robot represents the robotic device by which
the picture is drawn and/or the (robot-like) schematicity of the product.
The specialized meaning is neither available nor computable without
technical knowledge.
For those with sufficient background knowledge about the denotation of
these four [NN]N compounds, their interpretation offers no difficulties. It
is not the case for a child, when not yet exposed to them, or for a speaker
of Portuguese as a foreign language, especially if there are no directly
corresponding expressions in their native language.
Let us now consider the novel compounds in (11).
(11) a. bébé-brinquedo lit. ‘baby-toy’

b. mãe-polícia lit. ‘mother-police’
The computation task of understanding novel compounds such as (11)

highlights that (a) several possible semantic dimensions can be gathered for
the understanding of the whole and that (b) the conventionalized meaning
of each word is not necessarily reached without the help of the co(n)text
and/or world knowledge. A bebé-brinquedo can be a bebé ‘baby’ used as
a ‘toy’ by its brothers, or a bebé that evokes a toy because of their way of
moving or playing, or even a toy, like a doll, imitating a real ‘baby’. A mãe-
polícia can be a mãe ‘mother’ whose professional activity is a policewoman,
but also a controlling and/or excessively protective mother. In both cases,
more than one thematic role can be involved and the precise meaning of
the compound results from the interaction between linguistic information
and extralinguistic sources of knowledge. When a compound has several
readings, multiple semantic roles must be used to describe it. Conceived

dynamically and in interaction with world knowledge, thematic relations
can provide the compound with semantic plausibility.
2.4 Conclusions on compounding

The semantic structure of each compound reflects the ‘maximal semantic
frame’ associated with each of the constituents, as well as the plausible
semantic and grammatical relations relying on them.13 Coindexation, in
accordance with a semantic plausibility principle, assures the maximal
compatibility between the meanings involved.
The specific meaning of each compound is due to the semantic struc-
ture of both constituents, in articulation (a) with semantic/conceptual
templates governing the relation between the components and (b) with
referential, pragmatic or figurative constraints. The floating adjustments
in the meaning of compounds are mainly governed by referential and/or
pragmatic motivations. Figurative mechanisms provide semantic coherence
when denotational or objective tools are overlooked.
From a production point of view, a compound is a construction
whose meaning is anchored, to a variable degree, to the meaning of its
constituents. But referential and/or pragmatic reasons may lead to idi-
omaticity and opacity, and for this reason the interpretation is sometimes
only weakly compositional. Due to the LCS frames associated with each
compound member and due to the semantic features that result from
their combination and by reference and plausibility needs, the range of
semantic possibilities of the final meaning is quite broad, though not
unlimited.
A theory of lexical (de)composition must incorporate a textured set of
dimensions and procedures that cut across lexico-conceptual representa-
tions, coindexation, referential coercion, semantic shifts and figurative
devices of meaning production.
3 CONCLUSIONS
Are there radical differences between the semantic processing of deriva-

tions and compounds?
1. Derived words, as well as compounds, can present compositional and

idiomatic meanings. In both cases the whole can display a compositional
meaning or an idiomatic one, not computable from the meaning of the
parts.
2. In both cases semantic coindexation is responsible for the construction

of the meaning, supported by the articulation between the semantics of
the units and their maximal semantic frame.
3. In both cases the meaning of the whole respects semantic coindexation
guidelines of maximal or minimal compatibility between the features of
the lexical units involved and the most prototypical or least representa-
tive meanings and nouns of the paradigm. In both cases a straightfor-
ward compatibility leads to a transparent noun; the more complex the
meaning construction is, the higher the idiomaticity.
4. Nevertheless, the set of LCS features associated with a lexeme and, a
fortiori, with a relation between two lexemes is potentially wider than the
set of features associated with a suffix. This difference opens up the uni-
verse of possible semantic and denotational codifications performed by
a compound. So, due to the fact that compound constituents represent
two LCS universes that articulate and enrich each other, the meaning
structure in compounding tends to be freer than in derivation. However,
the relation between the nouns is not totally free. By default, the meaning
of a derived word is less unpredictable than the meaning of a compound.
We could relate this observation to Jackendoff’s (2010: 422–3) proposal that

compounding is a relic of protolanguage, that is an architecture anterior
to language that would contain semantics and phonology but no syntax.
According to Jackendoff, compounding (but not derivation) displays some
properties of a ‘protolinguistic’ fossil, namely a rudimentary grammatical
structure that does not shape semantic interpretation. However, the alleged
freedom of compound meaning is indeed constrained by the semantics
of the units involved and by pragmatic and/or referential conditions of
compatibility between them. The question arising then is: if semantics,
as Jackendoff proposes, has a generative character, would that generative
character be present in the semantics of protolanguage? Would protolan-
guage semantics already have the structural complexity of the semantics of
fully developed language?
NOTES
1. The article is the result of close collaboration between the authors.

For academic purposes, Rodrigues is responsible for section 1 and
Rio-Torto for section 2. The remaining text is the joint responsibility
of both authors.
2. Contrary to Lieber (2004), we propose that coindexation is strictly
semantic and does not work with syntactic or argumental elements.
The concept of compatibility we adopt is also different from the one

proposed by Langacker (1987a), since his concept admits sentences
and properties of whatever nature.
3. This ‘maximal semantic frame’ refers to a conceptual universe related
to each lexeme (Langacker 1987a; Pustejovsky 1995a; Jackendoff 2002).
4. We do not say that the affix contains in itself the argument structure in
the case of event deverbal nouns, or a specific argument in the case of
agent deverbal nouns. We mean that the affix, because of its meaning,
functions as a constraint on the development of argument structure in
the deverbal product. This explains why deverbal nouns from the same
verb end up with different argument structure solutions according to
the affix that built them. For instance, apagão ‘power blackout’ does
not have any argument structure, while apagamento ‘switching off;
putting out’ does (from apagar ‘to switch off; to put out’).
5. Independently of their internal structure and of their gradual nature
(from a more to a less phrasal level), compounds are conceived as
lexical constructions (Booij 2009b) whose functioning is ruled by
morpho-semantic (not syntactic) principles (Lieber and Scalise 2007).
6. For the classifying (bomba atómica ‘atomic bomb’) and qualifying (mau
feitio ‘bad temper’) functions of Portuguese adjectives see Rio-Torto
and Ribeiro (2009: 282–4).
7. Adapting Jackendoff’s (1997: 49) words about sentences, we suggest
that the meaning of a compound ‘may contain, in addition to concep-
tual contents of its LCSs, other material that is not expressed lexically,
but that must be present in conceptual structure either (i) in order to
achieve well-formedness in the composition of the LCSs into concep-
tual structure (coercion, to use Pustejovsky’s term), or (ii) in order to
satisfy the pragmatics of the discourse or extralinguistc content.’
8. For an overview of the relations between compositionality and idioma-
ticity, see Cruse (2004: 68–77).
9. According to Baroni et al. (2007), the combinatorial history of a noun
influences the interpretation of a novel phrase involving that noun.
That is, people use the distributional knowledge of how nouns have
previously been combined to interpret a novel combination.
10. Shamsfard and Mousavi (2008) work with seventeen roles: agent,
experiencer, patient, theme, time, location, cause, source, destination,
reason, topic, instrument, force, state, comparison, message and ben-
eficiary. Jackendoff (2002: 250) announced a repertoire of twenty roles
for English NN compounds. Jackendoff (2009, 2010) presents a system
generating relations on the basis of a list of fourteen semantic roles.
11. Portuguese compounds are typically left-headed: this explains why a
criança-soldado cannot denote a soldier that seems/behaves like a child.
12. As Jackendoff (2010: 429) emphasizes, one of the procedures involved

in combinatorial semantics of NN compounds is profiling or topicali-
zation, according to which a feature is picked out and selected as the
one to be referred to. We claim that the selected feature is present in
the semantic frame of the units involved, and is the one that optimally
fulfils the conditions of semantic compatibility between them.
13. Jackendoff (2002: 250) says that in compounding ‘the grammatical
principle involved is simply one of concatenating two nouns into a
bigger noun, and the semantic relation between them is determined by
a combination of pragmatics and memorization’. Despite the apparent
simplicity of grammatical relations of coordination, subordination or
modification between the nouns of a compound, pragmatic, referential
and/or conceptual information associated with each noun plays a
major role in the construction of the conventionalized meaning of the
whole.
chapter 10
Deverbal nominalizations in
English: an LMBM approach
Maria Bloch-Trojnar1
D espite extensive research in the area, a comprehensive account

of deverbal nominalizations remains a challenge due to the fact
that category changing operations (transpositions) cannot be explained
without addressing fundamental questions regarding the overall struc-
ture of the grammar and the interaction between components, the
nature of the linguistic sign, the structure of the lexicon and the role of
morphology. For almost fifty years, English deverbal nominalizations
have served as a testing ground for various theoretical models, which
with varying success attempted to accommodate their systematic as well
as idiosyncratic properties.2 The main objective of this chapter is to test
the predictive and explanatory potential of the framework of Lexeme
Morpheme Base Morphology (henceforth LMBM) put forward by
Beard (1995).
As no theory evolves in a vacuum, I will also make reference to and
comparison with accounts grounded in the lexicalist tradition (Malicka-
Kleparska 1988; Cetnarowska 1993), Event Structure theory developed by
Grimshaw (1990) and the Parallel Architecture advocated by Jackendoff
(2002). The lexicalist analyses, which I regard as the point of departure for
further considerations, do not adopt the ‘one affix – one rule approach’ of
Aronoff (1976: 89) but rather subscribe to a categorial view of morphologi-
cal processes whereby one derivational category may subsume more than
one derivational type (Szymanek 1985, 1988). Word Formation Rules
(WFRs) are viewed as generative mechanisms producing regular (also
potential) forms and analytic devices, i.e. redundancy statements linking
sets of forms which are formally and semantically related (Jackendoff
1975).
The chapter is organized as follows. In section 1 I present the data with
a view to pinpointing problems inherent in their analysis. Section 2 serves
deverbal nominalizations in english 181
as a theoretical background in which basic tenets of LMBM are expli-

cated. In section 3 I deploy the analytical tools of LMBM for an in-depth
examination of the grammatical, semantic and formal aspect of deverbal
noun formation in English. It will be argued throughout that the proper-
ties of action nominalizations in English call for an approach to morphol-
ogy which is not sign-based. Paradoxically, an approach which separates
form and meaning in complex words provides a more integrated account
which strikes the balance between phonological and semantic issues and
offers a deeper insight into mutual interactions between grammatical
modules.
1 THE DATA AND INHERENT PROBLEMS
An account of the category of deverbal nominalizations which aspires

to being called comprehensive must span all derivational types without
arbitrarily excluding one or the other. English deverbal nominalizations
are characterized by the following exponents: -ing, the so called Latinate
suffixes: -(at)ion (e.g. celebration), -ment (e.g. development), -ance/-ence
(e.g. acceptance), -age (e.g. leakage), -al (e.g. arrival), -ure (e.g. exposure), -y
(e.g. delivery)3 and zero (e.g. kick).4
The suffix -ing is frequently omitted from the scope of discussion
and relegated to the limbo of inflection-derivation borderline cases (e.g.
Szymanek 1989: 136–7). However, numerous arguments for distinguish-
ing the regular -ing nominalization from the syntactically derived gerund
can be found in, for instance, Chomsky (1970), Schachter (1976) and
Malicka-Kleparska (1988: 83–94). In this chapter, focus is placed on those
-ing forms which can fill the slot of the head of an NP, i.e. those which can
be modified by determiners, adjectives, an of-phrase or a relative clause as
in (1).5
(1) a. the annual gathering of the South Pacific Forum

b. her endless nagging, which drove him away from home
The following subsections are devoted to the formal and semantic char-
acteristics of deverbal nouns which defy a straightforward explanation.
Section 1.1 addresses the question of productivity and the scope of applica-
tion of particular exponents. It identifies theoretical problems which arise
if we want to subsume all exponents under one WFR. Section 1.2 indicates
the need for semantic distinctions which are more fine-grained than the
traditional process/result dichotomy. It also points to a close link between
semantics and countability.
182 maria bloch-trojnar
1.1 Productive exponents and the scope of their application

In terms of productivity and scope of application the relevant exponents
cannot be placed on the same footing. In order to discover the categories
relevant to the linguistic system and rules which are capable of generating
an infinite number of grammatically well-formed words we must concen-
trate on productive morphological processes (Aronoff 1976, 1980; Bauer
2001, 2003; Plag 1999). Morphological rules are characterized by a cline
of productivity with unproductive and fully productive as extremes, e.g.
the suffix -al can rarely be used to form new words and the list of dever-
bal action nominalizations in the OED amounting to thirty-five items is
unlikely to be expanded (Haspelmath 2002: 41). The number of derivatives
cannot of course be regarded as a yardstick for productivity due to the fact
that a high type frequency may merely be an indicator of past productivity.
The suffix -ment is a case in point. Bauer (2001: 181) observes that the suffix
‘appears to have been productive between the mid-sixteenth century and
the mid-nineteenth century’ and according to Haspelmath (2002: 109) only
four twentieth-century neologisms with -ment are attested in the OED.
Bauer (2001: 183) contends that in synchronic terms there are two ways
of forming nominalizations in English: conversion and suffixation in -ion.6
Notably, Bauer (2001: 177–99) excludes -ing nominalizations from the scope
of his analysis. Since it is based on a sample from the OED, it does not come
as a surprise that -ing forms, being totally regular and predictable in formal
and semantic terms, are not listed. Malicka-Kleparska (1988), in turn,
concentrates solely on suffixed nominalizations, since in her view ‘conver-
sions should be looked upon as a phenomenon with a different status than
suffixal nominalisations’ (Malicka-Kleparska 1988: 12). She argues that -ing
is a truly productive suffix, while the remaining suffixed nominalizations
are mostly lexicalized and can only be described in terms of redundancy
statements. This does not apply to nominalizations in -ment corresponding
to be-, eN-verbs (e.g. bedevilment, bereavement, enforcement, embezzlement)
and nominalizations with the suffix -(A)tion whose a llomorphic variants
appear regularly with verbs ending in -ize, -ate and -ify (e.g. privatization,
evaluation, intensification) (Malicka-Kleparska 1988: 165).
Malicka-Kleparska (1988) stresses the fact that the process of -ing suf-
fixation is characterized by high generality since only two groups of verbs
fall outside its scope.7 The first group includes various stative verbs such as
those denoting relations and verbs of emotion and cognition, for instance
belong, hold, believe, admire, love, know (Lees 1960: 66). Postpositional verbs
such as abide by, depend on constitute the second (Marchand 1969: 249).8
There is a strong tendency for -ing to occur with transitive activity verbs
(Malicka-Kleparska 1988: 103). As far as bases for conversion are concerned,
Bauer (2001:181) points to morphologically simple verbs and phrasal verbs

(e.g. jump, walk, think, check up). Adams (2001: 29) observes that bare
nominals are derived predominantly from native items and occur mostly in
complex predicates, in informal registers.9 According to Brinton (1998: 49),
verbs denoting all situation types qualify as potential input to the rule of
verb-to-noun conversion, i.e. verbs denoting states (e.g. love, hate, delight),
activities (e.g. run, walk, chat), accomplishments (e.g. escape), achievements
(e.g. collapse, defeat, round-up) and semelfactives (e.g. bite, kiss, pinch).
In sum, we can observe that three morphophonological operations are
available, i.e. -ing, conversion and -ion (and its variants) added to verbs in
-ize, -ate, -ify. There is a great deal of overlap in the scope of application of
exponents which may lead to more than one derivative based on the same
stem, as illustrated in (2).
(2) a. refusal – refusing, civilization – civilizing, equipment – equipping

b. drawing – draw, launching – launch, looking up – look up
c. movement – moving – move, reservation – reserving – reserve,
commandment – commanding – command
d. transferal/transference – transferring – transfer
Latinate bases primarily accept Latinate suffixes but also admit -ing, as
in (2a). The suffix -ing attaches to activity verbs, which are also input to
conversion providing that the base verb is morphologically simple (prefer-
ably native) or a phrasal verb, as in (2b). As a result doublets are the order
of the day. There are also occasional triplets, as in (2c) or even quadruplets
based on the same verb, as in (2d).
High productivity (profitability) of morphological processes used for
transpositional purposes is something to be expected (Bauer 2001: 208).
However, the existence of diverse forms based on the same stem exhibit-
ing an actional reading is bound to create problems for an analysis where
the existence of one rule subsuming both suffixed and converted nomi-
nalizations is envisaged (see Cetnarowska 1993). The interaction between
competing morphological processes may be twofold. In cases where
constraints strictly delimit complementary scopes of application, affixa-
tion rules are unordered (parallel affixation) (Malicka-Kleparska 1985;
Plag 1999). Another option is to order rules from the most to the least
specific, the domain of application of one affix curtailing the domain of
application of another (co-functional affixation). This approach is reflected
in Szymanek’s (1985) disjunctive ordering and van Marle’s (1985, 1986)
Domain Hypothesis. Ordering implies no doublets, let alone triplets. Lack
of ordering requires complementary domains. Neither is satisfied if all
nominalization processes are combined into one WFR.
The data also pose a problem for the mechanism of blocking (Aronoff
1976; van Marle 1986; Rainer 1988). In his discussion of the thief – stealer
case, Bauer (1988: 66) argues that blocking prevents institutionalization of
potential forms to the effect that they are not accepted into general use or
listed in dictionaries. He regards -ing as the default nominalizing marker
for freshly coined verbs such as Unseav – a verb formed by conversion from
the acronym United Nations Special East Asian Volunteers (Bauer 2001:
90), as in (3).
(3) We wish to protest in the strongest terms against the Unseaving of the border
between North and South Korea.
However, -ing is attested on native as well as Latinate bases and we encoun-

ter pairs such as civilization/civilizing or equipping/equipment, i.e. there is
no blocking effect.10 Malicka-Kleparska (1988: 165) admits that -ing forms
may sound slightly clumsy but there are abundant examples to be found in
the OED, such as those in (4a–b) and (4c–d) below.
(4) a. to attempt the civilization of the Australian aborigines

b. the civilizing of the Highlands of Scotland . . .
c. for the endowment and equipment of a chair of Anatomy
d. the equipping of two such armaments
The occurrence of -ing alongside special cases, i.e. Latinate suffixes, means
that it cannot be regarded merely as the general or elsewhere case, in the
sense of van Marle (1985). Therefore, in Malicka-Kleparska’s analysis
-ing affixation is granted the status of an independent rule, which does not
interact with other affixation operations. It belongs to a separate block.
If zero derivation is co-functional with Latinate suffixation, as argued
by Cetnarowska (1993), we can explain the non- occurrence of bare
nominalizations in the actional reading if there is a corresponding Latinate
nominalization, e.g. reserve receives no actional interpretation due to the
existence of reservation, as in (5a–b). Since -ing belongs to a separate block,
reserving is generated independently of reservation and can be used in the
actional reading, as in (5c).
(5) a. This guarantee ensures the reservation of your room after 7 p.m.11
b. *This guarantee ensures the reserve of your room after 7 p.m.
c. the reserving of parking spaces12
With native verbs, we expect no blocking effects between -ing and zero
derivatives and it should be possible for -ing forms to replace the zero
derived nominalizations in the actional reading. We find doublets such as

beating – beat, transferring – transfer, drawing – draw but they cannot be used
interchangeably like forms in (4) above. The meanings of -ing and zero nom-
inals are actional but not synonymous, as the opposition in (6) illustrates.
(6) a. I love you with every beat of my heart.13

b. *I love you with every beating of my heart.
It transpires that -ing is generated independently of Latinate nominals

whereas bare nominals are co-functional with them. However, the relation-
ship between -ing and zero derivatives is not the same as that between -ing
and Latinate nominalizations.
1.2 Regular and lexicalized meanings

Due to lexicalization, the meaning of action nouns, regardless of the
derivational type they belong to, exhibits a tendency to gravitate towards
denoting concrete referents (Marchand 1969: 303). Malicka-Kleparska
(1988) perceives lexicalization as the process of incorporation of the most
object-like thematic role of the verb into the meaning of a given nominali-
zation while other thematic roles (if there are any left) are deleted. Notably,
the lexicalized result reading is not always countable in the case of suffixed
nominalizations, whereas lexicalized zero derivatives always are, as illus-
trated in Table 10.1.
The actional meaning of suffixed nominals always goes hand in hand
with uncountability, whereas the concrete reading may be associated with
countable nouns. For every count noun there seems to be a corresponding
non-count noun but not vice versa (Bloch-Trojnar 2007: 51–2). In Cobuild
(2009), performance has two glosses associated with a count and a non-count
noun, i.e. ‘a performance involves entertaining an audience by doing some-
thing such as singing, dancing or acting’ and ‘the performance of a task is
the fact or action of doing it’. Nominals such as corrosion and condensation
are listed only as non-count nouns.
Table 10.1 Countability distinctions in lexicalized deverbal nominals
-ing Latinate suffixes Ø
building(s) possession(s) cut(s)
filling(s) in a tooth solution(s) find(s)
warning(s) government(s) exhibit(s)
stuffing condensation deposit(s)
roofing disposal
attendance
Two more facts have not received due attention, namely a systematic
difference in the meaning and grammatical category of -ing and zero deriv-
atives and the ambiguity of Latinate nominalizations. Nominalizations in
-ing are uncountable and interpreted as ‘action or process of V-ing’ (e.g.
building, walking, transferring), whereas the nomen acti reading, i.e. ‘a single
instance of V-ing’, is prevalent in countable zero derivatives (e.g. a look,
a kick, a gasp) (Cetnarowska 1993: 112–13; Adams 2001: 28–9). Malicka-
Kleparska (1988: 30) argues cogently that Latinate nominalizations can
refer to actions in their entirety and their meaning is not restricted to the
traditionally recognized process vs. result contrast. The examples in (7)
come from Pustejovsky (1995a: 170), who also regards Latinate nominals
as ambiguous between process, event and result readings.
(7) a. The house’s construction was finished in two months. (event)

b. The construction was interrupted during the rains. (process)
c. The construction is standing on the next street. (result)
To recapitulate, Latinate nominals denote a process, event and result and

are non-count on an actional interpretation. Formations in -ing refer to
processes and are uncountable, whereas zero derivatives refer to events
and are countable. Suffixed nominalizations show variable countability in
lexicalized senses.
2 THEORETICAL BACKGROUND
This section is a résumé of the major tenets of LMBM. I will characterize

the model of grammar developed by Beard (1995) and pay special atten-
tion to those aspects which differentiate it from other frameworks and are
conducive to analysing our data, i.e. the modular structure of grammar, the
realizational approach to morphology (the separation hypothesis) and the
morpholexical status of nominal number.
2.1 Model of grammar

LMBM advocates a strictly modular structure of grammar because of
the ubiquitous lack of isomorphism between grammatical modules.
Semantic classes, which are the reflection of conceptual structures in the
cognitive system, do not map isomorphically onto grammatical categories.
For example, the grammatical distinction between count and mass cuts
across a wide range of semantic classes. The system of determination and
quantification will not discriminate between dog and indication on the one
hand and toast and disrespect on the other, even though they do not share a
common aspect of meaning (Willim 2006: 31).
There are numerous examples of misalignments between semantic and
syntactic structures. For example, not all arguments at the level of Lexical
Conceptual Structure (LCS)14 will be realized as syntactic arguments of
a predicate and vice versa. Jackendoff (2010: 16–17) explains that there
may be verbs with supernumerary syntactic arguments which contrib-
ute nothing to the semantics, e.g. reflexive arguments of perjure oneself,
avail oneself of X alongside synonymous verbs which feature in different
syntactic configurations, e.g. replace X with Y, substitute Y for X. Such
misalignments are also recognized in generative semantics. Pustejovsky
(1995a: 63–4) argues that only true arguments are necessarily mapped
onto/expressed as syntactic constituents (8a), whereas default arguments
(8b) and shadow arguments (8c) are not.
(8) a. John arrived late. (true argument)

b. John built a house out of bricks. (default argument)
c. Mary buttered her toast with an expensive butter. (shadow argument)
Most conspicuous is the lack of isomorphism between lexical and gram-

matical categories and their phonological manifestation. One affix may
express a whole range of grammatical and semantic functions (e.g. -er:
agent noun, instrument noun, inhabitant noun, comparative adjective),
and conversely one function may be expressed by several exponents
(e.g. agent nouns in English: -er/or, -ant, -ee, -ist and Ø). Szymanek
(1989: 108–19) stresses the fact that the lack of one-to-one correspond-
ence between form and meaning in morphology received due attention
in Slavic studies on word-formation but was ‘overlooked or neglected
in most Western studies’. The English suffix -ing is one of at least eight
co-functional affixes realizing nomina actionis, and it can be linked to a
derivational function of subjective adjective and inflectional present par-
ticiple. Polyfunctionality is rampant in affixes, whereas in lexemes it is an
exception rather than the rule. In contradistinction to affixes which can
be both synonymous and polysemous like -ing or -er in English, lexemes
either have multiple meanings (polysemy) or share the same meaning
(synonymy), i.e. the misalignment of form and meaning is relatively rare
and unidirectional (Beard 1981: 105).
Robert Beard can be regarded as the most articulate champion of the
‘separation hypothesis’, i.e. separating the functional-semantic and formal
aspects of word coining. Similar views were voiced earlier by Slavic lin-
guists such as Dokulil (1962), Grzegorczykowa (1979) or Grzegorczykowa
and Puzynina (1979). In the generative thrust of research the first to
consider this possibility was Jackendoff (1975). It is also argued for and
implemented by Laskowski (1981), Szymanek (1985, 1988), Malicka-

Kleparska (1985, 1988), Aronoff (1994) and Bloch-Trojnar (2006).
Therefore, in LMBM there are strict boundaries between the LEXICON
(the storehouse of lexemes), the GRAMMAR (structural relations of syntax
and a set of morphological categories), the SEMANTIC MODULE and
the MORPHOLOGICAL SPELLING COMPONENT (the component
mapping grammatical function to phonology).
Beard (1995: 381) envisages the interaction between components in the
following way:
The only place in language, then, where semantic, grammatical,

and phonological representations are directly related to each other,
is the lexicon. The direct relation of these representations defines
the lexeme. Elsewhere, at the syntactic, semantic, and phonological
levels, information from one domain must be translated into the
representations of any other domain which employs that information.
Morphology does all the translating.
In other words, grammatical categories (derivational and inflectional ones)

provide a link with semantics whereas affixation rules map grammatical
functions to phonology.15
There is a rigid distinction between LEXEMES and MORPHEMES.
All open classes are lexical and thus housed in the lexicon; all closed classes
(including free morphemes) are grammatical and hence belong to the realm
of the Morphological Spelling Component (MS-Component). In Beard’s
model a lexeme is conceived of as a mutually implied triplet, p 4 g 4 r
where p stands for phonological representation or matrix, g for grammati-
cal representation or feature inventory and r for semantic representation or
feature inventory. Lexical and syntactic rules are abstract operations which
apply to the grammatical representation of a lexeme, i.e. to features such
as +Singular, –Plural, +Feminine, –Masculine.16 Grammatical features
(G-features) subsumed by inherent (morpho)lexical categories are internal
to (morpho)syntactic inflectional categories.
Affixation and other formal operations (prosodic variation, internal
modification, Ø, etc.) are post-syntactic and take place in an autonomous
Morphological Spelling Component. In this model no zero morphemes per
se are recognized (Beard 1984: 53). We are dealing with a process (a deriva-
tional or inflectional one) without an overt phonological reflex which may
contrast with a set of exponents used to mark the same function.
Figure 10.1 graphically represents ‘how derived lexemes with directly
related sound and meaning may be generated without any direct relation
Lexeme Operations on Responsible

lexemes grammatical
component
R Semantic SEMANTICS
operations
G Lexical derivation LEXICON

Inflectional SYNTAX
derivation
P Spelling operations MS-COMPONENT

Phonological PHONOLOGY
operations
Figure 10.1 Derivational and spelling operations in LMBM (Beard 1995: 49)
between the rules of derivation and those of morphological spelling, or

between the meanings and forms over which they operate’ (Beard 1995: 49).
2.2 The grammatical representation of the lexeme and

transposition
Beard’s classification of features as either morpholexical or morphosyn-
tactic hinges on the result of the application of three tests: the peripheral
affix test, the free analog test and the arbitrariness criterion.17 As for the
cardinal verbal categories Beard’s findings converge with other classifica-
tion systems, i.e. Verb Class18 and Transitivity are morpholexical features,
whereas Modality, Mood, Aspect, Voice, Tense, Person and Agreement
are morphosyntactic. However, the application of his tests leads to differ-
ent results with respect to nouns. Number, Gender and Noun Class19 are
morpholexical nominal categories, while Agreement and Case are inflec-
tional. Number distinctions are encoded by the features [± Singular] and
[± Plural]. Different value settings of these features correspond to various
classes of nouns, as in Table 10.2.20
Table 10.2 +/− Singular and +/− Plural as operators
+Singular –Singular
+Plural COLLECTIVE PLURAL NOUNS
–Plural COUNT NOUNS MASS NOUNS
In Beard’s model all operations on number properties (including

pluralization)21 are derivational. We can easily envisage feature value
switches involved in pluralization and recategorization phenomena, which
refer to a shift from a count noun to a mass noun as in (9a) and vice versa
as in (9b).
(9) a. We had lamb for dinner.

b. I’d like a coffee please.
In my analysis, number is regarded as a morpholexical feature since it

is lexically determined whether a given noun is count, mass or pluralis
tantum. However, I do not share Beard’s view that pluralization is always
derivational (Bloch-Trojnar 2012a).22
In LMBM ‘the Lexicon may transpose any member of any major lexical
class (N, V, A) to any other major lexical class by providing it only with the
lexical G-features of the target class and neutralizing (but not deleting) the
inherent G-features of the base’ (Beard 1995: 177).23
2.3 The lexicon and lexicalization

Beard (1981, 1987) argues that the lexicon should be conceived of as two
subcomponents: a functional subcomponent, which subsumes L-derivation
rules (Lexical Extension Rules), and the storage, which contains noun,
verb, adjective and adverb stems. Rules define L-derivations abstractly,
i.e. without any reference to affixation.24 They produce lexemes with
transparent meanings and are automatically applied to lexemes in lexical
storage. There are also Lexical Stock Expansion Rules which provide new
stems for the storage subcomponent rendering new names for objects,
activities and qualities in the world.
According to Beard (1987) ‘semantic drift’ affects stored items in both
principled and random ways. Semantic irregularity may be the offshoot
of evolution from primary transparent meanings (e.g. construction, paint-
ing) or an idiomatic meaning may be added to the output of a regular
process (e.g. transmission ‘gearbox’).25 An affected item ‘disengages from
the productive L-derivation rule which generates it’ (Beard 1987: 26) and
enters the lexicon, which has two consequences. Firstly, the generation of
another form is necessary to fulfil the original function. Thus well-being
was coined to replace wealth. Secondly, the stored item can serve as input
to L-derivation. Thus hateful is no longer regarded as a derivative of hate,
and hateful-ness, hateful-ly arise. When the meaning of a lexical derivate
drifts, it attracts its form with it (e.g. holiday, ghastly) or at least isolates it
from its derivational origin (e.g. awful, lovely).
3 ANALYSIS IN THE FRAMEWORK OF LMBM
Figure 10.2 gives an overview of the representations of regular and lexical-

ized deverbal nominalizations from the vantage point of LMBM.
The functional subcomponent of the lexicon generates abstract lexical
items whose grammatical features must be interpreted by other gram-
matical modules. Section 3.1 is devoted to syntactic ramifications of verb to
noun transposition. In section 3.2 an attempt is made to untie the Gordian
knot of semantic interactions between regular and lexicalized senses.
Section 3.3 deals with constraints governing the distribution of formal
exponents.
VERB
± Transitive
Verb Class
(native/Latinate)
Transposition
Non-count NOUN Count NOUN Plural

(Process) (Event)
0Transitive 0Transitive 0Transitive

0Verb Class 0Verb Class 0Verb Class
–Sg +Sg –Sg
–Pl –Pl +Pl
Noun Class X, Y Noun Class Z Noun Class Z
(native/Latinate) (native)
civilizing civilization beat
equipping equipment look
observing observation swim
destroying destruction dump
describing description LEXICALIZATION
beating
building
RESULT/OBJECT RESULT/OBJECT
–Sg +Sg –Sg

–Pl –Pl +Pl
equipment beat dump
destruction civilization
stuffing observation DERIVATION
description
building INFLECTION
LEXICALIZATION
Figure 10.2 LMBM representation of deverbal nominalizations in English

3.1 Grammatical interpretation of deverbal transpositions

A nominalizing L-derivation rule provides the morpholexical features of
substantives. As nouns in English do not bear morphological gender fea-
tures, the only category that is available to grammatical processes of trans-
position is number and we expect two major subclasses of nouns, count
nouns specified as [+Singular; –Plural] and non-count nouns [–Singular;
–Plural]. The information of the native/Latinate character of the base
is reflected by the division into noun classes. The syntactic structures
with derived nominals resemble structures with morphologically simple
nouns in all respects, i.e. they are preceded by appropriate determiners,
adjectives, and followed by prepositional phrase complements. In regular
nominalizing processes, the argument structure information of the verb is
preserved but neutralized, which means that LCS participants of verbs can
be realized in the complement and adjunct positions of a nominal phrase
marker, but they are by no means obligatory (see Grimshaw 1990). Since
nominal modifiers are generally optional we find suffixed and bare actional
nominals with or without accompanying NPs and PPs, as illustrated in
(10a–c) and (10d–f) respectively.26
(10) a. The enemy’s destruction of the city

b. John’s proving of the theorem27
c. the change of the climate by man’s emission of greenhouse gases28
d. But it’s not the sex I miss so much – it’s the kissing.
e. In hospital she’ll be under observation all the time.
f. The climb took twenty minutes.
The operation of pluralization can only apply to verbal nominals specified

as [+Singular; –Plural]. Count nouns as a rule may take plural morphology
and co-occur with cardinal numbers and enumerative determiners, as in
(11).29
(11) a. Can I have two guesses?

b. I took several rides in his car
The picture presented above ties in with the widely held view that there is
an analogy between count entities and events, on the one hand, and mass
entities and activities on the other (see Mourelatos 1978; Langacker 1987b;
Jackendoff 1991; Krifka 1992; Brinton 1998; Willim 2006; and the refer-
ences therein).
Brinton (1998) investigates the mapping of Aktionsart properties of a
verb to a corresponding nominalization and argues that, unlike Latinate
suffixes, -ing nominals and Ø-derivatives fail to preserve the Aktionsart

of the verb. Whereas the -ing suffix ‘has the effect of converting the situa-
tion into an activity, of making the situation durative, atelic and dynamic’
(Brinton 1998: 48), conversion ‘is a means of converting the situation into
an event (an accomplishment, achievement, or semelfactive) by adding the
feature of telicity’ (Brinton 1998: 50).30
In the process of lexicalization the verbal features are deleted, nominals
are viewed as non-derived objects and there is only a vague connection
with the verb due to phonological shape. Non-count nouns may remain so
(e.g. destruction), may acquire a referential meaning (e.g. stuffing) or may be
turned into [+Singular; –Plural] nouns capable of pluralizing (e.g. descrip-
tion, filling (in a tooth)). Count nouns in non-actional lexicalized meanings
will refer not to actions but to entities of some sort (e.g. permit, dump). The
seemingly arbitrary mode of lexicalization of suffixed nominals as either
countable or uncountable can partly be predicted from their aspectual
characteristics (Brinton 1998: 47) in the sense that state and activity verbs
give rise to mass nominalizations, as in (12a), whereas accomplishment,
achievement and semelfactive verbs are lexicalized as countable, as in
(12b).31
(12) a. the survival, a lot of guidance, some leakage, some resemblance, much
astonishment
b. a refusal, an arrangement, a lot of marriages, an entrance, an appearance,
two dismissals, a few conquests
3.2 Semantic interpretation of deverbal nominalizations and the

lexicon
The complexity of linguistic reality is a natural corollary of the interaction
between products of the functional subcomponent of the lexicon and listed
lexicalized items. In the account developed here, productively formed
nominalizations in -ing and zero derivatives are generated while the major-
ity of Latinate nominalizations are listed.
If Latinate nominalizations are listed, the generation of -ing forms as
their replacements in actional senses becomes essential, hence doublets
such as civilization and civilizing. There is some evidence to argue that
Latinate nominalizations are gradually becoming isolated from their origin
(they no longer have verbal features in their representation) because they
can serve as input to further derivation. Relational adjectives in English are
derived from Latinate nouns (e.g. industrial, atomic, cellular). Szymanek
(1985, 1989: 213–25) notes that nomina actionis do not give rise to this cat-
egory unless they have lexicalized senses (e.g. governmental, adaptational,
motivational, transformational). Regular and lexicalized nominalizations

behave in a different way with respect to further derivation.
How many entries are there then? No formal difference distinguishes
the regular and irregular uses of transmission. In the separationist model
of LMBM we can argue that two different categories happen to have a
homophonous realization. In the sense of ‘gearbox’, transmission is listed
in the storage, whereas transmission ‘act/action of V-ing’ is generated by
rule. We shall argue that Latinate nominalizations are logically polysemous
items along the lines proposed by Pustejovsky (1995a).32
Pustejovsky (1995a) rejects the approach in which the lexicon has multi-
ple listings of words annotated with separate meanings (Sense Enumerative
Lexicon). Pustejovsky and Anick (1988) introduce the notion of a lexical
conceptual paradigm (lcp), which allows a lexical item to be regarded as a
metaentry. There are three senses available to an lcp constructed from two
base types, as depicted in (13).
(13) process.result_lcp = {process.result, process, result}
Just like the noun construction in (7) above, the noun merger could allow ref-
erence to the entire event of the merging, in addition to a possible process
or result interpretation (Pustejovsky 1995a: 93).
Pustejovsky (1995a: 167–71) argues that only nominals which have the
process and result reading can have the additional event interpretation.
That is why -ing nominalizations as in (14) cannot be polysemous, because
they are not interpreted as the result of an event.33
(14) The launching of the space shuttle occurred at 10.30.
To this we could add that there is a zero-derived noun based on the same
verbal base which has the semantics to fill the remaining slots in the lcp.
According to Cobuild (2009) the nominal launch has the event reading (the
dot object) as in (15a) and the object reading as in (15b).
(15) a. This morning’s launch of the space shuttle Colombia has been delayed.
b. The captain was on the deck of a launch, steadying the boat for the pilot.
Pustejovsky (1995a) does not consider zero derived nominals in any

detail, but his observation with reference to the nominal purchase ties in
with the interpretation proposed above. He notes that purchase behaves
like a dot object which is logically polysemous between the event and the
object involved in the transaction (Pustejovsky 1995a: 179). Thus -ing
nominals (e.g. purchasing, pushing) refer to a process and contrast with
bare nominals (e.g. purchase, push) which denote an event, as illustrated

in (16).
(16) a. the purchasing of office supplies by State Agencies in Mississippi34

b. the purchase by India of howitzers from the Swedish firm of Bofors35
c. For hours there was pushing of the cart by Jones.36
d. Information is called up at the push of a button.
However, -ing nominals such as launching, purchasing, pushing should be

contrasted with nominals such as painting or building. Since the latter
display the result reading, the event interpretation should also be available,
which immediately prevents an event interpretation of a bare deverbal
noun, as seen in (17).37
(17) a. Brown’s painting of his daughter is taking a long time.

b. Brown’s painting/*paint of his daughter took one evening.
c. Brown’s paintings of his daughter
The situation is predictable in cases where, in addition to the - ing

nominal, there is a Latinate nominalization and a zero derivative based
on the same verbal stem. As expected, zero derivatives will not have an
event interpretation since it is already rendered by lexically listed Latinate
nominals, whereas an - ing nominal will be restricted to the process
reading, as in (18).
(18) the formation/forming/*form of clouds from water vapour38
This generalization will not hold if the Latinate nominal has become totally
isolated from its derivational source and shows no actional reading what-
soever, as in (19a). Products of regular processes will then fill the vacated
slots of an lcp in question, as in (19b–d).
(19) a. The Ten Commandments

b. a suite of tools for the commanding of planetary rovers
c. the struggle for command of the air
d. In computing, a command is an instruction that you give to a computer
In sum, lexicalization affects individual lexical items (not classes), which

explains why all types of nominals regardless of the formal marker may
show the result reading (building, civilization, launch). Actional readings
are of two types: process and event. Since only nominals which display
the result reading in addition to the process reading can be three-way
ambiguous, this ambiguity characterizes listed Latinate nominalizations

and a fraction of -ing nominals. It follows that nominals which denote
events and results (zero derivatives) cannot refer to processes. With native
roots a bare nominal and an -ing nominal are part of one lcp and show an
aspectual contrast (launching ‘atelic’ – launch ‘telic’), whereas Latinate
nominals block bare nominals in the event reading as long as their actional
interpretation is possible.
3.3 Morphological spelling

When the spelling operations begin to apply, the first operation can only
modify the phonological base, since this is the only phonological represen-
tation available. The interpretation of the g, p and r features (see section
2.1) builds outwards from the base, responding to each feature or set of
features that serve as conditions on its operations.39 A cluster of features
with [–Singular; –Plural] is marked with -ing. If the base is a Latinate
verbal stem one of the Latinate affixes is selected.40 A verbal root marked
[+Singular; –Plural] receives no marking at all, unless we are dealing with
a Latinate base.
4 CONCLUSION
The verb lexeme is subject to operations on its morpholexical features

which produce a lexical conceptual paradigm of deverbal nominal forms
with categories such as process, event and result. The first two are gener-
ated by regular WFRs (Lexical Extension Rules) whereas the latter category
encompasses products of lexicalization phenomena (Lexical Expansion
Rules). Process nominalizations are uncountable, whereas event nominali-
zations are countable when the base verb is native and uncountable when
the base verb is Latinate. Three semantic categories are mapped onto
three grammatical categories which in turn are mapped onto three types
of MS-spelling operations. However, there is no neat one-to-one mapping
between semantics, grammar and phonology. Suffixed nominalizations
denote processes, in the event reading zero contrasts with Latinate suffixes.
Idiosyncratic readings are observed in items with all formal exponents, as
summarized in Table 10.3.
On this view, morphology is interwoven with other components. In
the words of Beard (1995: 379–80), it is ‘the glue which holds the various
levels of language together. Grammatical functions map lexical classes and
syntactic categories onto semantics. The MS-module maps grammatical
functions to phonology.’
Table 10.3 An LMBM analysis of deverbal nominalizations in English

SEMANTIC RULE TYPE FEATURES IN THE EXPONENT
CATEGORIES GRAMMATICAL REPRESEN-
TATION OF THE NOUN
PROCESS LEXICAL [[V] +Latinate [–Singular; –Plural]] Latinate suffixes
{process} EXTEN- [[V] –Singular; –Plural] -ing
RULES
EVENT SION [[V] +Latinate [–Singular; –Plural]] Latinate suffixes
{process • result} [[V] –Latinate [+Singular; –Plural]] no modification
RESULT/ [+Latinate [–Singular; –Plural]], Latinate suffixes
OBJECT [+Latinate [+Singular; –Plural]]
LEXICAL
{result} [–Latinate [+Singular; –Plural]] no modification

EXPAN-
RULES
SION
Listed items [+Singular; –Plural] -ing
NOTES
1. I would like to thank Anna Malicka-Kleparska for comments on pre-

liminary drafts of this article as well as the participants at the workshop
‘Meaning and Lexicalization of Word Formation’ in April 2010 in
Budapest for valuable feedback. I am extremely grateful to the editors
of the volume Pius ten Hacken and Claire Thomas, whose suggestions,
both stylistic and substantive, were of immense help in articulating my
ideas.
2. Various approaches to nominalization include Transformationalist,
Lexicalist and Neo-transformationalist accounts, the Configurational
Hypothesis and Event Structure theory. For a concise overview
of the most important theoretical trends the reader is referred to
Koptjevskaja-Tamm (1993) and Rozwadowska (1997).
3. Apart from these there are a number of isolated irregular verb –
noun pairs such as complain – complaint, pursue – pursuit, fly – flight,
grow – growth, laugh – laughter, sell – sale, lose – loss (Marchand 1969;
Szymanek 1989; Adams 2001).
4. At this stage I leave unresolved the vexed question whether we are
dealing here with the affixation of an unpronounced element or no
affixation at all. In morphological theory, this phenomenon is referred
to as zero derivation or conversion. In Slavic linguistics the term
paradigmatic derivation is employed, whereas in cognitive linguistics
it is regarded as semantic extension. The problems besetting various
approaches and conclusions drawn by linguists of different persuasions
are extensively discussed in Cetnarowska (1993: 14–19), Twardzisz
(1997: 63–85) and especially Bauer and Valera (2005).
5. Unless otherwise indicated, examples cited throughout this paper come

from the following dictionaries: Collins Cobuild English Dictionary,
Collins Cobuild Advanced Dictionary of English and The Oxford English
Dictionary (OED).
6. The suffixing operation involves allomorphs, whose distribution is
sensitive to the syntagmatic properties of the base (see Aronoff 1976:
95–105). The form -cation is restricted to bases ending in -ify (e.g. elec-
trify – electrification). The allomorph -ation is selected if the base ends
in -ize (e.g. globalize – globalization). The variant -ion is added pre-
dominantly to verbs ending in -ate (e.g. terminate – termination).
7. There is a general constraint precluding modal and auxiliary verbs as
possible bases for any nominalizing operation (Wasow 1981: 308–9),
hence they are not mentioned separately here.
8. Phrasal verbs are possible, yet infrequent (Chomsky 1970: 193; Palmer
1974: 218; Malicka-Kleparska 1988: 101–2), as in (i–ii) below.
(i) The laying down of tools was unexpected.
(ii) The cutting off of the electricity was a shock for us.
9. She considers cases like those in (iii–iv) below as unlikely but not
impossible.
(iii) have a consider
(iv) give the patient an examine
10. Chomsky (1970: 215) regards (v) and (vi) as equivalent.
(v) John’s refusal of the offer
(vi) John’s refusing of the offer
11. http://www.etaphotel.com/gb/frequently-asked-questions/reservati
on. html#reserver10 (accessed April 2010)
12. http://bloomingdaleneighborhood.blogspot.com/2010/02/
reserving-cleaned-street-parking-spaces.html (accessed March 2010)
13. Google returns almost one hundred thousand hits for this sentence.
14. A detailed discussion of the framework of Lexical Conceptual
Semantics is offered in Jackendoff (1983, 1990).
15. Beard’s model (1995) in some respects converges with Jackendoff’s
Parallel Architecture (2010). Jackendoff (2010) recognizes three inde-
pendent generative components, i.e. phonology, syntax and semantics,
‘each with its own primitives and principles of organization’ and inter-
face rules. He is not explicit about the role of morphology in his model.
Commenting on the separation of morphological rules (M-rules) and
semantic rules (S-rules) proposed in his (1975) paper, Jackendoff
(2010: 57) says that ‘this separation of M-rules and S-rules dissolves
the lockstep of phonology, syntax and semantics and quietly begins
to lay the groundwork for the Parallel Architecture. Restated in terms
of the Parallel Architecture, the M-rules are (morpho)phonology-to-
(morpho)syntax interface rules, and the S-rules are (morpho)-syntax-

to-semantics interface rules.’ Beard introduces an autonomous level
of morphological categories (as does Aronoff 1994), i.e. grammatical
functions which interpret semantic functions algorithmically. The
two models also diverge as far as the status of morphemes is con-
cerned. Despite its obvious drawbacks in handling non-concatenative
mechanisms, Jackendoff adopts an analytical rather than procedural
treatment of regular morphology, and morphemes are placed in the
lexicon as long-term memory associations of phonological, syntactic
and semantic information (Jackendoff 2010: 19). In Beard’s model,
morphemes belong to grammar. They are not determinants of meaning
but merely clues which signal a relationship.
16. Aronoff (1984) pointed out that there seem to be no lexical rules of
the type V S N, where any verb of any lexical (sub)category is trans-
formed into a noun of any lexical (sub)category. Instead, verbs serve
as bases for the derivation of, for instance, subjective, instrumental or
feminine nouns, but not nouns capable of fulfilling all those functions
simultaneously. It is features which define the categories, rather than
the N, V, A labels (see Beard 1988: 35).
17. For a detailed discussion the reader is referred to Beard (1995: 102–54).
18. Verb Classes determine stems and are more or less arbitrary groupings
of words associated with different sets of inflections.
19. For Beard gender is synonymous with natural gender and as such is a
morpholexical feature of sexed beings. Natural gender is kept distinct
from grammatical gender, which in Beard is referred to as ‘Noun
Class’.
20. The two features [±Singular] and [±Plural] in the grammatical repre-
sentation of nouns (Beard 1982) seem to map directly onto the features
±b(ounded) and ±i(nternal structure) at the level of LCS (Jackendoff
1991). A more detailed presentation of the interaction between lexical-
conceptual semantics of nouns and linguistic countability is offered in
Bloch-Trojnar (2012a).
21. Beard (1982, 1995: 111–15) is not the sole advocate of the possible
lexeme status of nominal plurals (see Tiersma 1982; Dimmendaal
1987). According to Corbett (1999: 10) the interaction of number
with the Animacy hierarchy has the result that ‘the status of number
as an inflectional category is much less straightforward than generally
imagined. [. . .] It really is not a simple inflectional feature (+/–plural)
ready to have a role in syntax.’
22. According to Corbett (2000) in English, subject noun phrases and
predicate verbs (in the present tense) have a singular-plural system,
and normally they match by agreement. For the class of nouns where
number contrast is present, its role in contextual inflection is regular.

However, it is only the count nouns that are involved in the number
system, whereas non-count nouns fall outside. Agreement should not
be conceived of as one-to-one copying since mismatches demonstrate
that the system of the controller is substantially different from that of
the target. Therefore what we call inflection and what derivation is not
a universal feature of categories or values but rather a matter of para-
metric choices of which paradigmatic oppositions are grammaticalized.
23. In LMBM, lexical derivation (L-derivation) subsumes transposition,
feature value switches and functional and expressive derivation.
24. The two subcomponents correspond to the Conditional and Permanent
Lexicons of Allen (1978) and Malicka-Kleparska (1985).
25. In the automotive sense transmission cannot have evolved slowly from
the primary, transparent meaning ‘process of transmitting’. Carstairs-
McCarthy (1992: 185) explains that ‘there is nothing in either the deri-
vation itself or our experience of the world, which might tell us that it
relates to the transmission of power from the engine to the wheels in
cars, rather than (say) the transmission of a message from the speaker
to the hearer in a telephone conversation (whereby transmission might
mean “handset”), or the transmission of an inherited characteristic
from one generation to another (whereby transmission might mean
“gene”).’ Thus the application of this term to a particular referent
is the result of a conscious choice of a speaker and its subsequent
institutionalization.
26. The distinction into Complex event and Simple event/result nomi-
nals introduced by Grimshaw (1990) and emphasized in the ‘event
structure’ and ‘structure theory’ of nominalizations (Alexiadou and
Grimshaw 2008) cannot be maintained. Alexiadou (2009) points to the
need for the dissociation of the presence of Argument Structure (AS)
from verbalizing morphology and the optionality of the licensing of
AS in the nominal system. This, however, is tantamount to undermin-
ing one of the cornerstones of the syntacticocentric morpheme-based
approach to deverbal nominals. For a more detailed discussion of these
issues see Bloch-Trojnar (2011). Also see Soares Rodrigues and Rio-
Torto (this volume).
27. Examples (10a–b) come from Chomsky (1970).
28. http://www.astralweb.co.uk/smart-meters-glossary.html (accessed
February 2010). Admittedly, structures in which bare nominals occur
with NPs and PPs corresponding to arguments of the base verb are hard
to come by and additionally may seem artificial, awkward or marked.
Rozwadowska (1997: 13) notes that ‘nominalizations are structures that
are rare in everyday conversation, especially those with a full array of
satellites’. According to Herbst (1988: 297) nominal phrases laden with

PPs and infinitival clauses are heavy and stylistically awkward. Hence,
in spoken language they are avoided and replaced with their sentential
analogues. Stylistic considerations, however, do not offer a satisfactory
explanation as to why suffixed nominals such as those in (10a–b) above
show a greater propensity to take modifying phrases. It can be argued
that the reluctance of bare nominals to occur with satellite NPs and
PPs has to do with their aspectual characteristics and the presence of
suffixed nominals based on the same root (Bloch-Trojnar 2011).
29. Examples in (11) come from Cetnarowska (1993: 43).
30. For a critical evaluation of approaches which insist on direct relation-
ships between aspectuality, countability and morphological marking in
action nominalizations see Bloch-Trojnar (2012b).
31. The examples in (12) are taken from Brinton (1998: 47).
32. For a more detailed exposition of Pustejovsky’s approach the reader is
referred to Thomas (this volume).
33. The suffixed Latinate nominalizations can be replaced by -ing forma-
tions only in the regular actional sense. No such replacement is pos-
sible in the lexicalized sense, e.g. the ancient civilizations/*civilizings
(Malicka-Kleparska 1988: 165).
34. http://www.procurement.msstate.edu/officesupplies.pdf (accessed
April 2010).
35. Example taken from Cetnarowska (1993: 71).
36. This example comes from Mourelatos (1978: 427), who regards the
-ing nominal as a lexical means of marking process (activity) predica-
tion. It is ‘mass-quantified’ and does not have a terminus or closure,
which defines event predications.
37. The examples in (17) are modelled on Quirk et al. (1985: 1290–1),
who distinguish a number of intermediate stages between the count
noun some paintings of Brown’s and the participle in a finite verb phrase
Brown is painting his daughter.
38. Google returns about 576,000 hits for the phrase the formation of
clouds and about 16,800 for the forming of clouds (June 2013). The
phrase the form of clouds is widespread, but it does not have an actional
interpretation.
39. The choice of exponents may depend on the phonological, morpho-
logical, syntactic or semantic properties of the base. Importantly,
conditions on affixation are ‘absolutely irrelevant in the determination
of the derivative’s meaning’ (Beard 1995: 51).
40. The block of affixation rules along the lines proposed by Malicka-
Kleparska (1988) should be understood as a morphomic function
(Aronoff 1994), a function providing nominal forms of verbal stems
with the lexical diacritic [+Latinate]. The feature [+Latinate] triggers

the application of affixation rules which, as Malicka-Kleparska (1988)
demonstrates, depend on various structural properties of the base, e.g.
the base for -(a)tion terminates in -ize, -ify, -ate or requires an intran-
sitive base, -ment attaches to be-, eN-verbs and is attached to transitive
bases. Derivatives with -al and -ance/ence have to be ordered first and
lexically marked. The choice of the most productive exponents (-ing
and zero) depends on a variety of factors, number marking being just
one of them (Bloch-Trojnar 2012b).
chapter 11
Degrees of lexicalization in Ancient

Greek deverbal nouns Lexicalization in Ancient Greek deverbal nouns
Germana Olga Civilleri
T he aim of this chapter is to show lexicalization phenomena in deverbal

nouns (DNs) as processes that make their compositional semantics
opaque. I use the label of DNs to refer to morphological nominalizations,
i.e. nominalizations produced by applying morphological means to a verbal
base, as in (1a), in opposition to syntactic nominalizations, as in (1b).
(1) a. destruction of the city.

b. destroying the city.
It will be shown how such phenomena lie at different points on the lexi-
calization scale. In order to do that, it is necessary to start with those nouns
which have compositional semantics by identifying the specific semantic
value which is normally associated with certain morphological schemas, i.e.
their core meaning. Once we make this core meaning clear, the lexicalized
meanings displayed by some nouns can be explained on the basis of their
relationship to it. In general, such a relationship is motivated by metonymic
and metaphoric shifts.
Thanks to the rich articulation of its morphological level, Ancient Greek
(AG) is a particularly interesting language in which to study derivation
rules: I analyse the formation of AG DNs by trying to associate a semantic
content to the patterns of morphological derivation. My corpus will be
composed of the Homeric poems which represent the beginnings of Greek
literature (see also Civilleri, submitted). In order to find the data within this
corpus, the computer query system Diogenes (Heslin 1999–2007) was used
for browsing the Thesaurus Linguae Graecae (TLG 2001). Using this com-
puter query system gives us many benefits, since it includes morphological
search tools and the well-known Liddell and Scott Greek–English Lexicon
(Liddell and Scott [1843] 1992).
204 germana olga civilleri
1 DNs IN ANCIENT GREEK: MORPHOLOGICAL STRATEGIES
The two main strategies for the derivation of nouns from verbal1 stems in
AG are suffixation and apophony. The former is the most common among
world languages (see Malkiel 1978; Kerleroux 1996; Naumann and Vogel
2000), probably thanks to its iconic character (see Dressler 1985). Suffixes,
indeed, represent ‘formal expansions of simpler form’ (Naumann and
Vogel 2000: 932) and, by adding phonetic content, they also carry semantic
information, as the examples in (2) show (Lehrer 2002: 525).
(2) a. Noun/Adjective + -ize ‘to make/cause’.

b. modernize X ‘to make X modern’.
c. winterize X ‘to get X ready for winter’.
Hence, it is more likely that suffixes have an intrinsically clear semantics

that modifies the value of the verbal stem. Not every suffix has a clear
inherent meaning, but there is an internal gradation in the semantic
potential of each suffix. We can represent such a condition by means of a
continuum, as in Figure 11.1.
At one extreme, we find the suffixes with inherent lexical meaning (I call
them semantically full suffixes), and at the other extreme, we have the suf-
fixes with no inherent meaning (semantically abstract suffixes).2 I bracketed
argument structure in Figure 11.1 because this feature does not describe the
whole category of nouns with semantically full suffixes. Such nominals may
or may not have argument structure. In any case arguments in nominals –
at least in AG nominals – are never obligatory.
The semantic value of the suffix -σι(ς) [-si(s)]3 is illustrated in (3).
(3) ἀνά- βα- σις

aná- basis
up go proc.f:nom.sg.
‘going up’
Semantically Semantically
full suffixes abstract suffixes
-sis, -mis -ia -mis -mone
more productive, less productive, resulting

more verbal, nouns are nearer to fully
(argument structure) referential nouns
Figure 11.1 Continuum of the semantic values of AG suffixes

lexicalization in ancient greek deverbal nouns 205
Here the suffix -sis, which has a process meaning (perhaps similar to the
English suffix -tion), is added to the prefixed verbal stem, marking the
category status of the DN and giving it a process meaning. The prefix
aná-is added to the verbal stem before suffixation: in fact, it is a so-called
preverb, typically affixed to verbal forms (see among others Booij and
Van Kemenade 2003 for such a terminology). It is an interesting question
whether the value of the preverb in verbs is the same as their value in DNs.
Several works claim that the main value of the AG preverb is to make atelic
predicates telic (see Cuzzolin 1995; Romagno 2004; Pompei 2010). As far as
-sis nominals are concerned, they are not only marked by the [+ durative]/
[+ process] feature but on the whole they seem to be [− telic] too, because
the event is not described as concluded in time. On one hand the higher
frequency of preverbed -sis DNs in comparison with non-preverbed cor-
responding forms is quite remarkable, e.g. ἀνάβασις [aná-basis], πρόβασις
[pró-basis], ἔκβασις [ék-basis] versus the simple form βάσις [básis] which
does not occur in the Homeric corpus, but appears in texts from later stages
of the language. On the other hand, the [+ durative] and [− telic] value of
suffixes like -sis is the most prominent feature in the derivation of the noun,
so the possible telic value of the verbal theme gets lost.4 As far as verbal
bases undergoing such a procedure are concerned, we can reconstruct the
historical track in (4).
(4) simple verbal base (atelic) > prefixed verbal base (telic) > suffixed noun
(atelic).
Generally speaking, we could claim that the function of the preverbs

in DNs is quite different from the function they have in verbs. Indeed
derivation rules modify the relationship between preverb and verbal base.
Furthermore another fact should be considered even more important:
when the preverbed verbal stem is selected for the formation of the DN,
what is selected is the whole lexical entry in which the semantic value of
the preverb is more or less incorporated into the semantics of the verb (see
Pompei 2010 for such an idea of incorporation of the preverb). Therefore
it cannot be an aspectual matter, but should rather be analysed as a merely
lexical one, i.e. in ἀνάβασις [aná-basis] ‘going up’ the preverb does not
have an aspectual value but merely plays a role in the meaning of the lexical
item which, in turn, the word is built on. However the problem should be
more thoroughly investigated.
Let us now have a look at (5).
ῥοχ-
(5) μός
roch- mós
slide def.proc.m:nom.sg.
‘landslide’
The suffix -μό(ς) [-mó(s)] has a process meaning too, but in the sense of
Simone (2003) probably encodes a more definite processuality than -si(s),
i.e. an event concluded in time (see Benveniste [1948] 1975). This means
that -mós nominals are not only [+ durative], but also [+ telic]. In fact,
Simone (2003) – borrowing terminology from the Arabic grammarians –
distinguishes between indefinite process nouns illustrated in (6a), definite
process nouns (6b) and nouns of once (6c).
(6) a. distruzione il nuotare5 (Italian)

‘destruction’ ‘(the) swimming’
b. nuotata (Italian)
‘swim’
c. sorso bracciata (Italian)
‘sip’ ‘stroke’
This distinction depends on the grade of definiteness of the described

event. The type of noun illustrated in (6a) denotes an event unless there is
reference to its conclusion, the type in (6b) refers to an event represented
as definite in time, and in (6c) the event is not only definite in time but
also denotes a single (reiterable) act. This sequence also reflects what Sasse
(2001) and others call a verbiness/nouniness rank. Nouns such as (6a) are
indeed more verbal, whereas (6b) and (6c) are increasingly nearer to the
prototypical noun side (see also section 5 below). Hence, in line with the
description of the feature of the suffixes -sis and -mós above, we will con-
sider -sis nominals as indefinite process nouns and -mós nominals as definite
process nouns or, in some cases – e.g. ὑλαγμός [hylagmós] ‘barking’ in (14)
below – even nouns of once.
In other cases, represented on the right-hand side of the continuum in
Figure 11.1, the suffix is not able to provide the lexeme with a clear seman-
tic value, since it only marks the category transposition V > N. A noun of
this type could be (7).
δύνᾰ-
(7) μις
dýna- mis
to be able f.:nom.sg.
‘power’.
I do not mean that such suffixes as those in (7) do not have any semantic
value at all, but synchronically their semantics is no longer productive.
Furthermore, the lower level of productivity of these suffixes (both in the
quantitative sense of frequency and in the qualitative sense of regular use)
prevents us from referring to them as grammaticalized suffixes, because
normally the trigger of a grammaticalization process is just the frequency
of the element, which undergoes a semantic bleaching and is generalized to
new contexts in a regular way, becoming grammatical.
As far as apophony is concerned, it represents the other productive
strategy which builds nouns from verbal bases. It is the alternation of the
stem vowel involving normal, zero and o-grades. For example, the DN in
(8) is formed without any specific suffix, only by displaying the stem λεγ-
[leg-] at the o-grade.
(8) λόγ- oς
lóg- os
say m:nom.sg.
‘word, speech’
Indeed, the suffix -os is an inflectional marker, not a derivational suffix.

On the basis of the iconicity principle, we should expect apophony
to have limited semantic potential. However, some elements lead us to
suppose that apophonic alternations (at least the o- grade) have some
semantic function too. Due to the specific thematic constraints of this
chapter this matter cannot be dealt with in more detail here.6 In fact, since
the aim of the present chapter is to show how the compositional meaning
of the DNs can be made opaque through lexicalization processes, it is
more useful to focus on those examples in which the semantic value of
the derivation rules is more evident. Therefore only nouns which show
semantically full suffixes will be taken into account here.
2 COMPOSITIONAL MEANING AND LEXICALIZATION
When other limiting factors do not occur, the meaning of the derived
noun normally results from the meaning of the base plus the derivational
rule7 (Kuryłowicz 1956) in a compositional way. For example, the nouns
ἀνάβασις [anábasis] and ῥοχμός [rochmós] have an entirely compositional
meaning, as shown in (3) and (5). However, there are many cases in which
the core rule normally associated with the suffix is not productive and the
word’s compositional meaning is lost. First of all, the semantics can change
depending on the phrasal context,8 as in the English example in (9) taken
from Aronoff and Fudeman (2005: 130), where two alternative readings (as
a place or as people) are possible.
(9) a. The president and his family live in the White House.
b. The White House announced yesterday that the peace talks will continue.
Secondly, there are cases of polysemy. Indeed, through metaphoric and

metonymic shifts (see Lakoff and Johnson 1980; Lakoff 1987; Taylor 1995;
Lehrer 2002, 2003), lexemes – and single suffixes too (see Lehrer 2002,
2003) – can develop a new meaning which is connected to the o riginal one.
Finally, many words are lexicalized, i.e. come into the lexicon, as whole
units. The focus of this chapter will be on the so-called lexicalized words,
since I will be analysing their relationship with polysemy.
The term lexicalization has at least two senses in which it should be
understood. Firstly, lexical is opposed to grammatical, i.e. what is lexical is
optional, irregular and holistic, whereas what is grammatical is obligatory,
regular and analytical (see Jakobson 1959; Lehmann 2002, 2004;9 Thomas,
this volume). Alternatively, lexicalization means admission into the lexicon
(see lexémisation in Fruyt 2000: 28). In general, nominalizations have a
marginal role within the lexicon, because they do not need to be stored and
can be produced and decoded on the basis of other lexical elements which
are already in the lexicon and certain rules stored in the lexicon as well
(Schwarze 2001). However, the DNs which undergo a lexicalization process
are no longer analysable in such a way and must be stored as whole units. So
this second meaning of lexicalization includes the most common uses of the
term in the linguistic literature according to Himmelmann (2004), namely
univerbation (see Lehmann 2004: 14, ‘loss of internal structure, thus of
compositional motivation’), idiomatization (as diachronic lexicalization, in
other words, according to Naumann and Vogel 2000, the reverse of gram-
maticalization) and fossilization/ceasing of productivity (see Naumann and
Vogel 2000: 930 – ‘word formations lose their transparency, regularity and
compositionality up until the point where they have become partly or fully
demotivated’; see also Aronoff and Fudeman 2005: 105).
Relying on both senses, I will explain the lexicalized lexemes as cases
of polysemy, in which metaphoric and metonymic shifts from the original
compositional meaning play a role (see Taylor 1995 for such a use of poly-
semy). According to the traditional cognitive perspective, the foundation
of which is the famous work by Lakoff and Johnson (1980),10 (conceptual)
metaphor is a cognitive process stimulated by resemblance between con-
cepts belonging to different conceptual domains while, on the other hand,
metonymy is a cognitive process which relates contiguous concepts within
the same conceptual domain.
It seems to me that lexicalization works according to the same pro-

cesses as those governing polysemy and, since its semantic result is the
creation of a new meaning as well, lexicalization can easily be described
as a special case of the larger phenomenon called polysemy. In fact, as in
polysemy, the new meaning arises by metaphoric and metonymic shifts
starting from the core meaning, which in the case of derived words is
compositional.
By analysing the data elicited from the Homeric Greek corpus (Iliad
and Odyssey) I will show how the lexicalized nouns have been lexicalized to
different degrees. In order to point out these facts more clearly, I will use
those words that show semantically full suffixes, since they seem to be more
productive and have a more evident core/original meaning from which the
lexicalization process moves. Furthermore, before analysing the data, it
seems relevant to me to underline that from the point of view of the verb–
noun continuum it is going to be evident that the analysed lexicalized forms
are also less verbal than non-lexicalized ones. That is to say, they share a
minor range of inherited verbal features such as event structure and argu-
ment structure, whereas they share some of the prototypical noun features
such as lack of argument structure and pluralizability (see the difference
between features of Result-Nominals and Complex Event-Nominals in
Alexiadou and Grimshaw (2008: 3)). In other words, in the DNs lexicaliza-
tion also corresponds with loss of verbal features. This fact can easily be
interpreted as loss of transparency and compositionality too.
3 DATA
In this section, we will see a number of words, some of which have trans-
parent, compositional semantics (section 3.1) while others undergo more or
less complete lexicalization processes (section 3.2).
3.1 Non-lexicalized DNs

I will start with the forms in which the relationship between the semantics
of the suffix and the semantics of the resulting DN is regular, i.e. composi-
tional. In addition to (3) and (5) above, an entirely compositional meaning
is shown in (10).
(10) ἀμφί- βα- σις

amphí- basis
around go proc.f:nom.sg.
‘surrounding’
The meaning of the word is perfectly derivable from the meaning of the
verbal base plus the meaning of the suffix, so the features of the verbal base
itself are more likely to be preserved. Indeed, ἀμφίβασις [amphíbasis]
preserves the event structure of the verbal base and exhibits argument
structure encoding,11 e.g. a subjective genitive in (11).
(11) ἀμφίβασιν . . . Τρώων (Il. 5.623)
amphíbasin . . . Trōṓn
surrounding:acc.sg Troian:gen.pl.
‘the surrounding by the Trojans’12
Another good example of an entirely compositional DN is (12).

(12) λύ- σις
lý- sis
dissolve proc.f:nom.sg.
‘deliverance (from)’
The possibility of having an argument structure shows the high degree of

verbiness of this noun. In (13), the argument is an objective genitive.
(13) θανάτου λύσιν (Od. 9.421)
thanátou lysin
death:gen.sg. dissolution:acc.sg.
‘deliverance from the death’
However, the so-called ablative value of the genitive is also likely, as

Murray’s translation underlines. After all, the Greek genitive is the mor-
phological case which covers most of the functions of the Indo-European
ablative (see, among others, Luraghi 1987). Therefore, the fact that the
nouns in (10) and (12) have an argument structure is further evidence of
their verbal nature and of their semantic compositionality. The presence of
an objective (or ablative) genitive in (13) is even stronger evidence for this
than the subjective genitive in (11). While a subjective genitive can often
easily be read as a merely relational genitive (i.e. not an argument of the
predicative noun, but a genitive which also pure nouns may exhibit), this
ambiguity does not arise for the objective genitive.13
Let us now have a look at some examples of compositional nouns ending
in suffixes other than -sis.
(14) ὑλαγ- μός
hylag- mós
bark def.proc.m:nom.sg.
‘barking’14
In (14) we find again the suffix -mós which bundles the action of barking
and makes it definite in time (in the sense above, see (5) in section 1). In
particular this noun encodes a so-called noun of once (Simone 2003; see
section 1), because it denotes a single event which can be repeated.
(15) ὀρχησ- τύς

orches- týs
dance abstr.proc.f.:nom.sg.
‘art of dancing’
In (15) the suffix -τύς [-týs], carrying a more abstract meaning than
-sis, according to Benveniste ([1948] 1975: 74) encodes ‘la disposition et
l’aptitude, l’exercice de la notion comme vocation et capacité de celui qui
l’accomplit, en un mot la « destination » subjective et en général la « fonction »
au sens propre, l’exercice de la notion étant considéré comme la « fonction »
de celui qui la pratique.’15 However, the degree of productivity of such
forms is lower than that of the forms with -sis and -mós. In the Homeric
corpus, I found eighteen -týs types with a total token frequency less than
fifty, whereas there are thirty-eight -sis types and twenty-seven (plus
seven non-deverbal) -mós types, each of which has a total token frequency
numbering in the hundreds. Besides, it must be said that the class of -týs
action nominals, which is a very early category in AG, is recessive in the
later stages of the language (see Meillet and Vendryes 1966: 370; Schwyzer
1953: 506). Furthermore, in the corpus that I analysed in Civilleri (2012)16
this category is totally lacking.
Another typical example of this class of DNs is (16), where the value of
the suffix is still clear.
(16) τᾰνυσ- τύς

tanys- týs
draw abstr.proc.f.:nom.sg.
‘drawing (the bow)’
The fact that the -týs and -mós nominals I found in Homer do not have
arguments also underlines their lower degree of verbiness compared to
the -sis nominals, since I consider argument structure to be a prototypical
verbal feature depending on the event structure of the predicate, which
implies participants.
3.2 Lexicalized DNs: an internal gradation

The cases presented in section 3.1 are prototypical examples of the most
productive DNs, in which the semantic relationship between suffix and
verbal base is transparent. Showing this is important in order to clarify the
various possible stages of lexicalization that this kind of DN can display
starting from the core meaning of the derivation rule. A first stage on the
path towards lexicalization is represented in (17).
(17) δεσ- μός

des- mós
bind def.proc.m:nom.sg.
‘imprisonment, bond, chain’
In fact, sometimes such a noun has the compositional sense of ‘imprison-

ment’ shown by (18) and (19). The translation ‘bonds’, as a collective, in
both cases is equivalent to ‘imprisonment’ in its meaning (see Liddell and
Scott [1843] 1992).
(18) ἀλλά σφωε δόλος καὶ δεσμὸς

allá sphōe dólos kaì desmòs
but pron:3pl.acc. snare:nom.sg. and imprisonment:nom.sg.
ἐρύξει (Od. 8.317)

erýxei
hold.fut:3sg.
‘but the snare and the bonds shall hold them . . .’
(19) χαλεπὸς δέ ἑ δεσμὸς

chalepòs dé he desmòs
grievous:nom.sg. ptcl. pron:3sg.acc. imprisonment:nom.sg.
ἐδάμνα (Il. 5.391)

edámna
force.impf:3sg.
‘his grievous bonds were overpowering him’
On the other hand the noun is widely used in the lexicalized sense of ‘bond’
as shown by the sentences in (20) and (21).
(20) περὶ χερσὶ δὲ δεσμὸν ἴηλα χρύσεον

perì chersì dè desmòn iēla chrýseon
about hand:dat.pl. ptcl. bond:acc.sg. cast.aor:3sg. golden:acc.sg.
ἄρρηκτον (Il. 15.19)

árrēkton
unbroken.Vadj.
‘and about thy wrists cast a band of gold that might not be broken’.
(21) ἀμφὶ δὲ δεσμοὶ τεχνήεντες ἔχυντο

amphì dè desmoì technḗentes échunto
about ptcl. bond:nom.pl. cunning.part.pres:nom.pl. cling.aor:3sg.
πολύφρονος Ἡφαίστοιο (Od. 8.296)
polýphronos Hephaístoio
wise:gen.sg. Hephaestus:gen.sg.
‘and about them clung the cunning bonds of the wise Hephaestus’
The lexicalized meaning of the noun in the latter two examples is clearly
underlined in (20) by the apposition χρύσεον [chrýseon] ‘golden’ specify-
ing its concrete, physical sense, and in (21) by its being plural. Therefore
in (20) and (21) the noun has a merely referential value, i.e. its semantic-
pragmatic property is reference (Croft ([1990] 2003: 184ff.).
The relationship between the compositional meaning and the lexicalized
concrete one is intuitively clear: it is a metaphorical relationship, because
the bond can be interpreted as a metaphor of the imprisonment by a shift
from the abstract17 domain to the concrete domain. But it can also be
interpreted as a metonymic relationship. Within the domain imprison-
ment, the bond represents a contiguous concept to the process of imprison-
ing since it is an imprisoning instrument. This case shows that sometimes
it is not easy to identify whether we are dealing with a metaphor or with
metonymy, but arguably, making this distinction clear is not central to our
discussion. What is more important in the example above is that the shift
from one sense to the other is very easy and motivated. More generally, we
can also describe such a shift, as well as the other semantic shifts we will
analyse, by using the idea of a lexical cycle, well described by Simone (2000).
Namely, as far as δεσμός [desmós] is concerned, the lexical cycle is ‘process
> object’. The notion of a cycle evokes the possibility for the process to go
back to the original function (for further details see Simone 2000).
Furthermore in most of the occurrences of the noun δεσμός [desmós]
the sense of the lexeme is the second (concrete) one. This means that the
concrete sense of the word, which is not compositional, is lexicalized as a
new unit at least partially independent from the abstract one. Hence the
same word has two different meanings which are related to each other, i.e. it
is a polysemous word. This new lexeme is nearer to prototypical nouns, i.e.
nouns with a purely referential function (I call them fully referential nouns).18
The degree of lexicalization is higher when the existing occurrences

never show the compositional meaning, as with (22).
(22) χύ- σις

chý- sis
pour proc.f:nom.sg.
‘pile’
In fact, the meaning expected on the basis of the core rule ‘V+ -sis = process
of V’, i.e. ‘process of pouring’, does not exist in the Homeric corpus (nor
in later stages of the language). The meaning ‘pile’ can be interpreted as
coming from a metonymic shift from the ‘process itself of pouring’ to ‘what
is poured’ (or, in other words, the result of the process itself). But this is
not the last stage of the lexicalization scale, because χύσις [chýsis] often
has an argument structure which, in Homer, is mostly represented by the
genitive plural φύλλων [phýllōn] ‘of leaves’ as in (23).
(23) φύλλων γὰρ ἔην χύσις ἤλιθα

phýllōn gàr éēn chýsis ḗlitha
leaf:gen.pl. indeed be.impf:3sg. pile:nom.sg. very much
πολλή (Od. 5.483)
pollḗ
much
‘for fallen leaves (litt. ‘pile of leaves’) were there in plenty’
(24) χύσιν δ’ ἐπεχεύατο φύλλων (Od. 5.487)

chýsin d’ epecheýato phýllōn
pile:acc.sg. ptcl. heap up.impf:3sg.med. leaf:gen.pl.
‘(he) heaped over him the fallen leaves’
According to Benveniste ([1948] 1975), in the Homeric examples above

φύλλων [phýllōn] is an objective genitive, but Bruno (2000) argues that
this interpretation is impossible since the notion of object is not compat-
ible with the process type which is described by the DN. On the contrary,
such a genitive should have a value comparable with the subject of a
medial predicate, with which it shares some features, e.g. lack of control
of the event, affectedness and inagentivity. Actually, if the word χύσις
[chýsis], on which the genitive φύλλων [phýllōn] syntactically depends,
is lexicalized (hence nearer to fully referential nouns), speaking about real
arguments in reference to φύλλων [phýllōn] makes no sense. Prototypical
nouns, in fact, having no event structure, cannot take arguments because
arguments are participants in the event described by predicative elements
(verbs, event nominals, etc. . . .). So without an event, arguments stricto

sensu cannot exist. However, χύσις [chýsis] is not a pure, non-constructed
noun (a mot non construit in Corbin’s 1987 terms). It derives from a syn-
chronically attested verbal base and probably preserves at least part of its
predicative force. So interpreting the genitives in (23) and (24) as purely
relational is plausible, but we cannot exclude a real argument reading (be it
objective or subjective-medial) inherited from the verbal base with which
a relationship (though not completely clear from a semantic point of view)
still exists. So this kind of noun is less verbal than (17), but more verbal
than (25), which – as we are going to see – is nearer to a pure noun.
(25) θυ- μός

thy- mós
toss def.proc.m:nom.sg.
‘soul, heart’
In (25), in fact, the relationship with the verbal stem θυ-[thu-] ‘toss, be
restless’ is even more opaque, since θυμός [thumós] denotes what meta-
phorically tosses within the heart, so something that gives the human being
their energy (there are many co-occurrences with the word ψῡχή [psychḗ]
‘life, breath’), the soul, the spirit as the principle of life, feeling and thought,
the mind, the temper, courage, sometimes anger, but also – in a concrete
sense – the heart (note the occurrences with κραδίη [kradíē] ‘heart – in a
physical sense’). If we move from the compositional (not attested) sense of
‘(definite) process of tossing’ (see section 1), i.e. something like ‘perturbed
movement’, the abstract senses ‘soul, life, spirit, etc.’ come up through
a sequence of metaphoric shifts from one conceptual domain to another.
Firstly, breath can be interpreted as a sort of internal agitation implying a
physical movement (from the domain natural phenomena to interior).
Secondly, it is well-known that the ancient Greek culture identifies life
with breath, that is breath tosses within human beings giving them life.
Thus we explain the chain ‘perturbed movement > breath > principle of
life’. Moreover, life is a complex concept which comprises not only mere
existence but also interior life including feeling and thoughts. Finally a
metonymic shift from the ‘activity’ to the ‘location of the activity’ creates
the concrete meaning of ‘heart’, which is where the activities just described
take place.
We can represent such relations by means of a structure where a
core meaning exists and a series of related meanings arise from it, as in
Figure 11.2. The star in front of ‘perturbed movement’ underlines that
this sense is not attested. The circle including ‘life’ and ‘soul, spirit, etc.’
within the scheme underlines the fact that the meaning ‘heart’ comes from
courage
soul, spirit
*perturbed breath life mind heart
movement temper
anger
Figure 11.2 Semantic chain of θυμός [thumós]
the whole of these meanings, the heart being the site of all those activities
according to ancient Greek culture.
As we can see, the relationship between the meaning ‘heart’ and the
original one is not immediate, but it exists. The lexeme is completely lexi-
calized. At the last stage of lexicalization there are also lexemes like (26).
(26) βω- μός

bō- mós
go def.proc.m:nom.sg.
‘altar’
Here the lexicalized meaning comes up through a metonymic shift from

the ‘(definite) process of going’ to a ‘place to which one goes’, an ‘altar’
(process > location). No trace of the compositional semantics exists at a
synchronic level.
Furthermore, whereas for every lexeme seen above it is clear from which
stem it derives, in some words even the formal relation is quite hard to
recognize both synchronically and diachronically. Let us have a look, for
example, at the words in (27) and (28).
(27) ποτα- μός

pota- mós
‘river’
(28) ὀφθαλ- μός

ophthal- mós
‘eye’
According to Chantraine ([1968–70] 1999), (27) is likely to be related

to the stem πετ-[pet-] ‘fall’ (at the o-grade), (28) to the form ὀφθῆναι
[ophthḗnai] (aorist passive infinitive from the stem ὁπ-[hop–] ‘see’).
However, in both words there are also further formal elements which
make the connection less transparent (-a-and -al-respectively). The link
between (27) and (28) and their respective verbal bases is not only formally
but also semantically unclear. If Chantraine’s etymology were right, we
could reconstruct it by considering ‘river’ as ‘something which flows’ or
‘place where something flows’ (by metonymic shift from the process of
flowing – akin to the concept of falling), and ‘eye’ as ‘something by which
one sees’ (by metonymic shift from the process of seeing to the sight instru-
ment). In these cases the likely more definite nature of the kind of process
encoded by the suffix -mós (section 1) may be the reason why these nouns
are more easily lexicalized, often referring to places and instruments, i.e.
inanimate participants in the event.
Finally, Homer gives us only two examples of lexicalization of lexemes
carrying the suffix -týs, i.e. (29) and (30).
(29) ἐδη- τύς

edē- týs
eat abstr.proc.f:nom.sg.
‘meat, food’
(30) κλει- τύς

klei- týs
incline abstr.proc.f:nom.sg.
‘slope’
Both cases can be explained as metonymic shifts from the process to the
object affected by the process taking place, which in the latter case is also
the result of the process itself.
In fact, the lexicalization status of the noun ἐδητύς [edētýs] is a little
more complex, since the process of lexicalization is not complete; the coor-
dination of this word with the noun πόσις [pósis] ‘drink, drinking’, which
also undergoes a process of lexicalization, in the formulaic verse (31) in
both poems may be taken as evidence for that.
(31) αὐτὰρ ἐπεὶ πόσιος καὶ ἐδητύος

autàr epeì pósios kaì edētýos
but when drink:gen.sg. and food:gen.sg.
ἐξ ἔρον ἕντο (Il. 1.469; Od. 17.99, etc.)
ex éron hénto
out (of) desire:acc.sg. sendaor.med.:3pl.
‘But when they had put from them the desire for food and drink’
Murray’s translation in this case opts for the lexicalized meaning of both
words, but apart from this translation it is still possible to maintain their
compositional meaning.19 On the other hand there are also occurrences in

which the lexicalized meaning ‘meat, food’ is clearer, as in (36).
(32) πλησάμενος δ’ ἄρα θυμὸν ἐδητύος

plēsámenos d’ ára thymòn edētýos
fillpart.aor.:nom.sg. ptcl. then heart:acc.sg. meat:gen.sg.
ἠδὲ ποτῆτος (Od. 17.603)
ēdè potḗtos.
and drink:gen.sg.
‘But when he had satisfied his heart with meat and drink’
This dual disposition of the lexeme also characterizes the later stages of
the language, although the lexicalized sense seems to be preferred. The
gloss for the lemma ἐδητύς [edētýs] by the ancient grammarian Aelius
Herodianus (Perì klíseōs onomátōn 3,2.762.35 σημαίνει δὲ τὴν βρῶσιν
[sēmaínei dè tḕn brṓsin] ‘means tḕn brṓsin’) is not decisive. Neither is the
gloss by the later grammarian Aesychius (Lexicon 465.1 τροφή, βρῶσις
[trofé, brṓsis]).20 The grammarians in fact explain the word by using a
synonym, βρῶσις [brṓsis], which is itself a semi-lexicalized word, and
τροφή [trofé], a word which means both ‘food’ and ‘nurture, educa-
tion’. This means perhaps that ancient grammarians did not perceive the
difference to be so considerable.
The position of (30), only once attested in Homer in (33), is clearer.
(33) ἐς κλειτὺν ἀναβὰς Od. 5.470

es kleitỳn anabàs
to slope:acc.sg. go uppart.aor:nom.sg.
‘(if I) climb up the slope’
Non-Homeric occurrences, found by querying the digital corpus TLG,

show that the lexicalized meaning gradually replaces the compositional
one. Furthermore, the frequent plural forms of the noun seem to be strong
evidence of this.21 Further evidence is provided again by Aelius Herodianus
(Perì klíseōs onomátōn 3,2.535.20) who writes that κλειτύς, σημαίνει δὲ τὰ
ἐξέχοντα μέρη τῶν ὀρῶν [kleitýs sēmaínei dè tà exéchonta mérē tōn orṓn]
‘kleitýs means the projecting parts of the mountains’. This case shows that
sometimes the perspective of the ancient grammarians may be a helpful
tool for understanding to which extent AG words were perceived by native
speakers. Nevertheless, the analysis of the phrasal context is an important
tool too, whose potential should not be underestimated. Although our
analysis cannot be verified on the basis of native speakers’ intuitions, the
phrasal context gives us a surprisingly rich source of information.
4 LEXICALIZATION AND FREQUENCY
It seems to me that there is a link between lexicalization phenomena and

frequency. The statistically more frequent deverbal Homeric nouns tend
to show a less clear semantic relation with their verbal base, more easily
losing their compositional semantics. Actually, the more common a lexeme
is the more it undergoes semantic shifts determined by its use in a larger
variety of contexts, so the high frequency of the lexeme is not just the result
of the process of conventionalization, but it is also ‘a primary contributor
to the process’ (Bybee 2003: 1; see also Schwarze 2001). I use the term
conventionalization in a sense which includes – together with lexicaliza-
tion – the concept of grammaticalization (see Himmelmann 2004: 38). In
agreement with Lehmann (2002, 2004) and Himmelmann (2004), in fact, I
do not consider grammaticalization and lexicalization as mirror processes
in opposition to each other (see, among others, Lehmann 1989; Heine et
al. 1991). On the contrary they have much more in common than meets
the eye.
The correlation between lexicalization and frequency is clear, for
instance, in the case of (25), which is the most frequent among the suffixed
DNs in Homer – being what moves the warrior to act and makes him think
and have emotions and feelings.
What is more, many studies on linguistic change confirm that irregular-
ity (hence loss of compositional meaning as well) often concerns the most
frequent items (see Bybee 1985, 2003). This behaviour is only one of the
features that lexicalization shares with grammaticalization, both of them
being processes of conventionalization (Himmelmann 2004). Thus both of
them are processes arranging certain values: in the case of lexicalization the
value is fixed in the lexicon, whereas in the case of grammaticalization the
value is in the grammar.
5 CONCLUSIONS AND FUTURE RESEARCH
In conclusion, we can represent the results set out in section 3 by means

of the continuum in Figure 11.3 whose poles are the forms that show a
amphibias
thymós
desmós
kleitýs
edētýs
chýsis
Transparent Lexicalized
forms forms
Figure 11.3 Continuum between transparent and lexicalized forms

transparent relation between verbal semantics and derivation rule and the
lexicalized forms for which this relation becomes opaque.
We represent this opposition by means of a continuum because, as
already seen, the various nouns may be more or less transparent and
more or less lexicalized. Thus it seems that the nouns which more easily
undergo lexicalization processes, among the nouns we analysed, are the
-mós nominals and perhaps the explanation for such a condition lies
precisely in the more definite processuality encoded by the suffix -mós
(section 1). In fact, it is intuitively clear that finite, telic processes are
more likely to be connected with the nouniness sphere. According to
Cognitive Grammar a verb is conceptualized as an abstract region having
a temporal profile, i.e. it consists of various stages in which a Trajector is
profiled with respect to a Landmark. In contrast, the corresponding nomi-
nalization makes the predicate bounded, which is holistically perceived
as a whole preserving within itself the inherited conceptual structure. In
such a way, nominalizations lose their internal sequential scanning, so
they are atemporal – the various facets of a situation being examined in a
cumulative fashion (Langacker 1987b; see also Gaeta 2002: 104). This also
corresponds to what Givón (1979, 2001) claims about nouns. According
to him the most prominent feature of prototypical nouns is the so-called
time stability:
The most time-stable percepts, the ones that change slowly over
time, the ones that are likely to be identical to themselves (in terms of
properties), are lexicalized as nouns. The least time-stable percepts,
events and actions, which involve rapid changes in the universe, are
lexicalized as verbs [. . .] (Givón 1979: 321–2)
As for the reasons why some words get lexicalized, the tendency towards
the lexicalization of DNs might be explained as a sort of attempt by
language users to normalize such marked nominal forms by gradually
removing the inherited prototypical verbal features and attributing
prototypical nominal features. Indeed, these forms, though derived from
verbal bases, in terms of syntactic categorization are first of all nouns.
However, this is just a tendency by which the syntactic and semantic
behaviour of those nouns is partially governed. Many other factors
(frequency, syntagmatic relations, paradigmatic relations with related
words, etc.) can stimulate or prevent lexicalization, making every case
history different.
Furthermore, to claim that lexicalized lexemes totally lose their compo-
sitional semantics on the basis of the Homeric corpus is an oversimplifica-
tion. To some extent we should consider the possibility that a word which
in a certain stage of the language is lexicalized as a whole unit can recover

its compositional semantics in later stages. The notion of lexical cycle
introduced in section 3.2 clearly implies such a possibility.
In order to analyse such a topic in more depth, it would be important
to enlarge the corpus with diachronically different data. Analysing the
development of lexicalization throughout the history of a noun could
provide us with a more exact view of the question. For each of the DNs
placed on the lexicalization continuum it would be interesting to see how
many occurrences are lexicalized and how many occurrences are compo-
sitional (if any) and above all to what extent such occurrences are spread
diachronically.
Even a rapid comparison with later AG data, analysed in Civilleri
(2012), clearly shows noteworthy changes in the semantic values of the suf-
fixes mentioned above and in their reciprocal relationships. For instance,
the process value of the -sis nominals is well fixed during the classic stage
of AG thanks to the use of this suffix in the philosophers’ prose in opposi-
tion to the suffix -ma (see Chantraine [1933] 1979: 284–5). On the other
hand the suffix -týs is recessive. The productivity degree of the suffix -mós
seems to be reduced as well. Therefore, as far as this topic is concerned,
a diachronic analysis seems to be the most compelling task for future
research.
ABBREVIATIONS
abstr.proc abstract process

acc accusative
aor aorist
dat dative
def.proc definite process
f feminine
gen genitive
impf imperfect
m masculine
nom nominative
part. pres present participle
pl plural
proc process
ptcl particle
sg singular
Vadj verbal adjective
3 third person
NOTES
1. Strictly speaking, the term verbal has to be read as predicative. While

acknowledging that the predicative function is not an exclusive prop-
erty of the verb (sometimes other word classes can be predicative as
well), in this chapter I will refer to predicative entities as verbal enti-
ties. (As far as the function of predication is concerned see Croft 1991
and Gross 1994.)
2. For further analysis of the specific semantic value of each of the suf-
fixes in Figure 11.1 see Civilleri (2012).
3. Bracketed graphemes represent inflectional (as opposed to deriva-
tional) morphemes, with a purely grammatical function. They are not
interesting for our purpose, so in the discussion of the data in section
3 I will use the derivational suffix including the inflectional suffix, i.e.
-sis, -mós, etc.
4. In Civilleri (2012) I show that whatever the verbal base to which the
suffix -sis is added (continuative, resultative, transformative and –
under certain conditions – stative as well), the resulting noun is nor-
mally a process noun, i.e. [+ durative], which does not encode telicity
(at least as far as the nouns preserving compositional semantics are
concerned).
5. This is a nominal infinitive, namely an infinitive nominalized by means
of the definite article il.
6. On the basis of a relatively small group of words – set up with the most
common and productive stems (see Civilleri 2009) – I hypothesize that
in DNs the normal/zero grades keep the semantic value of the stem
while the o-grade modifies the semantic value of the stem. At the same
time, the o-grade – representing perfective/resultative actionality –
may be a feature that points out proximity to the prototypical nouni-
ness. For example, two different nouns derive from the durative stem
λεγ-[leg-] ‘to say’: λέξις [léxis] (< leg-sis) ‘discourse’ and λόγος [lóg-
os] ‘word, discourse’. The former denotes the ‘discourse’ as a durative
process, whereas the latter encodes a more definite processuality. In
fact, λόγος [lógos] is the term used to refer to written discourses, for
example the famous dialogues by Plato.
7. Talking about derivational rules instead of simply suffixation allows us
to include a wider range of (morphological) phenomena, encompassing
both affixation and apophony.
8. See Civilleri (2010, 2012) for examples taken from an AG (post-
Homeric) corpus.
9. ‘Given an object of cognition of some complexity, the human mind has
two ways of accessing it. The analytic approach consists in consider-
ing each part of the object and the contribution that it makes to the
assemblage by its nature and function, and thus to arrive at a mental
representation of the whole by applying rules of composition to its
parts. The holistic approach is to directly grasp the whole without
consideration of the parts’ (Lehmann 2002: 2).
10. Before Lakoff and Johnson (1980) metaphor was considered a mere
communicative tool of languages. In contrast, according to Lakoff and
Johnson metaphor does not work only at a linguistic level but takes
place at the conceptual level because it is a way of representing the
world (see Evola 2008).
11. According to Grimshaw (1990) only nouns which have an event struc-
ture can also have argument structure.
12. All translations for the Iliad are from Murray (1924) and for the
Odyssey from Murray (1919).
13. Examples of such ambiguity may easily be found not only in AG (e.g.
Plato Symp. 218b2 τῆς φιλοσόφου μανίας [tēs filosófou manías] ‘phi-
losopher’s interior agitation’, Epictetus Ench. 13 τὴν προαίρεσιν τὴν
σεαυτοῦ [tēn proaíresin tēn seautoù] ‘your moral choice’) but also in
other languages which encode arguments of DNs by genitive(-like)
expressions, because it is the notion of subjective genitive itself that
has a more uncertain status (since it is harder to be distinguished from
a merely relational genitive than objective genitive).
14. In ὑλαγμός [hylagmós], formed on the basis of ὑλα- [hyla-], -g- is
a linking morpheme, like -s-in (15), (16) and (17) below. The verb
derived from the same base is indeed [hyláō] ὑλάω. Considering these
elements as linking morphemes allows us to keep the uniformity of the
suffix.
15. ‘Disposition and aptitude, practice of the notion as vocation and ability
of the person accomplishing it, in short the subjective destination and
in general the function in the proper sense – the practice of the notion
being considered as the function of the person’ (my translation, COG).
16. This corpus consists of three philosophical texts: the Symposium by
Plato, the De prisca medicina by Hippocrates and the Encheiridion by
Epictetus.
17. The conceptual domains will be emphasized by small capitals.
18. As far as the prototypical referential function of nouns (vs. predicative
function of verbs) is concerned, see among others Croft ([1990] 2003:
184ff.).
19. For example, the Italian translation by Privitera (1989) ‘quando poi
si furono tolta la voglia di bere e di mangiare’ (‘after they satisfied the
desire of drinking and eating’) chooses to underline the compositional
sense of the DNs.
20. Sometimes it is very useful to compare semantic descriptions of words

by ancient grammarians because they may provide helpful information
about how the AG speakers did perceive the meaning of words.
21. One of the features of the event nominals is that in general they are not
pluralizable (see Table 1 in Alexiadou and Grimshaw 2008: 3).
chapter 12
How many factors influence

the meaning of denominal and
deadjectival verbs? The case of
Modern Greek verbs in -(ι)άζω The meaning of denominal and deadjectival verbs
Angeliki Efthymiou
T he aim of this chapter is to examine the factors involved in Modern

Greek verb forming processes. My evidence comes from the Modern
Greek causative suffix -(ι)άζω [(i)ázo], which usually carries an evalua-
tive connotation. After a presentation of the suffix in section 1, section 2
discusses various issues concerning the allomorphic variation of the suffix.
Section 3 describes the principal meanings of -(ι)άζω verbs. Section 4 deals
with the role of the meaning of the base in the creation of the meaning.
Section 5 discusses the relation between the phonetic shape of the suffix
and its evaluative meaning. In section 6, I discuss how the role of the word
formation process and the rivalry with other suffixes influences the meaning
and the productivity of -(ι)άζω. The last section briefly summarizes the
main findings of the article.
1 INTRODUCTION
The suffix -(ι)άζω usually attaches to nominal and adjectival [−learned]

bases and derives [−learned] verbs which express a whole range of
concepts:1 causative/resultative ‘cause to become x’ (e.g. κομματιάζω
[komatjázo], ‘to break/tear into pieces’), ornative ‘provide with x’ (e.g.
ντροπιάζω [dropjázo], ‘to disgrace’), locative ‘put into x’ (e.g. τσουβαλιάζω
[tsuvaljázo], ‘to bundle into a sack’), instrumental ‘use x’ (e.g. νυχιάζω
[nixjázo], ‘to scratch with one’s nails’), inchoative ‘be provided with many
and usually unwanted x’ (e.g. ρυτιδιάζω [ritidjázo], ‘to wrinkle, become
wizened’) (see Efthymiou 2011a).
Given that the most robust semantic pattern of -(ι)άζω derivatives is the
226 angeliki efthymiou
inchoative meaning ‘be provided with many and usually unwanted endoge-
nous entities’ (see Efthymiou 2011a), I address the following four questions:
1. What is the role of the meaning of the base? Is the evaluative (or cumu-
lative) meaning assigned by the base of the derivative or by the suffix?
For example, in the case of ρυτιδιάζω [ritidjázo] (‘to wrinkle, become
wizened’) the negative meaning is already expressed in the base noun
ρυτίδα [ritída] (‘wrinkle’).
2. What is the role of the word formation process in which -(ι)άζω par-
ticipates in the creation of the meaning? How can we distinguish the
meaning of these verbs from the meaning of other Modern Greek
verb forming suffixes (see λασπώνω [laspóno], ‘to cover with mud’
vs. λασπιάζω [laspjázo], ‘to become mash’, both from λάσπη [láspi],
‘mud’)? How can we distinguish the evaluative connotation of -(ι)άζω
verbs from their underlying causative/resultative semantic structure?
3. Is the phonetic shape of the suffix related to its evaluative and cumula-
tive meaning? Is it a coincidence that the sequence [glide (j) +á] is found
also in other Modern Greek suffixes like -ιά [iá] and -ιάρης [iáris],
which form [−learned] derivatives that express pejorative or collective
meanings (e.g. κοκαλιάρης [kokaljáris], ‘skinny person’, ζητιανιά [zit-
janjá], ‘beggarhood, typical behaviour of a beggar’) (see Anastassiadis-
Symeonidis 1997; Efthymiou 1999)?
4. Does the evaluative/expressive meaning of the suffix and the [−learned]
register of its derivatives affect its productivity? Does the rivalry with
other suffixes influence the meaning or the productivity of -(ι)άζω (see,
for example, ρυτιδιάζω [ritidjázo], ‘to wrinkle’ (intransitive) [−learned]
vs. ρυτιδώνω [ritidóno] ‘to wrinkle’ (transitive and intransitive) [+/−
learned])? How does this correlate with the fact that -(ι)άζω seems to
be the prevailing default verb forming suffix in Modern Greek for the
interpretation ‘become provided with many unwanted x’ in Modern
Greek (see Efthymiou 2011a)?
Elaborating on these questions I show that the computation of the meaning

of a word formation process is rather complex, since it is influenced by
various factors.
In order to give answers to the questions raised above, I take as a
starting point the analysis of 313 verbs in Efthymiou (2011a). My data
were extracted from Anastassiadis-Symeonidis (2002) Reverse Dictionary
of Modern Greek (RDMG). From the resulting list of 2,260 verbs the
following forms were removed: (a) those that did not feature the suffix
-(ι)άζω and (b) those that were derived by prefixation, composition or
parasynthesis. For the analysis of the data, I used the theory of lexical con-
the meaning of denominal and deadjectival verbs 227
ceptual semantics developed by Jackendoff (1983, 1990) and the semantic

categories established by Plag (1999).
2 -( ι)άζω: ONE OR TWO SUFFIXES? ETYMOLOGY AND

ALLOMORPHY
Before analysing the semantic behaviour of the -(ι)άζω verbs, let me first
present the etymology and form of the suffix. In Modern Greek grammars
and dictionaries the relationship of -άζω to -ιάζω has traditionally been
regarded as unclear. In Triandafyllidis’s Grammar (1941) and Dictionary
(INS 1998), -ιάζω and -άζω appear in allomorphic variation. According to
INS (1998), the Modern Greek suffix -(ι)άζω developed from the Ancient
Greek suffixes -άζω/-ιάζω and -ιϖ. In most cases -ιάζω derived from
reanalysis, i.e. from the attachment of the suffix -άζω to stems ending in -ι.
Moreover, INS has two different homonymous lemmas, one for the form
[ázo/jázo], and another for the learned variant [ázo/iázo] (see also section
2.2 for a discussion of the term learned). In the first lemma, the suffix follows
the so-called glide formation (or synizesis) rule2 (i.e. [ia] is pronounced as
one syllable, and the consonant of the base or [i] is palatalized: e.g. ρυτιδιάζω
[ritidjázo], from ρυτίδα [ritída], τεμπελιάζω [tebeljázo] ‘to laze’, from
τεμπέλης [tebélis] ‘lazy’), whereas in the second lemma [ia] is pronounced
as two syllables (e.g. διπλασιάζω [diplasiázo], ‘to double’). Before I accept
any of these positions, I will examine some data about the phonological
properties of -άζω and -ιάζω verbs and their distribution in my corpus.
2.1 Phonological remarks on -άζω and -ιάζω forms

The form -άζω usually appears with certain inflection classes as illustrated
in (1).
γιορτάζω [jortázo]
(1) a. γιορτ(ή) [jortí]
‘to celebrate’ ‘celebration, saint’s day’
b. ονομάζω [onomázo] όνομ(α) [ónoma]
‘to denominate’ ‘name’
c. ωριμάζω [orimázo] ώριμ(ος) [órimos]
‘to mature’ ‘mature’
In (1) and further examples, nominal or adjectival inflection is presented in

parentheses. (1a) represents a feminine nominal base taking the i nflectional
suffix -ή, (1b) a nominal base with the inflectional suffix -α and (1c) an
adjectival base taking (in its masculine form) the inflectional suffix -ος.
Furthermore, -άζω attaches primarily to consonant-final bases but avoids

bases ending in consonant clusters that do not contain liquids (i.e. l or r),
as in (1a).
On the other hand, -ιάζω attaches primarily to consonant-final bases,
to nominal feminine bases taking the inflectional suffix -α, as in (2a), to
neutral bases in /á + consonant + i/, as in (2b), to feminine nouns in /iá/,
as in (2c), and to imparisyllabic masculine nouns, as in (2d).
ρυτιδιάζω [ritidjázo]
(2) a. ρυτίδ(α) [ritída]
‘to wrinkle’ ‘wrinkle’
b. κομματιάζω [komatjázo] κομμάτ(ι) [komati]
‘to brake/tear into pieces’ ‘piece’
c. αγκαλιάζω [angaljázo] αγκαλι(ά) [angaljá]
‘to embrace’ ‘arms’
d. νταλκαδιάζω [dalkadjázo] νταλκ(άς) [dalkás]
/νταλκάδεςNOM.PL
‘to desire, crave’ ‘desire, heartache’
When the base is imparisyllabic (and exhibits stem allomorphy), -ιάζω

always selects the extended (i.e. the plural) stem, as in (2d). Secondly,
when the vowel of the preceding syllable is /a/ then the suffix always has
the form -ιάζω. On the basis of these observations, one could propose that,
since -άζω seems to be prohibited when the vowel of the preceding syl-
lable is /a/, then -ιάζω and -άζω can be analysed as variants of the same
suffix. Before accepting this position, I will examine in the next section how
-(ι)άζω verbs are distributed in my corpus according to their register status.
2.2 The distribution of -άζω and -ιάζω in my corpus and in

Printed School Modern Greek
In this section I will investigate the distribution of -ιάζω and -άζω forms
in my corpus and in Printed School Modern Greek, i.e. a small corpus based
on Efthymiou et al. (2010). The material of Printed School Modern Greek
was collected from the corpus of 3rd grade primary school textbooks and
contains 7,773 tokens and 1,705 types of Modern Greek suffixed words3
(see also section 6.1. for more details).4
Before addressing the distribution of -(ι)άζω verbs, I will provide a
brief explanation of the terms (or features) [+learned], [+/–learned] and
[–learned], which will be used in this section. Following Anastassiadis-
Symeonidis and Fliatouras (2003), I use the feature [+learned] in order to
characterize words that (a) come from Ancient Greek, (b) constitute artificial
formations of ‘katharevousa’ (i.e. an artificial, ancient-looking form of Greek
42
271
Figure 12.1 The distribution of -άζω and -ιάζω verbs in RDMG
105
-learned
+/-learned
208
Figure 12.2 The distribution of [+/–learned] and [–learned] -(ι)άζω verbs in

RDMG
developed by a group of literary people in the nineteenth century) or (c) are

used only in refined or written speech. On the other hand, words character-
ized as [–learned] are words that either do not originate from Ancient Greek
or are used in informal or spoken (or colloquial or vulgar) speech. Finally,
the feature [+/–learned] serves to characterize all words that are unmarked
in use or origin (i.e. they are neither learned nor non-learned).
In Figure 12.1, -(ι)άζω verbs are presented according to their allomor-
phic variation (i.e. -άζω or -ιάζω), whereas in Figure 12.2 -(ι)άζω verbs are
presented according to their register variation.
As indicated in Figures 12.1 and 12.2, the proportion of -άζω forms is
very small and most -(ι)άζω forms are [–learned]. All -άζω forms can be
characterized as [+learned] or [+/−learned]. I also note that in sixty-three
of the [+/– learned] forms, [i] is actually part of the base, e.g. διπλασιάζω
[diplasiázo] ‘to double’ (διπλάσιος [diplásios] ‘double’). Most of these
verbs are derived from numerals. Finally, if we compare these findings to
those in Table 12.1, analysed in Efthymiou et al. (2010), we arrive at the
following remarks (see also section 6.2).
If we focus on what the figures in Table 12.1 can reveal about suffixes, it
can be suggested that the form -άζω is neither productive nor very frequent
Table 12.1 Verb-forming suffixes: token frequency in printed school Modern Greek
-ίζω -(ι)άζω -εύω -ώνω -άζω -ιάζω
9.5% 6.3% 5.6% 5.6% 3.9% 2.5%
in Modern Greek. On the other hand, Table 12.1 supports the hypothesis
as to the [–learned] character of -ιάζω forms: -ιάζω hardly appears in the
written register. This lack of preference is also supported by the findings
of linguistic experiments (see Rytting 2005), which suggest that Greek
speakers have an awareness of the connection between glide formation (i.e.
palatalization) and informality. Thus, based only on morphophonological
criteria, one could get the impression that -ιάζω and -άζω appear in (almost)
allomorphic variation, i.e. that they can be analysed as variants of the same
suffix. On the other hand, taking into account stylistic and pragmatic criteria
as well, one would opt for an analysis that views Modern Greek as having
two different homonymous suffixes, one available in informal speech, namely
the [−learned] form [jázo], the other frequent in written Greek, namely the
[+/−learned] form [ázo/iázo]. Therefore I accept the INS analysis, but I
also believe that the learned suffix should be lemmatized as -άζω, and that
further research is needed in order to explain the distribution of these forms.
In the rest of the chapter I will focus on the [−learned] forms.
3 PRINCIPAL MEANINGS OF THE – ιάζω VERBS
In this section, I will examine the meanings expressed by -ιάζω verbs.

As already noted in Efthymiou (2011a), these verbs show a wide range
of polysemy. Often, verbs formed with -ιάζω mean ‘cause to become x’
(causative/resultative). Such examples are given in (3).
κομματιάζω κομμάτι
(3) a.
komatjázo komáti
‘to break/tear into pieces’ ‘piece’
b. κουρελιάζω κουρέλι
kureljázo kuréli
‘to cut into shreds’ ‘rag’
Interestingly, the vast majority of the -ιάζω verbs refer to events of modi-
fication of the state of an entity, whereas the base noun identifies the final
state of the process which affects the entity projected to the direct object or
subject position (see also (4) and (6) below).
In most cases, -ιάζω verbs mean ‘be saturated by x/ be covered by many
unwanted x’ (inchoative-ornative).5 Some examples are given in (4).
σκουριάζω
(4) a. σκουριά
skurjázo skurjá
‘to rust’ ‘rust’
ρυτιδιάζω
b. ρυτίδα
ritiδjázo ritíδa
As the glosses of the examples in (4) indicate, verbs tend to denote internally
caused states, i.e. the cause of the change of state event is linked to properties
inherent to the argument undergoing change (for the meaning of this term,
see also Alexiadou et al. 2006). In all these cases, like in (3), the verbs appear
to denote the modification of the state of an entity. Moreover, the majority
of these verbs are intransitive, and the most representative meaning of -ιάζω
derivatives is ‘be provided with usually unwanted endogenous x’, i.e. they
also display negative/evaluative and cumulative meanings.
In other cases, -ιάζω verbs mean ‘provide with x’ (ornative). An example
of this meaning is given in (5).
(5) λεκιάζω λεκές

lecjázo lecés
‘to stain’ ‘stain’
Furthermore, -ιάζω derivatives can also express the meaning ‘put into x’
(locative). Some examples are given in (6).
(6) a. τσουβαλιάζω τσουβάλι

tsuvaljázo tsuváli
‘to bundle into a sack’ ‘sack’
b. μπουντρουμιάζω μπουντρούμι
budrumjázo budrúmi
‘to put into a dungeon’ ‘dungeon’
As illustrated in (3–6), the vast majority of the -ιάζω verbs refer to events of
modification of the state of an entity. The base noun identifies the final state
of the process which affects the entity projected to the direct object or subject
position. In most cases, -ιάζω verbs denote internally caused states and display
negative semantics or pejorative and cumulative meanings. Moreover, the
majority of these verbs are intransitive, and the most representative meaning
of -ιάζω derivatives is ‘be provided with usually unwanted endogenous x or
become x’. In these cases the base nouns denote the end states or the final
positions in the causative act6 (as regards the pejorative meaning of -ιάζω
derivatives see also Efthymiou 2011a and Charitonidis 2011).
It is worth pointing out, however, that in my data I found only a
small number of verbs expressing other meanings, such as instrumental,
performative and similative. Some examples are given in (7–9).
(7) νυχιάζω νύχι

nixjázo níxi
‘to scratch with one’s nails’ ‘nail’
(8) κουβεντιάζω κουβέντα

kuvendjázo kuvénda
‘to chat, discuss’ ‘chat’
(9) γεροντοκοριάζω γεροντοκόρη

jerontokorjázo jerontokóri
‘to do things in a way ‘spinster, old-maid’
that is typical of a spinster/become a spinster’
As most of the derivatives of the types illustrated in (7–9) also allow causa-
tive readings and convey a pejorative meaning, I would like to propose that
-ιάζω verbs expressing instrumental, performative or similative meanings
are marginal cases, and cannot be considered central for determining the
role of the suffix within the system.
Finally, based on all these findings and in line with Gottfurcht (2008), I
would like to suggest that -ιάζω has developed a semantic category proto-
type related to the frequency of the meanings expressed by the derivatives.
(See also Tribout 2010, who suggests that, for each morphological process,
some semantic types appear to be licensed, privileged or prohibited.)
Therefore, the basic meanings of -ιάζω verbs can be ranked as follows:
1. inchoative-ornative (= more than two-thirds in the total number of
types); 2. ornative or causative (= less than two-thirds in the total number
of types); 3. instrumental, locative, performative or similative (= less than
one-third in the total number of types). In addition, as mentioned above,
all these verbs display negative semantics or pejorative meanings. Thus
this ranking means that the semantics of -ιάζω is such that a typical rep-
resentative of -ιάζω verbs should express both inchoative and pejorative
meanings (‘inchoative-ornative’) (see (4)). Ornative and resultative mean-
ings are expressed by less prototypical verbs and, finally, instrumental,
locative performative and similative meanings, which are the least frequent
of the corpus, are conveyed by non-prototypical -ιάζω verbs.
4 THE MEANING OF THE BASE
At the beginning of this chapter, I mentioned that the suffix -ιάζω com-
bines with adjectival and nominal bases, but that the majority of the deriva-
tives are derived from nouns. Going through the list of bases, one gets the
impression that -ιάζω is quite selective about the semantic categories of its
base. It usually attaches to [−learned] bases denoting something negative
or unpleasant. For example, many bases refer to an illness (10) or external
imperfections on the body (11).
(10) ψωριάζω ψώρα

psorjázo psóra
‘to become infected ‘scabies’
with scabies’
(11) σπυριάζω σπυρί

spirjázo spirí
‘to be covered with pimples’ ‘spot, pimple’
Furthermore, many bases denote unpleasant or dangerous things or

substances, as in (12).
(12) a. σκουληκιάζω σκουλήκι

skulicázo skulíci
‘to be wormy/wormeaten’ ‘worm’
b. μουχλιάζω μούχλα
muxljázo múxla
‘to be tainted ‘mildew, mould’
or to taint with mildew’
As illustrated in (10–12), nominal bases tend to denote either substances or

small entities usually encountered in quantities and seen as an amorphous
and homogeneous mass.
In some cases, the bases denote negative qualities or states, as in (13).
(13) a. τεμπελιάζω τεμπέλης

tembeljázo tembélis
‘to laze’ ‘lazy’
b. καραφλιάζω καραφλός/καράφλα
karafljázo karaflós /karáfla
‘to become bald’ ‘bald’/ ‘baldness’
There are, however, some cases, in which the base does not express
anything negative. For example, in some cases the base denotes contain-
ers where things are tied together so that they can be carried or stored, as
in (14).
(14) a. κασελιάζω κασέλα

kaseljázo kaséla
‘to put into a trunk/chest’ ‘trunk, chest’
b. τσουβαλιάζω τσουβάλι
tsuvaljázo tsuváli
‘to bundle into a sack’ ‘sack’
It is worth noting, however, that in these cases, the entities stored are seen
as a mass, i.e. they become spatially limited ‘stuffs’ composed of particles
which are not seen as significant enough for anybody to want to count them
or to focus on them as individual entities.
It is also worth pointing out that many [−learned] or [+/−learned] bases
are of Turkish or of Italian and Venetian origin.7 I suggest that, although
not all native speakers have etymological knowledge, in some of these cases,
one could argue that the choice of these bases is not only influenced by reg-
ister factors but also by phonological properties related to the etymology of
the base. For example, native speakers can recognize that the phonological
properties of μπουντρούμι ‘dungeon’ in (6a), which is of Turkish origin,
are different from those of φυλακή [filací] ‘jail’, which originates from
Ancient Greek and derives the unmarked quasi-synonymous -ίζω verb
φυλακ-ίζ(ω) [filacízo] ‘to jail’.
Interestingly, a closer look at the bases shows that most of them are
related to something (entity, substance, state or behaviour) perceived by
the senses, i.e. [+concrete]. In line with Efthymiou (1999), I suggest that
this could be partially attributed to the [−learned] character of the suffix.
As suggested by Efthymiou (1999), the fact that some [−learned] suffixes
tend to prefer concrete meanings (and bases) can be easily explained if we
think that these suffixes reflect common people’s everyday use and that
most people focus on things that are easily perceived by human senses and
relevant to human interests (see also Wierzbicka 1985: 155).
As seen in (10–14), although bases with negative connotations seem to
be the default for -ιάζω verbs, there are also some cases in which the base
does not express anything negative or unpleasant, as in (15a, c). (15b) is an
example where the base is ambiguous between a neutral and a negatively
connotated meaning.
(15) a. κομματιάζω κομμάτι

komatjázo komáti
‘to shred, break to pieces’ ‘piece’
b. παραμυθιάζω παραμύθι
paramithjázo paramíthi
‘to tell fairytales, cheat’ ‘fairytale, lie’
θρονιάζω θρόνος
c.
thronjázo thrónos
‘to enthrone’ (ironically) ‘throne’
As the glosses in (15) indicate, the meanings of the base and the suffix
match. The suffix selects the meaning of the base that best matches the
meaning of the derivation, i.e. a negative side of the meaning of the base,
and the base is sensitive to the meaning of the suffix.8 Since almost all
[−learned] derivatives refer to something unpleasant, I propose that the
pejorative meaning of the -ιάζω verbs is both selected and assigned by the
suffix. The suffix seems to add a connotation to the meaning of the deriva-
tive that creates a negative or ironic effect. A typical example that shows
the ironic connotation of these derivatives is (15c), which is mostly used
in the passive form with the meaning ‘be enthroned, sit, stay longer than
expected or wanted’, as in (16). Note that the ironic meaning is absent from
the unmarked [learned] rival parasynthetic verb εν-θρον-ίζω [enthronízo]
‘enthrone’.
(16) Ο Γιάννης ήρθε για δυο μέρες στο σπίτι μας αλλά θρονιάστηκε και δε θέλει
να φύγει.
Giánnis írthe giá dío méres sto spíti mas allá throniástike ke de thélei na fígei
‘Giannis came to our house for two days, but he is_as_if_he_was_enthroned
(=stayed) and does nοt want to go.’
Finally, it is worth mentioning that some derivatives express both pejora-

tive and cumulative meanings (i.e. ‘become saturated by many unwanted
entities’), as in (11) and (12a). I assume that in these cases the cumulative
reading is motivated by pragmatic factors, but it also reveals the inter-
play between the meaning of the base, the suffix and the intention of the
speaker. When the nominal bases denote small unwanted entities which
naturally appear in homogeneous groups (i.e. they tend to co-occur, like for
example pimples on a face), the derivative always has the meaning ‘become
covered by (a large number of) x’. Moreover, as mentioned in Efthymiou
(1999) and Ricca (2005), since the notion of collectivity implies low identi-
fiability of the individual, collective and pejorative meanings cannot always
be treated as independent. Note, however, that the exact meaning assigned
to the derivative is also related to the actual intentions or the emotive atti-
tude of the individual speaker. For example, whereas individual entities
like wrinkles and pimples are in most cases too insignificant for anyone to
talk about if they do not appear in great quantities, I think that one could
still use σπυριάζω ‘be covered with pimples’ in order to refer to a situation
where someone has one or two pimples on their face. In this case, the suffix
seems to intensify pragmatic effects already expressed by the base or the

derivational process.
5 THE PHONETIC SHAPE OF THE SUFFIX
Interestingly, the [−learned] phonetic shape and the negative connotation

of the suffix -ιάζω is found also in other Modern Greek [−learned] suf-
fixes such as -ιά [iá] and -ιάρης [iáris] (see also Efthymiou forthcoming).
The negative/pejorative meanings of these suffixes, which prefer (also)
attaching to bases denoting something negative, unpleasant or undesirable,
are illustrated in (17) (for the suffix -ιάρης, see Anastassiadis-Symeonidis
1997; for the suffix -ιά, see Efthymiou 1999).
(17) a. κοκαλιάρης κόκαλο

kokaljáris kókalo
‘skinny person’ ‘bone’
b. ζητιανιά ζητιάνος
zitjanjá zitjános
‘beggarhood’ ‘beggar’
‘typical behaviour of a beggar’
c. γαϊδουριά γαϊδούρι
γajδurjá γajδúri
‘typical action of a donkey’ ‘donkey’
Given the [−learned] phonetic shape and the negative connotation of

-ιάζω, -ιά and -ιάρης, as well as the fact that the speakers are aware
of their special stylistic status (see also section 2.2), it can be sug-
gested that the distribution of the [j + a] sequence is not accidental,
and that the negative connotation of the suffixes is related to their
[−learned] phonetic shape (for a more detailed discussion, see Efthymiou
2013).
Moreover, it seems that the ‘sound iconicity’ of the -ιάζω verbs is part of
a rich paradigm which includes expressive suffixes of many languages and
involves palatalisation (see Dressler and Barbaresi 1994). Native speakers
are sensitive to the fact that the phonetic make-up of this suffix differs from
the phonetic shape of [+learned] or [+/−learned] suffixes and use it in
order to denote expressive or negative meanings.
It is also worth noticing that the [−learned] negative connotation of the
suffix not only appears in the verbs of my corpus but also in Modern Greek
parasynthetic verbs of removal (see Efthymiou 2001, 2002, 2011b),9 as in
(18).
(18) ξεδοντιάζω[−learned] [kseδondjázo] ‘to take one’s teeth out’10

(ξε-[kse] ‘privative prefix’ + δόντι [dóndi] ‘tooth’ + -ιάζω)11
As the example in (18) shows, Modern Greek parasynthetic verbs in -ιάζω

appear to provide additional evidence for the claim that native speakers are
sensitive to the [−learned] and pejoratively connoted phonetic make-up of
-ιάζω. Interestingly, in these verbs, the suffix -ιάζω combines only with
the [−learned] negative-privative prefix ξε-.
6 THE ROLE OF THE WORD FORMATION PROCESS
In this section, I will discuss the role of the word formation process in
which -ιάζω participates in the creation of the meaning. I will suggest that
although Modern Greek verb-forming suffixes seem to share the same
underlying conceptual structure, each suffix seems to develop its own
semantic category prototype. I will also show that the [−learned] feature of
the suffix -ιάζω affects its frequency and its productivity.
6.1 The meanings of -ίζω, -ώνω, -εύω, -αίνω, -άρω derivatives and

-ποιώ formations
Apart from the suffix -(ι)άζω, Modern Greek has six verb-forming suffixes
and one main semi-suffix, namely the element -ποιώ [pió]. These are listed
and illustrated in (19).
(19) a. -ίζω [ízo]: βουρτσίζω [vurtsízo] ‘to brush’

b. -ώνω [óno]: βουτυρώνω [vutiróno] ‘to butter’
c. -εύω [évo]: προεδρεύω [proedrévo] ‘to chair, preside’
d. -αίνω [éno]: χοντραίνω [xondréno] ‘to get/grow fat, thicken’
e. -άρω [áro]: στρεσάρω [stresáro] ‘to stress’
f. -ποιώ [pió]: γραμματικοποιώ [gramatikopió] ‘to grammaticalize’.
As shown by Efthymiou (2011a), these derivatives show a wide variety

of meanings, such as causative, resultative, inchoative, ornative, locative,
instrumental, performative, similative, etc. In Efthymiou (2011a), follow-
ing Plag (1999), Lieber (2004) and Gottfurcht (2008), I use the theory of
lexical conceptual semantics developed by Jackendoff (1983, 1990) in order
to suggest that all Modern Greek verb forming processes share the same
underlying semantic structure, given in (20).
(20) CAUSE [x BE y LOC z]12

I also suggest (see Efthymiou 2011a) that the semantic interpretation of a

given verb depends upon two factors, first the extent to which the Lexical
Conceptual Structure (LCS) is fully expressed and second which argument
is filled by the noun base. Thus, for a resultative interpretation, the y argu-
ment in (20) is filled by the noun base. An example is given in (21).
(21) απλοποιώ [aplopió] ‘to simplify’ (cause to become απλός [aplós] ‘simple’)
Inchoative and similative/stative-essive interpretations are achieved when

the noun base is the y argument in (20) and the CAUSE x portion is not
realized. Examples for inchoative and similative-stative interpretations are
given in (22).
(22) a. χοντραίνω [xondréno] ‘to get/grow fat, thicken’

(become χοντρός [xondrós] ‘fat’)
b. προεδρεύω [proedrévo] ‘to preside’
(be/behave like πρόεδρος [próedros] ‘president’ for a certain period)
c. αριστοτελίζω [aristotelízo] ‘to imitate Aristotle’
(be/behave like Aristotle)
For a performative interpretation, the noun base is the only internal argu-
ment and the BE portion in (20) is not realized. Note that for the formali-
zation of this semantic interpretation, I have followed Gottfurcht’s (2008)
proposal. For Gottfurcht, performative is the mirror image of similative.
This is illustrated in (23).
(23) ταξιδεύω [taksidévo] ‘to travel’ (make a ταξίδι [taksídi] ‘trip’)
Ornative interpretations result from the full expression of the structure. In

this case, the base noun is the x argument co-indexed with the y argument
in (20). An example is given in (24).
(24) λαδώνω [ladóno] ‘to oil, bribe’ (provide with λάδι [ládi] ‘oil’)
For a locative interpretation, the base noun is the z argument in (20). An

example is given in (25).
(25) φυλακίζω [filacízo] ‘to jail’ (to put in φυλακή [filací] ‘jail’)
Finally, for the instrumental interpretation I follow Gottfurcht’s (2008)

proposal of an additional WITH predicate. This predicate follows [LOC z]
in (20) and has the noun base as its argument. An example is given in (26).
(26) καρφώνω [karfóno] ‘to nail’ (use καρφί [karfí] nail)
It is worth mentioning, however, that the hypothesis that all denomi-

nal verb formation processes share the same underlying structure is
not accepted by all morphologists (see, for example, Tribout 2010).
Furthermore, I suggest that the LCS proposed in (20) needs to be further
improved in order to account more satisfactorily for some problematic
issues, like for example the difference between similative, essive or
stative meanings. In what follows, I will present the principal meanings
related to each Modern Greek word formation process. It will be shown
that although these derivatives show a wide variety of meanings, they all
express at least a causative meaning. It will also become clear that for each
morphological process, some semantic types appear to be privileged or
prohibited.
In particular, the principal meanings of -ίζω derivatives can be described
as ‘(cause) to become x’, ‘imitate x’, ‘put in (to) x’, ‘perform/do/make x’,
‘provide with x’ and ‘use x’. Note, however, that the most frequent mean-
ings in these derivatives seem to be the similative, instrumental, performa-
tive and resultative meanings (see Efthymiou 2011a). Examples for -ίζω
derivatives are given in (27).13
(27) a. μαυρίζω [mavrízo] ‘to blacken’ μαύρος [mávros] ‘black’

b. πιθηκίζω [pithicízo] ‘to imitate πίθηκος [píthikos] ‘ape’
ape’s behaviour’
c. φυλακίζω [filacízo] ‘to jail’ φυλακή [filací] ‘jail’
d. σφουγγαρίζω [sfugarízo] ‘to sponge/ σφουγγάρι [sfugári] ‘sponge’
to mop’
The meanings of -ώνω derivatives can be described as ‘provide with x’,

‘(cause to) become x’, ‘use x’ and ‘put into x’. As shown by Efthymiou
(2011a), the ornative meaning seems to be one of the most frequent mean-
ings for these derivatives. Note also that no similative or performative
meanings are attested for these verbs. Examples of -ώνω derivatives are
given in (28).
(28) a. λαδώνω [ladóno] ‘to oil, bribe’ λάδι [ládi] ‘oil’

b. μαλακώνω [malakóno] ‘to soften’ μαλακός [malakós] ‘soft’
c. καρφώνω [karfóno] ‘to nail’ καρφί [karfí] ‘nail’
Derivatives in -εύω mean ‘carry out the official activities of x’, ‘become
x’, ‘do x’, provide with x’, ‘put into x’ and ‘use x’. However, it is worth
pointing out that the stative-essive ‘carry out the official activities of x’ and
the inchoative meanings seem to be the most frequent meanings for these
derivatives. Examples of these meanings are given in (29).
(29) a. προεδρεύω [proedrévo] ‘to chair, preside’ πρόεδρος [próedros] ‘president’

b. αγριεύω [agriévo] ‘to make/become fierce/ άγριος [ágrios] ‘fierce, wild’
roughen’
c. ταξιδεύω [taksidévo] ‘to travel’ ταξίδι [taksídi] ‘trip, journey’
d. παγιδεύω [pajidévo] ‘to trap’ παγίδα [pajída] ‘trap’
The meaning of the suffix -αίνω is quite restricted. Derivatives in -αίνω

mean ‘cause to become x’ and ‘provide with x’, as illustrated in (30).
(30) a. χοντραίνω [xondréno] ‘to get/grow fat χοντρός [xondrós] ‘fat, thick’
thicken’
b. λιπαίνω [lipéno] ‘to lubricate, fertilize’ λίπος [lípos] ‘fat, oil’
On the other hand, the meanings of -άρω derivatives can be described as

‘provide with x’, ‘do x’, ‘put into x’, ‘use x’ and ‘act as/be x’. Such exam-
ples are given in (31).
(31) a. πουδράρω [pudráro] ‘to powder’ πούδρα [púdra] ‘powder’

b. ζουμάρω [zumáro] ‘to zoom’ ζουμ [zum] ‘zoom, a zooming
camera shot’
c. πακετάρω [pacetáro] ‘to put into πακέτο [pacéto] ‘packet’
packet, pack’
d. φρενάρω [frenáro] ‘to brake’ φρένο [fréno] ‘brake’
Finally, -ποιώ verbs mean ‘cause to become x’, ‘put into x’ and ‘provide
with x’. Examples of these meanings are given in (32). As observed by
Efthymiou (2011a), no similative or performative meanings are attested for
-ποιώ formations. Furthermore, Mela-Athanasopoulou (2007) observes
that inchoative meanings are only possible with the passive voice of –ποιώ
verbs.
(32) a. απλοποιώ [aplopió] ‘to simplify’ απλός [aplós] ‘simple’

b. περιθωριοποιώ [perithoriopió] ‘to marginalize’
περιθώριο [perithório] ‘margin’
c. μορφοποιώ [morfopió] ‘to form’ μορφή [morfí] ‘form’
Elaborating on Efthymiou (2011a), I propose that the meanings of these

Modern Greek verb-forming processes can be summarized in Table 12.2.
In Table 12.2, ‘passive’ means that this meaning appears only in the
Table 12.2 The meanings of -ίζω, -(ι)άζω, -ώνω, -εύω, -αίνω, -άρω derivatives and
-ποιώ formations
-ίζω -(ι)άζω -ώνω -εύω -αίνω -άρω -ποιώ
cause to become x ¸ ¸ ¸ ¸ ¸ ¸ ¸
become x/be provided with x ¸ ¸ ¸ ¸ ¸ ¸ ¸passive
be provided with many unwanted x ¸
make x go to/in/on something ¸ ¸ ¸ ¸ ¸ ¸ ¸
make something go to/in/on x ¸ ¸ ¸ ¸ ¸ ¸
do x ¸ ¸ ¸ ¸
do /act like x ¸ ¸
use x ¸ ¸ ¸ ¸ ¸
carry out the official activities of x ¸
passive voice. As far as the types of base selected by these verbs are con-
cerned, it seems that Modern Greek suffixes do not behave in the same
way. For example, -ίζω is the only suffix among the suffixes of my corpus
that attaches to onomatopoetic words. Moreover, -εύω is the only suffix
that attaches to stage-level nouns denoting offices of persons,14 that is
nouns that denote temporary characteristics of their referents, in order
to derive verbs with the meaning ‘carry out the official activities of x for a
certain period’ (see also Efthymiou 2011a).
To sum up, based on all these findings, I suggest in line with Gottfurcht
(2008) that, although Modern Greek verb-forming suffixes seem to share
the same underlying structure (20), each suffix seems to develop a semantic
category prototype related to the frequency of the meanings expressed
by the derivatives. Therefore the realization of the underlying structure
depends on the preferences, the restrictions and the diachrony of each
suffix.
6.2 Frequency, productivity and suffixal rivalry

In spite of the relevance of frequency and productivity to assess the status
of word formation patterns (see Baayen 2008; Bauer 2001; Plag 1999), there
are no systematic investigations into the frequency and productivity of
Modern Greek suffixes. Because of the absence of reliable data for Modern
Greek, two kinds of empirical data have been investigated for this study,
namely on the one hand the existing -ίζω, -(ι)άζω, -ώνω, -εύω, -αίνω,
-άρω and -ποιώ verbs listed in RDMG (see Efthymiou 2011a), and on the
other -ίζω, -(ι)άζω, -ώνω, -εύω, -αίνω and -άρω verbs which are present
in Printed School Modern Greek (as investigated in Efthymiou et al. 2010).
It is not hard to think of reasons why the choice of the Reverse Dictionary
of Modern Greek and Printed School Modern Greek as text sources is not fully
Table 12.3 Verb forming processes

Verbs in Raw data Scrutinized data
-ίζω 3,507 650
-(ι)άζω 2,260 313
-ώνω 2,106 508
-εύω 1,207 325
-άρω 547 150
-ποιώ 252 200
-αίνω 687 113
Source: Data extracted from RDMG.
justified from a methodological point of view. First, dictionaries are not

accurate ways of estimating productivity. Second, Printed School Modern
Greek contains material from textbooks and is therefore not balanced for
text types or speech registers (see Plag 1999; Gaeta and Ricca 2003; Lieber
2010). However, although my data cannot give a comprehensive picture
of Modern Greek verb derivation from a quantitative point of view, they
seem to yield some interesting preliminary results. We can see the number
of -ίζω, -(ι)άζω, -ώνω, -εύω, -αίνω, -άρω and -ποιώ verbs as attested in the
RDMG in Table 12.3.
In the scrutinized data as counted in Table 12.3, I removed the follow-
ing forms: (a) those that did not feature the suffix -ίζω, -ώνω, etc. (such
as borrowings); (b) those that were derived by prefixation, composition or
parasynthesis; (c) all deponent verbs; (d) those that are -αρίζω formations
via the aorist of verbs in -άρω (see κονσερβάρω /κονσερβαρίζω [kon-
serváro/ konservarízo] ‘to can, tin’). This explains why the proportion of
scrutinized data as compared to the raw data is particularly low for -(ι)άζω.
In addition, Table 12.3 shows that -ίζω is more productive (or frequent)
than -ώνω, -εύω and -(ι)άζω.
The second corpus is based on Efthymiou et al. (2010). In this study
fifty-four Modern Greek suffixes were investigated. As mentioned in
section 3, the material collected from the corpus of 3rd grade primary
school textbooks contains 7,773 tokens and 1,705 types of Modern Greek
suffixed words. Note, however, that the semi-affix -ποιώ is not included
in this study. Moreover, -άζω and -ιάζω were analysed as variants of the
same suffix. Consequently, the -ιάζω category also includes [+/−learned]
forms, where /i/ is actually part of the base.15 My data from Printed School
Modern Greek are presented in Tables 12.4 and 12.5.
These tables present the ten most frequent suffixes in Printed School
Modern Greek. As already mentioned above, I do not claim that my data
provides a faithful picture of the ideal competence of a Modern Greek
educated speaker.16 Nevertheless, although the results in Tables 12.4 and
Table 12.4 Token frequency in printed school MG

-ικός -ση -ία -ίζω -α -ώνω -εύω -μα -άζω -ιάζω
(adj.) (n.) (n.) (v.) (adv.) (v.) (v.) (n.) (v.) (v.)
12% 11.1% 9.6% 9.5% 8% 5.6% 5.6% 4.9% 3.9% 2.5%
Table 12.5 Type frequency in printed school MG

-ικός -α -ση -ία -μα -ίζω -ώνω -εύω -άζω -ιάζω
(adj.) (adv.) (n.) (n.) (n.) (v.) (v.) (v.) (v.) (v.)
13.1% 11.3% 10.6% 8.4% 6.6% 5.9% 5% 3.8% 1.3% 1.3%
12.5 do not assure us that the frequency data obtained could be generalized
to any kind of textual typology, a number of observations can be made here.
Firstly, as expected according to the literature on productivity, Modern
Greek suffixes seem to differ considerably in their type and token frequency.
Secondly, as discussed by Efthymiou et al. (2010), it seems that a handful
of derivatives covers a large percentage of the overall token frequency of a
given suffix. Thirdly, the differences in token and type frequency confirm
the assumption about the [−learned] character of -ιάζω forms: -ιάζω hardly
appears in the written register. As expected (see, for example, Lieber
2010), the [−learned] (non-cultivated) pragmatic effect of the -ιάζω verbs
has obvious consequences on their frequency and productivity.17 It seems
that Greek native speakers associate the meaning and the form of the suffix
-ιάζω with something negative or [pejorative], and thus they choose -ιάζω
verbs in order to express something unpleasant. Therefore, [−learned]
-ιάζω verbs are absent from, for instance, scientific terminology or highly
refined usage of language (see Efthymiou et al. 2012a for similar remarks).
6.3 Doublets
I will finally turn to some doublets, which reveal that some verb-forming suf-
fixes are in competition in some semantic domains. As mentioned in the lit-
erature (Plag 1999; Gottfurcht 2008), a doublet occurs when two rival suffixes
are semantically and phonologically licensed. This is illustrated in (33–34).
(33) a. λασπώνω λάσπη

laspóno láspi
‘to cover with mud/ become mash’ ‘mud’
b. λασπιάζω λάσπη
laspjázo láspi
‘to become mash’ ‘mud’
(34) a. ρυτιδιάζω ρυτίδα

ritidjázo ritída
b. ρυτιδώνω ρυτίδα
ritidóno ritída
As the glosses indicate, there is some competition between -ιάζω and

-ώvω, mostly in the ornative and inchoative domains, but -ιάζω verbs
always select the [−learned], derogatory and intransitive reading. While
λασπώνω in (33a) has both causative and ornative meanings, λασπιάζω in
(33b) appears to accept only an inchoative reading. Moreover, ρυτιδιάζω in
(34a) appears to be [−learned] and intransitive, whereas ρυτιδώνω in (34b)
is [+/−learned] and can have both transitive and intransitive readings. The
glosses of the doublets reveal the prototypical and most frequent meaning
for each suffix, and thus, the suffixes are not similar enough to exhibit true
rivalry.
7 CONCLUSION
To sum up, I have shown that the computation of the meaning of -ιάζω
verbs is influenced by various factors, such as the semantic and struc-
tural properties of the base, the evaluative connotation of the suffix and
its derivatives and the productivity of the word formation process. The
results of my study also reveal the major role of pragmatic factors in
word formation. In particular, I suggested that Greek native speakers are
sensitive to the [−learned] phonological make-up of the suffix -ιάζω and
associate its meaning and form with something negative or [pejorative].
I also proposed that -ιάζω has developed a semantic category prototype
related to the frequency of the meanings expressed by the derivatives and
that a typical representative of -ιάζω verbs should express both inchoative
and pejorative meanings. It was also shown that the meanings of the base
and the suffix match and that the suffix seems to intensify pragmatic
effects already expressed by the base or the derivational process. The
suffix selects the meaning of the base that best matches the meaning of
the derivation, i.e. a negative side of the meaning of the base, and the
base is sensitive to the meaning of the suffix. Moreover, it was shown that
-ιάζω hardly appears in the written register and that the [−learned] (non-
cultivated) pragmatic effect of the -ιάζω verbs has obvious consequences
for its frequency and its productivity. Finally, I suggested that, although
there is some competition between -ιάζω and its rival suffixes in some
semantic domains, -ιάζω verbs always select the [–learned], derogatory

and intransitive reading. Furthermore, it was shown that the glosses of
the existing doublets reveal that the rival suffixes are not similar enough to
exhibit true rivalry.
NOTES
1. For the semantic description of the derived verbs I will use the labels
and glosses found in Plag (1999), namely causative/resultative ‘cause
to become x/turn into x’, ornative ‘make x go to/in/on something’,
locative ‘make something go to/in/on x’, inchoative ‘become x’, per-
formative ‘do x/perform x’, similative ‘act or be like x’, instrumental
‘use x’ and stative ‘be x’ (see also Lieber 2004). In order to account
for all Greek denominal verbs Ι need to add the following labels
and glosses: stative-essive ‘carry out the official activities of x’ and
inchoative-ornative ‘be saturated/covered by many unwanted x’.
2. For the glide formation rule see, among others, Kazazis (1968), Setatos
(1974), Warburton (1976), Nyman (1981), Rytting (2005).
3. Note that the affixoid -ποιώ [pió], which appears to be a moderately
productive verb-forming element in Modern Greek (see Efthymiou et
al. 2012b), was not included in this study.
4. The subjects of the textbooks included in the study were: language
and literature, history, mathematics, religion and environmental
education.
5. In this chapter I will not discuss the alternations in which the derived
verbs participate. For discussion of Modern Greek data, see Alexiadou
and Anagnostopoulou (2004) and Charitonidis (2005).
6. Sometimes, these verbs accept more than one interpretation (e.g.
μουχλιάζω [muxljázo], 1. inchoative ‘be tainted with mildew’, ‘mildew,
mould’, 2. ornative ‘taint with mildew’ (μούχλα [múxla] ‘mildew)).
In such cases the base noun is also interpreted as something trans-
ferred by the action (i.e. a theme). Furthermore, in certain cases (e.g.
μελανιάζω [melajázo] ‘bruise, become bruised’), it would be arbitrary
to decide whether the -ιάζω verb is derived from the nominal (i.e.
μελανιά [melaɲá] ‘bruise, bruising’) or from the adjectival base (i.e.
μελανός [melanós] ‘inky’).
7. The bases νταλκάς ‘desire’, λεκές ‘stain’, τσουβάλι ‘sack’, μπουντρούμι
‘dungeon’ in (2–6) have a Turkish origin. The base κασέλα ‘trunk,
chest’ in (6) is of Italian origin.
8. Nevertheless, this does not mean that the bases are commonly used
with negative connotation.
9. Following Corbin’s (1987) model, in Efthymiou (2001, 2002) I charac-

terize the segment -ιάζω in such cases as a categorical marker serving
to indicate verbal category.
10. The verb is not used in formal speech (or in cases where the teeth are
decayed and painful).
11. Neither *δοντιάζω nor *ξεδόντι is available as a base word for the
prefixation of ξε-or the suffixation with -ιάζω.
12. Following Gottfurcht (2008), I assume that in this structure the verb
has three arguments (x, y, z) and makes use of the semantic primitives
CAUSE, BE, LOC. LOC indicates an underspecified location between
two arguments. Note that this formalism differs from the one used by
Jackendoff (1983, 1990) although it uses some of the same labels.
13. For the semantics of -ίζω derivatives see also Charitonidis (2005).
14. For the distinction between stage-level and individual-level predicates
see Carlson (1977), Aronoff and Cho (2001) and Trips (2009).
15. This implies that on closer inspection, the actual number of [−learned]
-ιάζω verbs in this corpus would be even smaller.
16. Interestingly, the suffix -ιάζω has proven to be quite unproductive in
Efthymiou et al. (2012a), i.e. a corpus study of 4,143,583 words.
17. The influence of register on productivity has been repeatedly men-
tioned in the literature. See, for example, Plag et al. (1999).
chapter 13
Analysing en- and its Romance

equivalents in Jackendoff’s
Conceptual Structure
Jessica Forse
T his chapter examines one of the many interesting aspects of the often
overlooked yet highly significant relationship between morphology
and semantics. This critical oversight, which has led to a focus on the form
of words at the expense of their meaning, is all the more surprising given
that, as Levin and Rappaport Hovav (1998) point out, a morpheme is often
considered to be a minimal Saussurean sign relating form and meaning:
it is an arbitrary phonological form which represents a certain concept.
These authors suggest that the lack of research into the relation between
lexical semantics and morphology stemmed in part from the absence of
a comprehensive theory of lexical semantic representation that could
provide a framework within which to study such a relation. Over the last
four decades, Ray Jackendoff has devised and refined such a theory, which
he has called Conceptual Structure (CS). Jackendoff (2009) discusses the
treatment of compounds in the formalism, and in doing so, he offers some
suggestions as to how to encode derivational word formation processes,
such as the formation of actor nouns from verbs.
Here, this line of investigation is pursued further by using CS to analyse
the semantic changes brought about by one word formation process in par-
ticular: the formation of verbs using the prefix en-(and its allomorph em-) in
English, and its counterparts in some Romance languages, namely French,
Spanish and Portuguese. As a background, section 1 introduces Jackendoff’s
theory, highlighting the fact that he uses the term CS to refer to two separate
but related concepts. I make the distinction between the mental representa-
tion of CS and the formalism of CS, and then focus on the latter, outlining
the elements of the formalism that are used to encode the semantic changes
brought about by the word formation process of prefixation with en-.
In section 2, I present my analysis of the prefix en-, classifying the mean-
ings of en-verbs into semantic types, and then analysing these types within
248 jessica forse
the formalism of CS. In doing so, I demonstrate that en-has a core meaning
which can be represented by a single basic CS. This is a stable semantic
representation that can vary in several well circumscribed ways to give rise
to the specific variations in meaning in the vast repertoire of en- verbs.
In section 3, we turn to the analysis of two prefixes in French, Spanish
and Portuguese which bring about similar changes in meaning to English
en-, namely Romance en- and a-. I investigate whether they follow the same
semantic typology as the English prefix.
Finally, in section 4, I summarise my findings which demonstrate that
the six semantic types in my classification of English en-verbs all conform
to one basic CS, and that the different types can be grouped into three
clusters, demonstrating the marginal variation between them. As will be
seen, Romance en-verbs also have correlates for these types. There is some
semantic overlap between the Romance prefixes en- and a-, and, as such,
a-verbs correlate with some of the same semantic types as en-verbs. A sys-
tematic semantic difference between some en- and a-verbs is characterised
by the distinction between the INCH and GO functions in Jackendoff’s
formalism.
1 JACKENDOFF’S THEORY OF CONCEPTUAL STRUCTURE
Jackendoff’s theory of lexical semantics, Conceptual Structure, has evolved

over many years, as documented in several of his foundational publica-
tions, most notably Jackendoff (1983, 1990, 2002). Jackendoff (2002) sets
out a system for the study of language based on a parallel architecture in
which phonological, syntactic and conceptual structures exist side by side
and are linked by rules that constitute the entries of the mental lexicon.
Jackendoff’s Parallel Architecture (PA), shown in Figure 13.1, represents a
speaker/hearer’s mental grammar.
Each of the three autonomous, but interrelated, structures has its
own primitives and principles of combination. As Jackendoff (1990: 17)
explains, each structure is described by a set of formation rules that gener-
ates the well-formed structures of the level, while interface rules link the
different structures to one another. There are also interfaces between the
linguistic and non-linguistic domains. For instance, in order for us to be
able to describe things that we see, there must be a level which is compat-
ible with both linguistic and non-linguistic domains. This level is concep-
tual structure. Indeed, as Jackendoff (1983: 17) states, conceptual structure
is ‘a single level of mental representation [. . .] at which linguistic, sensory,
and motor information are compatible.’
Where does the lexicon fit into this system? Jackendoff (2002: 131)
Analysing en- and its Romance equivalents 249

Formation Rules Formation Rules Formation Rules
Interfaces to Interfaces to
hearing and perception
Structure Structure Structure
vocalisation and action
Interface Interface
Rules Rules
Interface
Rules
Figure 13.1 The Tripartite Parallel Architecture (adapted from Jackendoff 2002:
125, by permission of Oxford University Press)
argues that ‘the function of lexical items is to serve as interface rules, and
the lexicon as a whole is to be regarded as part of the interface components.’
This is quite obvious for words, but Jackendoff extends this to other expres-
sions stored in the lexicon. For a more detailed explanation of the position
of the lexicon in relation to Figure 13.1, see ten Hacken (this volume).
The formalism of conceptual structure is an algebraic notation used to
encode and/or analyse the meaning of concepts that are contained within
the mental representation of conceptual structure. This ambiguity in the
usage of terminology is reminiscent of and parallel to the acknowledged
ambiguity of the term grammar in Chomskyan linguistics, as highlighted
by ten Hacken (2007: 67). As Chomsky and Halle (1968: 3) state, ‘We use
the term “grammar” with a systematic ambiguity. On the one hand, the
term refers to the explicit theory constructed by the linguist and proposed
as a description of the speaker’s competence. On the other hand, we use
the term to refer to this competence itself.’ Chomsky (1986) introduces the
term I-language, which replaces the second sense of grammar and avoids
the confusion caused by such ambiguity in terminology usage. In the case
of conceptual structure, I take it that this ambiguity is not problematic and
will continue to use it in both senses.
The formalism of CS is made up of conceptual constituents, each of
which belongs to one of a small set of ontological categories, such as Thing,
Place and State. Each of these categories has a different function-argument
structure, which represents how a conceptual constituent belonging to a par-
ticular category can be decomposed in terms of its functions and arguments.
Functions vary in the number of arguments they have and in the ontological
category of the argument(s). I will begin by presenting some examples of
ontological categories with the lowest number of arguments, i.e. zero.
250 jessica forse
The ontological category Thing can be decomposed into a function and

zero arguments, as in (1).
(1) [THING] ‡ [Thing BOY]
The constituent itself, in this case BOY, serves as the function.1 The
category Property has the same structure in that it is decomposed into a
function with zero arguments, with the constituent serving as the function.
In (2), this is illustrated for HAPPY.
(2) [PROPERTY] ‡ [Property HAPPY]
The category Place has a function, such as IN, and one argument, which
belongs to the category of Thing, as in (3).
(3) [PLACE] ‡ [Place IN ([Thing HOUSE])]
The CS in (3) encodes the meaning of the phrase in the house.

The category State has a function, such as BE, and two arguments, the
first of which is a Thing and the second is a Place, as in (4).
(4) [STATE] ‡ [State BE ([THING], [PLACE])]
The Place argument in (4) decomposes further as we have already seen in

(3). This is also illustrated in the example in (5), which is the CS for the
sentence The boy is in the house.
(5) [State BE ([Thing BOY], [Place IN ([Thing HOUSE])])]
Jackendoff (1990: 43) states that the category Event can be elaborated using
three Event-functions, GO, STAY and CAUSE. The examples in (6) are
adapted from Jackendoff (1990: 44).
(6) a. EVENT ‡ [Event GO ([THING], [PATH])]

b. EVENT ‡ [Event STAY ([THING], [PLACE])]
c. [EVENT] ‡ THING
CAUSE , [EVENT]
EVENT
Event
The elaboration I am most concerned with is (6c), since this is the one I
will be using to encode the semantic changes brought about by the word
formation process of prefixation with en-.2
Analysis in conceptual structure can have different degrees of granu-

larity. The function CAUSE, as in (6c), can be further analysed into
the CS function on the thematic tier,3 and the AFF (‘affect’) function
on the macrorole tier, as proposed by Jackendoff (1990, 2007).4 I will
not need this further analysis of causation in the present study, and
will encode the semantics of en-verbs using the more coarse-grained
CAUSE function.
In some cases, it is not immediately apparent which function should be
used to most accurately encode the meaning of certain expressions in the
formalism of conceptual structure. When encoding the semantic changes
that en-brings about, it is important to be clear about the difference
between two functions in particular, GO and INCH. For Jackendoff, the
function GO encodes continuous transition: ‘whatever the particulars of a
path, GO expresses the traversal of every point of it’ (1983: 174). He argues
against a treatment of GO which reduces it to a succession of two states.
Although in this earlier work Jackendoff does not name this ‘succession
of two states’, this description marries up nicely to the INCH function he
introduces in his subsequent work (Jackendoff 1990: 92). The difference
between the two functions can be observed in (7).
(7) a. The metal cooled.

b. The metal cooled for hours.
(7a) can be interpreted as there being two discrete points on a metaphorical

scale: initially, when the metal was ‘warmer than cool’ (for argument’s sake
I will call this ‘hot’, although this could be any temperature higher than its
end-state) and afterwards, when the metal was cool. Such a succession of
two states would be encoded in CS using the INCH function. In contrast,
when (7a) is modified by a durative expression such as for hours, as in (7b),
the sentence can only express a continuous process of the metal becoming
gradually cooler, which would be encoded in CS using the GO function.
The distinction between (7a) and (7b) is best represented pictorially as in
(8a) and (8b) respectively.
hot cool
(8) a.
hot cool
b.
hot cool
hot cool
Having presented the relevant theoretical background to Jackendoff’s
framework and formalism, I now turn to the central focus of this chapter:
the analysis of en-. The elements of the formalism that have been intro-
duced here will be used in the following sections to encode and analyse
252 jessica forse
the semantic changes that the word formation process of prefixation with
en-brings about.
2 ANALYSING en-
The English prefix en-has attracted special attention in generative mor-

phology since it behaves differently to most other prefixes in terms of
both form and meaning. As an example of the discussion it has generated,
Trommelen and Zonneveld (1986) compare three theories on headed-
ness put forward by Williams (1981), Lieber (1980) and Selkirk (1982) on
the basis of Dutch morphology. Each of these theories offers a different
treatment of en-. In form-oriented approaches, headedness is generally
thought of as a matter of syntactic category, with the head of a derived
form determining the category it belongs to. English is considered to be
predominantly right-headed. However, the prefix en-is one of a small
group of English category-changing prefixes which violate the Right-hand
Head Rule (RHR) as proposed by Williams (1981). Whereas prefixes such
as counter- and un-adhere to the RHR, as in counterattack and unhappy,
the exceptional prefix en-determines the syntactic category of the derived
form, creating verbs such as entomb, enslave and enrich. In Trommelen
and Zonneveld’s (1986) and Williams’ (1981) approach, en-is marked
exceptionally as a left-hand head.
The distinction between prefixes such as counter- and un-and those
such as en-can be expressed in semantic as well as syntactic terms. Indeed,
the greater syntactic contribution that en-exhibits through changing
the category of derived forms correlates with the prefix having a greater
semantic content. This can be represented in the conceptual structures
in (9), where un-only contributes the meaning NOT to the derived form
unhappy.
(9) Word formation process: prefixation with un-

a. Input: happy ‡ [Property HAPPY]
b. Output: unhappy ‡ [Property NOT [HAPPY]]
I assume here that word formation processes take a base as their input and
change the meaning of that base in some way. The output of the word for-
mation process in (9), unhappy, belongs to the same ontological category as
the input, happy. This means that the input constitutes the main concept
of the derived form.
The prefix en-affects the concept of the input much more strongly,
modifying its ontological category, as represented in (10).
(10) Word formation process: prefixation with en-

a. Input: tomb ‡ [Thing TOMB]
b. Output: entomb (X entombs Y) ‡
[Event CAUSE ([Thing X], [Event INCH
([State BE ([Thing Y], [Place IN ([Thing TOMB])])])])]
The CSs in (10) show that the input, tomb, is a Thing, while the output,
entomb, is an Event. Therefore, in contrast to (9), it is the prefix, en-, which
contributes more strongly to the derived form and not the base it attaches
to. The stronger semantic contribution of en-can be attributed to the
fact that it determines the highest function of the conceptual structure,
CAUSE, as in (10b). The semantic contribution of the input tomb is more
deeply embedded.
The second Event-function in the CS in (10b) is INCH. As (8) repre-
sents, INCH encodes a succession of two states, whereas GO expresses
continuous transition. In the semantic analysis of en-, it is the change of
state that the prefix brings about that is important, not the actual transi-
tion. Therefore INCH is used to encode the meaning of en-verbs in the
formalism.
The objective of this study, then, is to demonstrate that the word forma-
tion process of prefixation with en-brings about systematic semantic shifts,
and that these shifts in meaning can be encoded in terms of Jackendoff’s
formalism of conceptual structure. In the same way as Lieber and Baayen
(1993) argue that unitary LCSs can be ascribed to the Dutch verbal prefixes
ver-, be- and ont-, it is my aim to show that en-has a core meaning which
can be represented by a single basic CS.
I base my analysis of the English prefix en-on data from the online
Oxford English Dictionary (OED).5 All entries beginning with em- and
en-were selected and entered into a spreadsheet of nearly 2,500 entries.
In my preliminary analysis, I searched for data that would illustrate and
contribute to the word formation rule that is in a current speaker’s com-
petence. I therefore put aside any irregular cases that did not match this
criterion, such as non-verbs, obsolete verbs and verbs that were derived
before borrowing, such as enter. Following this initial selection, I was left
with 264 current en-verbs where both the prefix and the base belong to the
mental lexicon of a present-day speaker of English. In the more in-depth,
semantic analysis, I classified these verbs into semantic types and analysed
them using the formalism of CS.
En-verbs in English have two types of etymology. Firstly, they origi-
nated in Middle English loans from French (Marchand 1960: 113) from
the end of the fourteenth century onwards (OED). For example, endeavour
was adapted into English from the French en+devoir, meaning ‘to make it
254 jessica forse
one’s duty to do something’.6 In other words, *deavour does not exist in the
mental lexicon of a current English speaker, nor did its previous forms exist
in the lexicons of past English speakers; the verb endeavour was adapted
and borrowed as a whole, not formed by a productive word formation
process in the English language. Towards the end of the sixteenth century,
English speakers began to use this word formation process productively,
using the prefix en-to form new words in English, in strict accordance
with the Latin analogies. An example of this second type of en-verb is
endear, which first appears in English language literature in 1580 (OED).
Both parts of the verb, en- and dear, belong to the lexicon of the speaker.
Furthermore, endear does not exist in French, providing unambiguous
proof that it was formed by the word formation process in English. It is this
second type of en-verb that I have analysed in order to ascertain the current
word formation rule.
Marchand (1960: 114) classifies the semantics of en-verbs into the types
given in (11).
(11) a. Type encage ‘put in . . .’

b. Type enslave ‘make into . . .’, type enfeeble ‘make . . .’
c. Type enwrap ‘wrap in, wrap up’
One of the shortcomings of Marchand’s classification is that verbs which

do not share the same semantic schema are grouped together. For example,
Marchand includes verbs of the type ‘put a . . . on a person or thing’, such
as encrown, within his first type, (11a), although it does not conform to
the same ‘put in . . .’ structure as encage. Furthermore, there are en- verbs
that do not fit easily into any of these three categories, such as enwall.
Therefore, I reclassified en-verbs into the six semantic types listed in (12).7
(12) a. Type 1 – Put (something/someone) in/on X (n), e.g. entomb, enthrone

b. Type 2 – Put X (n) in/on (something/someone), e.g. engem, empoison
c. Type 3 – Surround (something/someone) with X (n), e.g. enwall
d. Type 4 – Make (something/someone) into X (n), e.g. enslave
e. Type 5 – Make (something/someone) (more) X (a), e.g. embitter
f. Type 6 – X (v) (something/someone) in, e.g. enwrap
Marchand’s classification in (11) does not make the distinction between

Types 1 and 2 in (12). Both are subsumed under his first semantic type
in (11a). However, there is a clear difference between the two types of
verbs in that the base and the Experiencer swap positions. Therefore it is
necessary to distinguish between these two types of verbs and classify them
separately.
Type 3 verbs, such as enwall, do not easily fit into any of Marchand’s
classes in (11). There are fifteen examples of Type 3 verbs in current use,
including encincture, encloud and enhalo. Therefore it is necessary to create
a new category to incorporate verbs of this type.
Types 4 and 5 in (12d–e) together correspond to Marchand’s second
type in (11b). I have separated it into two types since they differ both
syntactically and semantically. Type 4 verbs mean ‘to make into X’,
where X is a noun, whereas Type 5 verbs are derived from adjectives.
The syntactic distinction coincides with a semantic distinction. Type
4 verbs encode a discrete change of state, whereas, being derived from
adjectives, Type 5 verbs rather encode a change on a scale. This distinc-
tion is lost in Marchand’s classification but is expressed in my revised
typology.
Finally, Type 6 verbs mean ‘to X in’, where X is a verb, as in (12f). This
corresponds to Marchand’s third type in (11c).
Let us now consider how each of the six semantic types can be analysed
in terms of the formalism of CS. The examples in (13) show how some
verbs belonging to Type 1 can be expressed in CS.
(13) a. The high priest entombed Tutankhamun.

[Event CAUSE ([Thing HIGH PRIEST], [Event INCH
([State BE ([Thing TUTANKHAMUN], [Place IN ([Thing TOMB])])])])]
b. The Messiah enthroned the King.
[Event CAUSE ([Thing MESSIAH], [Event INCH
([State BE ([Thing KING], [Place ON ([Thing THRONE])])])])]
(13a) can be paraphrased as ‘The high priest causes the Event which ends
with Tutankhamun being in the tomb’, and (13b) can be paraphrased as
‘The Messiah causes the Event which ends with the King being on the
throne.’ A noteworthy difference between the CSs in (13) is that (13a) has
the Place-function IN while (13b) has the Place-function ON. However,
the distinction between IN and ON is rather English-specific, and not a
cross-linguistic generality.8 Therefore it should not be the basis for sepa-
rating verbs into different semantic types. For this reason, verbs such as
entomb and enthrone, which share the same conceptual patterning bar the
Place-function, are classified within the same semantic type, Type 1. The
question of IN versus ON and other language-specific phenomena will
have to be considered systematically within a broader context.
The examples in (13) illustrate why INCH is the most appropriate
function to use to encode the semantics of en-verbs. If GO were used in
these CSs, it would encode the actual physical process, or in Jackendovian
terms, the Path, of entombing Tutankhamun or enthroning the King. It is
256 jessica forse
only the end result that these en-verbs bring about that is highlighted, and
therefore INCH encodes the semantics of these verbs well.
The CSs in (13) both conform to the same basic CS shown in (14), with
only a minor adjustment of the Place-function.
(14) [Event CAUSE ([Thing X], [Event INCH

([State BE ([Thing Y], [Place IN/ON ([Thing Z])])])])]
Whereas (14) and the lexical entries with the base specified at Y or Z
are part of the speaker’s competence, (13) exemplifies the use of them in
particular sentences, i.e. the performance. What (14) does not specify is
the position of the base. The positions which are not occupied by the base
constitute the arguments. It is important to distinguish between the posi-
tion of the base and that of the arguments in (14), since the word formation
process operates on the lexicon, retrieving the base and prefix to form the
en-verb, while the lexical entries which constitute the arguments simply
occupy syntactic positions and are not part of the word formation process.
In (13), a particular distribution is adopted. This shows that Type 1 verbs
have their base at Z, with the arguments occupying the X and Y positions.
At this point, it is interesting to consider to what extent the other seman-
tic types also conform to the basic CS in (14). Type 2 en-verbs have the
meaning ‘to put X in or on something’, as in (15).
(15) a. The jeweller engemmed the ring.

[Event CAUSE ([Thing JEWELLER], [Event INCH
([State BE ([Thing GEM], [Place ON ([Thing RING])])])])]
b. The regime empoisoned the minds of the young.

[Event CAUSE ([Thing REGIME], [Event INCH
([State BE ([Thing POISON], [Place IN ([Thing MINDS OF THE
YOUNG])])])])]
The CSs in (15) conform to the basic CS for en-verbs given in (14).
However, the base and arguments occupy different positions to Type 1
verbs. For Type 2 verbs, the base occupies the Y position and the argu-
ments are positioned at X and Z. As with Type 1, the Place-function varies
between ON in (15a) and IN in (15b). The ‘minds of the young’ in (15b)
are considered as a Thing. Although the poison in (15b) is not a physical
poison, it is a metaphor. For these reasons, (15b) is encoded in the Spatial
field.
Type 3 en-verbs mean ‘to surround with X’, for example, enwall, as in
(16).
(16) The Trojans enwalled the city of Troy.

[Event CAUSE ([Thing TROJANS], [Event INCH
(State BE ([Thing WALL], [Place AROUND ([Thing CITY OF TROY])])])])]
The CS in (16) also follows the pattern of the core CS for en-verbs, with
a change of the Place-function, in this case, to AROUND. As with Type
2, the base of Type 3 verbs occupies the Y position and the arguments are
positioned at X and Z.
Type 4 verbs mean ‘to make into X’. Two examples are given in (17).
(17) a. The master enslaved the boy.

[Event CAUSE ([Thing MASTER], [Event INCH
([State BEIdent ([Thing BOY], [Place ATIdent ([Property SLAVE])])])])]
b. The spy enraged the king.

[Event CAUSE ([Thing SPY], [Event INCH
([State BEIdent ([Thing KING], [Place ATIdent ([Property RAGE])])])])]
Until this point, all of the examples have been encoded in the Spatial field,
meaning that they represent the end result of a spatial movement of Things.
The notable difference is that the examples in (17), and indeed all Type 4
verbs, are encoded in a different conceptual field, the Identificational field.
The Identificational field is used to encode the attribution of Properties as
opposed to spatial movements. Following Gruber (1965), Jackendoff (1983:
188) proposes to use the Spatial field as a model for the other domains.
Indeed, the Identificational field has parallel lexical patterning to the Spatial
field, but where a Thing is located at a Place in the Spatial field, in the
Identificational field, a Thing is attributed a Property (Jackendoff 1990:
25–6). In CS, functions belonging to the Identificational field are marked
with a subscript Ident, as shown in (17), and all of the functions apply in the
same way as in the Spatial field. There is also a change of ontological category
to express the fact that RAGE is a Property, and the Place-function changes
to AT. The category of the base is Property, since the object does not change
into something else; it simply takes on another property. For example, the
boy does not cease to be a boy because he is now a slave; he merely takes on
the property of being a slave.9 As with Type 1 verbs in (13), Type 4 verbs also
have their base at Z, with the arguments occupying the X and Y positions.
Type 5 verbs mean ‘to make (more) X’, for example, embitter, as in (18).
(18) The divorce embittered Michael.

[Event CAUSE ([Thing DIVORCE], [Event INCH
(State BEIdent ([Thing MICHAEL], [Place ATIdent ([Property BITTER])])])])]
258 jessica forse
The CS in (18) shows that Type 5 verbs also conform to the basic CS for en-
verbs, notwithstanding some systematic variations. As was the case in (17),
the Place-function is changed to AT and the ontological category is changed
to Property. Again, this example is encoded in the Identificational field,
meaning that the relevant functions are marked with the subscript Ident.
Type 5 verbs follow the same pattern as Types 1 and 4, insofar as they have
their base at Z, with the arguments occupying the X and Y positions.
It seemed likely that the change in meaning brought about by Type
5 verbs could be represented by a cline and therefore encoded using the
GO function, given that they are derived from adjectives. However, upon
closer examination, this is not the case. As with the other types, the change
of state brought about by Type 5 verbs is best represented by a succession
of two states, and therefore the INCH function. It is the end-result, of
being embittered for example, that is important, and not the transition,
or Path, leading to this embitterment. In other words, the ‘more’ in (12e)
is not a significant part of the meaning of this type. The examples in (19)
illustrate this point further.
(19) a. *Michael was embittered, but he was not bitter given his cheerful
disposition.
b. The metal cooled slightly, but it was still too hot to touch.
(19a) does not make sense, while (19b) is a perfectly valid statement. This is
because the meaning of embitter cannot be represented by a cline, whereas
the meaning of the verb cool can. In other words, the end-result of embit-
ter is always that the Experiencer is bitter whereas the end-result of cool is
not always cool; in some cases, as in (19b), it is merely cooler than it was
before.10 Therefore INCH should be used to encode en-verbs, including
Type 5, where the end-result of the process is the relevant part of the verb’s
meaning.
Type 6 verbs mean ‘to X in’, for example, enwrap means ‘to wrap in’ and
is used predominantly in a figurative sense and in the passive, as in (20).11
(20) The girl was enwrapped by the fascinating tale.

[Event CAUSE ([Thing FASCINATING_TALE], [Event INCH
(State BE ([Thing GIRL], [Place IN ([Thing WRAP])])])])]
As with Types 1, 4 and 5, Type 6 verbs have their base at Z with the argu-
ments occupying the X and Y positions.
Having analysed examples from all six of the semantic types in terms of
the formalism, I conclude that all semantic types for en-conform to one
basic CS, given in (21).
Table 13.1 Variables by semantic type

Type Y Z α β Semantic field
1 Experiencer Base IN/ON Thing Spatial
6 Experiencer Base IN Thing Spatial
4 Experiencer Base ATIdent Property Identificational
5 Experiencer Base ATIdent Property Identificational
2 Base Experiencer IN/ON Thing Spatial
3 Base Experiencer AROUND Thing Spatial
(21) [Event CAUSE ([Thing X], [Event INCH

([State BE ([Thing Y], [Place α ([β Z])])])])]
The variation between the different types of verbs can be reduced to the
following variables:
• whether the base is positioned at Y or Z;

• the Place-Function, α;12
• the ontological category of Z, β;
• the semantic field of the predicate.
Each type has a combination of these four variables (the first argument, or
Agent, always occupies the X position), as summarised in Table 13.1. The
types in Table 13.1 have been purposely rearranged so as to facilitate com-
parison between certain types and highlight the similarities between them.
We can now distinguish three clusters:
• Although the semantic schemas of Types 1 and 6, as in (12), are quite

different, a closer examination of the CSs in (13) and (20), and the vari-
ables presented in Table 13.1, shows striking similarities between them.
Indeed, it could be argued that Type 6 is a subset of Type 1, since α is
always IN for Type 6 verbs; that is, it is one of the two options for α for
Type 1 verbs. Given this semantic similarity, the difference between this
pair of types is rather expressed at a syntactic level, since Type 1 verbs
are derived from nouns (entomb) and Type 6 from verbs (enwrap).
• Types 4 and 5 form a second cluster of semantic types. In the analysis
of the semantics of Type 5 verbs we saw that, although they are derived
from adjectives, they do not encode a cline. It is the end-result of the
Event that is important, and not the transition leading to it. Therefore
Types 4 and 5 are conceptually identical. As with Types 1 and 6, the dif-
ference between 4 and 5 is expressed at a syntactic rather than a semantic
level. The nuances of meaning follow from these syntactic properties.13
260 jessica forse
• Together, Types 2 and 3 form the third cluster. The only difference
between these types is the Place-function α. They are noticeably dif-
ferent from the other types in syntactic terms, since the base and
Experiencer swap positions.
In conclusion, this classification comprises six semantic types for en- verbs,
as presented in Table 13.1.14 By encoding these different types in CS, the
relationship between them can be expressed precisely. By examining the
variables in Table 13.1, it is also possible to group the types into three
clusters: (1) Types 1 and 6; (2) Types 4 and 5; and (3) Types 2 and 3.15 This
shows that the variation between the different types, while expressible in
CS, is only marginal.
3 ROMANCE EQUIVALENTS
The prefix en-came into English from Latin via French, as described by
Marchand (1960: 113), and therefore there is a historic link between the
English prefix and Romance en-. In the Romance languages, however,
there is a semantic overlap between the prefixes en- and a-, which has not
been taken over into English. It is interesting to consider whether verbal
derivatives with these prefixes in French, Spanish and Portuguese conform
to the same semantic types as English en- verbs.
For French, I used the online Trésor de la Langue Française (TLF)
and the fourth edition of the Oxford Hachette French Dictionary (2007)
as my data sources. For Spanish, I used the online dictionary of the Real
Academia Española and the third edition of the Oxford Spanish Dictionary
(2003). For my Portuguese data, I used the third edition of the Collins
Portuguese Dictionary (2007).
For my analysis of Romance en-, I took the semantic types for English
en-verbs as a basis and investigated whether French, Spanish and
Portuguese have correlates for these types. (22) shows examples of French
en-verbs that correspond to each semantic type.
(22) a. Type 1 embouteiller ‘put in bottle’ bouteille ‘bottle’

b. Type 2 empoisonner ‘put poison in’ poison ‘poison’
c. Type 3 encadrer ‘surround with frame’ cadre ‘frame’
d. Type 4 embouler ‘make into a ball’ boule ‘ball’
e. Type 5 embrunir ‘make (more) brown’ brun ‘brown’
f. Type 6 enfermer ‘shut in’ fermer ‘shut’
Spanish en-verbs also correlate with the same semantic types, as in (23).
(23) a. Type 1 embotellar ‘put in bottle’ botella ‘bottle’

b. Type 2 empapelar ‘put paper on’ papel ‘paper’
c. Type 3 enmarcar ‘surround with frame’ marco ‘frame’
d. Type 4 enviudar ‘make into a widow(er)’ viudo/a ‘widow(er)’
e. Type 5 enrojecer ‘make red, redden’ rojo ‘red’
f. Type 6 encerrar ‘shut in’ cerrar ‘shut’
(24) shows examples of Portuguese correlates.
(24) a. Type 1 engarrafar ‘put in bottle’ garrafa ‘bottle’

b. Type 2 envenenar ‘put poison in’ veneno ‘poison’
c. Type 3 emoldurar ‘surround with frame’ moldura ‘frame’
d. Type 4 enviuvar ‘make into a widow(er)’ viuvo/a ‘widow(er)’
e. Type 5 emagrecer ‘make thin’ magro ‘thin’
f. Type 6 encerrar ‘shut in’ cerrar ‘shut’
As (22–24) illustrate, en-verbs in all three languages have correlates for

the same six semantic types set up originally for English. Because of their
semantic similarity, arguably all of these correlates conform to the same
basic CS for English en-verbs, given in (21). This is illustrated for Type 1
verbs in the parallel examples in (25), all meaning ‘The wine producer puts
the wine in bottles’.
(25) a. Le viticulteur embouteille le vin.

[Event CAUSE ([Thing VITICULTEUR], [Event INCH
([State BE ([Thing VIN], [Place IN ([Thing BOUTEILLE])])])])]
b. El vinicultor embotella el vino.

[Event CAUSE ([Thing VINICULTOR], [Event INCH
([State BE ([Thing VINO], [Place IN ([Thing BOTELLA])])])])]
c. O vinicultor engarrafa o vinho.

[Event CAUSE ([Thing VINICULTOR], [Event INCH
([State BE ([Thing VINHO], [Place IN ([Thing GARRAFA])])])])]
This generalisation across the languages is due to the etymological relation-

ship between the different en-prefixes, as described in section 2.
By looking at French translations of English en-verbs, I found that the
Romance prefix a-can also bring about some of the same semantic changes
as the English prefix en-. I investigated this correspondence in a systematic
way, using the same data sources that I used for my analysis of Romance
en-. A-verbs are much less frequent than en-verbs, and the absence of
262 jessica forse
certain types in the data seems to be at least in part accidental. (26) gives
examples of the two types found in French.
(26) a. Type 1 alunir ‘put on moon’ lune ‘moon’

b. Type 5 assouplir ‘make softer, soften’ souple ‘soft’
Spanish a-verbs also correlate with some of the same semantic types for
en-verbs, as in (27).
(27) a. Type 2 acuchillar ‘put knife in, stab’ cuchillo ‘knife’

b. Type 3 acordonar ‘surround with cord, cordon off’ cordón ‘cord’
c. Type 5 ablandar ‘make softer, soften’ blando ‘soft’
(28) shows some Portuguese a-verbs which correlate with some of the
semantic types for en- verbs.
(28) a. Type 4 abagunçar ‘make into mess’ bagunça ‘mess’

b. Type 5 amolecer ‘make softer, soften’ mole ‘soft’
The examples in (29) demonstrate how Type 5 a-verbs are encoded in CS.
(29) a. Obama assouplit la politique américaine envers Cuba.

‘Obama softens the American policy on Cuba’.
[Event CAUSE ([Thing OBAMA], [Event GOIdent
([Thing POLITIQUE_AMERICAINE_ENVERS_CUBA],
[Path TOWARDSIdent ([Property SOUPLE])])])])]
b. El Ayuntamiento ablanda las restricciones de tráfico.

‘The Council softens traffic restrictions’.
[Event CAUSE ([Thing AYUNTAMIENTO], [Event GOIdent
([Thing RESTRICCIONES_DE_TRAFICO], [Path TOWARDSIdent
([Property BLANDO])])])])]
c. A idade amolece o ditador.

‘Age softens the dictator’.
[Event CAUSE ([Thing IDADE], [Event GOIdent
([Thing DITADOR], [Path TOWARDSIdent ([Property MOLE])])])])]
It is interesting to note that, in contrast to Type 5 en-verbs, the examples

in (29) are best encoded using the GO function. It is clear that the sense of
the verb is ‘make more X’ rather than ‘make X’, as with embitter. Obama’s
policy on Cuba is not soft; rather, he has relaxed it somewhat, but it is still
tough. In Jackendovian terms, the Experiencer is moving towards the end-

result, but there is no implication that this end-result has been reached. At
least for Type 5 verbs, there is a systematic semantic difference between
en- and a-verbs in this respect.
There is no direct correlation between Romance a-and English en-
since they are not etymologically related. However, there is some semantic
overlap, especially between Romance en- and a-. As (22) and (26) show,
both French prefixes have verbs belonging to Types 1 and 5, meaning
that they compete for some of the same meanings. However, verbs
formed using a- and en-also have semantic idiosyncrasies. For example,
the prefix a-forms derivatives belonging to other semantic types, such
as ‘get on X’, for example agenouiller means ‘to get on (one’s) knee(s)’,
i.e. ‘to kneel’. The exhaustive list of semantic types for a-verbs would
be another interesting topic to explore further. Furthermore, French
en-verbs can have moral connotations, which can be observed when you
compare ennoblir, meaning ‘to make noble’ in a moral sense, with anoblir
‘to make noble’ in the proper sense (TLF). Similarly, French a-verbs can
have a relative sense, which can be observed when you compare baisser
‘to lower’ in an absolute sense, with abaisser ‘to lower’ in a relative sense.
There are two hypotheses for how these differences in meaning came to
be. Firstly, the semantic contrast between the verbs could be attributed
to the word formation processes having developed specialisations of
meaning. That is, it could be argued that prefixation with en-carries with
it moral connotations, and prefixation with a-attributes a relative sense to
its derivatives. An alternative – and perhaps more plausible – explanation
is that this is simply an effect of name-giving. In such an onomasiological
perspective, there is a need to name a concept, and therefore a naming
process is required. If the most natural choice of process has already been
used to name another concept, an alternative process has to be used. In
the absence of further evidence of this type, the second hypothesis seems
the most likely.
4 CONCLUSION
In light of the preceding analyses, the following conclusions can be drawn:
• The six semantic types for English en-verbs given in Table 13.1 all
conform to one basic CS, given in (21).
• The variation between the types can be reduced to the four variables in
Table 13.1. By analysing the types in this way, the relationship between
the different types could be expressed in precise terms. On this basis,
264 jessica forse
it was found that the types can be grouped into three clusters. This
demonstrates that the variation between the different types is not only
systematic, but also marginal.
• Romance en-verbs have correlates for the same six semantic types as
English en-verbs. Given this semantic correlation, it can be argued that
they have the same core CS. This is not surprising in view of the etymo-
logical relationship between the prefixes.
• The Romance prefix a-can bring about some of the same semantic
changes as the English prefix en-. However, there is no direct correlation
since they are not etymologically related.
• The change in meaning brought about by en-verbs is represented by a
succession of two discrete states and is therefore most accurately encoded
using the INCH function. In contrast, it seems that the semantic change
caused by a-verbs can be represented by a cline and is consequently best
encoded using the GO function.
This chapter has shown how Jackendoff’s theory of Conceptual Structure

can be used as a framework within which to investigate the relationship
between lexical semantics and morphology. In particular, we have seen
that the formalism is a useful tool for encoding and analysing the semantic
changes brought about by derivational word formation processes. It is
hoped that this study will serve as a useful point of departure for continued
exploration into the conceptual semantics of word formation.
NOTES
1. In this framework, human beings are not distinguished from other

Things since they are also physical objects.
2. The curly brackets in (6c) denote that the first argument of CAUSE is
either a Thing or an Event.
3. The CS function is not to be confused with the abbreviation for
Conceptual Structure, CS.
4. This finer- grained analysis of causation was first introduced in
Jackendoff (1990: 126–50). In this earlier version, the macrorole tier
was called the action tier. Jackendoff (2007: 204–6) expands the action
tier to some of the perception verbs and replaces the name with mac-
rorole tier.
5. The OED data was collected in April 2010. Some data may have
changed since this time due to the continual revision of the OED.
6. The first documented appearance of endeavour in the English language
was c.1400 (OED).
7. In describing my classification, I use X to refer to the base followed by

the syntactic category in brackets and ‘something/someone’ to refer to
the object of the resulting verb.
8. This is illustrated by comparing the English phrase on the bus with the
equivalent phrase in French, dans le bus, which literally translates as ‘in
the bus’.
9. Jackendoff (1983: 194) treats properties indicated by nouns, as in (17a),
as Thing Type rather than Property. This distinction could be main-
tained here with a slight complication of the ensuing generalization.
10. For a verb such as enlarge, the contrast is perhaps less obvious, since it
is not an exact science how big an object has to be in order to be con-
sidered large. However, a minimum size is implied; an enlarged object
would be bigger than normal. An enlarged photograph, for instance,
would be bigger than the standard size. Therefore it is still the end-
result of being large that needs to be encoded and not the transition of
becoming large.
11. If something is expressed in the passive voice, it is possible that one of
the arguments (the agent) may remain underspecified, but the overall
conceptualisation of a phrase is not affected by voice. If the example in
(20) were rephrased in the active voice (albeit marked in this case) as
‘The fascinating tale enwrapped the girl’, the conceptualization is the
same, and therefore the CS would also be the same.
12. For Types 1 and 2, α is either IN or ON. As argued with reference to
(13), the distinction between IN and ON is specific to English and is
not a cross-linguistic generality. For this reason, verbs which share the
same conceptual patterning, bar this distinction in Place-function, are
classified within the same semantic type.
13. This approach to the distribution of tasks between syntax and seman-
tics is in line with Culicover and Jackendoff’s (2005) general assump-
tion that it should depend on empirical considerations rather than
theoretical preconceptions which aspects should be covered by which
component.
14. Despite the reordering of the types in Table 13.1, the original number-
ing will be maintained for ease of comparison throughout the chapter.
15. It is interesting to note that the three clusters do not correspond
exactly to Marchand’s three semantic types in (11).
chapter 14
Semantics of diminutivization:
evidence from Russian
Renáta Panocová
P revious cross-linguistic research into diminutivization and augmen-

tativization, also commonly referred to as evaluative morphology, has
pointed to the nearly universal nature of nominal diminutives and augmen-
tatives (Bauer 1997; Dressler and Barbaresi 1994). It has also been observed
that diminutives and augmentatives express not only their basic meaning of
diminution and augmentation but also that of intensification, politeness or
rudeness strategies, etc. (Dressler and Barbaresi 1994: 27).
Grandi (2011: 7) classifies languages into four types (A, B, C and D)
based on the presence or absence of diminutives and augmentatives in a
language. Type A includes languages with diminutives only, type B with
both diminutives and augmentatives, type C without diminutives and aug-
mentatives. Type D, a language with augmentatives only, is not attested.
Slavic languages belong to type B, which implies they have morphological
devices at their disposal to derive diminutives and augmentatives. It may
be stated that there is a direct link between the typological characteristics
of a language and the presence of morphologically formed diminutives and
augmentatives. As indicated by Grandi (2011: 8), inflectional languages,
e.g. Romance, Slavonic and Baltic languages, tend to belong to type B
whereas analytic languages, e.g. Germanic languages, belong either to type
A (presence of diminutives only) or type C (no evaluative morphology).
Russian is an East-Slavic and inflectional language especially rich in its
inventory of diminutives. The word classes that can be morphologically
diminutivized in Russian include nouns, adjectives, verbs and adverbs.
Denominal diminutivization is a highly productive process in Russian.
Kempe et al. (2003: 473) claims with respect to Russian that ‘almost any
concrete noun can be diminutivized’. As the data in section 2 reveal, certain
abstract nouns in Russian can also have diminutive forms.
The aim of the present chapter is to investigate how diminutive nouns
semantics of diminutivization 267
in Russian get their meaning from the perspective of the onomasiological

approach. Section 1 briefly outlines Horecký’s multi-level model of word
formation as it represents the theoretical point of departure of this chapter.
Section 2 applies the model to Russian diminutive nouns and focuses on
the description of the semantic level. The main aim of this section is to
describe a complete semantic definition of diminutive nouns in Russian,
including semantic features, relations between them and their hierarchy.
Section 3 then summarizes the conclusions. The data exemplifying strings
of semantic features are presented in Tables 14.1, 14.2 and 14.3 in the
appendix.
1 THEORETICAL BACKGROUND
In this section I will first present Horecký’s (1983) onomasiological model

of word formation. Section 1.2 will clarify what is meant by a complete
semantic definition of diminutive nouns.
1.1 Horecký’s model of word formation

Horecký’s (1983) multi-level model of word formation represents one of
several models that can be set within the onomasiological framework, in
which the study of word formation departs from meaning rather than form.
Horecký’s model involves a particular object of extra-linguistic reality,
and includes conceptual, semantic, onomasiological, onomatological and
phonological levels. The conceptual level is also referred to as the pre-
semantic component (Horecký 1994: 12) and describes an object of extra-
linguistic reality by means of logical predicates. Based on the Saussurean
concept of the linguistic sign, the semantic level constitutes the signifié,
while the onomasiological, onomatological and phonological levels combine
to form the signifiant.
First, the formal levels will be briefly outlined. The onomasiological
structure is expressed by the onomasiological base and the onomasiological
mark. The former also includes a set of categories, e.g. word class and
related categories. Both the onomasiological base and mark are expressed by
morphemes (formants) at the onomatological level. The phonological level
specifies the phonological features of a particular naming unit and applies
the appropriate phonological rules.
Let us now turn to a more detailed description of the semantic level,
which plays the most prominent role in Horecký’s multi-level model of
word formation. In addition, Horecký’s elaborate approach to the semantic
level represents the theoretical starting point for the analysis of Russian
268 renáta panocová
diminutive nouns in section 2. Horecký (1994) differentiates between four

types of meaning: categorial, invariant, specific and lexical. As pointed out
by Štekauer (2005b: 211), ‘the first three meanings as a whole are labelled
as the “structural” meaning (given by the interrelation between onoma-
siological base and mark), and underlie the lexical meaning’. Horecký
(1980: 84) provides the example of the Slovak word hovädzina (‘beef’) to
illustrate the four meanings. The categorial meaning of hovädzina (‘beef’)
is denominal noun, the invariant meaning is defined by a string of seman-
tic features, –HUM +CONCR +RES –INS +MAT –FIN +ORIG (the
meaning of the individual features will be explained in section 1.2). The
specific meaning is ‘meat from a certain animal’ and its lexical meaning is
‘meat from beef cattle’. The specific meaning specifies that it is a particular
kind of meat and it represents a model denoting the whole class of different
kinds of meat. Only the lexical meaning adds up the information that the
meat comes from beef cattle.
Horecký (1994) sets out a list of semantic distinctive features, explains
their relations and finally provides their hierarchical arrangement. He
points to the fact that a set of semantic distinctive features constitutes the
semantic level of the language and describes a particular word formation
field. It is the word class of the word formation base and the resulting
naming unit (deadjectival adjectives, deverbal nouns, denominal verbs,
etc.) that define a corresponding word formation field.
Horecký (1994: 20) uses the term complete semantic definition for the
string of semantic distinctive features describing a particular word forma-
tion field, specified relations between the features and their hierarchical
arrangement in the form of a tree-diagram. He emphasizes that the meaning
of derived words may not be deduced from the meaning of the base or the
formant, but only from the derived word itself (Horecký, 1980: 85). That
the meaning of a derivation is not entirely compositional is expected. This
is illustrated by the example in Figure 14.1.
Horecký (1994: 34) considers the semantic distinctive feature of sta-
tivity (STA) to be the most abstract and hierarchically highest semantic
feature of deadjectival verbs in Slovak. Verbs like veseliť sa (‘enjoy oneself’)
+ veselit’ sa (‘enjoy oneself’)

STA + bielit’ (‘whiten/bleach’)
FAC
– +
rohovatiet’ (‘keratinize’)
MAT
–
beliet’ (‘pale’)
–
Figure 14.1 Complete semantic definition of deadjectival verbs in Slovak by

Horecký (1994: 34)
derived from veselý (‘merry’) are assigned the semantic feature +STA.
Verbs with –STA fall into factives +FAC and non-factitives –FAC. The
former are exemplified by bieliť (‘whiten’). The latter are subdivided based
on whether or not they denote a change of state of material (MAT). Verbs
like rohovatieť (‘keratinize’) are assigned the semantic feature +MAT as the
base or what is called the motivating word refers to a substance, but verbs
like belieť (‘to pale’) are assigned –MAT.
Deverbal nouns, denominal nouns, deadjectival nouns, deverbal adjec-
tives, denominal adjectives, deadjectival verbs, denominal verbs and
deverbal verbs in contemporary literary Slovak were described in a similar
way by Horecký, but he does not analyse them in detail. Kačmárová
(2010) analyses Slovak diminutives, but she uses a different framework.
Therefore examining Russian diminutives is interesting because it gives
a new phenomenon and a new language. I will examine Russian diminu-
tives from the perspective of Horecký’s approach to the description of the
semantic level.
1.2 The semantics of diminutive nouns in Russian

Diminutive nouns in Russian represent a subset of denominal nouns,
deverbal nouns and deadjectival nouns. It is assumed that the derivation of
diminutives modifies the meaning of the base words by adding a modifying
feature of diminution. Dokulil (1962: 47) distinguishes three onomasiologi-
cal categories: mutational (relational), transpositional and modificational.
The mutational category is the most basic type. The concepts of one cat-
egory constituting the onomasiological base are specified by the concepts
of either the same or a different category representing the onomasiological
mark. For instance, the category of SUBSTANCE is determined by its
relation to the concept of the category of SUBSTANCE (toothpaste) or
ACTION (teacher).The meaning of a derived word substantially differs
from the meaning of the base.
In the transpositional category a phenomenon understood as a mark
dependent on a SUBSTANCE is generalized (abstracted) and becomes
independent of the SUBSTANCE. An example is the objectification of
QUALITY (playful – playfulness). It is only the word class that changes,
not the lexical meaning of the base.
The modificational category differs from the other two in the fact that
a modifying feature (mark) is added to a concept of a particular category.
The modification can be of several types including not only diminutives
but also augmentatives, names of the young, collective nouns and changes
of gender. The category of the names of the young refers to animate beings
that are not adult, e.g. kitten, child.
To my knowledge, diminutives in Russian have not been examined

before within the above framework. Typical treatments of diminutives are
exemplified by the Academic Russian Grammar edited by Švedova (1982)
and the books on Russian word formation by Bratus (1969), Townsend
(1975) and Vinogradov (1986). They provide lists of masculine, feminine
and neuter diminutive suffixes and their corresponding allomorphs and
specify the conditions that must be met for their use.
The fundamental issue to be addressed when describing the semantics
of derived words within the onomasiological approach is the systematic
description in terms of semantic distinctive features. In this framework it is
assumed that the meaning of derived words is given by an entire string of
semantic features, i.e. the invariant meaning of a resulting word. Horecký
(1994: 39) emphasizes that such an analysis points to ‘a reciprocal relation
between bases (motivating words) and formants at least in the sense that
certain bases combine only with certain formants or vice versa’. This also
means that particular strings of semantic features may be assigned, for
instance, to denominal nouns derived by certain specific formants (suffixes)
and not by certain other suffixes to denominal nouns. The starting point for
derivation would not be assigning a suffix to a noun but a noun to a suffix.
As mentioned in section 1.1, Horecký’s model of word formation
involves a particular object of extra-linguistic reality, and includes concep-
tual, semantic, onomasiological, onomatological and phonological levels. I
will now briefly outline the relation of semantic level to the formal levels.
The so-called onomasiological structure comprising onomasiological base
and mark is specified at the onomasiological level. It is based on the string
of semantic features defined at the semantic level. The onomasiological
structure is fleshed out by formants at the onomatological level. It may be
stated that the approach from meaning to form reveals the relation between
semantic and onomatological structure. Horecký (1994: 52) points out that
different formants can be found in the same string of semantic features and
conversely, the same formant can be found in several different strings of
semantic features, e.g. the morpheme -stvo realizes two strings of semantic
features: –HUM –CONCR +QUAL illustrated by priateľstvo (‘friendship’)
and –HUM –CONCR –QUAL exemplified by učiteľstvo (‘teachership’).
The contrast in meaning is that učiteľstvo (‘teachership’) is a profession or a
group of people whereas priateľstvo (‘friendship’) is a relationship.
Horecký’s (1974: 129) systematic description of semantic features is
based on two types of criteria, traditional and systematic. Traditional
criteria stem from derivational definitions whereas systematic criteria rely
on the selection of specific semantic distinctive features leading to a hier-
archically arranged system of these features. Horecký (1974: 131) points
out several principles governing the whole selection process. Firstly, it is
necessary to select the semantic features applicable to all possible members

of a particular word formation field, for instance denominal nouns based on
the method of binary division into positive and negative features. Secondly,
it seems appropriate to proceed from general semantic distinctive features
to more specific ones. Last but not least, their hierarchical ordering must
be determined in order to describe a system of semantic features. Horecký
(1994: 36–7) lists thirty-nine semantic features and subsequently applies
them in describing semantics of deverbal nouns, denominal nouns, dead-
jectival nouns, deverbal adjectives, denominal adjectives, deadjectival
verbs, denominal verbs and deverbal verbs in contemporary Slovak.
Horecký’s (1994) inventory of semantic distinctive features is also used
here to describe a complete semantic definition of Russian diminutive
nouns. Diminutive nouns in Russian can be found in deverbal nouns,
denominal nouns and deadjectival nouns, and the list of semantic features
applied in their complete semantic definition did not include all thirty-nine
of Horecký’s features, but twenty of them, namely ADH, AFF, AGN,
CONCR, CONT, EFF, ERG, HAB, HUM, ‘HUM’, INS, LOC, MAT,
MOD, OFF, ORIG, POS, RES, REZ, STAT. All the semantic distinctive
features are explained and exemplified in sections 2.1, 2.2 and 2.3. The
names of all the features are derived from words of Latin origin.
2 RUSSIAN DIMINUTIVE NOUNS
For the purpose of the analysis it was necessary to collect a wide range of
diminutivized Russian nouns. The data were collected from the above-
mentioned sources. Some diminutive forms were verified in Dal’ (2005)
and by searching in the Russian national corpus (http://ruscorpora.ru/
search-main.html#). The sources with complete bibliographical data are
included in the references.
The diminutive nouns taken from the above sources served as a basis for
a more detailed analysis. The next step was to classify the collected diminu-
tive forms of Russian nouns with respect to their categorial meaning, i.e.
denominal, deverbal and deadjectival nouns. These classes were divided
into sets including names of persons, names of places, names of instru-
ments, names of quality, etc. Then each of the sets was described based
on common semantic features. In other words, each diminutive noun was
assigned the appropriate strings of semantic distinctive features. A full
overview of the strings of semantic features with examples of Russian
diminutives is given in Tables 14.1, 14.2 and 14.3 in the appendix to the
chapter. The final step was to present the results of the analysis in tree
diagrams, which will serve as our basis for the discussion in the sections
below. Throughout the discussion I will follow Horecký’s tree diagrams.

Where the Russian data require me to expand it, this will be discussed in
more detail.
2.1 Denominal diminutive nouns

As mentioned above, Dokulil (1962) characterizes the category of diminu-
tives as a type of a modificational onomasiological category. It is modifi-
cational in the sense that it adds a modifying feature to the content of a
particular concept and it points to the complex character of the modifying
feature of diminution. This is illustrated in (1).
бутылка: бутылочка: butylka: butylochka: (‘bottle’: ‘small bottle’)

(1) a.
b. писатель: писателишка: pisateľ: pisatelishka: (‘writer’ pejorative)
внук: внучонок: vnuk: vnuchonok: (‘grandson’: small grandson and/or
c.
dear grandson)
It may refer to quantity, i.e. size of a concept, smaller than usual (1a),
quality, i.e. emotional evaluation, appreciative or depreciative (1b), and
quantity and quality (1c). If a modifying feature is emotional or expres-
sive, it may vary in its intensity. Various degrees of intensity may often
be expressed by repeated application of the diminutivization process,
commonly referred to as diminution of the first and second degree and
illustrated in (2).
(2) цвет → цветок → цветочек

cvet → cvetok → cvetochek
‘flower → flowerDIM → flowerDIM-DIM’
Diminutives of the second degree tend to denote either smaller size than
diminutives of the first degree, or they intensify the emotional or expres-
sive value of the first degree diminutives.
The notion of modificational category is essential for the description of
semantic definition and therefore indicated at the top of the hierarchy in
all three tree diagrams as MOD (see Figures 14.2–14.4) and the linear
representations in the appendix to this chapter. As the modifying feature
of diminution is only one of the possible modifications covered by the
modificational category, it is also necessary to include it in the tree diagrams
and linear representations as DIM. It should be emphasized that DIM
is understood and used in this chapter as a modifying feature denoting
quantity, quality and a combination of both.
Figure 14.2 presents the hierarchy of the complete semantic definition of
MOD
+DIM
+HUM –HUM
+APPEL –APPEL +CONCR –CONCR

Nadja :
Nadeňka
+EFF –EFF +RES –RES +‘HUM’ +TEMP –TEMP

kotenok : godik : jumorok :
kitten DIM year DIM humour DIM
+CONCR –CONCR +AFF –AFF +INS –INS +POS –POS

stoljarik : muzykantik : ogorodnichek : nozhichek : dolinka :
joiner DIM musician DIM gardener DIM knife DIM valley DIM
+ADH –ADH +MAT –MAT +CONT –CONT

butylochka : golovka :
bottle DIM head DIM
+HAB
+LOC –LOC –HAB +ORIG –ORIG +LOC –LOC
brjuchanchik:
nemchishka: vnuchonok: druzhok govjadinka : bumahka : shkolka: rubashechka:
man with a
German DIM grandson DIM friend DIM beef DIM paper DIM school DIM shirt DIM
big belly DIM
Figure 14.2 Denominal diminutive nouns in Russian

denominal diminutive nouns in Russian. More examples illustrating a par-

ticular string of semantic features are given in Table 14.1 in the Appendix.1
Traditional semantic classifications of Russian nouns divide them into
two classes: animate and inanimate. This classification is based on the
basic division into names of persons (animate) and names of things/objects
(inanimate). Due to the fact that names of animals are considered inani-
mate, it seems reasonable to top a hierarchy with the semantic distinctive
feature ±HUM (from Latin humanus) differentiating between names of
persons and things.
Russian nouns with semantic feature +HUM that can be diminutivized
may be further divided according to the feature appellative (±APPEL, from
Latin appellativus). The term appellative is used to denote common nouns
(Concise Oxford Dictionary of Linguistics 2005). Horecký (1994) does not use
this feature, but it was necessary to introduce it to encode the difference
between proper and common diminutive nouns in Russian. The process of
diminution applies to a large number of Russian common nouns of differ-
ent types and therefore a more detailed specification in terms of semantic
distinctive features is required. On the other hand, proper names in Russian
are also frequently diminutivized, e.g. Ванюшка: Vanjushka, Марфушка:
Marfushka, Васенька Vaseňka, Наденька: Nadeňka. Although some lin-
guistic authorities (Dressler and Barbaresi 1994: 84; Volek 1987) exclude
hypocoristics from the category of diminutives, they are included in a com-
plete semantic definition of Russian denominal diminutives in this chapter
because diminutivized proper names represent a large subset of Russian
nominal diminutives. In addition, proper names are part of the lexicon and
as such they should be accounted for.
Common diminutive nouns marked as +HUM may be broadly divided
into two groups based on the feature effector ±EFF (from Latin efficio). The
main difference is illustrated by the examples +EFF столяр: столярик:
stoljar: stoljarik (‘joiner’), сапожник: сапожничек: sapozhnik: sapozh-
nichek (‘shoemaker’) who literally produces furniture or shoes, and some-
body like –EFF рыбак: рыбачок: rybak: rybachok (‘fisherman’) who does
not literally produce fish. The semantic distinctive features ±CONCR
+EFF put emphasis on the concrete or abstract nature of the object of
the action performed by the person. The diminutive nouns –EFF can
be assigned the semantic feature affector ±AFF (from Latin afficio). This
feature points to the fact of whether a person takes care of something
or not. An example of +AFF is огородник: огородничек: ogorodnik:
ogorodnichek (‘gardener’).
–AFF diminutive nouns are further divided by the feature ADH (adher-
ence). Diminutive nouns +ADH denote location +LOC, e.g. немец:
немчишка: nemec: nemchishka (‘German’) or adherence to a particular
group or organization (family), e.g. сестра: сестрица: sestra: sestrica

(‘sister’). Bearers of a certain quality with –ADH have a semantic feature
+HAB (from Latin habeo) which points to the presence of an obvious phys-
ical characteristic feature, e.g. брюхo (‘big belly’), лоб (‘high forehead’),
e.g. лобан: лобанчик: loban: lobanchik (‘man with a high forehead’),
брюхан: брюханчик: brjuchan: brjuchanchik (‘man with a big belly’) as
opposed to those with –HAB, e.g. господин: господинчик: gospodin:
gospodinchik (‘gentleman’), друг: дружок: drug: druzhok (‘friend’).
Russian common diminutive nouns with the semantic feature –HUM
are assigned the hierarchically topmost semantic distinctive feature
±CONC (concrete). In many languages, e.g. German, abstract nouns
cannot be diminutivized. In Russian, however, many examples of diminu-
tivized abstract nouns can be found, e.g. теория: теорийка: teorija:
teorijka (‘theory’). A large group of abstract diminutives can be described
by a semantic feature of time TEMP, e.g. утро: утречко: utro: utrechko
(‘morning’). The semantic feature TEMP is not used by Horecký, but
I introduced it because of the nature of a relatively large set of Russian
diminutives.
A relatively large class of –HUM +CONCR diminutives is represented
by the names of young animals, e.g. котенок: котеночек: kotenok: kote-
nochek (‘kitten’). Typically, these can be assigned a distinctive semantic
feature –‘HUM’. Horecký (1974, 1994) emphasizes the difference between
HUM and ‘HUM’. The former indicates the contrast between names
of persons and things while the latter describes changes of gender that
belong to a modificational category. For Horecký (1994: 53), in Slovak
+‘HUM’ marks changes of gender, e.g. chirurg: chirurgička (‘surgeon’)
(‘woman surgeon’), člen: členka (‘member’) (‘woman member’) as opposed
to –‘HUM’ which refers to the names of the young e.g. holub: holubica
(‘pigeon’) (‘dove’), tiger: tigríča (‘tiger’) (‘tiger cub’). The changes of
gender +‘HUM’ are not included as a separate path in the present semantic
definition of Russian diminutive nouns due to the fact that not all changes
of gender are diminutivized. For instance, американка: amerikanka
(‘American woman’) has a diminutive form американочка: amerikano-
chka whereas докторша: doktorsha (‘female doctor’) is not diminutivized.
The former is described by more relevant semantic features which can be
found under a different path of the tree diagram.
The feature –‘HUM’ points to the fact that young animals are smaller,
perhaps less strong and not human beings. Interestingly, diminutivized
names of young animals seem to combine two modifying features within
Dokulil’s modificational category, namely the feature young and the feature
small.
Denominal diminutive nouns +CONCR may be further subdivided
based on the feature materiality RES (from Latin res ‘thing, matter’).
The diminutive nouns marked +RES usually denote names of things and
instruments. The semantic feature +INS then may be assigned to words
like нож: ножичек: nozh: nozhichek (‘knife’). Within –INS diminu-
tives there is a special subgroup denoting materials (MAT), e.g. золото:
золотишко: zoloto: zolotishko (‘gold’), meaning a valuable yellow metal
used for making jewellery or things made of gold. A diminutive form
emphasizes a positive attitude. A group of diminutives +MAT might be
characterized by a feature of origin +ORIG, e.g. говядина: говядинка:
govjadina: govjadinka (‘beef’), телятина: телятинка: teljatina: teljatinka
(‘veal’). In contrast, diminutives –ORIG do not specify their origin in
terms of direct information present in the base, e.g. серебро: серебришко:
serebro: serebrishko (‘silver’), сукнo: суконце: sukno: sukonce (‘cloth’).
–MAT Diminutives are typically associated with a particular place +LOC,
e.g. школа: школка: shkola: shkolka (‘school’). Those with the feature
–LOC include e.g. рубашка: рубашечка: rubashka: rubashechka (‘shirt’).
Russian –RES diminutives fall into another two subgroups, one with the
nouns referring to a position within a particular space POS, e.g. долина:
долинка: dolina: dolinka (‘valley’). The group marked –POS splits into
two based on the semantic feature content (CONT). +CONT indicates
places that store things, tools, materials or substances, e.g. солонка:
солоничка: solonka: solonichka (‘salt box’). The feature –CONT charac-
terizes the following diminutive nouns, e.g. голова: головка: головушка:
golova: golovka: golovushka (‘head’), which may only point to storing
something metaphorically.
2.2 Deverbal diminutive nouns

As opposed to the complete semantic definition of denominal diminutives,
that of deverbal diminutive nouns in Russian seems less complex. A tree
diagram of deverbal diminutive nouns is given in Figure 14.3. More exam-
ples are provided in Table 14.2 in the appendix to the chapter.
The topmost division of deverbal diminutive nouns is represented by
the semantic feature of ergativity (ERG). In Horecký’s (1973: 267) under-
standing, the feature ERG points to an instigator of an action. Deverbal
diminutive nouns marked +ERG include names of persons, instruments
and materials, while −ERG nouns denote places, actions and results of
actions. The next important distinctive feature of +ERG deverbal diminu-
tive nouns is ±HUM. Nouns marked +ERG and +HUM can also be
assigned the feature of agentivity (+AGN). Diminutive nouns marked as
+AGN and +OFF (from Latin officium, i.e. service, duty) denote persons
performing a particular activity as their profession, e.g. купец: купчишка:
MOD
+DIM
+ERG –ERG
+HUM –HUM +STA –STA
+AGN +INS +LOC +REZ –REZ

meshalka : gostinica : risunok : instruktazh :
meshalochka : gostinichka : risunochek : instruktazhik :
stirrer : stirrer hotel : hotel DIM drawing : drawing instructing :
DIM DIM instructing DIM
+OFF –OFF
rabotnik : lgun : lgunishka :
rabotnichek : liar : liar DIM
worker : worker
DIM
Figure 14.3 Deverbal diminutive nouns in Russian
kupec: kupchishka (‘merchant’), работник: работничек: rabotnik: rabot-

nichek (‘worker’), while those with –OFF refer to persons performing
any action, not only their occupation or profession e.g. лгун: лгунишка:
lgun: lgunishka (‘liar’), покупатель: покупателик: pokupateľ: pokupatelik
(‘buyer’). However, some agentive diminutives may be ambiguous, e.g.
водитель: водительчик: voditeľ: voditeľchik (‘driver’) as these can refer
to somebody who drives a vehicle or to somebody whose job it is to drive a
vehicle. Diminutives with the value –AGN do not seem to occur. None of
the examples in the sources indicated falls into this category and formations
such as, for example, рыхлитель: *рыхлительничек: rychliteľ: *rych-
litelnichek (‘ripper’), which would have this feature sound unnatural. The
question arises of whether this is a logical implication of semantic features
or a property of diminutives in Russian. Inanimate words characterised
by –HUM can be assigned the semantic feature instrument (+INS), e.g.
мешалка: мешалочка: meshalka: meshalochka (‘stirrer’). They denote
tools and devices used to perform a particular action. Nouns with –INS
were not found in the sample.
Nouns marked –ERG fall into three groups. In order to characterize

them I use the semantic features STA, REZ and LOC. Horecký uses them
as privative features, but here they are used as binary ones.
Diminutive deverbal nouns –ERG include words denoting places
+LOC, e.g. спальня: спальнюшка: spaľnja: spaľnjushka (‘bedroom’),
гостиница: гостиничка: gostinica: gostinichka (‘hotel’), and those
expressing the feature result of action +REZ,2 e.g. рисунок: рисуночек:
risunok: risunochek (‘drawing’), отпечаток: отпечаточек: otpechatok:
otpechatochek (‘imprint’).
Diminutives describing names of action are assigned the semantic
feature –REZ, e.g. инструктаж : инструктажик: instruktazh: instruk-
tazhik (‘instructing’), просьба: просьбочка: prosba: prosbochka (‘plea’).
Names of actions tend to emphasize the course of action, whereas names
of places where actions occur focus on circumstances and state. Therefore
a distinctive feature ±STA (state) is applied to capture the difference.
Names of places are marked +STA while names of action and results of
action are described as –STA.
2.3 Deadjectival diminutive nouns

Figure 14.4 outlines semantic features of deadjectival diminutive nouns in
Russian. More examples are presented in Table 14.3 in the appendix to the
chapter.
The basic distinctive semantic feature of deadjectival diminutive nouns is
±HUM. Diminutive deadjectival nouns tend to denote a particular quality
or, more precisely, a bearer of that quality, e.g. толстяк: толстячок: tol-
stjak: tolstjachok (‘fat person’). The nouns marked –HUM refer to places,
MOD
+DIM
+HUM –HUM
+QUAL +LOC
tolstjak : tolstjachok : svetlica : svetlichka :
fat man : fat man DIM light room : light room
DIM
Figure 14.4 Deadjectival diminutive nouns in Russian

e.g. больница: больничка: boľnica: boľnichka (‘hospital’). The adjective

from which the above noun and subsequently diminutive noun are derived
is больнoй: boľnoj (‘ill’).
3 CONCLUSION
The aim of the present chapter was to explore the process of how diminu-
tive nouns in Russian get their meaning from the perspective of the
onomasiological approach. Horecký’s (1983) multi-level model of word
formation was taken as a point of theoretical departure. The semantic level
represented the main concern because it also plays the most prominent role
in this model. Horecký (1980: 85) emphasizes that the meaning of derived
words may not be deduced from the meaning of the base or the formant,
but only from the derived word itself. Horecký (1994: 20) sets out a list of
semantic distinctive features characterizing derived words in contempo-
rary literary Slovak, explains their relations and gives their hierarchical
arrangement in tree diagrams. Here, I applied Horecký’s model to Russian
diminutive nouns. The collected examples of Russian diminutive nouns
point to the fact that diminutivization takes place not only in the word
formation field of denominal nouns but also in that of deadjectival nouns
and deverbal nouns. Thus all three categories were specified in terms of
appropriate semantic distinctive features and hierarchical relations among
them as graphically represented in Figures 14.2, 14.3 and 14.4.
The specific nature of diminution in Russian required the introduction
of three additional semantic features DIM, APPEL and TEMP, which
were not specified in Horecký’s description of Slovak. The analysis of
Russian diminutives also revealed that the semantic description of Russian
diminutive nouns differs from that of Russian nouns in general. This
may be illustrated by the fact that not all nouns can be diminutivized, e.g.
female gender nouns like американка: amerikanka (‘American woman’)
and милиционерка: milicionerka (‘police woman’) have a diminutive
form американочка: amerikanochka, милиционерочка: milicionerochka
(‘police woman’) whereas учительница: uchitelnica (‘female teacher’) or
докторша: doktorsha (‘female doctor’) are not diminutivized.
In line with Horecký’s observations on Slovak, also in Russian diminu-
tives, one diminutive morpheme may refer to several strings of distinctive
semantic features. In other words, the mapping between form and meaning
is one to many. For instance, the diminutive suffix -ик: -ik is related to
the string of semantic features MOD +HUM +APPEL +EFF –CONCR:
музыкант: музыкантик: muzykant: muzykantik (‘musician’) from the
field of denominal nouns and also to MOD +ERG –HUM –ACT +INS:
самовар: самоварик: samovar: samovarik (‘device traditionally used to

heat and boil water for tea’) from deverbal nouns. From the perspective of
further research, it might be interesting to examine the relations between
the semantic description of Russian diminutives and their linguistic
expression at the onomatological level.
APPENDIX
Table 14.1 Denominal diminutive nouns in Russian

Russian base Transliteration Russian Transliteration of Gloss of R. base
of R. base diminutive R. diminutive
Ваня Vanja Ванюшка Vanjushka –
Марфа Marfa Марфушкa Marfushka –
Вася Vasja Васенька Vaseňka –
Надя Nadja Наденька Nadeňka –
MOD +DIM +HUM –APPEL
столяр stoljar столярик stoljarik joiner
сапожник sapozhnik сапожничек sapozhnichek shoemaker
MOD +DIM +HUM +APPEL +EFF +CONCR
музыкант muzykant музыкантик muzykantik musician
помощник pomoshchnik помощничек pomoshchnichek helper
ремонтник remontnik ремонтничек remontnichek repairer
MOD +DIM +HUM +APPEL +EFF –CONCR
рыбак rybak рыбачок rybachok fisherman
моряк morjak морячок morjachok sailor
милиционер milicioner милиционерчик milicionerchik policeman
гусляр gusljar гуслярик gusljarik violinist
MOD +DIM +HUM +APPEL –EFF +AFF
китаец kitajec китайчонок kitajchonok Chinese
американка amerikanka американочка amerikanochka American
woman
немец nemec немчишка nemchishka German
француз francuz французик francuzik Frenchman
земляк zemljak землячок zemljachok fellow
countryman
MOD +DIM +HUM +APPEL –EFF –AFF +ADH +LOC
мама mama маменька mameňka mother
папа papa папенька papeňka father
внук vnuk внучонок vnuchonok grandson
сестра sestra сестрица sestrica sister
тётя teta тётенька teteňka aunt
комсомолка komsomolka комсомолочка komsomolochka member of
Komsomol
(woman)
MOD +DIM +HUM +APPEL –EFF –AFF +ADH –LOC
горбун gorbun горбунчик gorbunchik hunchback
пузан puzan пузанчик puzanchik man with a big
belly
губан guban губанчик gubanchik man with big lips
лобан loban лобанчик lobanchik man with high
forehead
брюхан brjuchan брюханчик brjuchanchik man with a big
belly
MOD +DIM +HUM +APPEL –EFF –AFF –ADH +HAB
Table 14.1 (continued)

друг drug дружок druzhok friend
дружочек druzhochek
господин gospodin господинчик gospodinchik gentleman
хозяин xozjain хозяйчик xozjajchik master
MOD +DIM +HUM +APPEL –EFF –AFF –ADH –HAB
юмор jumor юморoк jumorok humour
теория teorija теорийка teorijka theory
смерть smerť смертушка smertushka death
правда pravda правдочка pravdochka truth
забава zabava забавка zabavka fun
MOD +DIM –HUM –CONCR –TEMP
год god годик godik year
день deň денёк denek day
вечер vecher вечерок vecherok evening
ночь noch ночка nochka night
ноченька nocheňka
утро utro утречко utrechko morning
понедельник ponedeľnik понедельничек ponedeľnichek Monday
зима zima зимушка zimushka winter
MOD +DIM –HUM –CONCR +TEMP
змея zmeja змейка zmejka snake
собака sobaka собачка sobachka dog
ёж jezh ёжонок jezhonok hedgehog
заяц zajac зайчик zajchik hare
червяк chervjak червячок chervjachok worm
телёнок telenok телёночек telenochek calf
котенок kotenok котеночек kotenochek kitten
жеребёнок zherebenok жеребёночек zherebenochek foal
MOD +DIM –HUM +CONCR –‘HUM’
долина dolina долинка dolinka valley
долиночка dolinochka
долинушка dolinushka
низина nizina низинка nizinka lowland
вершина vershina вершинка vershinka top
середина seredina серединка seredinka centre
MOD +DIM –HUM +CONCR –RES +POS
нож nozh ножичек nozhichek knife
вилка vilka вилочка vilochka fork
ложка lozhka ложечка lozhechka spoon
ключ kljuch ключик kljuchik key
игла igla иголка igolka needle
дробовик drobovik дробовичок drobovichok rifle
MOD +DIM –HUM +CONCR +RES +INS
Table 14.1 (continued)

телятина teljatina телятинка teljatinka veal
говядина govjadina говядинка govjadinka beef meat
медвежатина medvezhatina медвежатинка medvezhatinka bear meat
MOD +DIM –HUM +CONCR +RES –INS +MAT +ORIG
бумага bumaga бумажка bumazhka paper
золото zoloto золотишко zolotishko gold
серебро serebro серебришко serebrishko silver
сукнo sukno суконце sukonce cloth
MOD +DIM –HUM +CONCR +RES –INS +MAT –ORIG
школа shkola школка shkolka school
университет universitet университетик universitetik university
кино kino киношка kinoshka cinema
театр teatr театрик teatrik theater
библиотека biblioteka библиотечка bibliotechka library
аптека apteka аптечка aptechka chemist’s
квартира kvartira квартирка kvartirka flat
город gorod городок gorodok town
MOD +DIM –HUM +CONCR +RES –INS –MAT +LOC
кольцо koľco колечко kolechko ring
стол stol столик stolik table
салфетка salfetka салфеточка salfetochka napkin
окнo okno оконце okonce window
рубашка rubashka рубашечка rubashechka shirt
брюки brjuki брючки brujchki trousers
MOD +DIM –HUM +CONCR +RES –INS –MAT –LOC
графин grafin графинчик grafinchik carafe
бутылка butylka бутылочка butylochka bottle
банка banka баночка banochka tin
ваза vaza вазочка vazochka vase
кружка kruzhka кружечка kruzhechka jug
чашка chashka чашечка chashechka cup
MOD +DIM –HUM +CONCR –RES –POS +CONT
голова golova головка golovka head
головушка golovushka
жила zhila жилка zhilka vein
жилочка zhilochka
рука ruka ручка ruchka ruchonochka hand
ручоночка
MOD +DIM –HUM +CONCR –RES –POS –CONT
Table 14.2 Deverbal diminutive nouns in Russian

купец kupec купчишка kupchishka merchant
писатель pisateľ писателишка pisatelishka writer
Пастух pastux пастушок Pastushok shepherd
работник rabotnik работничек Rabotnichek worker
MOD +DIM +ERG +HUM +AGN +OFF
Лгун lgun лгунишка lgunishka liar
ассистент assistent ассистентик assistentik assistant
Бездельник bezdeľnik бездельничек bezdeľnichek lazybone
защитник zashchitnik защитничек zashchitnichek defender
Крикун Krikun крикунишка krikunishka crybaby or sb who
shouts
хвастун xvastun хвастунишка xvastunishka sb who boasts
эмигрант emigrant эмигрантик emigrantik emigrant
MOD +DIM +ERG +HUM +AGN –OFF
мешалка meshalka мешалочка meshalochka stirrer
пилка pilka пилочка pilochka nail file
самовар samovar самоварик samovarik device traditionally
used to heat and
boil water for tea
зажигалка zazhigalka зажигалочка zazhigalochka cigarette lighter
MOD +DIM –ERG –HUM +INS
Спальня spaľnja спальнюшка spaľnjushka bedroom
Гостиница gostinica Гостиничка gostinichka hotel
Колыбель kolybeľ колыбелькa kolybeľka cradle
MOD +DIM –ERG +STA +LOC
Подарок Podarok Подарочек podarochek gift
Рисунок risunok Рисуночек risunochek drawing
Отпечаток otpechatok отпечаточек otpechatochek imprint
Кипяток Kipjatok кипяточек kipjatochek boiling/hot water
MOD +DIM –ERG –STA +REZ
инструктаж instruktazh инструктажик instruktazhik instructing
массаж massazh массажик massazhik massage
просьба prosba просьбочка prosbochka plea
MOD +DIM –ERG –STA –REZ
Table 14.3 Deadjectival diminutive nouns in Russian

толстяк tolstjak толстячок tolstjachok fat person
бедняга bednjaga бедняжка bednjazhka poor person
милуша milusha милушка milushka darling
простак prostak простачок prostachok simple person
ленивец lenivec ленивчик lenivchik lazybones
дурак durak дурачок durachok fool
старик starik старичок starichok old man
MOD +DIM +HUM +QUAL
больница boľnica больничка boľnichka hospital
светлица svetlica светличка svetlichka light room
темница temnica темничка temnichka dungeon
MOD +DIM –HUM +QUAL +LOC
NOTES
1. In order to save space, it was necessary to simplify the examples in

Figure 14.2–14.4. Only the transliteration of the diminutive form fol-
lowed by a gloss is included. The Russian original can be found in the
Appendix in the corresponding table.
2. The semantic feature of result is spelled REZ in order to distinguish it
from the semantic feature materiality RES.
Semantics of Word Formation and Lexicalization
Notes on contributors
Maria Bloch-Trojnar is Associate Professor in the Department of

Celtic Studies, John Paul II Catholic University of Lublin. Her major
research interests include morphology and its interfaces with other gram-
matical components, in particular deverbal nominalizations, the inflection-
derivation distinction, lexicology, English, Celtic and Slavic languages. She
is the author of Polyfunctionality in Morphology – A Study of Verbal Nouns
in Modern Irish (Lublin: Wydawnictwo KUL, 2006), editor of Perspectives
on Celtic Languages (Lublin: Wydawnictwo KUL, 2009) and co-editor
with Anna Bloch-Rozmej of Modules and Interfaces (Lublin: Wydawnictwo
KUL, 2012). She has published among others in Éigse: A Journal of Irish
Studies and Journal of Celtic Linguistics.
Germana Olga Civilleri received her PhD in Linguistics at Roma Tre

University with a dissertation on deverbal nouns in Ancient Greek, pub-
lished in 2012 as the 38th volume of the series ‘Studies in Indo-European
Linguistics’ (Munich: LINCOM). Her research topics are Classical lan-
guages, historical linguistics, case systems, the noun–verb continuum,
deverbal nouns, and word formation and the lexicon. Among her major
publications, she is contributing to the forthcoming Encyclopedia of Ancient
Greek Language and Linguistics (Leiden: Brill) by writing two entries about
abstract nouns and mass/collective nouns.
Angeliki Efthymiou is a tenured Assistant Professor of Linguistics

at Democritus University of Thrace, Greece. She studied Greek at
the University of Thessaloniki, Greece, and specialized in Linguistics
(MA and PhD) at the University of Lille III, France. Her research and
teaching interests fall within the areas of lexicology, lexical semantics,
morphology, lexicography, educational linguistics and language teaching.
notes on contributors 287
She co-authored with Georgia Fragaki and Angelos Markos the article
‘Productivity of verb forming suffixes in Modern Greek: a corpus-based
study’ in Morphology (2012). She is also the author of Le suffixe -iá en
grec moderne. La manifestation d’un degré maximal d’anthropocentricité
(Saarbrücken: Éditions Universitaires Européenes, 2012).
Jessica Forse recently completed her PhD in Translation at Swansea

University. The focus of her research is to examine to what extent
Jackendoff’s theory of lexical semantics, Conceptual Structure, can be
applied to the semantics of word formation processes, with an emphasis on
Romance languages, particularly French and Spanish.
Pius ten Hacken is a Professor at the Institut für Translationswissenschaft

of Innsbruck University. Until recently he was at Swansea University.
His research interests include morphology, terminology, and the phi-
losophy and history of linguistics. He is the author of Defining Morphology
(Hildesheim: Olms, 1994) and of Chomskyan Linguistics and Its Competitors
(London: Equinox, 2007), and the editor of Terminology, Computing and
Translation (Tübingen: Narr, 2006). He has published in various journals
including the International Journal of Lexicography, International Journal
for the Semiotics of the Law and Linguistische Berichte.
Ewelina Kwiatek is a Research Assistant in the English Department at

the Pedagogical University of Krakow, Poland. She completed her PhD
studies in Translation at Swansea University in 2012. Her research inter-
ests include terminology, specialised translation, corpus linguistics and
CAT tools. She is the author of Contrastive Analysis of English and Polish
Surveying Terminology (Newcastle: CSP, 2013).
Renáta Panocová currently teaches morphological theory, comparative

morphology, intercultural communication and speech communication at
the Department of British and US Studies, Faculty of Arts, P. J. Šafárik
University in Košice, Slovakia. She received her Master’s degree in
English language and literature – Russian language and literature from
P. J. Šafárik University in Prešov. She continued her studies to earn a PhD
degree in linguistics, specializing in Slavistics and Slavic languages, at
Prešov University in Slovakia. Her research interests include morphology,
terminology and communication.
Kaarina Pitkänen-Heikkilä works as a post-doctoral researcher in the

Department of Finnish, Finno-Ugrian and Scandinavian Studies at the
University of Helsinki, on the Bank of Finnish Terminology in Arts and
288 semantics of word formation and lexicalization
Sciences, a project for creating a virtual research infrastructure of Finnish

scholarly terms. Her research interests include the morphology, terminol-
ogy and history of written Finnish. Her doctoral thesis in 2008 dealt with
the development of botanical terminology in Finnish. In addition, she has
written around ten scientific articles on Finnish vocabulary and terminol-
ogy, mostly in Finnish. She has also studied non-fiction translation pro-
cesses in nineteenth-century Finnish.
Graça Rio-Torto is full professor at the University of Coimbra. Her PhD

in Linguistics (University of Coimbra, 1993) focuses on word formation,
semantics and the morphology of evaluatives. Her habilitation analyses the
interface between the lexicon and grammar. The main area of her research
is the morphology and semantics of Portuguese word f ormation. She is the
author of the monographs Morfologia derivacional (Porto Editora, 1998)
and Verbos e nomes (Coimbra: Almedina, 2004), as well as several articles
(e.g. ‘Portuguese Compounds’, in the Probus 2012 special issue on Romance
compounds), and is the editor of the Léxico de la Ciencia: tradición y moder-
nidad (Munich: Lincom, 2012) and the Derivational Grammar of Portuguese
Language (Coimbra: Coimbra University Press, in press).
Maria Rosenberg received her PhD in French linguistics at Stockholm

University, Sweden, in 2008, after which she had research funding for
four years. Since 2012 she has been assistant professor in languages and
language didactics at Umeå University, Northern Sweden. Her main fields
of interest include morphology and lexical semantics, as well as contras-
tive and corpus linguistics. She has published various articles on word
formation, adopting synchronic, diachronic and contrastive perspectives.
Her ongoing research project deals with nominal compounds and construc-
tions, based on a French-Swedish parallel corpus. She is also involved in
research on L1 acquisition of Swedish.
Martin Schäfer received his DPhil from the Universität Leipzig. He

currently holds a position in English Linguistics at the Friedrich-Schiller-
Universität Jena. A major focus of his work is the syntax and semantics
of adjectives and adverbs, be it in adverbial modification or in complex
nominal structures. He is currently completing work on a monograph with
the title Positions and Interpretations: German Adverbial Adjectives at the
Syntax-Semantics Interface.
Barbara Schlücker is a lecturer in linguistics at the department of

German and Dutch at the Freie Universität Berlin. She holds a doctoral
degree in German linguistics from the Humboldt- Universität Berlin
notes on contributors 289
(2006). In 2012, she received her habilitation degree (German and Dutch
linguistics) from the Freie Universität Berlin. She works on German with a
particular focus on the relationship between German and Dutch. Her spe-
cialist areas are lexical semantics, word formation and grammatical theory.
She has published in various journals including Lingua, Italian Journal of
Linguistics and Linguistische Berichte.
Alexandra Soares Rodrigues is Associate Professor at the Escola

Superior de Educação of the Instituto Politécnico de Bragança. In her
PhD she analyses Portuguese deverbal noun formation. Her main area of
research is Portuguese word formation. She has published Formação de
Substantivos deverbais sufixados em Português (München: Lincom, 2008)
and A construção de postverbais em Português (Porto: Granito Editores, 2001)
as well as the paper ‘Portuguese converted deverbal nouns: constraints on
their bases’, Word Structure, Vol. 2, No. 1 (2009), pp. 69–107.
Claire Thomas completed her PhD in Translation at Swansea University.

In it, she investigates the contributions made by Jackendoff’s Conceptual
Structure, Pustejovsky’s Generative Lexicon and Lieber’s theory of
Morphology and Lexical Semantics to the characterization of polysemy in
French and English deverbal nouns.
Bibliography
Abney, Steven Paul (1987) The English Noun Phrase in Its Sentential Aspect. PhD dissertation,
Massachusetts Institute of Technology.
Adams, Valerie (2001) Complex Words in English. London: Longman.
Alexiadou, Artemis (2009) ‘On the role of syntactic locality in morphological processes: the
case of (Greek) derived nominals’, in A. Giannakidou and M. Rathert (eds), Quantification,
Definiteness and Nominalization. Oxford: Oxford University Press, pp. 253–80.
Alexiadou, Artemis and Anagnostopoulou, Elena (2004) ‘Voice morphology in the causative-
inchoative alternation: evidence for a non- unified structural analysis of unaccusatives’,
in A. Alexiadou, E. Anagnostopoulou and M. Everaert (eds), The Unaccusativity Puzzle.
Oxford: Oxford University Press, pp. 114–36.
Alexiadou, Artemis and Grimshaw, Jane (2008) ‘Verbs, nouns and affixation’, in SINSPEC
(Working Papers of the SFB 732), 1: 1–16.
Alexiadou, Artemis, Anagnostoloulou, Elena and Schäfer, Florian (2006) ‘The properties of
anticausatives crosslinguistically’, in Mara Frascarelli (ed.), Phases of Interpretation, Studies
in Generative Grammar 91. Berlin and New York: Mouton de Gruyter, pp. 187–211.
Allen, Margaret Reece (1978) Morphological Investigations. PhD dissertation, University of
Connecticut.
Amiot, Dany (2005) ‘Between compounding and derivation. Elements of word-formation cor-
responding to prepositions’, in W. U. Dressler, D. Kastovsky, O. E. Pfeiffer and F. Rainer
(eds), Morphology and Its Demarcations. Amsterdam and Philadelphia: John Benjamins,
pp. 183–95.
Anastassiadis-Symeonidis, Anna (1997) ‘On Modern Greek denominal adjectives’, in G. Booij,
A. Ralli and S. Scalise (eds), Proceedings of the 1st Mediterranean Morphology Meeting,
University of Patras, pp. 29–40.
Anastassiadis-Symeonidis, Anna (2002) Αντίστροφο Λεξικό της Νέας Ελληνικής [Reverse
Dictionary of Modern Greek]. Thessaloniki: Institouto Neoellinikon Spoudon.
Anastassiadis-Symeonidis, Anna and Fliatouras, Asimakis (2003) Η διάκριση [ΛΟΓΙΟ] και
[ΛΑΪΚΟ] στην Ελληνική γλώσσα. Ορισμός και ταξινόμηση [‘The features [learned] and
[–learned] in Greek: definition and classification’], in Proceedings of the 6th International
Conference on Greek Linguistics, University of Crete, http://www.philology.uoc.gr/
conferences/6thICGL/gr.htm.
Anderson, Stephen R. (1992) A-Morphous Morphology. Cambridge: Cambridge University
Press.
bibliography 291
Arntz, Reiner, Picht, Heribert and Mayer, Felix (2009) Einführung in die Terminologiearbeit, 6th
edn. Hildesheim: Olms.
Aronoff, Mark (1976) Word Formation in Generative Grammar. Cambridge, MA: MIT Press.
Aronoff, Mark (1980) ‘The relevance of productivity in a synchronic description of word
formation’, in J. Fisiak (ed.), Historical Morphology. The Hague: Mouton de Gruyter,
pp. 71–82.
Aronoff, Mark (1984) ‘Word formation and lexical semantics’, Quaderni di Semantica, 5: 45–9.
Aronoff, Mark (1992) ‘Stems in Latin verbal morphology’, in M. Aronoff (ed.), Morphology
Now. Albany, NY: State University of New York Press, pp. 5–32.
Aronoff, Mark (1994) Morphology by Itself. Stems and Inflectional Classes. Cambridge, MA: MIT
Press.
Aronoff, Mark and Cho, Sungeun (2001) ‘The semantics of -ship suffixation’, Linguistic Inquiry,
23: 167–73.
Aronoff, Mark and Fudeman, Kirsten (2005) What is Morphology? Oxford: Blackwell.
Baayen, Harald (2008) ‘Corpus linguistics in morphology: morphological productivity’, in
A. Lüdeling and M. Kyto (eds), Corpus Linguistics. An International Handbook. Berlin:
Mouton de Gruyter, pp. 899–919.
Bannister, Arthur, Raymond, Stanley and Baker, Raymond (1998) Surveying, 7th edn. Harlow:
Prentice Hall.
Barner, David and Bale, Alan (2002) ‘No nouns, no verbs: psycholinguistic arguments in favor
of lexical underspecification’, Lingua, 112: 771–91.
Baroni, Marco, Guevara, Emiliano and Pirreli, Vito (2007) ‘NN compounds in Italian: modeling
category induction and analogical extension’, Lingue e Linguaggio, VI (2): 263–90.
Baroni, Marco, Matiasek, Johannes and Trost, Harald (2002) ‘Predicting the components of
German nominal compounds’, in Proceedings of the 15th European Conference on Artificial
Intelligence, ECAI’2002, pp. 470–4.
Bauer, Laurie (1983) English Word Formation. Cambridge: Cambridge University Press.
Bauer, Laurie (1988) Introducing Linguistic Morphology. Edinburgh: Edinburgh University
Press.
Bauer, Laurie (1990) ‘Be-heading the word’, Journal of Linguistics, 26 (1): 1–31.
Bauer, Laurie (1997) ‘Evaluative morphology: in search of universals’, Studies in Language, 21:
533–75.
Bauer, Laurie (1998) ‘When is a sequence of two nouns a compound in English?’, English
Language and Linguistics, 2: 65–86.
Bauer, Laurie (2001) Morphological Productivity. Cambridge: Cambridge University Press.
Bauer, Laurie (2003) Introducing linguistic morphology, 2nd edn. Edinburgh: Edinburgh
University Press.
Bauer, Laurie and Valera, Salvador (eds) (2005) Approaches to Conversion/Zero-Derivation.
Münster: Waxmann Verlag.
Beard, Robert (1981) The Indo-European Lexicon: A Full Synchronic Theory. Amsterdam: North
Holland.
Beard, Robert (1982) ‘Plural as a lexical derivation’, Glossa, 16: 133–48.
Beard, Robert (1984) ‘Generative lexicalism’, Quaderni di Semantica, 5: 50–7.
Beard, Robert (1987) ‘Lexical stock expansion’, in E. Gussmann (ed.), Rules and the Lexicon:
Studies in Word Formation. Lublin: Redakcja Wydawnictw KUL, pp. 24–41.
Beard, Robert (1995) Lexeme Morpheme Base Morphology. Albany, NY: State University of New
York Press.
Beard, Robert (1998) ‘Derivation’, in A. Spencer and A. M. Zwicky (eds), The Handbook of
Morphology. Oxford: Blackwell, pp. 44–65.
Bensaude-Vincent, Bernadette and Stengers, Isabelle (1995) Histoire de la chimie. Paris: La

Découverte.
Benveniste, Émile ([1948] 1975) Noms d’agent et noms d’action en Indo-Européen. Paris: Librairie
d’Amérique et d’Orient.
Benveniste, Émile (1966) ‘Formes nouvelles de la composition nominale’, Bulletin de la Société
de linguistique de Paris, 61: 82–95.
Berman, Ruth A. and Clark, Eve V. (1989) ‘Learning to use compounds for contrast: data from
Hebrew’, First Language, 9: 247–70.
Bielec, Dana (1998) Polish: An Essential Grammar. London: Routledge.
Bisetto, Antonietta and Scalise, Sergio (2005) ‘The classification of compounds’, Lingue e
linguaggio, 2: 319–30.
Bloch, Bernard and Trager, George L. (1942) Outline of Linguistic Analysis. Baltimore, MD:
Linguistic Society of America.
Bloch-Trojnar, Maria (2006) Polyfunctionality in Morphology: A Study of Verbal Nouns in
Modern Irish. Lublin: Wydawnictwo KUL.
Bloch-Trojnar, Maria (2007) ‘Even more remarks on nominalization’, in P. Stalmaszczyk
and I. Witczak-Plisiecka (eds), PASE Studies in Linguistics. Łódź: Łódź University Press,
pp. 47–56.
Bloch-Trojnar, Maria (2011) ‘A morphologist’s perspective on “event structure theory” of
nominalizations’, in M. Pawlak and J. Bielak (eds), New Perspectives in Language, Discourse
and Translation Studies. Berlin: Springer, pp. 61–76.
Bloch-Trojnar, Maria (2012a) ‘The category of nominal number in English and the
inflection-derivation distinction’, in E. Cyran, B. Szymanek and H. Kardela (eds), Sound,
Structure and Sense: Studies in Memory of Edmund Gussmann. Lublin: Wydawnictwo KUL,
pp. 29–53.
Bloch-Trojnar, Maria (2012b) ‘The interaction of aspectuality and valency with countability
and morphological marking in the process of verb to noun transposition’, in M. Bloch-
Trojnar and A. Bloch-Rozmej (eds), Modules and Interfaces. Lublin: Wydawnictwo KUL.
Bloomfield, Leonard (1933) Language. New York: Henry Holt.
Bolinger, Dwight (1967) ‘Adjectives in English: attribution and predication’, Lingua, 18: 1–34.
Bolinger, Dwight (1975) Aspects of Language, 2nd edn. New York: Harcourt Brace
Jovanovich.
Booij, Geert (2005) The Grammar of Words. Oxford: Oxford University Press.
Booij, Geert (2007) The Grammar of Words, 2nd edn. Oxford: Oxford University Press.
Booij, Geert (2009a) ‘Phrasal names: a constructionist analysis’, Word Structure, 2: 219–40.
Booij, Geert (2009b) ‘Compounding and construction morphology’, in R. Lieber and P. Štekauer
(eds), The Oxford Handbook of Compounding. Oxford: Oxford University Press, pp. 201–16.
Booij, Geert and van Kemenade, Jaap (2003) ‘Preverb: an introduction’, in G. Booij and J. van
Kemenade (eds), Yearbook of Morphology 2003. Dordrecht: Kluwer, pp. 1–11.
Bosch, Peter (1983) Agreement and Anaphora – A Study of the Roles of Pronouns in Discourse and
Syntax, Cognitive Science Series. London and New York: Academic Press.
Bratus, B. V. (1969) The Formation and Expressive Use of Diminutives in Russian. Cambridge:
Cambridge University Press.
Brinton, Laurel J. (1998) ‘Aspectuality and countability: a cross-categorial analogy’, English
Language and Linguistics, 2: 37–63.
Brinton, Laurel J. and Traugott, Elizabeth Closs (2005) Lexicalization and Language Change.
Cambridge: Cambridge University Press.
Browne, Wayles (1974) ‘On the topology of anaphoric peninsulas’, Linguistic Inquiry, 5 (4):
619–20.
bibliography 293
Bruno, Carla (2000) ‘Fra nome e verbo: osservazioni sulla sintassi dei nomi in -σις’, Studi e Saggi
Linguistici, 38: 129–67.
Bücking, Sebastian (2009) ‘How do phrasal and lexical modification differ? Contrasting
adjective-noun combinations in German’, Word Structure, 2: 184–204.
Bücking, Sebastian (2010) ‘German nominal compounds as underspecified names for kinds’, in
S. Olsen (ed.), New Impulses in Word-Formation. Hamburg: Buske, pp. 253–81 (= Linguistische
Berichte Sonderheft, 17)
Bybee, Joan (1985) Morphology: A Study of the Relation Between Meaning and Form. Amsterdam:
John Benjamins.
Bybee, Joan (2003) ‘Mechanisms of change in grammaticization: the role of frequency’, in
B. D. Joseph and R. D. Janda (eds), The Handbook of Historical Linguistics. Oxford: Blackwell,
pp. 602–23.
Cabré, M. Teresa (1999) Terminology: Theory, Methods and Applications, trans. Janet Ann
DeCesaris, ed. Juan C. Sager. Amsterdam: John Benjamins.
Cadiot, Pierre (1993) ‘À entre deux noms: vers la composition nominale’, Lexique, 11:
193–240.
Carlson, Greg (1977) Reference to Kinds in English. Amherst, MA, PhD thesis, University of
Massachusetts.
Carstairs-McCarthy, Andrew D. (1992) Current Morphology. London: Routledge.
Cetnarowska, Bożena (1993) The Syntax, Semantics and Derivation of Bare Nominalisations in
English. Katowice: Uniwersytet Śląski.
Chantraine, Pierre ([1933] 1979) La formation des noms en grec ancien. Paris: Klincksieck.
Chantraine, Pierre ([1968–70] 1999) Dictionnaire étymologique de la langue grecque. Paris:
Klincksieck.
Charitonidis, Chariton (2005) Verb Derivation in Modern Greek. Frankfurt: Peter Lang.
Charitonidis, Chariton (2011) Making Verbs Happen: Interviews on Greek Verb Endings. Munich:
Lincom Europa.
Chomsky, Noam (1957) Syntactic Structures. The Hague: Mouton.
Chomsky, Noam (1964) Current Issues in Linguistic Theory. The Hague: Mouton.
Chomsky, Noam (1965) Aspects of the Theory of Syntax. Cambridge, MA: MIT Press.
Chomsky, Noam (1970) ‘Remarks on nominalization’, in R. A. Jacobs, and P. S. Rosenbaum
(eds), Readings in English Transformational Grammar. Waltham, MA: Ginn, pp. 184–221;
repr. in Chomsky (1972: 11–61).
Chomsky, Noam (1972) Studies in Semantics in Generative Grammar. The Hague: Mouton.
Chomsky, Noam (1980) Rules and Representations. New York: Columbia University Press.
Chomsky, Noam (1981) Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam (1986) Knowledge of Language: Its Nature, Origin, and Use. Westport, CT:
Praeger.
Chomsky, Noam (2000) ‘Minimalist inquiries: the framework’, in R. Martin, D. Michaels and
J. Uriagereka (eds), Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik.
Cambridge, MA: MIT Press, pp. 89–155.
Chomsky, Noam and Halle, Morris (1968) The Sound Pattern of English. Cambridge, MA: MIT
Press.
Civilleri, Germana Olga (2009) L’apofonia come strategia di marcatezza nel greco antico: i nomi
deverbali. PhD dissertation, Roma Tre University.
Civilleri, Germana Olga (2010) ‘Deverbal nouns in Ancient Greek: criteria and methodo-
logical remarks for a morpho-semantic analysis’, in Proceedings of the International Conference
‘Lexique, normalisation, transgression’, Cergy Pontoise (Paris), 7 September.
Civilleri, Germana Olga (2012) Continuum nome/verbo e nominalizzazioni morfologiche: i nomi
deverbali in greco antico, LINCOM Studies in Indo-European Linguistics 38. Munich:

LINCOM Europa.
Civilleri, Germana Olga (submitted) ‘Nomi di processo in greco antico’, Archivio Glottologico
Italiano.
Clark, Eve V. and Berman, Ruth A. (1987) ‘Types of linguistic knowledge: interpreting and
producing compound nouns’, Journal of Child Language, 14: 547–67.
Cobuild (2009) Collins Cobuild Advanced Dictionary of English. Glasgow: HarperCollins.
Collins (1986) Collins Dictionary of the English Language, 2nd edn, editor in chief Patrick Hanks.
Glasgow: Collins.
Collins Cobuild English Dictionary (1995) London: HarperCollins.
Corbett, Greville G. (1999) ‘Prototypical inflection: implications for typology’. in G. Booij and
J. van Marle (eds), Yearbook of Morphology 1998. Dordrecht: Kluwer, pp. 1–22.
Corbett, Greville G. (2000) Number. Cambridge: Cambridge University Press.
Corbin, Danielle (1987) Morphologie dérivationnelle et structuration du lexique, 2 vols. Tübingen:
Niemeyer.
Corbin, Danielle (1992) ‘Hypothèses sur les frontières de la composition nominale’, Cahiers de
grammaire, 17: 25–55.
Corum, Claudia (1973) ‘Anaphoric peninsulas’, Chicago Linguistic Society, 9: 89–97.
Craigie, William A. and Onions, C. T. (1933) ‘Historical introduction’, A New English
Dictionary on Historical Principles, Supplement. Oxford: Clarendon Press, pp. vii–xx.
Croft, William ([1990] 2003) Typology and Universals, Cambridge Textbooks in Linguistics.
Croft, William (1991) Syntactic Categories and Grammatical Relations: The Cognitive Organization
of Information. Chicago: University of Chicago Press.
Croft, William (2000) ‘Part of speech as language universals and as language-particular catego-
ries’, in P. M. Vogel and B. Comrie (eds), Approaches to the Typology of Word Classes. Berlin:
Cruse, Alan D. (1986) Lexical Semantics. Cambridge: Cambridge University Press.
Cruse, Alan D. (2004) Meaning in Language. An Introduction to Semantics and Pragmatics.
Oxford: Oxford University Press.
Culicover, Peter W. and Jackendoff, Ray (2005) Simpler Syntax. Oxford: Oxford University
Press.
Cuzzolin, Pierluigi (1995) ‘A proposito di sub vos placo e della grammaticalizzazione delle
adposizioni’, Archivio Glottologico Italiano, 80: 122–43.
Cysouw, Michael and Wälchli, Bernhard (2007) ‘Parallel texts. Using translational equivalents
in linguistic typology’, STUF – Language Typology and Universals, 60 (2): 95–9.
Dahl, Östen (2004) The Growth and Maintenance of Linguistic Complexity. Amsterdam and
Philadelphia: John Benjamins.
Daľ, Vladimir Ivanovič (2005) Tolkovyj slovar’ russkogo jazyka: sovremennoje napisanije. Moscow:
Astrel’.
Di Sciullo, Anna Maria and Williams, Edwin (1987) On the Definition of Word. Cambridge, MA:
MIT Press.
Dimmendaal, Gerrit J. (1987) ‘Drift and selective mechanisms in morphological changes: the
Eastern Nilotic case’, in A. Giacalone-Ramat, O. Carruba and G. Bernini (eds), Papers from
the 7th International Conference on Historical Linguistics. Amsterdam: Benjamins, pp. 193–210.
Dokulil, Miloš (1962) Tvoření slov v češtine I: Teorie odvozování slov. Praha: Nakladatelství
ČSAV.
Downing, Pamela (1977) ‘On the creation and use of English compound nouns’, Language, 53
(4): 810–42.
bibliography 295
Dowty, David R. (1989) ‘On the semantic content of the notion “thematic role”’, in B. Partee,
G. Chierchia and R. Turner (eds), Properties, Types and Meaning, Vol. II. Dordrecht:
Kluwer, pp. 69–139.
Dowty, David R. (1991) ‘Thematic proto-roles and argument selection’, Language, 67 (3):
547–619.
Dressler, Wolfgang U. (1985) ‘On the predictiveness of natural morphology’, Journal of
Linguistics, 21: 321–37.
Dressler, Wolfgang U. (2006) ‘Compound types’, in G. Libben and G. Jarema (eds), The
Representation and Processing of Compound Words. Oxford: Oxford University Press,
pp. 23–44.
Dressler, Wolfgang U. and Barbaresi, Lavinia Merlini (1994) Morphopragmatics. Diminutives
and Intensifiers in Italian, German and Other Languages. Berlin: Mouton de Gruyter.
Drodz, L. and Seibicke, W. (1973) Deutsche Fach-und Wissenschaftssprache: Bestandsaufnahme,
Theorie, Geschihte. Wiesbaden: Brandstetter.
Εfthymiou, Angeliki (1999) Le suffixe -iá en grec moderne. La manifestation d’un degré maximal
d’anthropocentricité. Lille: Presses Universitaires du Septentrion; 2nd edn (2012) Saarbrücken:
Éditions Universitaires Européenes.
Efthymiou, Angeliki (2001) ‘Το νεοελληνικό πρόθημα ξε-: οι έννοιες της απομάκρυνσης και
της αλλαγής κατάστασης’ [‘The Modern Greek prefix kse-. The concepts of move away and
change of state’], Studies in Greek Linguistics, 2000, pp. 202–13.
Efthymiou, Angeliki (2002) ‘Σημασιολογικές παρατηρήσεις για τα νεοελληνικά προθήματα
ξε-, εκ-, από-’ [‘Semantic considerations on the Modern Greek Prefixes kse-, ek-, apo-’],
Studies in Greek Linguistics, 2001, pp. 199–209.
Efthymiou, Angeliki (2011a) ‘The Semantics of Verb Forming Suffixes in Modern Greek’,
Proceedings of the 19th International Symposium of Theoretical and Applied Linguistics, 3–5
April 2009, School of English, Aristotle University of Thessaloniki, pp. 174–84.
Efthymiou, Angeliki (2011b) ‘Are there any principles governing the order and combination of
affixes in Modern Greek parasynthetic verbs?’, paper read at the 3rd Vienna Workshop on
Affix Order, Vienna, 15–16 January.
Efthymiou, Angeliki (2013) ‘On the interaction between semantics and phonetic iconicity in
evaluative morphology’, SKASE Journal of Theoretical Linguistics 10: 152–66.
Efthymiou, Angeliki, Fragaki, Georgia and Markos, Angelos (2012a) ‘Productivity of verb
forming suffixes in Modern Greek: a corpus-based study’, Morphology, 22/4: 515–43.
Efthymiou, Angeliki, Fragaki, Georgia and Markos, Angelos (2012b) ‘Exploring the produc-
tivity of -pió in Modern Greek: a corpus-based study’, in Z. Gavrilidou, A. Efthymiou, E.
Thomadaki and P. Kambaki-Vougioukli (eds), Selected Papers from the 10th international
Conference of Greek Linguistics, Komotini, 1–4 September, Democritus University of Thrace
http://www.icgl.gr, pp. 243–54.
Efthymiou, Angeliki, Gavanozi, Vaso and Havou, Eleni (2010) ‘The frequency of Modern
Greek suffixes in the primary school textbooks’, Ζητήματα διδακτικής της γλώσσας [Issues of
Language Teaching], Democritus University of Thrace, Komotini, 7–8 May.
Eisenberg, Peter (2002) ‘Struktur und Akzent komplexer Komposita’, in D. Restle and
D. Zaefferer (eds), Sounds and Systems. Studies in Structure and Change. A Festschrift for Theo
Vennemann. Berlin and New York: De Gruyter, pp. 349–65.
Embick, David and Noyer, Rolf (2007) ‘Distributed morphology and the syntax-morphology
interface’, in G. Ramchand and C. Reiss (eds), The Oxford Handbook of Linguistic Interfaces.
Encyclopedia Britannica (1986) The New Encyclopaedia Britannica in 32 volumes, eds P. W. Goetz
and M. Sutton. London: Encyclopaedia Britannica.
Engel, Ulrich (1996) Deutsche Grammatik. Heidelberg: Groos.

Eriksson, Olof (1997) Språk i kontrast. En jämförande studie av svensk och fransk meningsstruktur.
Göteborg: Akademiförlaget.
Evola, Vito (2008) ‘La metafora come carrefour cognitivo del pensiero e del linguaggio’, in
C. Casadio (ed.), Vie della metafora: linguistica, filosofia, psicologia. Chieti: Editore Prime
Vie-Sulmona.
Fahim Elsayed, Mohammed Salah (1977) Untersuchungen zum Modell substantivischer Komposita
mit einem Primäradjektiv als erster unmittelbarer Konstituente. Dissertation, Leipzig.
Felber, Helmut (1984) Terminology Manual. Paris: UNESCO and Wien: Infoterm.
Fillmore, Charles J. (1977a) ‘Scenes-and-frames semantics’, in A. Zampolli (ed.), Linguistic
Structures – Studies in Linguistics, Computational Linguistics, and Artificial Intelligence. New
York: North-Holland, pp. 55–82.
Fillmore, Charles J. (1977b) ‘The case for case reopened’, in P. Cole and J. Sadock (eds), Syntax
and Semantics 8: Grammatical Relations. New York: Academic Press, pp. 59–81.
Fradin, Bernard (2003) Nouvelles approches en morphologie. Paris: Presses Universitaires de
France.
Fradin, Bernard (2005) ‘On a semantically grounded difference between derivation and
compounding’, in W. U. Dressler, D. Kastovsky and F. Rainer (eds), Morphology and its
Demarcations. Amsterdam and Philadelphia: John Benjamins; online at http://www.llf.cnrs.
fr/Gens/Fradin/FRA_3-double.pdf.
Fradin, Bernard (2009) ‘IE, Romance: French’, in R. Lieber and P. Štekauer (eds), The Oxford
Handbook of Compounding. Oxford: Oxford University Press, pp. 417–35.
Fruyt, Michèle (2000) ‘La création lexicale: généralités appliquées au domaine latin’, in
C. Nicolas, and M. Fruyt (eds), La création lexicale en latin, Table ronde sur ‘La création
lexicale’ organisée par M. Fruyt au 9ème Coll. intern. de Linguistique latine, Madrid, avril
1997. Paris: Presses de l’Université Paris-Sorbonne, pp. 11–48.
Gaeta, Livio (2002) Quando i verbi compaiono come nomi. Pavia: Franco Angeli.
Gaeta, Livio and Ricca, Davide (2003) ‘Frequency and productivity in Italian derivation: a com-
parison between corpus-based and lexicographical data’, Rivista di Linguistica, 15 (1): 63–98.
Gaeta, Livio and Ricca, Davide (2009) ‘Composita solvantur: compounds as lexical units or
morphological objects?’, in L. Gaeta and M. Grossmann (eds), Compounds Between Syntax
and Lexicon, Special Issue of Italian Journal of Linguistics / Rivista di Linguistica, 21: 35–70.
Gagné, Christina L. and Shoben, Edward J. (1997) ‘Influence of thematic relations on the com-
prehension of modifier-noun combinations’, Journal of Experimental Psychology: Learning,
Memory, and Cognition, 23 (1): 71–87.
Gagné, Christina L. and Spalding, Thomas L. (2006) ‘Conceptual combination’, in G. Libben
and G. Jarema (eds), The Representation and Processing of Compound Words. Oxford: Oxford
University Press, pp. 145–68.
Giegerich, Heinz J. (2004) ‘Compound or phrase? English noun-plus-noun constructions and
the stress criterion’, English Language and Linguistics, 8: 1–24.
Giegerich, Heinz J. (2005) ‘Associative adjectives and the lexicon-syntax interface’, Journal of
Giegerich, Heinz J. (2006) ‘Attribution in English and the distinction between phrases and
compounds’, in P. Rösel (ed.), Englisch in Zeit und Raum – English in Time and Space:
Forschungsbericht für Klaus Faiss. Trier: Wissenschaftlicher Verlag Trier, pp. 10–27.
Givón, Talmy (1979) On Understanding Grammar. New York: Academic Press.
Givón, Talmy (2001) Syntax, Vol. 2. Amsterdam: John Benjamins.
Gottfurcht, Carolyn (2008) Denominal Verb Formation in English. PhD dissertation,
Northwestern University, Evanston, Illinois.
bibliography 297
Grandi, Nicola (2011) ‘Renewal and innovation in the emergence of Indo-European evaluative
morphology’. Lexis, 6: 5–25.Online at: http://lexis.univ-lyon3.fr/IMG/pdf/Lexis_6.pdf.
Grimshaw, Jane (1990) Argument Structure. Cambridge, MA: MIT Press.
Gross, Gaston (1994) ‘Classes d’objets et description des verbes’, Langages, 115 (28): 15–30.
Gruber, Jeffrey S. (1965) Studies in Lexical Relations. PhD dissertation, MIT.
Grzega, Joachim (2009) ‘Compounding from an onomasiological perspective’, in R. Lieber and
P. Štekauer (eds), The Oxford Handbook of Compounding. Oxford: Oxford University Press,
pp. 217–32.
Grzegorczykowa, Renata (1979) Zarys Słowotwórstwa Polskiego. Słowotwórstwo Opisowe.
Warszawa: PWN.
Grzegorczykowa, Renata and Puzynina, Jadwiga (1979) Słowotwórstwo Współczesnego Języka
Polskiego. Rzeczowniki Sufiksalne Rodzime. Warszawa: PWN.
Grzegorczykowa, Renata, Laskowski, Roman and Wróbel, Henryk (eds) (1999) Gramatyka
współczesnego języka polskiego: Morfologia. Warszawa: PWN.
Guevara, Emiliano and Scalise, Sergio (2009) ‘Searching for universals in compounding’,
in S. Scalise and A. Bisetto (eds), Universals of Language Today. Dordrecht: Springer,
pp. 101–28.
Gunkel, Lutz and Zifonun, Gisela (2009) ‘Classifying modifiers in common names’, Word
Structure, 2: 205–18.
Gunkel, Lutz and Zifonun, Gisela (2011) ‘Klassifikatorische Modifikation im Deutschen
und Französischen’, in E. Lavric, W. Pöckl and F. Schallhart (eds), Comparatio delectat:
Akten der VI. Internationalen Arbeitstagung zum romanisch-deutschen und innerromanischen
Sprachvergleich: Innsbruck, 3.–5. September 2008. Frankfurt am Main: Lang, pp. 549–62.
Habert, Benoît and Jacquemin, Christian (1993) ‘Noms composés, termes, dénominations com-
plexes: problématiques linguistiques et traitements automatiques’, Traitement Automatique
des Langues, 2: 5–41.
ten Hacken, Pius (1994) Defining Morphology: A Principled Approach to Determining the
Boundaries of Compounding, Derivation, and Inflection. Hildesheim: Olms.
ten Hacken, Pius (1999) ‘Motivated tests for compounding’, Acta Linguistica Hafniensia, 31:
27–58.
ten Hacken, Pius (2003a) ‘Phrasal elements as parts of words’, in E. Hajičová, A. Kotěšovcová
and J. Mírovský (eds), Proceedings of CIL17, CD-ROM. Praha: Matfyzpress, MFF UK.
ten Hacken, Pius (2003b) ‘Phrases in words’, in C. Tschichold (ed.), English Core Linguistics.
Bern: Lang, pp. 185–203.
ten Hacken, Pius (2007) Chomskyan Linguistics and Its Competitors. London: Equinox.
ten Hacken, Pius (2008) ‘Prototypes and discreteness in terminology’, in E. Bernal and
J. DeCesaris (eds), Proceedings of the XIII Euralex International Congress, IULA-UPF,
Barcelona, pp. 979–87.
ten Hacken, Pius (2009a) ‘Early generative approaches’, in R. Lieber and P. Štekauer (eds), The
Oxford Handbook of Compounding. Oxford: Oxford University Press, pp. 54–77.
ten Hacken, Pius (2009b) ‘What is a dictionary? A view from Chomskyan linguistics’,
International Journal of Lexicography, 22: 399–421.
ten Hacken, Pius (2010a) ‘The tension between definition and reality in terminology’, in
A. Dykstra and T. Schoonheim (eds), Proceedings of the XIV Euralex International Congress.
Ljouwert: Fryske Akademy/Afuk, pp. 915–27.
ten Hacken, Pius (2010b) ‘Synthetic and exocentric compounds in a parallel architecture’,
Linguistische Berichte Sonderheft, 17: 233–51.
ten Hacken, Pius (2010c) ‘Creating legal terms: a linguistic perspective’, International Journal
for the Semiotics of the Law, 23: 407–25.
ten Hacken, Pius and Panocová, Renáta (2011) ‘Individual and social aspects of word forma-
tion’, Kwartalnik Neofilologiczny, 58: 283–300.
Hajičová, Eva (1995) ‘Prague school syntax and semantics’, in E. F. K. Koerner and R. E. Asher
(eds), Concise History of the Language Sciences: From the Sumerians to the Cognitivists. Oxford:
Pergamon/Elsevier Science, pp. 253–62.
Halle, Morris and Marantz, Alec (1993) ‘Distributed morphology and the pieces of inflection’,
in K. Hale and S. J. Keyser (eds), The View from Building 20: Essays in Linguistics in Honor
of Sylvain Bromberger. Cambridge, MA: MIT Press, pp. 111–76; repr. in F. Katamba (ed.)
(2004), Morphology: Critical Concepts in Linguistics (6 vols). London: Routledge, Vol. 1,
pp. 381–440.
Hankamer, Jorge and Sag, Ivan (1976) ‘Deep and surface anaphora’, Linguistic Inquiry, 7:
391–428.
Harley, Heidi (2009) ‘Compounding in distributed morphology’, in R. Lieber and P.
Štekauer (eds), The Oxford Handbook of Compounding. Oxford: Oxford University Press,
pp. 129–44.
Harley, Heidi and Noyer, Rolf (2003) ‘Distributed Morphology’, in L. Cheng and R. Sybesma
(eds), The Second Glot International State-of-the-Article Book: The Latest in Linguistics.
Berlin: Mouton de Gruyter, pp. 462–96.
Harris, Alice C. (2006) ‘Revisiting anaphoric islands’, Language, 82 (1): 114–30.
Harris, Zellig S. (1951) Methods in Structural Linguistics. Chicago: University of Chicago Press;
repr. (1960) as Structural Linguistics.
Haspelmath, Martin (1992) ‘Grammaticization theory and heads in morphology’, in M. Aronoff
(ed.), Morphology Now. Albany, NY: State University of New York Press, pp. 69–82.
Haspelmath, Martin (2002) Understanding Morphology. London: Arnold.
Heine, Bernd, Claudi, Ulrike and Hünnemeyer, Friederike (1991) Grammaticalization: A
Conceptual Framework. Chicago: University of Chicago Press.
Herbst, Thomas (1988) ‘A valency model for nouns in English’, Journal of Linguistics, 24:
265–301.
Heslin, Peter (1999–2007) Diogenes. Computer query system, version 3.1.6.
Heyvaert, Liesbet (2009) ‘Compounding in cognitive linguistics’, in R. Lieber and P. Štekauer
(eds), The Oxford Handbook of Compounding. Oxford: Oxford University Press, pp. 233–54.
Himmelmann, Nikolaus P. (2004) ‘Lexicalization and grammaticalization: opposite or
orthogonal?’, in W. Bisang, N. P. Himmelmann and B. Wiemer (eds), What Makes
Grammaticalization? A Look from Its Fringes and Its Components. Berlin: Mouton de Gruyter,
pp. 21–42.
Hockett, Charles F. (1942) ‘A system of descriptive phonology’, Language, 18: 3–21.
Hockett, Charles F. (1958) A Course in Modern Linguistics. New York: Macmillan.
Hoekstra, Teun and van der Putten, Frans (1988) ‘Inheritance phenomena’, in M. Everaert,
A. Evers, R. Huybregts and M. Trommelen (eds), Morphology and Modularity. In Honour of
Henk Schultink. Dordrecht: Foris, pp. 163–86.
Homer, Iliad, trans. August Taber Murray. London: Loeb Classical Library, 1924.
Homer, Odyssey, trans. August Taber Murray. London: Loeb Classical Library, 1919.
Honeybone, Patrick (2005) ‘N. S. Trubetzkoy’, in S. Chapman and C. Routledge (eds), Key
Thinkers in Linguistics and the Philosophy of Language. Edinburgh: Edinburgh University
Press, pp. 267–8.
Horecký, Ján (1973) ‘Sústava deverbatívnych substantív v spisovnej slovenčine’, Slovenská reč,
38: 264–9.
Horecký, Ján (1974) ‘Sústava desubstantívnych substantív v spisovnej slovenčine’, Slovenská
reč, 39: 129–36.
bibliography 299
Horecký, Ján (1980) ‘Invariantný význam odvodených slov’, Slovo a slovesnost, 41: 81–5.
Horecký, Ján (1983) Vývin a teória jazyka. Bratislava: ŠPN.
Horecký, Ján (1994) Semantics of Derived Words. Prešov: Filozofická fakulta v Prešove
Univerzity P. J. Šafárika v Košiciach.
INS (Institute of Hellenic Studies, Manolis Triandafyllidis Foundation) (1998) Λεξικό
της Κοινής Νεοελληνικής [Standard Modern Greek Dictionary]. Aristotle University of
Thessaloniki.
Isabelle, Pierre (1984) ‘Another look at nominal compounds’, in Proceedings of the 10th
International Conference on Computational Linguistics and the 22nd Annual Meeting of the
Association for Computational Linguistics. Stanford, CA, pp. 509–16.
Jackendoff, Ray (1975) ‘Morphological and semantic regularities in the lexicon’, Language, 51:
639–71.
Jackendoff, Ray (1983) Semantics and Cognition. Cambridge, MA: MIT Press.
Jackendoff, Ray (1990) Semantic Structures. Cambridge, MA: MIT Press.
Jackendoff, Ray (1991) ‘Parts and boundaries’, Cognition, 41: 9–45.
Jackendoff, Ray (1997) The Architecture of the Language Faculty. Cambridge, MA: MIT Press.
Jackendoff, Ray (2002) Foundations of Language: Brain, Meaning, Grammar, Evolution. Oxford:
Oxford University Press.
Jackendoff, Ray (2007) Language, Consciousness, Culture: Essays on Mental Structure. Cambridge,
MA: MIT Press.
Jackendoff, Ray (2008) ‘Construction after construction and its theoretical challenges’,
Language, 84: 8–28.
Jackendoff, Ray (2009) ‘Compounding in the Parallel Architecture and Conceptual Semantics’,
in R. Lieber and P. Štekauer (eds), The Oxford Handbook of Compounding. Oxford: Oxford
University Press, pp. 105–28.
Jackendoff, Ray (2010) Meaning and the Lexicon: The Parallel Architecture 1975–2010. Oxford:
Oxford University Press.
Jacobs, Joachim (1993) ‘Integration’, in M. Reis (ed.), Wortstellung und Informationsstruktur.
Tübingen: Niemeyer, pp. 63–116.
Jacobs, Joachim (1999) ‘Informational Autonomy’, in P. Bosch and R. van der Sandt (eds),
Focus. Linguistic, Cognitive, and Computational Perspectives. Cambridge and New York:
Cambridge University Press, pp. 56–81.
Jakobson, Roman (1959) ‘Boas’ view of grammatical meaning’, in W. Goldschmidt (ed.), The
Anthropology of Franz Boas: Essays on the Centennial of His Birth. Menasha, WI: American
Anthropological Association.
Jakobson, Roman (1960) ‘Closing statement: linguistics and poetics’, in T. A. Sebeok (ed.),
Style in Language. Cambridge, MA: MIT Press, pp. 350–77.
Jarema, Gonia (2006) ‘Compound representation and processing’, in G. Libben and G. Jarema
(eds), The Representation and Processing of Compound Words. Oxford: Oxford University
Press, pp. 45–70.
Jespersen, Otto (1922) A Modern English Grammar on Historical Principles. Heidelberg: Winter.
Jespersen, Otto (1942) A Modern English Grammar. Part VI, Morphology. Copenhagen: Ejnar
Munksgaard.
Johnston, Michael and Busa, Federica (1999) ‘Qualia structure and the compositional interpre-
tation of compounds’, in E. Viegas (ed.), Breadth and Depth of Semantic Lexicons. Dordrecht:
Kluwer, pp. 167–89.
Josefsson, Gunlög (2005) Ord. Lund: Studentlitteratur.
Joseph, John (2002) From Whitney to Chomsky: Essays in the History of American Linguistics.
Amsterdam: Benjamins.
Junczys-Dowmunt, Marcin (2008) ‘Influence of accurate compound noun splitting on bilin-

gual vocabulary extraction’, in Text Resources and Lexical Knowledge. Text, Translation,
Computational Processing (TTCP) 8. Proceedings of Konvens 2008. Berlin: Mouton de Gruyter,
pp. 91–105.
Kačmárová, Alena (2010) ‘A short survey of diminutives in Slovak and English’, in A. Kačmárová
(ed.), English Matters: A Collection of Papers by the Department of English Language and
Literature. Prešov: Prešovská univerzita v Prešove, pp. 17–21.
Kallas, Krystyna (1999) ‘Przymiotnik’, in R. Grzegorczykowa, R. Laskowski and H. Wróbel
(eds), Gramatyka współczesnego języka polskiego: Morfologia. Warszawa: PWN, pp. 469–523.
Kastovsky, Dieter (1982) ‘Word-formation: a functional view’, Folia Linguistica, 16: 181–98.
Kazazis, Kostas (1968) ‘Sunday Greek’, in CLS 4, Department of Linguistics, University of
Chicago, pp. 130–40.
Kempe, Vera, Brooks, Patricia. J., Mironova, Natalija and Fedorova, Olga (2003)
‘Diminutivization supports gender acquisition in Russian children’, Journal of Child
Language, 30: 471–85.
Kerleroux, Françoise (1996) La coupure invisible. Villeneuve d’Ascq: Presses Universitaires du
Septentrion.
Koptjevskaja-Tamm, Maria (1993) Nominalizations. London: Routledge.
Koptjevskaja-Tamm, Maria (2009) ‘Proper-name compounds in Swedish between syntax and
lexicon’, Rivista di Linguistica/Italian Journal of Linguistics, 21: 119–48.
Koskenniemi, Kimmo (1983) Two-Level Morphology: A General Computational Model for Word-
Form Recognition and Production. University of Helsinki, Department of General Linguistics
Publications No. 11.
Krifka, Manfred (1992) ‘Thematic relations as links between nominal reference and temporal
constitution’, in I. Sag and A. Szabolcsi (eds), Lexical Matters. Stanford, CA: Center for the
Study of Language Information, pp. 29–53.
Kuryłowicz, Jerzy (1956) L’apophonie en indo-européen. Wrocław: Polska Akademia NAUK.
Kwiatek, Ewelina (2013) Contrastive Analysis of English and Polish Surveying Terminology.
Newcastle upon Tyne: Cambridge Scholars Press.
Labov, William (1973) ‘The boundaries of words and their meanings’, in C.-J. N. Bailey and
R. W. Shuy (eds), New Ways of Analyzing Variation in English. Washington, DC: Georgetown
University Press, pp. 340–73; repr. in B. Aarts et al. (eds) (2004) Fuzzy Grammar: A Reader.
Lahav, Ran (1989) ‘Against compositionality: the case of adjectives’, Philosophical Studies, 57:
261–79.
Lahav, Ran (1993) ‘The combinatorial-connectionist debate and the pragmatics of adjectives’,
Pragmatics and Cognition, 1: 71–88.
Lakoff, George (1987) Women, Fire, and Dangerous Things: What Categories Reveal about the
Mind. Chicago: University of Chicago Press.
Lakoff, George and Johnson, Mark (1980) Metaphors We Live By. Chicago: University of
Chicago Press.
Lakoff, George and Ross, John (1972) ‘A note on anaphoric islands and causatives’, Linguistic
Inquiry, 3: 121–5.
Lang, Friedrich (1998) ‘Eugen Wüster – his life and work until 1963’, in E. Oeser and
C. Galinski (eds), Eugen Wüster (1898–1977): Leben und Werk – Ein österreichischer Pionier
der Informationsgesellschaft. Vienna: TermNet, pp. 13–26.
Langacker, Ronald W. (1987a) Foundations of Cognitive Grammar, Volume 1: Theoretical
Prerequisites. Stanford, CA: Stanford University Press.
Langacker, Ronald W. (1987b) ‘Nouns and Verbs’, Language, 63 (1): 53–94.
bibliography 301
Langacker, Ronald W. (1991) Foundations of Cognitive Grammar, Volume 2: Descriptive

Application. Stanford, CA: Stanford University Press.
Langacker, Ronald W. (2000a) Grammar and Conceptualization. Berlin: Mouton de Gruyter.
Langacker, Ronald W. (2000b) ‘A dynamic usage-based model’, in M. Barlow and S. Kemmer,
(eds), Usage Based Models of Language. Stanford, CA: CSLI.
Langacker, Ronald W. (2008) Cognitive Grammar: A Basic Introduction. Oxford: Oxford
University Press.
Lapata, Maria (2002) ‘The disambiguation of nominalizations’, Computational Linguistics, 28:
357–88.
Laskowski, Roman (1981) ‘Derywacja słowotwórcza’, in J. Bartmiński (ed.), Pojęcie Derywacji
w Lingwistyce. Lublin: Uniwersytet Marii Curie-Skłodowskiej, Zakład Języka Polskiego,
pp. 107–26.
Lees, Robert B. (1960) The Grammar of English Nominalizations. The Hague: Mouton de
Gruyter.
Lehmann, Christian (1989) ‘Grammatikalisierung und Lexikalisierung’, Zeitschrift für Phonetik,
Sprachwissenschaft und Kommunikationsforschung, 42: 11–19.
Lehmann, Christian (2002) ‘New reflections on grammaticalization and lexicalization’, in
I. Wischer and G. Diewald (eds), New Reflections on Grammaticalization. Amsterdam: John
Benjamins, pp. 1–18.
Lehmann, Christian (2004) ‘Theory and method in grammaticalization’, Zeitschrift für
Germanistische Linguistik, 32 (2): 152–87.
Lehrer, Adrienne (2002) ‘Semantic relations of derivational affixes’, in D. A. Cruse,
F. Hundnurscher, M. Job and P. R. Lutzeier (eds), Lexikologie – Lexicology: Ein interna-
tionales Handbuch zur Natur und Struktur von Wörtern und Wortschätzen – An international
handbook on the nature and structure of words and vocabularies. Berlin: Walter De Gruyter,
pp. 525–32.
Lehrer, Adrienne (2003) ‘Polysemy in derivational affixes’, in B. Nerlich, Z. Todd, D. D. Clarke
and V. Herman (eds), Polysemy: Flexible Patterns of Meaning in Mind and Language. Berlin:
Lesselingue, Chrystèle (2003) ‘Les noms composés [NN]n « holonymiques »: illustration de la
spécificité sémantique des unités construites morphologiquement’, in B. Fradin, G. Dal,
N. Hathout, F. Kerleroux, M. Plénat and M. Roché (eds), Les unités morphologiques, Vol. 3,
Silexicales. Villeneuve d’Ascq: SILEX: CNRS and Université Lille 3, pp. 100–7.
Levi, Judith N. (1975) The Syntax and Semantics of Non-Predicating Adjectives in English. PhD
dissertation, University of Chicago.
Levi, Judith N. (1977) ‘The constituent structure of complex nominals, or, That’s funny, you
don’t look like a noun!’, in W. A. Veach, S. E. Fox and S. Philosoph (eds), Papers from the
Thirteenth Regional Meeting: Chicago Linguistic Society. Chicago: Chicago Linguistic Society,
pp. 325–38.
Levi, Judith N. (1978) The Syntax and Semantics of Complex Nominals. New York: Academic
Press.
Levin, Beth and Rappaport Hovav, Malka (1998) ‘Morphology and lexical semantics’, in
A. Spencer and A. M. Zwicky (eds), The Handbook of Morphology. Oxford: Blackwell,
pp. 248–71.
Li, Charles (1971) Semantics and the Structure of Compounds in Chinese. Berkeley, CA: University
of California dissertation.
Libben, Gary (2006) ‘Why study compound processing? An overview of the issues’, in
G. Libben and G. Jarema (eds), The Representation and Processing of Compound Words.
Libben, Gary, Gibson, Martha, Yoon, Yeo Bom and Sandra, Dominiek (2003) ‘Compound
fracture: the role of semantic transparency and morphological headedness’, Brain and
Liddell, Henry George and Scott, Robert ([1843] 1992) A Greek–English Lexicon. Oxford:
Clarendon Press.
Lieber, Rochelle (1980) On the Organization of the Lexicon. PhD dissertation, Massachusetts
Institute of Technology, distributed by Indiana University Linguistics Club, Bloomington,
IN, 1981.
Lieber, Rochelle (1983) ‘Argument linking and compounds in English’, Linguistic Inquiry, 14:
251–85.
Lieber, Rochelle (2004) Morphology and Lexical Semantics. Cambridge: Cambridge University
Press.
Lieber, Rochelle (2009) ‘A lexical semantic approach to compounding’, in R. Lieber and
P. Štekauer (eds), The Oxford Handbook of Compounding. Oxford: Oxford University Press,
pp. 78–104.
Lieber, Rochelle (2010) Introducing Morphology. Cambridge: Cambridge University Press.
Lieber, Rochelle and Baayen, Harald (1993) ‘Verbal Prefixes in Dutch: A Study in Lexical
Conceptual Structure’, in G. Booij and J. van Marle (eds), Yearbook of Morphology 1993.
Dordrecht: Kluwer Academic, pp. 51–78.
Lieber, Rochelle and Scalise, Sergio (2007) ‘The lexical integrity hypothesis in a new theoretical
universe’, in G. Booij, L. Ducceschi, B. Fradin, E. Guevara, A. Ralli and S. Scalise (eds),
On-line Proceedings of the Fifth Mediterranean Morphology Meeting, Fréjus, 15–18 September
2005, University of Bologna; online at http://www.morbo.lingue.unibo-it/.
Lieber, Rochelle and Štekauer, Pavol (2009) ‘Introduction: status and definition of compound-
ing’, in R. Lieber and P. Štekauer (eds), The Oxford Handbook of Compounding. Oxford:
Oxford University Press, pp. 3–18.
Lönnrot, Elias (1860) Suomen kasvisto. Flora Fennica. Koelma. SKST 24. Helsinki: Finnish
Literature Society.
Luraghi, Silvia (1987) ‘Patterns of case syncretism in Indo-European languages’, in A. Giacalone
Ramat, O. Carruba and G. Bernini (eds), Papers from the VIIth International Conference on
Historical Linguistics. Amsterdam: John Benjamins, pp. 355–71.
Lyons, John (1963) Structural Semantics: An Analysis of Part of the Vocabulary of Plato. Oxford:
Blackwell.
McEnery, Tony, Xiao, Richard and Tono, Yukio (2006) Corpus-Based Language Studies: An
Advanced Resource Book. London and New York: Routledge.
McKoon, Gail, Ward, Gregory, Ratcliff, Roger and Sproat, Richard (1990) M orphosyntactic
and Pragmatic Manipulations of Salience in the Interpretation of Anaphora. Evanston, IL and
Murray Hill, NJ: Northwestern University and AT&T Bell Laboratories, MS.
McKoon, Gail, Ward, Gregory, Ratcliff, Roger and Sproat, Richard (1993) ‘Morphosyntactic
and pragmatic factors affecting the accessibility of discourse entities’, Journal of Memory and
Maienborn, Claudia (2003) ‘Event-internal modifiers: semantic underspecification and con-
ceptual interpretation’, in E. Lang, C. Maienborn and C. Fabricius-Hansen (eds), Modifying
Adjuncts. Berlin: Mouton de Gruyter, pp. 475–509.
Mail, Noor, O’Brien, Peter and Pang, Geordi (2007) ‘Lag correction model and ghosting analy-
sis for an indirect-conversion flat-panel imager’, Journal of Applied Clinical Medical Physics,
8: 137–46.
Malicka-Kleparska, Anna (1985) The Conditional Lexicon in Derivational Morphology. A Study of
Double Motivation in Polish and English. Lublin: Redakcja Wydawnictw KUL.
bibliography 303
Malicka-Kleparska, Anna (1988) Rules and Lexicalisations. Selected English Nominals. Lublin:
Redakcja Wydawnictw KUL.
Malkiel, Yakov (1978) ‘Derivational categories’, in Joseph H. Greenberg et al. (eds), Universals
of Human Language: Word Structure, Vol. 3. Stanford, CA: Stanford University Press,
pp. 125–49.
Marantz, Alec (1997) ‘No escape from syntax: don’t try morphological analysis in the privacy of
your own lexicon’, UPenn Working Papers in Linguistics, 4: 201–25.
Marchand, Hans (1960) The Categories and Types of Present-Day English Word-Formation: A
Synchronic-Diachronic Approach. Munich: Beck.
Marchand, Hans (1969) The Categories and Types of Present-Day English Word Formation: A
Synchronic-Diachronic Approach, 2nd edn. Munich: Beck.
van Marle, Jaap (1985) On the Paradigmatic Dimension of Morphological Creativity. Dordrecht:
Foris.
van Marle, Jaap (1986) ‘The domain hypothesis: the study of rival morphological processes’,
Matthews, Peter H. (1972) Inflectional Morphology. A Theoretical Study Based on Aspects of Latin
Verb Conjugation. Cambridge: Cambridge University Press.
Matthews, Peter H. (1993) Grammatical Theory in the United States from Bloomfield to Chomsky.
Matthews, Peter H. (2005) Oxford Concise Dictionary of Linguistics. Oxford: Oxford University
Press.
Meillet, Antoine and Vendryes, Joseph (1966) Traité de grammaire comparée des langues clas-
siques. Paris: Honoré Champion.
Mela-Athanasopoulou, Elizabeth (2007) ‘The polysemy of -ize derivatives and the Modern
Greek counterpart -pio’, in E. Agathopoulou, M. Dimitrakopoulou and D. Papadopoulou
(eds), Selected Papers on Theoretical and Applied Linguistics, 17th International Symposium.
Thessaloniki: Monochromia, pp. 157–67.
Millikan, Ruth (1984) Language, Thought, and Other Biological Categories. Cambridge, MA:
MIT Press.
Motsch, Wolfgang (2004) Deutsche Wortbildung in Grundzügen. Berlin: De Gruyter.
Mourelatos, Alexander (1978) ‘Events, processes, and states’, Linguistics and Philosophy, 2:
415–34.
Müller, L. P., Suffner, J., Wenda, K., Mohr, W. and Rommens, P. M. (1998) ‘Radiation expo-
sure to the hands and the thyroid of the surgeon during intramedullary nailing’, Injury, 29:
461–8.
Namer, Fiammetta (2005) La Morphologie Constructionnelle du Français et les Propriétés
Sémantiques du Lexique: Traitement Automatique et Modélisation, UMR 7118 ATILF. Nancy:
Université de Nancy 2.
Naumann, Bernd and Vogel, Petra M. (2000) ‘Derivation’, in G. Booij, C. Lehmann and J.
Mugdan (eds), Morphologie – Morphology: Ein Internationales Handbuch zur Flexion und
Wortbildung – An International Handbook on Inflection and Word-Formation (Vol. 1). Berlin:
Walter de Gruyter, pp. 929–43.
Newmeyer, Frederick J. (1986) Linguistic Theory in America, 2nd edn. New York: Academic
Press.
Nicoladis, Elena (2002) ‘When is a preposition a linking element? Bilingual children’s acquisi-
tion of French compound nouns’, Folia Linguistica, 36: 45–63.
Noailly, Michèle (1990) Le substantif épithète. Paris: Presses Universitaires de France.
Nyman, Martti (1981) ‘Paradigms and transderivational constraints: stress and yod in Modern
Greek’, Journal of Linguistics, 17: 231–46.
OED (1989) Oxford English Dictionary, 2nd edn, 12 vols. Oxford: Oxford University Press.
OED (2002) Concise Oxford English Dictionary, 10th edn. Oxford: Oxford University Press.
OED (2011) Oxford English Dictionary. Oxford: Oxford University Press; online at http://
www.oed.com/.
Omero, Odissea, trans. Aurelio Privitera. Milan: A. Mondadori (Fondazione Lorenzo Valla),
1989.
Ortner, Lorelies and Müller-Bollhagen, Elgin (1991) Deutsche Wortbildung. Typen und Tendenzen
in der Gegenwart. Vierter Hauptteil: Substantivkomposita. Berlin and New York: De Gruyter
(= Sprache der Gegenwart 79)
Palmer, Frank R. (1974) The English Verb. London: Longman.
Panagiotidis, Phoevos (2005) ‘Against category-less roots in syntax and word learning: objec-
tions to Barner and Bale (2002)’, Lingua, 115: 1181–94.
Partee, Barbara (1997) ‘Lexical semantics and compositionality’, in D. H. Osherson (ed.),
Language. An Invitation to Cognitive Science, 2nd edn. Cambridge, MA: MIT Press,
pp. 311–60.
Paul, Waltraud (2005) ‘Adjectival modiﬁcation in Mandarin Chinese and related issues’,
Linguistics, 43 (4): 757–93.
Paul, Waltraud (2008) ‘Adjectives in Mandarin Chinese: The Rehabilitation of a Much
Ostracized Category’. Manuscript.
Pearson, Jennifer (1998) Terms in Context. Amsterdam: John Benjamins.
Petropoulou, Evanthia (2009) ‘On the parallel between neoclassical compounds in English and
Modern Greek’, Patras Working Papers in Linguistics, 1: 40–58.
Picht, Heribert and Draskau, Jennifer (1985) Terminology. An Introduction. Guildford:
University of Surrey.
Pitkänen, Kaarina (2005) ‘Suomen kielen rikkaus ja ihanin ominaisuus. Elias Lönnrotin
johto-oppia’ [‘Elias Lönnrot’s views on word formation by derivation’], Virittäjä, 109:
52–82.
Pitkänen, Kaarina (2008) Suomi kasvitieteen kieleksi. Elias Lönnrot termistön kehittäjänä [The
Development of Botanical Finnish: Elias Lönnrot as the Creator of New Terminology],
Publications of the Finnish Literature Society 1162. Helsinki: Finnish Literature Society.
Plag, Ingo (1998) ‘The polysemy of -ize derivatives: on the role of semantics in word forma-
tion’, in G. Booij and J. van Marle (eds), Yearbook of Morphology 1997. Dordrecht: Kluwer,
pp. 219–42.
Plag, Ingo (1999) Morphological Productivity. Structural Constraints in English Derivation. Berlin
and New York: Mouton de Gruyter.
Plag, Ingo (2003) Word-Formation in English. Cambridge: Cambridge University Press.
Plag, Ingo, Dalton-Puffer, Christine and Baayen, Harald (1999) ‘Productivity and register’,
English Language and Linguistics, 3: 209–28.
Pompei, Anna (2010) ‘Space coding in verb-particle constructions and prefixed verbs’, in
G. Marotta (ed.) Atti del convegno ‘Space in Language’. Pisa, 8–10 October 2009.
Postal, Paul M. (1969) ‘Anaphoric islands’, in R. I. Binnick, A. Davison, G. M. Green and
J. L. Morgan (eds), Papers from the Fifth Regional Meeting of the Chicago Linguistic Society.
Chicago: University of Chicago, pp. 205–39.
Pustejovsky, James (1991) ‘The generative lexicon’, Computational Linguistics, 17 (4): 409–41.
Pustejovsky, James (1995a) The Generative Lexicon. Cambridge, MA: MIT Press.
Pustejovsky, James (1995b) ‘Linguistic constraints on type coercion’, in P. Saint-Dizier and
E. Viegas (eds), Computational Lexical Semantics. Cambridge: Cambridge University Press,
pp. 71–97.
Pustejovsky, James and Anick, Peter (1988) ‘On the semantic interpretation of nominals’, in
bibliography 305
Proceedings of the 12th conference on Computational Linguistics, Vol.2. Budapest: Association

for Computational Linguistics, pp. 518–23.
Quirk, Randolf, Greenbaum, Sidney, Leech, Geoffrey and Svartvik, Jan (1985) A Comprehensive
Grammar of the English Language. London and New York: Longman.
Rainer, Franz (1988) ‘Towards a theory of blocking. The case of Italian and German quality
nouns’, in G. Booij and J. van Marle (eds), Yearbook of Morphology 1988. Dordrecht: Foris,
pp. 155–85.
Rainer, Franz and Varela, Soledad (1992) ‘Compounding in Spanish’, Rivista di linguistica, 4:
117–42.
Rappaport Hovav, Malka and Levin, Beth (1992) ‘-Er nominals: implications for the theory of
argument structure’, in T. Stowell and E. Wehrli (eds), Syntax and the Lexicon. San Diego,
CA: Academic Press, pp. 127–53.
Rijkhoff, Jan (2004) The Noun Phrase, revised and extended paperback edition. Oxford: Oxford
University Press.
Rijkhoff, Jan (2008) ‘Descriptive and discourse-referential modifiers in a layered model of the
noun phrase’, Linguistics, 46: 789–829.
Rio-Torto, Graça and Ribeiro, Sílvia (2009) ‘Compounding in Portuguese’, Lingue e Linguaggio,
VIII (2): 271–91.
Rio-Torto, Graça and Ribeiro, Sílvia (2010) ‘Unidades pluriverbais: Processamento e ensino’,
in M. H. Moura Neves (ed.), As interfaces da Gramática. Araraquara, São Paulo: Editora da
UNESP, pp. 151–65.
Rio-Torto, Graça and Ribeiro, Sílvia (2013) ‘Composição’, in G. Rio-Tinto et al. (eds),
Gramática derivacional do Português. Coimbra: Coimbra University Press, pp. 385–431.
Rodrigues, Alexandra (2008) Formação de substantivos deverbais sufixados em português. Munich:
Lincom Europa.
Romagno, Domenica (2004) ‘Ancora su preverbazione e sistemi verbali. Il caso dei preverbi
greci’, Archivio Glottologico Italiano, 89: 165–80.
Rosch, Eleanor (1978) ‘Principles of categorization’, in E. Rosch and B. B. Lloyd (eds),
Cognition and Categorization. Hillside, NJ: Lawrence Erlbaum, pp. 27–48.
Rosenbach, Annette (2006) ‘Descriptive genitives in English: a case study on constructional
gradience’, English Language and Linguistics, 10: 77–118.
Rosenbach, Annette (2007) ‘Emerging variation: determiner genitives and noun modifiers in
English’, English Language and Linguistics, 11: 143–89.
Rosenberg, Maria (2007) ‘Classification, headedness and pluralization: corpus evidence from
French compounds’, Acta Linguistica Hungarica, 54 (3): 341–60.
Ross, John R. (1967) Constraints on Variables in Syntax. Doctoral dissertation, Massachusetts
Institute of Technology.
Rozwadowska, Bożena (1997) Towards a Unified Theory of Nominalizations. External and
Internal Eventualities. Wrocław: Wydawnictwo Uniwersytetu Wrocławskiego.
Russian National Corpus – available at http://ruscorpora.ru/search-main.html#.
Ryder, Mary Ellen (1994) Ordered Chaos: The Interpretation of English Noun-Noun Compounds.
Berkeley, CA: University of California Press.
Rytting, C. Anton (2005) ‘An iota of difference. Attitudes to jod in lexical and social contexts’,
Journal of Greek Linguistics, 6 (1): 145–79.
Sager, Juan (1990) A Practical Course in Terminology Processing. Amsterdam and Philadelphia:
John Benjamins.
Saussure, Ferdinand de (1916) Cours de linguistique générale, eds C. Bally and A. Sechehaye,
édition critique préparée par Tullio de Mauro. Paris: Payot, 1981.
Scalise, Sergio (1986) Generative Morphology, 2nd edn. Dordrecht: Foris.
Scalise, Sergio and Bisetto, Antonietta (2009) ‘The classification of compounds’, in R. Lieber
and P. Štekauer (eds), The Oxford Handbook of Compounding. Oxford: Oxford University
Press, pp. 34–53.
Scalise, Sergio and Vogel, Irene (eds) (2010) Cross- Disciplinary Issues in Compounding.
Amsterdam: John Benjamins.
Schachter, Paul (1976) ‘A non-transformational account of gerundive nominals in English’,
Linguistic Inquiry, 7: 205–41.
Schiller, Anne (2005) ‘German compound analysis with WFSC’, in Finite-State Methods and
Natural Language Processing, 5th International Workshop, FSMNLP, pp. 239–46.
Schreuder, Robert and Baayen, Harald (1997) ‘How complex simplex words can be’, Journal of
Memory and Language, 37: 118–39.
Schwarze, Christoph (2001) ‘Aspetti semantici della formazione delle parole’, in
Sonderforschungsbereich 471, Variation und Entwicklung im Lexicon. Fachbereich
Sprachwissenschaft der Universität Konstanz.
Schwyzer, Eduard (1953) Griechische Grammatik. Munich: G. H. Becksche
Verlagsbuchhandlung.
Selkirk, Elisabeth O. (1982) The Syntax of Words. Cambridge, MA: MIT Press.
Semenza, Carlo and Mondini, Sara (2006) ‘The neuropsychology of compound words’, in
G. Libben and G. Jarema (eds), The Representation and Processing of Compound Words.
Setatos, Michalis (1974) Phonology of Modern Greek Koine. Athens: Papazisis [in Greek].
Shamsfard, Mehrnoush and Mousavi, Maryan Sadr (2008) ‘Thematic role extraction using
shallow parsing’, International Journal of Computational Intelligence, 4 (2): 126–32.
Shoben, Edward J. (1991) ‘Predicating and nonpredicating combinations’, in P. J. Schwanenflugel
(ed.), The Psychology of Word Meanings. Hillsdale, NJ: Erlbaum, pp. 117–35.
Simone, Raffaele (2000) ‘Cycles lexicaux’, Studi italiani di linguistica teorica e applicata, 29:
259–87.
Simone, Raffaele (2003) ‘Masdar, ‘ismu al-marrati et la frontière verbe/nom’, in J. L Girón
Alconchel (ed.), Estudios ofrecidos al profesor J. Bustos de Tovar. Madrid: Universidad
Complutense de Madrid, pp. 901–18.
Simoska, Silvana (1999) ‘Die morphologische und semantische Vielfalt des Adjektiv + Nomen-
Kompositums’, Deutsche Sprache, 27: 156–87.
Simpson, J. M. Y. (1995) ‘Prague school phonology’, in E. F. K. Koerner and R. E. Asher
(eds), Concise History of the Language Sciences: From the Sumerians to the Cognitivists. Oxford:
Pergamon/Elsevier Science, pp. 247–53.
Sproat, Richard W. (1992) Morphology and Computation. Cambridge, MA: MIT Press.
SSA (1992–2000) Suomen sanojen alkuperä I–III. Etymologinen sanakirja [The Origin of Finnish
Words. Etymological Dictionary], Publications of the Research Institute for the Languages
of Finland 62, Publications of the Finnish Literature Society 556, executive eds Ulla-Maija
Kulonen and Erkki Itkonen. Helsinki: Finnish Literature Society/Research Institute for the
Languages of Finland.
Stearn, William T. ([1966] 2004) Botanical Latin: History, Grammar, Syntax, Terminology and
Vocabulary. Newton Abbot: David & Charles.
Štekauer, Pavol (1998) An Onomasiological Theory of English Word-Formation. Amsterdam: John
Benjamins.
Štekauer, Pavol (2005a) Meaning Predictability in Word Formation: Novel, Context-free Naming
Units, Studies in Functional and Structural Linguistics. Amsterdam: John Benjamins.
Štekauer, Pavol (2005b) ‘Onomasiological approach to word-formation’, in P. Štekauer and
R. Lieber (eds), Handbook of Word-Formation. Dordrecht: Springer, pp. 207–32.
bibliography 307
Stump, Gregory T. (2001) Inflectional Morphology: A Theory of Paradigm Structure. Cambridge:

Cambridge University Press.
Stump, Gregory (2011) ‘Morphology’, in P. C. Hogan (ed.), The Cambridge Encyclopedia of the
Language Sciences. Cambridge: Cambridge University Press, pp. 515–19.
Švedova, Natalija Ju (ed.) (1982) Russkaja Grammatika. Akademija Nauk SSSR, Institut
russkogo jazyka, Moscow: Nauka.
Sweetser, Eve (1999) ‘Compositionality and blending: semantic composition in a cognitively
realistic framework’, in T. Janssen and G. Redeker (eds), Cognitive linguistics: Foundations,
Scope, and Methodology. Berlin and New York: Mouton de Gruyter, pp. 129–62.
Szymanek, Bogdan (1985) English and Polish Adjectives: A Study in Lexicalist Word-Formation.
Lublin: Redakcja Wydawnictw KUL.
Szymanek, Bogdan (1988) Categories and Categorisation in Morphology. Lublin: Redakcja
Wydawnictw KUL.
Szymanek, Bogdan (1989) Introduction to Morphological Analysis. Warszawa: PWN.
Szymanek, Bogdan (2010) A Panorama of Polish Word Formation. Lublin: Wydawnictwo KUL.
Talmy, Leonard (1985) ‘Lexicalization patterns: semantic structure in lexical forms’, in T.
Shopen (ed.), Language Typology and Syntactic Description: Grammatical categories and the
lexicon. Cambridge: Cambridge University Press, pp. 57–149.
Taylor, John R. (1995) Linguistic Categorization: Prototypes in Linguistic Theory. Oxford:
Clarendon Press.
Tegelberg, Elisabeth (2000) Från svenska till franska. Kontrastiv lexikologi i praktiken. Lund:
Studentlitteratur.
Temmerman, Rita (2000) Towards New Ways of Terminology Description: The Sociocognitive
Approach. Amsterdam: Benjamins.
Thèses (1929) ‘Thèses présentées au Premier Congrès des philologues slaves’, Travaux du
Cercle Linguistique de Prague, 1: 5–29; reprinted in J. Vachek (1964), A Prague School Reader
in Linguistics. Bloomington, IN: Indiana University Press, pp. 33–58.
Thomas, George (1991) Linguistic Purism. London: Longman.
Tic Douloureux, P. R. N. (1971) ‘A note on one’s privates’, in A. Zwicky, P. Salus, R. Binnick
and A. Vanek (eds), Studies Out in Left Field: Defamatory Essays Presented to James D.
McCawley on the Occasion of his 33rd or 34th Birthday. Edmonton and Champaign, IL:
Linguistic Research, pp. 45–52.
Tiersma, Peter M. (1982) ‘Local and general markedness’, Language, 58: 832–49.
TLG (2001) Thesaurus Linguae Graecae. Digital library, CD-Rom, University of California,
Irvine.
Townsend, Charles E. (1975) Russian Word Formation. Columbus, OH: Slavica Publishers.
Triandafyllidis, Manolis ([1941] 1991) Νεοελληνική γραμματική [Modern Greek Grammar], 3rd
rev. edn. Thessaloniki: Manolis Triandafyllidis Foundation.
Tribout, Delphine (2010) Les conversions de nom à verbe et de verbe à nom en français. PhD thesis,
University of Paris Diderot (Paris 7).
Trips, Carola (2009) Lexical Semantics and Diachronic Morphology. The Development of -hood,
-dom and -ship in the History of English. Tübingen: Niemeyer.
Trommelen, Mieke and Zonneveld, Wim (1986) ‘Dutch morphology: evidence for the Right-
hand Head Rule’, Linguistic Inquiry, 17: 147–69.
Trubetzkoy, Nikolaj S. (1939) Grundzüge der Phonologie. Prague: Harassowitz.
Twardzisz, Piotr (1997) Zero Derivation in English. A Cognitive Grammar Approach. Lublin:
Wydawnictwo UMCS.
UDC Consortium (2010) Universal Decimal Classification (UDC). Online at http://udcc.org/
(last accessed 25 June 2011).
Ullmann, Stephen (1962) Semantics: An Introduction to the Science of Meaning. Oxford:

Blackwell.
Uriagereka, Juan (1998) Rhyme and Reason: An Introduction to Minimalist Syntax. Cambridge,
MA: MIT Press.
Vendler, Zeno (1967) Linguistics in Philosophy. Ithaca, NY: Cornell University Press.
Vinogradov, Viktor Vladimirovič (1986) Russkij jazyk. Grammatičeskoje učenije o slove. Moscow:
Nauka.
Volek, Bronislava (1987) Emotive Signs in Language and Semantic Functioning of Derived Nouns
in Russian. Amsterdam: Benjamins.
Warburton, Irene (1976) ‘On the boundaries of morphology and phonology: a case study from
Modern Greek’, Journal of Linguistics, 12: 258–78.
Ward, Gregory, Sproat, Richard and McKoon, Gail (1991) ‘A pragmatic analysis of so-called
anaphoric islands’, Language, 67 (3): 439–74.
Wasow, Thomas (1981) ‘Major and minor rules in lexical grammar’, in T. Hoekstra, H. van der
Hulst and M. Moortgat (eds), Lexical Grammar. Dordrecht: Foris, pp. 285–312.
Watt, William C. (1975) ‘The indiscreteness with which impenetrables are penetrated’, Lingua,
37: 95–128.
Whitney, William Dwight (1879) A Sanskrit Grammar, Including Both the Classical Language,
and the Older Dialects, of Veda and Brahmana. Leipzig: Breitkopf & Härtel.
Wierzbicka, Anna (1985) Lexicography and Conceptual Analysis. Ann Arbor, MI: Karoma
Publishers.
Wiese, Richard (1996) ‘Phrasal compounds and the theory of word syntax’, Linguistic Inquiry,
27: 183–93.
Williams, Edwin (1981) ‘On the notions “lexically related” and “head of a word”’, Linguistic
Inquiry, 12: 245–74.
Williams, Edwin (2007) ‘Dumping lexicalism’, in G. Ramchand and C. Reiss (eds), The Oxford
Handbook of Linguistic Interfaces. Oxford: Oxford University Press, pp. 353–81.
Willim, Ewa (2006) Event, Individuation, and Countability: A Study with Special Reference to
English and Polish. Kraków: Jagiellonian University Press.
Wray, Alison (2002) Formulaic Language and the Lexicon. Cambridge: Cambridge University Press.
Wright, Sue Ellen (2006) ‘Terminology standards for the language industry’, in P. ten Hacken
(ed.), Terminology, Computing and Translation. Tübingen: Narr, pp. 19–39.
Wright, Sue Ellen and Budin, Gerhard (eds) (1997) Handbook of Terminology Management,
Volume 1: Basic Aspects of Terminology Management. Amsterdam: John Benjamins.
Wüster, Eugen (1931) Internationale Sprachnormung in der Technik, besonders in der Elektrotechnik
[International Language Standardization, Especially Within Electrotechnics]. Dusseldorf:
VDI-Verlag.
Wüster, Eugen (1979) Einführung in die allgemeine Terminologielehre und terminologische
Lexikographie. Vienna: Springer.
Zimmer, Karl E. (1971) ‘Some general observations about nominal compounds’, Working Papers
on Language Universals. Stanford University, CA, 5, C1–C21.
Zwitserlood, Pienie (1994) ‘The role of semantic transparency in the processing and representa-
tion of Dutch compounds’, Language and Cognitive Processes, 81: 358–67.
Author index
Abney, Steven Paul, 43n3 Bauer, Laurie

Adams, Valerie, 183, 186, 197n3 1983, 68, 78
Alexiadou, Artemis 1988, 139n9, 184
2004 (and Anagnostopoulou), 245n5 1990, 104
2006 (et al.), 231 1997, 266
2008 (and Grimshaw), 200n26, 209, 1998, 123, 158n5
224n21 2001, 80, 182–4, 241
2009, 200n26 2003, 46, 182
Anagnosopoulou, Elena see Alexiadou, 2004, 2005 (and Valera), 197n4
2006 Beard, Robert, 180–202
Allen, Margaret Reece, 85, 200n24 1981, 187, 190
Amiot, Dany, 103 1982, 199n20, 199n21
Anastassiadis-Symeonidis, Anna, 226, 228, 1984, 188
236 1987, 190
Anderson, Stephen, 46, 143, 158n3 1988, 27n6, 199n16
Anick, Peter see Pustejovsky, 1988 1995, 12–14, 25, 180, 186–90, 196, 198n15,
Arntz, Reiner, 11, 91, 92 199n17, 199n21, 201n39
Aronoff, Mark 1998, 104
1976, 180, 182, 184, 198n6 Bensaude-Vincent, Bernadette, 5
1980, 182 Benveniste
1984, 199n16 1948/1975, 206, 211, 214
1992, 103 1966, 108
1994, 188, 199n15, 201n40 Berman, Ruth A., 122; see also Clark,
2001 (and Cho), 246n14 Eve V.
2005 (and Fudeman), 207, 208 Bielec, Dana, 90
Bisetto, Antonietta, 119n6, 171; see also
Baayen, Harald, 241; see also Plag, 1999; Scalise, 2009
Schreuder Bloch, Bernard, 6–7, 14–15
Baker, Raymond see Bannister Bloch-Trojnar, Maria, 25, 180–202, 286
Bale, Alan see Barner 2006, 12, 188
Bannister, Arthur, 91 2007, 185
Barbaresi, Lavinia Merlini see Dressler 1994 2011, 200n26, 201n28
Barner, David, 17 2012a, 190, 199n20
Baroni, Marco 2012b, 201n30, 202n40
2002 (et al.), 106 Bloomfield, Leonard, 6–9, 27n5, 123,
2007 (et al.), 178n9 138n1
Bolinger, Dwight Corbin, Danielle

1967, 131 1987, 23, 33, 42, 215, 246n9
1975, 102 1992, 105
Booij, Geert Corum, Claudia, 145
2003 (and van Kemenade), 205 Craigie, William A., 10
2005, 46 Croft, William
2007, 138n7 1990/2003, 213, 223n18
2009a, 136 1991, 222n1
2009b, 104, 178n5 2000, 139n12
Bosch, Peter, 147–8 Cruse, Alan D.
Bratus, B. V., 270 1986, 104
Brinton, Laurel J. 2004, 178n8
1998, 183, 192–3, 201n31 Culicover, Peter W., 43n3, 265n13
2005 (and Traugott), 45, 50, 51, 65 Cuzzolin, Pierluigi, 205
Brooks, Patricia J. see Kempe Cysouw, Michael, 105
Browne, Wayles, 145
Bruno, Carla, 214 Dahl, Östen, 131, 135
Bücking, Sebastian, 127, 138n2 Dal’, Vladimir Ivanovič, 271
Budin, Gerhard see Wright Dalton-Puffer, Christine see Plag, 1999
Busa, Federica see Johnston Di Sciullo, Anna Maria, 104
Bybee, Joan, 219 Dimmendaal, Gerrit J., 199n21
Dokulil, Miloš, 11–12, 39, 187, 269, 272, 275
Cabré, M. Teresa, 5, 92 Downing, Pamela, 39, 102, 108–12, 122, 136
Cadiot, Pierre, 119n9 Dowty, David R., 173
Carlson, Greg, 246n14 Draskau, Jennifer see Picht
Carstairs-McCarthy, Andrew D., 12–13, 103, Dressler, Wolfgang U.
200n25 1985, 204
Cetnarowska, Bożena, 180, 183–4, 186, 1994 (and Barbaresi), 236, 266, 274
197n4, 201n29, 201n35 2006, 83–4, 119n2
Chantraine, Pierre, 216–17, 221 Drodz, L., 67
Charitonidis, Chariton
2005, 245n5, 246n13 Efthymiou, Angeliki, 26, 225–46, 286
2011, 231 1999, 226, 234–6
Cho, Sungeun see Aronoff, 2001 2001, 236, 246n9
Chomsky, Noam, 7–10, 20 2002, 236, 246n9
1957, 9 2010 (et al.), 228–9, 241–3
1964, 9, 31 2011a, 225–6, 230–1, 237–41
1965, 7–9 2011b, 236
1968 (and Halle), 249 2012a (et al.), 243, 246n16
1970, 9, 31–2, 181, 198n8, 198n10, 200n27 2012b (et al.), 245n3
1980, 7–8 2013, 236
1981, 13–14 Eisenberg, Peter, 132
1986, 249 Embick, David, 16, 18
2000, 16 Encyclopedia Britannica, 91
Civilleri, Germana Olga, 26, 203–24, 286 Engel, Ulrich, 91
2009, 222n6 Eriksson, Olof, 106
2010, 222n8 Evola, Vito, 223n10
2012, 211, 221, 222n2, 222n4, 222n8
submitted, 203 Fahim Elsayed, Mohammed Salah, 150, 152
Clark, Eve V., 108; see also Berman Fedorova, Olga see Kempe
Claudi, Ulrike see Heine Felber, Helmut, 67–8
Cobuild, 185, 194, 198n5 Fillmore, Charles J., 161
Collins English Dictionary, 36, 40, 85, 86 Fliatouras, Asimakis see Anastassiadis-
Corbett, Greville G., 199n21, 199n22 Symeonidis, 2003
author index 311
Forse, Jessica, 26, 247–65, 287 Halle, Morris, 15–18; see also Chomsky, 1968
Fradin, Bernard Hankamer, Jorge, 142
2003, 104–5, 119n3 Harley, Heidi, 15–18
2005, 161 Harris, Alice C., 158n2
2009, 102 Harris, Zellig S., 7
Fragaki, Georgia see Efthymiou, 2012 Haspelmath, Martin
Fruyt, Michèle, 208 1992, 104
Fudeman, Kirsten see Aronoff, 2005 2002, 103, 182
Havou, Eleni see Efthymiou, 2010
Gaeta, Livio Heine, Bernd, 219
2002, 220 Herbst, Thomas, 201n28
2003 (and Ricca), 242 Heslin, Peter, 203
2009 (and Ricca), 136 Heyvaert, Liesbet, 14
Gagné, Christina L. Himmelmann, Nikolaus P., 208, 219
1997 (and Shoben), 108–12, 116–17, 173 Hockett, Charles F.
2006 (and Spalding), 109–10, 115, 119n2, 1942, 8
173 1958, 18
Gavanozi, Vaso see Efthymiou, 2010 Hoekstra, Teun, 162
Gibson, Martha see Libben, 2003 Homer, 26, 203–24, 223n19
Giegerich, Heinz J. Honeybone, Patrick, 5
2004, 123 Horecký, Ján, 26–7, 267–80
2005, 142, 144, 146, 158n5 1973, 276
2006, 123 1974, 270, 275
Givón, Talmy, 220 1980, 268, 279
Gottfurcht, Carolyn, 232, 237–8, 241, 243, 1983, 12, 267, 279
246n12 1994, 267–8, 270–1, 274–5, 279
Grandi, Nicola, 266 Hünnemeyer, Friederike see Heine
Greenbaum, Sidney see Quirk
Grimshaw, Jane, 162, 180, 192, 200n26, INS, 227
223n11; see also Alexiadou, 2008 Isabelle, Pierre, 108
Gross, Gaston, 222n1
Gruber, Jeffrey S., 257 Jackendoff, Ray, 9, 20–1, 28–44, 110–13,
Grzega, Joachim, 11, 39 247–65
Grzegorczykowa, Renata, 88, 187 1975, 9, 21, 32–3, 63, 180, 188, 198n15
Guevara, Emiliano, 135; see also Baroni, 2007 1983, 20, 28–9, 92, 198n14, 227, 237,
Gunkel, Lutz, 130–1, 135 246n12, 248, 251, 257, 265n9
1990, 20, 28, 198n14, 227, 237, 246n12,
Habert, Benoît, 114 248, 250–1, 257, 264n4
ten Hacken, Pius, 1–27, 28–44, 83–101, 249, 1991, 192, 199n20
287 1997, 29–30, 161–2, 170, 178n7
1994, 35, 43n8, 84–5, 138n2, 138n4, 2002, 20–1, 23, 28–35, 39, 42, 43n4, 44n8,
139n11, 158n6 51–2, 62–3, 104, 161–2, 164, 169, 173,
1999, 84–5 178n3, 178n10, 179n13, 180, 248–9
2003a, 35, 94 2007, 162, 251, 264n4
2003b, 35, 94, 138n4 2008, 37
2007, 7, 20, 36, 249 2009, 21–2, 24, 28, 34, 42, 51–2, 55, 103–4,
2008, 11 108–12, 118, 119n7, 119n8, 178n10,
2009a, 8, 10 247
2009b, 43n6 2010, 21, 23, 32, 34, 37–8, 41–2, 51–4,
2010a, 84 101n1, 135, 138n4, 177, 178n10,
2010b, 40, 42, 43n7, 89 179n12, 187, 198n15
2010c, 11 see also Culicover
2011 (and Panocová), 7 Jacobs, Joachim, 132
Hajičová, Eva, 4 Jacquemin, Christian see Habert
Jakobson, Roman Levi, Judith N

1959, 208 1975, 108
1960, 4 1977, 145–6, 158n5
Jarema, Gonia, 108, 116 1978, 9–10, 85, 87, 91, 108–9
Jespersen, Otto Levin, Beth, 247; see also Rappoport Hovav
1922, 108 Li, Charles, 108
1942, 137n1 Libben, Gary
Johnson, Mark see Lakoff, 1980 2003 (et al.), 149
Johnston, Michael, 118 2006, 104
Josefsson, Gunlög, 102 Liddell, Henry George, 203, 212
Joseph, John, 1 Lieber, Rochelle
Junczys-Dowmunt, Marcin, 106 1980, 252
1983, 19
Kačmárová, Alena, 269 1993, 253
Kallas, Krystyna, 90 2004, 19–20, 22, 23, 45–51, 54–5, 57, 63–5,
Kastovsky, Dieter, 46, 50, 62, 136 65n1, 108, 161, 164, 166, 177n2, 237,
Kazazis, Kostas, 245n2 245n1
van Kemenade, Jaap see Booij, 2003 2007 (and Scalise), 161, 170, 178n5
Kempe, Vera, 266 2009, 45, 47, 50
Kerleroux, Françoise, 204 2009 (and Štekauer), 83–4
Koptjevskaja-Tamm, Maria 2010, 242–3
1993, 197n2 Lönnrot, Elias, 23, 69–81
2009, 139n13 Luraghi, Silvia, 210
Koskenniemi, Kimmo, 43n5 Lyons, John, 103
Krifka, Manfred, 192
Kuryłowicz, Jerzy, 207 McEnery, Tony, 119n4
Kwiatek, Ewelina, 24, 83–101, 287 McKoon, Gail, 147, 148–9, 159n9; see also
2013, 91 Ward
Maienborn, Claudia, 138n5
Labov, William, 11 Mail, Noor, 41
Lahav, Ran, 124, 128 Malicka-Kleparska, Anna, 197n1
Lakoff, George 1985, 183, 188, 200n24
1972 (and Ross), 145 1988, 180–2, 184–6, 188, 198n8, 201n33,
1980 (and Johnson), 68, 208, 201n40
223n10 Malkiel, Yakov, 204
1987, 208 Marantz, Alec, 16–17; see also Halle
Lang, Friedrich, 5 Marchand
Langacker, Ronald W. 1960, 253–5, 260, 265n15
1987a, 14–15, 161, 178n2, 178n3 1969, 102, 182, 185, 197n3
1987b, 192, 220 Markos, Angelos see Efthymiou, 2012
1991, 14 van Marle, Jaap, 183–4
2000a, 25 Matiasek, Johannes see Baroni, 2002
2000b, 160n18 Matthews, Peter H.
2008, 14–15 1972, 223
Lapata, Maria, 118 1993, 5–6
Laskowski, Roman, 188 2005, 274
Leech, Geoffrey see Quirk Mayer, Felix see Arntz
Lees, Robert B., 8–9, 182 Meillet, Antoine, 211
Lehmann, Christian Mela-Athanasopoulou, Elizabeth, 240
1989, 219 Millikan, Ruth, 111
2002, 208, 219, 223n9 Mironova, Natalija see Kempe
2004, 208, 219 Mohr, W. see Müller
Lehrer, Adrienne, 204, 208 Mondini, Sara see Semenza, Carlo
Lesselingue, Chrystèle, 105 Motsch, Wolfgang, 135, 138n6
author index 313
Mourelatos, Alexander, 192, 201n6 Puzynina, Jadwiga see Grzegorczykowa,

Mousavi, Maryan Sadr see Shamsfard 1979, 1999
Müller, L. P., 41
Müller-Bollhagen, Elgin see Ortner Quirk, Randolph, 201n37
Murray, August Taber see Homer
Rainer, Franz
Namer, Fiammetta, 103, 118n1 1988, 184
Naumann, Bernd, 204, 208 1992 (and Varela), 103
Newmeyer, Frederick J., 8 Rappaport Hovav, Malka, 162; see also Levin
Nicoladis, Elena, 90–1, 108, 112, 118 Ratcliff, Roger see McKoon
Noailly, Michèle, 119n3 Raymond, Stanley see Bannister
Noyer, Rolf, see Embick; Harley, 2003 Ricca, Davide see Gaeta, 2003, 2009
Nyman, Martti, 245n2 Rijkhoff, Jan, 130
Rio-Torto, Graça, 25, 161–79, 288
O’Brien, Peter see Mail 2009 (and Ribeiro), 170, 178n6
OED 2010 (and Ribeiro), 170–1
1989/2011, 40, 182, 184, 198n5, 253–4, 2013 (and Ribeiro), 174
264n5, 264n6 Rodrigues, Alexandra, 162–9
2002, 72–3, 75, 78, 82n1 Romagno, Domenica, 205
Onions, C. T. see Craigie Rosch, Eleanor, 11
Ortner, Lorelies, 138n6 Rommens, P. M. see Müller
Rosenbach, Annette, 131
Palmer, Frank R., 198n8 Rosenberg, Maria, 24, 102–20, 288
Panagiotidis, Phoevos, 17 2007, 108
Pang, Geordi, see Mail Ross, John, 158n1; see also Lakoff, 1972
Panocová, Renáta, 26–7, 266–85, 287; see also Rozwadowska, Bożena, 197n2, 200n28
ten Hacken, 2011 Ryder, Mary, 104
Partee, Barbara, 131 Rytting, C. Anton, 230, 245n2
Paul, Waltraud, 142, 144, 146
Pearson, Jennifer, 5, 10 Sag, Ivan see Hankamer
Petropoulou, Evanthia, 95 Sager, Juan, 10, 67–8, 81
Picht, Heribert, 67; see also Arntz Sandra, Dominiek see Libben, 2003
Pirrelli, Vito see Baroni, 2007 Saussure, Ferdinand de, 1–4, 7–8, 13–14, 20,
Pitkänen-Heikkilä, Kaarina, 23–4, 66–82, 67, 247, 267
287–8 Scalise, Sergio
2005, 70 1986, 104
2008, 66, 68–9 2009 (and Bisetto), 170
Plag, Ingo 2010 (and Vogel), 170
1998, 51 see also Bisetto; Guevara; Lieber, 2007
1999, 165, 182–3, 227, 237, 241–3, Schachter, Paul, 181
245n1 Schäfer, Florian see Alexiadou, 2006
1999 et al., 246n17 Schäfer, Martin, 25, 84, 137, 138n3, 140–60,
2003, 46, 165 288
Pompei, Anna, 205 Schiller, Anne, 106
Postal, Paul, 25, 126, 140–2, 145, 147–8, Schlücker, Barbara, 24–5, 84, 121–39, 288–9
158n1, 158n2, 158n5, 159n8 Schreuder, Robert, 116
Privitera, Aurelio see Homer (Omero) Schwarze, Christoph, 208, 219
Pustejovsky, James Schwyzer, Eduard, 211
1988 (and Anick), 194 Scott, Robert see Liddell
1991, 58–60, 178n7 Seibicke, W. see Drodz
1995a, 21–2, 23, 25–6, 55–65, 118, 169, Selkirk, Elisabeth O., 15, 19, 104, 108, 143,
178n3, 186–7, 194 252
1995b, 59–60 Semenza, Carlo, 104, 119n2
van der Putten, Frans see Hoekstra Setatos, Michalis, 245n2
Shamsfard, Mehrnoush, 178n10 Trips, Carola, 246n14

Shoben, Edward J., 109; see also Gagné, 1997 Trommelen, Mieke, 252
Simone, Raffaele Trost, Harald see Baroni, 2002
2000, 213 Trubetzkoy, Nikolaj S., 3, 5
2003, 206, 211 Twardzisz, Piotr, 197
Simoska, Silvana, 138n6
Simpson, J. M. Y., 3 UDC Consortium, 91
Soares Rodrigues, Alexandra, 25, 161–79, Ullmann, Stephen, 67
289 Uriagereka, Juan, 36
Spalding, Thomas L. see Gagné, 2006
Sproat, Richard, 43; see also McKoon, Valera, Salvador see Bauer, 2005
Ward Varela, Soledad see Rainer, 1992
SSA, 71–3, 75–6 Vendler, Zeno, 166
Stearn, William T., 69, 81, 82n1 Vendryes, Joseph see Meillet
Štekauer, Pavol Vinogradov, Viktor V., 270
1998, 12 Vogel, Irene see Scalise, 2010
2005a, 82 Vogel, Petra M. see Naumann
2005b, 11, 102, 268 Volek, Bronislava, 274
see also Lieber, 2009
Stengers, Isabelle see Bensaude-Vincent Wälchli, Bernhard see Cysouw
Stump, Gregory T., 15 Warburton, Irene, 245n2
Suffner, J. see Müller Ward, Gregory, 25, 145–50, 158n6, 159n9;
Svartvik, Jan see Quirk see also McKoon
Švedova, Natalija Ju, 270 Wasow, Thomas, 198n7
Sweetser, Eve, 124 Watt, William C., 145, 147
Szymanek, Bogdan Wenda, K. see Müller
1985, 180, 183, 188, 193 Whitney, William Dwight, 5
1988, 12, 180, 188 Wierzbicka, Anna, 234
1989, 181, 187, 193, 197n3 Wiese, Richard, 35
2010, 88–90, 94, 96 Williams, Edwin
1981, 104, 252
Talmy, Leonard, 53, 61 2007, 17
Taylor, John R., 208 see Di Sciullo; Grimshaw, 1993
Tegelberg, Elisabeth, 105 Willim, Ewa, 187, 192
Temmerman, Rita, 11 Wray, Alison, 88
Thèses, 4–5, 11 Wright, Sue-Ellen, 10
Thomas, Claire, 1–27, 45–65, 201n32, 208, Wüster, Eugen
289 1931, 5
Thomas, George, 74 1979, 67
Tic Douloureux, P. R. N., 145
Tiersma, Peter M., 199n21 Xiao, Richard see McEnery
TLG, 203, 218
Tono, Yukio see McEnery Yoon, Yeo Bom see Libben, 2003
Townsend, Charles E., 270
Trager, George L. see Bloch Zifonun, Gisela see Gunkel, 2009, 2011
Traugott, Elizabeth C. see Brinton, 2005 Zimmer, Karl E., 122
Triandafyllidis, Manolis, 227 Zonneveld, Wim see Trommelen
Tribout, Delphine, 232, 239 Zwitserlood, Pienie, 149, 159n11
Subject index
abbreviations, 94, 96 chosen terms, 71–2

accepted terms, 71 classifying modification, 130–2, 134
action modality, 111–12 co-composition, 57
adherence, 274–5 coercion, 21–2, 89, 170, 174
affector, 274 Cognitive Grammar (Langacker), 14, 25, 26,
affix, 17, 19, 52, 165, 167–9, 187, 204, 270 220
affixation, 13, 33–5 collocation, 60
affixoid, 237, 245n3 competence, 7, 30, 41, 256
agent nouns, 162–5 complete semantic definition (CSD), 27, 267,
agreement, 199n22 268–9, 274, 276
Aktionsart, 192–3 complex event, 49
allomorphs, 198n6, 227, 229, 247, 270 compositionality, 46, 124, 170–1, 176, 203,
analogy, 80, 81 209–21
anaphoric island, 25, 86, 126, 140–60 compounding, 8, 9, 15, 17, 24, 25, 33, 35,
Ancient Greek, 26, 71, 73, 95, 203–24 83–91, 93–101, 103–20, 121–39, 140–60,
antecedent, 25, 86 170–7
apophony, 207 concepts, 10, 24, 84, 90, 131, 134, 135
appellative, 274 conceptual category, 12
arbitrariness, 3, 67 conceptual structure (Jackendoff), 23, 26, 28,
argument structure, 56–7, 162–4, 169, 204, 51, 53–4, 65, 110–11, 187, 247–65
209–10 construction morphology (Booij), 104
attributive modification, 122 conversion, 54, 182–5, 186, 193–6, 197n4
availability (disponibilité), 33, 37, 42–3 copulative compounds, 35
corpora, 10, 24, 25, 26, 91–3, 103, 105–6,
back formation, 32, 73 114, 153–8, 157–8, 228, 271
bahuvrihi compounds, 138n7 correspondence record, 92–3
blocking, 184–5 COSMAS II, 153
body (Lieber), 19, 47, 50 countability, 185, 186, 190, 192, 209
borrowing, 11, 39, 68, 69, 93–4, 96, 97–8, 101 created terms, 72–81
botany, 23, 66–82
definition, 11, 12, 24, 83–5, 103–5, 121–39,
calque, 74 170
cartography, 92–101 deictic compounds, 39, 136
causative, 225, 239, 244 derivation, 25–6, 35–7, 39, 70, 162–9, 176–7,
chemistry, 4, 67 180–246, 247–65
Chinese, 144–5 Deutsches Referenz Korpus (DeReKo), 153
diachronic linguistics, 2–3, 4, 45, 46, 51, 65 iconicity, 207, 236

diacritic feature, 37–8 identificational field, 257–9
dictionary, 226, 242; see also Oxford English idiom, 17, 18, 30
Dictionary idiomaticity, 170–2, 173, 176
diminutive, 27, 266–7 inbound anaphora, 158n5
discourse model, 148–9 inchoative, 226
Distributed Morphology (DM), 15–18 inflection, 6, 13–14, 15, 17, 18, 21, 33–4, 123,
dotted type, 57 189, 191
DP hypothesis, 43n3 Institut für Deutsche Sprache (IDS), 153
Dutch, 149 institutionalization, 200n25
integration, 127, 132
effector, 274 intersective modification, 127
empty element, 141 ISO/TC37, 5
entrenchment (Langacker), 25, 157 Italian, 26, 69, 234
ergativity, 276
etymology, 82, 227, 234, 252–3, 261 katharevousa, 228–9
European Parliament, 24, 105–6
evaluative meaning see pejorative L-derivation, 13, 190, 192–3, 196–7
connotation land surveying, 24, 91–101
evaluative morphology see diminutive langue, 3, 4, 7
event nouns, 165–9 Latin, 68–82, 95, 260
event structure, 56–7, 166–7, 200n26, 210, Latinate suffixes, 183–4, 186, 196–7
214–15 learned vocabulary, 26, 227, 228–30, 233–4,
exocentric compounds, 88–9, 152–3 243–5
lexeme, 103, 188
Finnish, 23, 66–82 Lexeme-Morpheme Base Morphology
formant, 267 (LMBM), 12–14, 25–6, 180, 186–90
formulaic expression, 88 lexical conceptual paradigm (lcp), 58, 60,
French, 24, 26, 53–4, 61–2, 69, 102–3, 105–8, 194–6
112–20, 247, 253–4, 260–4, 265n8 lexical-conceptual structure (LCS), 163, 169,
frequency, 219 171, 173, 176, 178n7, 187, 192, 199n20,
full-entry theory, 9, 32–3 237–9
function word, 17, 30–1 lexical cycle, 213
lexical entry, 8, 30–1, 51–2
gender, 275 lexical extension rules see L-derivation
generative lexicon (Pustejovsky), 21–2, 23, lexical inheritance structure, 56, 58–9
25, 55–65 Lexical Integrity Hypothesis, 143–4
generative semantics, 8, 9, 14, 25, 87 lexicalist hypothesis, 9, 15, 31–2, 42, 142–3
genitive, 89, 91, 96, 100, 210, 214–15; see also lexicalization, 7, 23, 26, 45–65, 102, 104–5,
Saxon genitive 125, 131, 185, 190, 191, 195–6, 203–24;
geodetic surveying, 92–101 see also semantic specialization
German, 24, 25, 121–39, 141–2, 147–8, lexicalization scale, 203
149–58 lexicology, 67
gerund, 181 lexicon, 7, 8, 9, 15, 16, 17, 21, 29–31, 104,
glide formation, 227, 230 190, 248–9, 253–4
GPS see satellite positioning system Linguistic Wars, 8, 20
grammaticalization, 219 linking elements, 88, 90–1, 108
Greek see Ancient Greek, Modern Greek loan shift, 74
loan translation, 72, 74, 75, 79, 96, 99, 101
head, 103–5, 115–16, 129, 130, 132, 144–5, Logical Form (LF), 16
150–2, 252 logical polysemy, 56, 60, 194
homonym, 6, 19
hypocoristics, 274 Mandarin see Chinese
hyponymy, 59, 130; see also subconcepts materiality, 276
subject index 317
maximal compatibility principle, 161, 165, pragmatics, 25, 146–8, 157–8, 244
168–9, 173–6 Prague School, 3–5, 12
maximal semantic frame, 163–4, 169, 170, predication, 222n1
176 prefixes, 252–64
metaphor, 39, 67, 68, 71–2, 73, 75, 78, 81, preverb, 205
171, 203, 208, 213, 215, 276 priming, 62–3, 65
metonymy, 39, 67, 71–2, 138n7, 153, 171, privative features, 38, 48
203, 208, 213–17 process-result alternation, 47–50, 55–65,
Modern Greek, 26, 225–46 167–9, 186, 194–6, 206
modificational type, 12, 269, 272, 275 processing, 104–5
morpheme, 6, 7, 8, 13, 15, 17, 21, 188 productivity, 4, 23, 26, 31–44, 102, 104–5,
morpholexical features, 189–90 112, 118, 121, 182–5, 207, 226, 229–30,
morphological family, 116–17 241–5
Morphological Spelling Component profitability (rentabilité), 33, 42–3, 183
(LMBM), 188 pro-form, 140–60
morphome, 201n40 projective conclusion space, 59, 64
motivation, 67–8 projective inheritance, 59
multi-word units, 30 pronoun see pro-form
mutational type, 12, 269 Proper Function (PF), 111–12
proper names, 39, 94, 136, 274
named language, 8, 36 protolanguage, 33–4, 177
naming, 4–5, 10–12, 38–43, 66–82, 139n9 prototype, 11, 15, 84
neoclassical word formation, 68, 78, 95–6, psycholinguistics, 148–9, 157–8, 173
99–100, 101, 118n1
nominalization, 25, 180–6, 191–7, 203–24 qualia structure, 22, 56–7, 59
non-referentiality, 135
nouniness, 206, 220, 222n6 Real Academia Española (RAE), 260
number, 189–90, 192 recoverably deletable predicates (RDPs), 9,
91
onomasiological approach, 10–12, 23, 26–7, recursion, 93–4, 272
39, 67, 91, 102, 263, 267–80 redundancy rules, 8, 9, 21, 32–3, 35, 180–1,
onomasiological base, ~ mark, 11, 267–9, 182
270 regularity, 7, 8, 9, 21, 33, 42–3
ontological class, 48 relational adjectives, 86–7, 89–90, 94, 96, 99,
opacity, 147, 149, 152, 176, 203, 220 100, 125, 131, 193–4
outbound anaphora, 140 rewrite rules, 8, 9, 15
Oxford English Dictionary (OED), 10, 253–4 right-headedness, 15, 252
rule ordering, 183
paradigm, 15, 44n8 Romanian, 69
Parallel Architecture (PA), 20–1, 23, 25, Russian, 27, 266–7, 269–85
26, 28–44, 51–5, 104, 110–11, 198n15, Russian National Corpus, 271
248–9
parole, 3, 7 S-structure, 13, 16
partial modification, 127 satellite positioning system, 92–101
patterns of lexicalization, 54 Saxon genitive, 86, 87, 94
pejorative connotation, 26, 226, 231–7, 244 selective binding, 57
performance, 7, 30, 41, 256 semantic coindexation, 161–2, 164–9, 171–2,
Phonetic Form (PF), 16 173–4, 176
phrasal constituents, 24 semantic drift, 190
phrasal verbs, 182–3 semantic specialization, 122, 124, 128, 131,
Polish, 24, 88–101 171–2
polysemy, 19, 22, 48, 55–6, 63, 171–2, 208–9, semantic transparency, 147–9, 153, 170, 203,
213 220
Portuguese, 25, 26, 161–79, 247, 260–4 semiproductivity, 23, 37–8, 42–3
Sense Enumerative Lexicon (SEL), 21, 56, telicity, 205

194 term extraction, 92
separation hypothesis, 187–8 terminology, 4–5, 10–11, 23–4, 66–82,
sign (Saussure), 3, 13, 14, 247, 267 91–101, 151
Simpler Syntax Hypothesis, 43n3 thematic roles, 173–6
skeleton (Lieber), 19, 47–50 tiers (PA), 164, 251
Slovak, 268–9, 271, 275, 279 topicality, 147–9, 157
Spanish, 26, 69, 247, 260–4 translation, 24, 92–3, 105–8, 114
specialized vocabulary, 11 transposition, 12, 180, 189–90, 191–3, 269
speech community, 7, 14, 24 Trésor de la Langue Française (TLF), 260,
stage-level predicates, 241 263
standardization, 5, 10–11 Turkish, 234
stimulus-response model, 6, 7 Two-Level Morphology, 43n5
stress, 88, 123, 132, 143 type coercion, 22, 57
subconcept, 133–4, 137 type constructor, 60
suffixation see affix
surveying see land surveying Uniform Theta Assignment Hypothesis
Swedish, 24, 69–74, 76–82, 102–3, 105–8, (UTAH), 43n3
112–20
symbolic unit, 14 Variable R condition, 85
synapsies, 108 verbiness, 206, 210, 211
syntax, 8, 121–39
synthetic compounds, 88 zero-derivation see conversion

0748689605

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

0748689605

Transféré par

Droits d'auteur :

Formats disponibles

The Semantics of

Word Formation and

Edinburgh University Press Ltd

ISBN 978 0 7486 8960 6 (hardback)

The right of the contributors

List of figures vii

1. Word formation, meaning and lexicalization 1

10. Deverbal nominalizations in English: an LMBM approach 180

Notes on contributors 286

1.1 Latin facio and conficio2

5.5 Compounds in each subfield of the English termbase 97

2.1 Adjective formation with -al and -ary36

13.1 Variables by semantic type 259

Word formation, meaning and

W ord formation is a rule-­based process for producing new words.

In understanding the historical development of linguistic thought in the

1.1 The European tradition

facio conficio Époque B

should describe such a relationship in terms of two successive stages. The

valency, topic-­focus articulation and the opposition between theme and

ment of a confusing collection of names for chemical substances, based on

1.2 The American tradition

The first comprehensive overview of linguistic analysis in the American

(1) For some languages, it is useful to divide the morphological constructions of

In a number of ways, (1) plays down the importance of the distinction,

As a final component of this overview of how Bloomfield’s theory

Current approaches to the meaning and lexicalization of word formation

2.1 The onomasiological approach in terminology

A more pertinent challenge to this mainstream is posed by the socio-

2.2 The onomasiological approach to lexical semantics

2.3 Lexeme-­Morpheme Base Morphology

2.4 Cognitive Grammar

sentence is less commonly automatized. However, the difference is one of

2.5 Distributed Morphology

The unit of operation in DM is the morpheme. Harley and Noyer

is no distinction between periphrastic and synthetic expressions of the

2.6 Other generative approaches

phonological syntactic conceptual

Figure 1.4 Parallel Architecture (after Jackendoff 2002: 125, by permission of

used in model-­theoretical semantics as a basis for the expression of quan-

(3) a. Last night, Anna started a new book.

Although Pustejovsky (1995a) does not directly address word forma-

3 OVERVIEW OF THE VOLUME

The contributions to the present volume give a multifaceted overview of

a basic meaning difference, such that compounds generally have a classify-

(1995a) theory of the generative lexicon can be invoked to explain some of

onomasiological model of word formation and the word formation process

1. ‘Language conceived of as a functional system.’ Our translation,

Semiproductivity and the place of

1 LEXICAL ENTRIES IN JACKENDOFF’S PARALLEL

Jackendoff’s Parallel Architecture is a model of language that is intended

(1) Anna has a new car.

(2) a. Annai hasj ak newl carm

(3) a. Swansea is a sunny city.

The representation of meaning in (2c) is underspecified. Only the concepts

The information necessary for generating (2) is in the lexicon. This is

(5) a. Carol made a genuine attempt to put Swansea on the map.

The example in (5a) is typical of the cases considered by Jackendoff (1997).

(6) a. putp onq ther maps

2 REPRESENTING MORPHOLOGICAL PRODUCTIVITY

hypothesis’, which implies ‘that derived nominals will correspond to

(9) a. Morphology is in the lexicon.

2.1 Full Entry Theory

(10) a. aggression – aggressive – aggressor

List of figures vii

1. Word formation, meaning and lexicalization 1

10. Deverbal nominalizations in English: an LMBM approach 180

Notes on contributors 286

1.1 Latin facio and conficio2

5.5 Compounds in each subfield of the English termbase 97

2.1 Adjective formation with -al and -ary36

13.1 Variables by semantic type 259

W ord formation is a rule-based process for producing new words.

valency, topic-focus articulation and the opposition between theme and

2.3 Lexeme-Morpheme Base Morphology

used in model-theoretical semantics as a basis for the expression of quan-

1 LEXICAL ENTRIES IN JACKENDOFF’S PARALLEL

Table 2.1 Adjective formation with -al and -ary

• There are open-ended sets of nouns ending in -tion and -ment.

3 WORD FORMATION AS THE FORMATION OF LEXICAL

(1) a. Access to attorneys might compromise Padilla’s ongoing interrogation by the

(3) [-material, dynamic ([ ], <base>)]

(9) a. V1-er2 = PERSON2α; [F1, (α, . . .)] (agentive -er)

(12) a. Political involvement in the appointment of judges would mean a more