Vous êtes sur la page 1sur 20




Curt Rice on
Generative metrics

Optimizing structure in context:

Scrambling and information
by Hye-Won Choi
reviewed by Helen de Hoop

Furthermore, unlike researchers

working on a natural language in
which they themselves lack
intuitions, those working on
metrical poetry rarely have the
opportunity for field work.

Event semantics of verb frame

alternations: A case study of
Dutch and its acquisition
by Angeliek van Hout
reviewed by William Philip

Recent issues in linguistics

Elan Dresher on
The Olde Yankee Grammarian

So first Ill lift the wh-phrase out

of its slot and lay it down on the
side. While Im doing that,
Howard, could you reach into the
knickknacks box and grab some
traces and indices?

Volume 2


Book notices
Five books on Chinese


Conference reports
by Joo Costa


Letters and replies

John Goldsmith and Geo(rey
Huck respond to Elan Dresher

Monthly Magazine
for Linguists

September 1996
[published September 1997]


ISSN 13813439

Price single issue:

Dfl 14,50


Glot International
appears monthly,
except June & July.

John Goldsmith & Geoffrey Huck

Elan Dresher takes a novel approach to
the stultifying task of reviewing
linguistics books in these pages [Glot
International 2/4, p. 11]. Hes got a new
technique, which evidently relieves him
of the burden of reading the book hes
reviewing. Instead, as he tells us, he
ponders the dust jacket, skims a few
pages, and then naps. We certainly see
the appeal in this strategy, especially
for a man whose legendary wit is of the
sort that goes straight for the jocular.
But it has its hazards. For example, if
he had followed the more conventional
route and read our book, Ideology and
Linguistic Theory: Noam Chomsky and
the Deep Structure Debates, which he
undertook to comment upon in his
column he would have discovered that
nowhere in it did we either raise or
imply an answer to a question that he
somehow imagined was central to it:
Since there was nothing wrong with

Generative Semantics Theory, and since

Chomskys arguments against it were
faulty, shouldnt linguistic theory today
follow the Generative Semantics
methodology (if not its theory) rather
than Chomskys? Except in Elans
dreams, we found no compelling evidence that one of the programs was to
be preferred over the other on any
grounds empirical, methodological, or
conceptual. He complains that we set
the bar rather high in deciding whether
an argument [against Generative
Semantics] is conclusive or not, but
loses interest when the question is
whether Interpretivists adequately
responded to Generativist arguments.
As we observed, in linguistics as
elsewhere, the tendency is to assume
that my arguments against you should
be taken much more seriously than your
arguments against me.
Copyright 1997 by
Holland Academic Graphics.
All rights reserved.


Glot International, Volume 2, Issue 7, September 1996 [1997]

Page 2

Lisa Lai-Shen Cheng
Rint Sybesma
Holland Institute of Generative Linguistics
address: Sinological Institute | Leiden University |
P.O. Box 9515 | 2300 RA Leiden | The Netherlands
| phone: +31 71 5272538 | fax: +31 71 5272615 |
e-mail: glot@rullet.leidenuniv.nl/llcheng@uci.edu


Conference reports and

announcements editor
Jan-Wouter Zwart
Groningen University | Department of linguistics |
Oude Kijk in t Jatstraat 26 | 9712 JK Groningen |
The Netherlands | e-mail: zwart@let.rug.nl

Goodies editor
Andrew Carnie
Department of Linguistics | Harvard University |
77 Dunster Street | Cambridge MA 02138 | USA |
e-mail: carnie@linguistlist.org

Regular contributors
Eullia Bonet (Barcelona) | Marcel den Dikken
(Vrije Universiteit Amsterdam) | Paula Fikkert
(Konstanz) | Bob Frank (Johns Hopkins University)
| Eric Hoekstra (P.J. Meertens Instituut, Amsterdam) | Helen de Hoop (Utrecht) | Sabine Iatridou
(University of Pennsylvania) | La Nash (Paris) |
Guido Vanden Wyngaerd (Brussels) | Shohei
Yoshida (Yokohama National University) | JanWouter Zwart (Groningen)

What do you mean question period?!

I havent even gotten to the point
yet! If I stop now, nobody will
be able to ask anything!

Elan Dresher (Toronto), Neil Smith (London)

Information for contributors

Linguists who want to contribute a State-of-the-Article
or a Review are invited to contact the Editors.
Linguists who have recently finished their dissertation are invited to forward a copy to the Editors.
Everybody is welcome to contribute to Glot International. But before you send in anything,
please, contact the Editors.
Glot International prints conference announcements,
calls for papers, conference programs, etc.

Subscription information
Glot International appears every month, except in
June and July. Subscriptions are for the printed or
the electronic (pdf) version.
Prices for Volume 3 (1998):
NLG 120.0
pdf only
NLG 035.0
printed version
NLG 050.0
surface mail
NLG 032.5
other countries
NLG 037.5
air mail
NLG 042.5
other countries
NLG 047.5
surface shipping in three batches
NLG 015.5
other countries
NLG 020.5
Subscriptions at the individual rate must be ordered
directly from the publisher.
Subscriptions at the institutional rate may be ordered from your bookseller or subscription agent,
or directly from the publisher.
Prices exclude 6% VAT (EC only).
Orders must be prepaid.
Payments preferably by giro, credit card (American
Express, Euro/Access/Master, Visa) or international money order. Credit card orders must include the complete cardholders address. There is
an additional charge of Dfl 17,50 for payments
by international bank cheque or bank transfer.

Holland Academic Graphics
Julius Rntgenstraat 15 | 2551 KS The Hague |
The Netherlands | phone +31 70 4480203 |
fax +31 70 4480177 | e-mail: mail@hag.nl |
http://www.hag.nl | bank acc. ABN AMRO 49 88 22
796 | postal giro 6261107 | VAT nr. NL

Contact the publisher

Printed in The Netherlands

Contributors to this issue

Dissertations reviewed

Hye-Won Choi
Department of Linguistics | University of Southern California | Los Angeles, CA 90089 | USA

Hye-Won Choi, Optimizing structure in context:

scrambling and information structure. Stanford
University. Supervisor: Joan Bresnan. Degree
date: September 1996. 256 pp. Downloadable
from Rutgers Optimality Archive #147: ROA
URL: http://ruccs.rutgers.edu/roa.html. For hard
copy, contact author.

Joo Costa
HIL/Leiden University | PO Box 9515 | 2300 RA
Leiden | The Netherlands
Elan Dresher
Department of Linguistics | University of Toronto | Toronto | Ontario M5S 1A1 | Canada
Helen de Hoop
Research Institute for Language and Speech |
Utrecht University | Trans 10 | 3512 JK Utrecht
| The Netherlands

Angeliek van Hout, Event Semantics of Verb

Frame Alternations. A case study of Dutch and its
acquisition. Tilburg University. Supervisor: Henk
van Riemsdijk. Degree date: February 1996. 296
pp. ISBN: 90-9009230-7. Tilburg Dissertations in
Language Studies; secretariaat.fdl@kub.nl.

Angeliek van Hout

IRCS - University of Pennsylvania | 3401 Walnut Street Suite 400C | Philadelphia, PA
19104-0357 | USA
William Philip
University of Maine at Farmington | Department of Humanities | University of Maine |
Farmington, Maine 04938
Curt Rice
University of Troms | School of Languages and
Literature | 9037 Troms | Norway

In the next issue

Norbert Hornstein on Control
Constraints on Subjects. An Optimality Theoretic
by Vieri Samek-Lodovici
reviewed by Peter Ackema
by Neil Smith

I am grateful to David Pesetsky for drawing my
attention to the fact that I made a mistake in
my description of the Stroop Effect in
Colourful Language.
Reading colour names written in the wrong
colour is not sigificantly harder than reading
them written in the congruent colour. What is
dramatically harder is naming the colour in
which an incongruent colour word is written.
Fortunately, the theoretical argument about
modularity still goes through on the corrected
Neil Smith


Glot International, Volume 2, Issue 7, September 1996 [1997]

Curt Rice
Generative metrics is the branch of phonology which is concerned with
metrical poetry. In the following article, Curt Rice introduces the most
salient issues in the field. Why are some lines of poetry metrically
ungrammatical? What is the monosyllabic word rule? How does OT turn
out to be useful? Thy edge should be blunter than appetite.

This article aspires to give the reader a sense
of current research in generative metrics, a research enterprise which is nearly as old as generative grammar itself, dating from the pioneering
work of Morris Halle and S.J. Keyser (1966,
1971a,b). We begin this selective overview of the
current state of the field with discussion of two
general issues relevant to conducting this research and then move on to consider a number of
cases which advance our understanding of the
grammars of metrical poetry and which have
significant implications for current research in
generative phonology.

Generative metrics and generative


2.1. Methodology
Generative metrics applies the central perspectives of generative grammar to the study of
metrical poetry; these include a descriptive orientation regarding the nature of the data, the distinction between grammatical and ungrammatical utterances, and the importance of native
intuitions. However, there are a number of methodological restrictions constraining research on
metrical poetry. Within generative metrics, there
is a heightened awareness that the construction
of a grammar is not for the metrical poetry of
some language, but rather for the metrical poetry
of some particular poet. One consequence of this
realization is that researchers cannot resort to
their own intuitions about grammaticality. Furthermore, unlike researchers working on a natural language in which they themselves lack intuitions, those working on metrical poetry rarely
have the opportunity for field work. The situation
is parallel to that of working with a dead language, or a closed corpus. Naturally, focus on the
grammar of an individual does not preclude
theories of universal poetics, with various parametric possibilities. Indeed, among the most
important work in the field are theoretically
insightful characterizations of di(erences between the grammars of individual poets; an example of this appears in 3.1.
2.2. Grammaticality
When working under such circumstances,
the applicability of the notion of grammaticality
requires reflection. Like generative grammarians
working in other domains, the phonologist studying metrical poetry is concerned to develop a
model which not merely produces those utterances that the native speaker does, but which also
crucially fails to produce those utterances that a
native speaker does not produce. A central premise that must be accepted if generative metrics is
to have any credibility is that a grammar of a
poets metrics can be constructed on the basis of
the poetry which that poet produced. For example, the collective utterances in Shakespeares
corpus are taken to present the entire range of
his possible constructions. Furthermore, patterns
which are absent from his corpus are not so accidentally. From this methodological assumption,
there emerges an unambiguous notion of
grammaticality: patterns found in the corpus are

grammatical, while those absent from the corpus

are ungrammatical. Hence, there is a tendency to
conduct research on poets with large corpora; a
large corpus increases the likelihood of both robust and subtle (un)grammaticality distinctions.
To elaborate on the relevance of the generativist notion of grammaticality to studies of metrical poetry, consider Halle & Keysers (1966, et
seq.) proposal that a metrical grammar be characterized with two properties: the relevant abstract
template and the appropriate mappings between
that template and actual lines of poetry. This
proposal highlights one central concern of even
the most recent research, e.g. Biggs (1997), Golston & Riad (1997), Hanson & Kiparsky (1996),
Helsloot (1995). Indeed, for readers new to research in generative metrics, one hastens to point
out that mapping between a line of poetry and an
abstract metrical template is not simply a matter
of aligning stresses in words with metrically
strong positions. One need only examine a few
lines from major poets to determine that mismatches of this type alone are inadequate to
make determinations of grammaticality; cf. the
examples in (1), which are drawn, respectively,
from Shakespeare, Sonnet I; Shakespeare, Sonnet
CXXXIII; Donne, Progress of the Soul; Hopkins,
Gods grandeur.

Mking a fmine where abndance lies

ws w s w
s w s w
Prson my heart in thy steel bsoms ward
w s w
s w
In vaine this sea shall enlrge, or enrugh
w s
w s
w s w
s w s
Gnertions have trod, have trod, have trod
w s ws

These lines are all grammatical in the familiar

sense defined above, i.e., they appear in the corpora of the poets. Yet they all contain mismatches between lexical stresses and the core template
for iambic pentameter. Leaving aside for the
moment the possibility of deviation from a one-toone correspondence between positions and syllables, there are two logically possible mismatches:
stressed syllables might appear in metrically
weak positions (e.g., the emboldened cases
above), and unstressed syllables might appear in
metrically strong positions (e.g., the unstressed
syllables of generations in (1d)). The distribution
of these two logical possibilities throughout the
work of English poets is quite uneven: unstressed
syllables in metrically strong positions are commonplace, while the appearance of a stressed
syllable in a metrically weak position is remarkable. In this sense, English poets leave metrically
strong positions unregulated: stressed and unstressed syllables freely appear in those
positions. In contrast, the poets strictly regulate
metrically weak positions; considerable work
within generative studies of English metrics has
been focused on determining the precise situations in which stressed syllables are allowed into
metrically weak positions. One central observation on this matter is that monosyllabic words
(e.g. steel in (1b)) appear in metrically weak positions much more freely than the stressed syllables of polysyllabic words (Magnuson & Ryder
1971). This observation becomes theoretically

Page 3

coherent through the defining insight of metrical

phonology, namely that stress is a relational phenomenon (Liberman 1976; Liberman & Prince
1977; van der Hulst 1995; Hayes 1995). With the
metrical understanding of stress as relative, the
behavior of monosyllabic words is not surprising:
the monosyllabic word rule (MWR) allows any
syllable lacking relative prominence within the
word to appear anywhere in the poetic line; i.e.,
the distribution of monosyllabic words is unrestricted. Only those syllables which are prominent relative to other syllables within the word
are prevented from appearing in metrically weak
positions (Kiparsky 1977). Other constraints, as
we will see below, may operate to allow rigidly
delimited exceptions.
The MWR is one central principle distinguishing grammatical lines from ungrammatical
constructs, cf. (2). It is indeed the case that for
Shakespeare, a line such as (2b), a construct, can
coherently be labeled ungrammatical, insofar as
his corpus is entirely devoid of examples in which
a syllable bearing stress in a non-phrase-initial
polysyllabic word appears in a metrically weak
position, as with private and widowers in (2b).
The polysyllabic words in (2a), from Sonnet IX,
all conform to the MWR. (The violations of the
MWR seen in (1a,b) are violations of a type which
Shakespeare allows, namely mismatches in
phrase initial position, so-called initial inversion,
the initial trochee being construed as an inverted
iamb, cf. the related discussion in 3.)

When very prvate wdow well may keep

w s w sw
w s
*When all prvate wdowers well may keep
w s
ws w
w s

We turn now to illustrations of central issues in

current research, starting with an analysis of a
contrast between Shakespeare and Milton, and
the relevance of that contrast to phonological

The prosodic hierarchy

Analyses of metrical poetry have been particularly useful in developing arguments for various
levels of the prosodic hierarchy, such as the mora, the foot, the prosodic word, and even the
prosodic phrase; for a recent example and discussion of the literature, cf. Helsloot (1995). Generative metrics can take less credit for uniquely
developing arguments for the syllable, since this
is assumed to be the basic component of a line of
metrical poetry in generative and non-generative
traditions alike.
To illustrate the contribution of metrical
analyses to argumentation for the prosodic hierarchy, we consider two di(erent analyses highlighting the relevance of the foot, one from earlier
work in generative metrics, and one representing
a central area of contemporary research. Both of
these arguments are framed here to highlight
di(erences between the iambic pentameter of
Shakespeare and Milton.
For the purpose of these arguments or analyses, the grammars of both Shakespeare and Milton are taken to include the monosyllabic word
rule, illustrated in 2.2. As noted, the constraint
is not absolute. Both poets allow violations of the
MWR in a line (or phrase) initial foot. The first
argument below addresses one additional possible violation of the MWR allowed by Milton, but
not by Shakespeare.
Specifically, Milton allows inversions in noninitial, phrase-medial positions. Nonetheless, his
grammar does not freely violate the MWR. As
argued in Kiparsky (1977), Milton allows MWR
violations (relatively stressed syllables in metrically weak positions) only when the lexical foot
dominating the inverted syllables coincides
exactly with the metrical foot, as in (3a), from
Paradise Lost, and (3b), from Paradise Regained.
To the extent that this relationship between
lexical and metrical feet is crucial to an analysis
of Miltons inversion possibilities, it constitutes
an argument for the foot in both of these do-

mains: lexically and metrically. (We assume with
Kiparsky 1977 that a lexical foot dominates the
entirety of the words showing inversion in the
examples in (3).)

To the Grden of Bliss, thy seat preprd

[w s] [w s] [w s]
[w s] [ w s]
To their night-wtches in wrlike Parde
[w s] [ w
s] [w s] [w s] [w s]

Such inversions are relatively rare in Miltons

corpus. Much more common di(erences between
Shakespeares and Miltons meters are unaddressed by this foot-based analysis of inversion. However, current research in generative
metrics sheds light on some of these additional
di(erences, again through consideration of the
role of the foot in each of the two grammars, but
now via a radical extension of the conception of
the template for the line.
Earlier work focused on characterizing mappings between positions in the ideal template and
syllables in the actual line. Contemporary work is
now exploring the possibility that prosodic units
other than the syllable can be mapped onto terminal positions of the abstract metrical feet
(Golston & Riad 1994, 1995, 1996, 1997; Hanson
1995; Hanson & Kiparsky 1996; Helsloot 1995).
To illustrate this development, consider the possibility that iambic pentameter optionally consists of a sequence of five pairings of two feet,
rather than five pairings of two syllables.
At least two frequently encountered points of
variation between Shakespeare and Milton can
be understood by claiming that the grammar of
Shakespeares iambic pentameter allows a metrical position to be occupied by as much as a
moraic trochee (one heavy syllable or two light
ones), while the grammar of Miltons iambic pentameter limits the mapping between the line and
the abstract metrical position, such that a position is maximally occupied by a single syllable
(Hanson 1995; Hanson & Kiparsky 1996).
The corpus of each of these poets contains
lines which cannot be fully parsed into the canonical 10 positions; i.e., each poet allows extrametrical material in a line. A line from Shakespeare, however, can have up to two surplus
unstressed syllables, while a line from Milton will
not show more than one, cf. (4a), from Shakespeares Comedy of Errors, 2.2.118, and (4b),
again from Miltons Paradise Lost, 1.98. In the
proposed analysis, each poets lines are constrained in the same way insofar as they tolerate
exactly one extrametrical position. Since Shakespeares position can be occupied by a moraic
trochee, it follows that two light syllables may be

Unless I spoke or lookt or toucht or carvd to thee

[w s] [w s] [w s] [w s]
[w s]
w w
And high disdin, from sense of njured mrit
s] [w s] [w
s] [w s][w

The analysis under consideration here also provides insight into the treatment of a challenging
lexical stress pattern from English, namely the
initial dactyl pattern seen in words such as
justification, which have a sequence of two unstressed light syllables between two stressed
syllables. If a metrical position is occupied by a
maximum of one syllable, words with this stress
pattern are unusable by the poet since they will
necessarily incur a violation of the MWR, as in

js ti fi c tion
w sw s w
s ws w s

Indeed, this is precisely the situation we find for

Milton; the initial dactyl portion of the lexical is
almost entirely unusable for him. The only position in which Milton can use words with this
stress pattern at all is phrase initially, where the
option of initial inversion facilitates a grammatical scansion, cf. (6a), Paradise Lost 12.296. Since

Glot International, Volume 2, Issue 7, September 1996 [1997]

a metrical position in a line of Shakespeares
poetry can be occupied by as much as a foot, the
two light syllables between the stressed ones can
be parsed into one position, allowing such words
to be used anywhere in the line, so long as they
conform to the MWR, cf. (6b), The Tempest
1.2.234. (A complete discussion of the possibility
of mapping two syllables onto one metrical position requires further consideration of the phenomenon traditionally known as resolution;
space considerations preclude such a discussion
here. The interested reader is referred to the
following works and references therein: Cable
1991; Dresher & Lahiri 1991; Fulk 1992; Hanson
1991, 1993, 1995; Hanson & Kiparsky 1996;
Stockwell & Minkova, to appear.)

Jstifiction towards God, and peace

[w s][ws]
And are upon the mditerrnean float
s] [w s] [w s][w s] [w s]

By expanding the possibilities for mapping between the abstract metrical template and the
actual line of poetry such that a terminal in the
template may be either a foot or a syllable, we
take a significant step forward in understanding
some crucial aspects of English poetry. As an
example, we have seen that variation between
Shakespeare and Milton with respect to restrictions on extrametricality and on the appearance
of initial dactyls are straightforward consequences following from variation in the nature of the
terminal element in the verse foot. Alongside this
increase in our understanding of the poetry, the
analyses can be taken as evidence for the existence of the foot in the prosodic hierarchy. (The
present discussion ignores many details from
Hanson 1995 and Hanson & Kiparsky 1996 and
deviates from others for expository reasons; the
interested reader is referred to the original resources for a more thorough discussion.)

Optimality theory
Readers of this journal are familiar with the
impact that Optimality Theory (McCarthy &
Prince 1993; Prince & Smolensky 1993) has had
in selected spheres of linguistic research (cf.
Burzio 1995, or even Dresher 1996). It comes as
no surprise, then, that current work in generative metrics also bears the mark of OT. As elsewhere, the most interesting examples of this
influence are claims that OT o(ers new insights,
especially insights which in principle cannot be
replicated in derivational approaches. In this
section, I present three di(erent issues from
recent papers, all of which achieve their results as a
consequence of adopting OTinspired strategies.
4.1. Typology
The characterization of a set of metrical patterns as following from a hierarchy of violable
constraints introduces the possibility of evaluating individual patterns as more or less optimal.
In this sense, such a strategy formalizes a typology. Statistical studies of a corpus might then be
undertaken; a case in which the typological characterization of most to least optimal is reflected
in the statistical frequencies of these types is a
case which reveals the superiority of a constraintbased analysis. As an example, consider an analysis of classical Arabic meters (Golston & Riad
Our presentation of Golston & Riads (1996)
study stays at a somewhat abstract or idealized
level. Readers interested in the detailed analysis
of various types of lines, or in examples thereof,
are referred to the literature on Arabic metrics,
extensive references to which can be found in
Golston & Riad (1996), Maling (1973), and Prince
Ancient Arabic meters are traditionally divided into eleven di(erent types. Studies of the
relative frequencies of these types reveal that the
four most common ones account for as much as
8090% of the poetry. Of these four, the single
most common one, known as the t awl, accounts

Page 4

for as much as half of all poetry, depending on

the corpus. The t awl is consistently the most
frequent type, across corpora; the kaml, wafir
and bast are consistently the next three, although their relative frequencies vary with the
In all of the eleven types, the line is a series
of metra, each of which consists of two verse feet,
which in turn consist of H, LL, or L. The typology
follows from assessing violations of two wellmotivated constraints: CLASH, which bans stress
on adjacent syllables, and LAPSE, which bans a
string of two unstressed moras. These can be
violated either within the verse foot (CLASH-FT,
LAPSE-FT) or within the metron (CLASH-MTN,
LAPSE-MTN). Local violations are the most costly, hence violations within the foot are worse
than those across feet.
The first major division, between the four
most common types on the one hand and the
seven very rare types on the other, follows from
LAPSE-FT. Specifically, the verse feet of the four
most common types are based on the iambic
structure [LH], while the uncommon types have
either the requirement or the possibility of instantiating the trochaic structure [HL]. On the
assumption that stress on a heavy syllable is on
the first of two moras, the [LH] iamb is a perfect
structure with respect to CLASH and LAPSE.
There is only one stress, hence no violation of
CLASH; the stress occurs on the medial of the
three moras in the foot, and there is therefore no
string of two unstressed moras, i.e. no violation of
LAPSE. The trochaic form [HL], on the other
hand, inherently violates LAPSE at the moraic
level, leaving the second mora of H and the lone
mora of L as a string of two unstressed moras.
The second major division is between the
most common iambic pattern, the t awl, and the
remaining three as a class. The analysis of these
lines is such that the t awl contains no inherent
violation of LAPSE or CLASH, either at the level of
the verse foot or at the level of the metron. In
this sense it is the most optimal of the patterns, a
typological consequence of the analysis made
more credible by its statistical dominance. The
other three types all contain an inherent violation of one constraint. To illustrate this point,
consider the following tableau (page 5) in which
violations inherent in the patterns are represented. To read this tableau, two important abstractions which Golston & Riad (1996) employ are
important; s represents either L or H, while f
represents either LL or H. The metron for the
bast pattern, for example, can be either [LH.LH]
or [HH.LH]; similarly, the metron for the kaml
pattern can be either [LLH.LH] or [HH.LH].
This tableau distinguishes the t awl from the
remaining three patterns, the latter group being
identifiable as those iambic meters whose very
pattern bears a violation of one of the constraints. The interpedal HL sequence in the
kaml and bast violate LAPSE-MTN, while the
Hf sequence of the wafir, regardless of whether
f is realized as H or LL, violates CLASH-MTN.
The t awl is clearly superior insofar it lacks
an inherently encoded violation of any single constraint under consideration. However, its superiority crucially depends on the abstraction of H
and L to s. When the t awl appears as LH.LH, it
commits the same violation as the kaml and the
bast patterns; when it is LH.HH it commits a
violation of both CLASH-FT and CLASH-MTN.
There are only two possible instantiations of the
taawil pattern; both incur some violation.
The observation that the abstraction of H
and L to s is crucial for the analysis, however,
underscores the importance of the very enterprise of generative metrics; the statistically dominance of the t awl is undisputed, and by bringing
the force of phonological theory to the matter, an
analysis of the Arabic verse types which reflects
the statistics becomes available. Since it is not
clear how alternative theories might o(er an
equally insightful analysis, the crucial use of a
methodology in which constraints are ranked and
violable lends credibility to OT.



t awl








Glot International, Volume 2, Issue 7, September 1996 [1997]



4.2. Factorial typology

One important topic of research within OT
involves the implications of ranking and re-ranking constraints. The power of the theory is partially derived from the re-ranking option; indeed,
any possible ranking (i.e., grammar) which is not
instantiated might suggest excessive power.
Hence, an achievement of some analyses is to
consider a factorial typology of the relevant constraints, and to associate some set of data with
every possible ranking.
In a forthcoming article, Hayes &
MacEachern (to appear) analyze a corpus of folk
verse with a set of constraints, and then consider
the implications of positing precisely those constraints by exploring the associated factorial
typology. They demonstrate that their analysis
predicts exactly those cases which are found in
the corpus, and indeed they pursue some analyses of statistical frequency in a manner conceptually akin to Golston & Riads (1996) strategy,
seen in the previous section. To illustrate the role
of the factorial typology and its relationship to a
corpus of data, I present the constraint which is
conceptually most central in Hayes & MacEacherns analysis, along with its three possible
instantiations, in a total of five di(erent combinations. (The examples in this section come from
various corpora; for references to these sources,
consult Hayes & MacEachern.)
Hayes & MacEachern claim that the basic
folk verse structure is a quatrain consisting of
two couplets; each couplet consists of two lines,
and each line consists of four beats. (The binarity
of the system breaks down in the mapping from
one line to four beats, although one might explore
the possibility that the four beats represent two
verse feet, each of which in turn has two constituents. Such an approach would make Hayes &
MacEacherns system thoroughly binary.) Quatrains may highlight various of these constituents; one may highlight the line, another the
couplet, yet another the quatrain itself. The basic
strategy for making a constituent salient is truncation, i.e., the non-realization of a beat. Di(erent verse types can be described, then, by positing three saliency constraints, di(ering only by
domain of application (line, couplet, quatrain).
Here again we see an important development
made possible through an OT analysis; earlier
work in generative metrics includes discussion of
complexity and various strategies for formalizing
its representation. None of these strategies can
represent a preference for leaving a position in
the template unfilled.
Three di(erent verse types are predicted by
ranking each of the three of these constraints
highest. Consider them now in turn. The constraint SALIENT(LINE) selects as optimal those
patterns which truncate each line. For example,
patterns in which only three of the four beats are
realized. There are two such patterns illustrated
below: lines in which the third beat is on the final
syllable of the line are labeled 3, while lines
with an o(beat after the third beat are labeled
3f, after the traditional terminology which dubs
lines with a final o(beat as feminine. The following example, in which each line is truncated
and therefore in accordance with SALIENT(LINE),
is a series of feminine lines missing the final
beat, i.e., 3f 3f 3f 3f, based on a contemporary
childrens song.
Willoughby, wallaby, wespen,
An elephant sat on Espen,
And Willoughby, wallaby, windre,
An elephant sat on Sindre.



The constraint SALIENT(COUPLET) will select

as optimal those verse structures in which the
final beat of each couplet is unrealized. The following example is 4343.
Theres two little brothers going to school.
The oldest to the youngest called:
Come go with me to the green shady grove
And Ill wrestle you a fall.

Verse which makes the quatrain salient is selected as optimal when SALIENT(QUATRAIN) is highly
ranked. The unity of the quatrain is emphasized
by having three equally long lines and then truncating the fourth, for example, 4443, as seen
One little, two little, three little Indians,
Four little, five little, six little Indians,
Seven little, eight little, nine little Indians,
Ten little Indian boys.

When two saliency constraints are highly ranked,

two of the three verse units are emphasized. This
possibility follows, again, from re-ranking. For
example, when both SALIENT(COUPLET) and SALIENT(LINE) are highly ranked, the optimal pattern should truncate each line, but the second
and fourth ones more than the first and third
ones. One such pattern could be 3f 3 3f 3, as in
the following example.
Last night as I lay on my pillow,
Last night as I lay on my bed,
Last night as I lay on my pillow
I dreamed little Bessy was dead.

Yet another di(erent pattern emerges as optimal

when the highly ranked constraints are
this case the optimal pattern should truncate
each line, but now the fourth one should be truncated more than the first three. The following
example is 3f 3f 3f 3.
Up Eliza, poor girl;
Hoot Eliza, poor girl;
Up Eliza, poor girl;
She died on the train.

Hayes & MacEacherns insight into truncation

and its role in the grammar of verse also makes
predictions about ungrammatical verse types.
The saliency constraints predict that we should
not find verse types with truncation patterns that
fail to highlight a genuine unit of the verse prosody. For example a quatrain with the pattern 3444
shows truncation only in the first line, which is to
say that on the whole no emphasis is given to the
quatrain, the couplet, or to the line (at least not
pervasively, as required by SALIENT(LINE)). Similarly, a verse with a 3434 structure is also ungrammatical, following the same type of reasoning. The 3434 case demonstrates that the mere
possibility of construing the quatrain as two
units of two lines each is not a su+cient strategy
for making the couplet salient. Saliency follows
exclusively from truncation at the end of the unit.
To convince the reader of the ungrammaticality
of these two patterns, consider the following
examples, respectively 3444 and 3434 (loosely
after Hayes & MacEachern).
*Ten little Indian boys.
*Nine little, eight little, seven little Indians,
*Six little, five little, four little Indians,
*Three little, two little, one little Indian.
*Hickory, dickory, dock,
*The frightened mouse ran up the clock,
*Hickory, dickory, dunn,
*The mouse ran down when the clock struck one.

Page 5

This presentation of Hayes & MacEacherns work

is greatly simplified and selective. The point here
is to illustrate their exploration of a factorial
typology for a set of constraints, and their strategy for formalizing a preference for unfilled templatic positions. Their study is an important contribution to work in OT insofar as the corpora
they are working with provide empirical instantiation of the grammars they predict. To give the
reader an idea of the scope of such an undertaking, I note that Hayes & MacEacherns analysis
includes a total of eleven constraints, three of
which are inviolable, and hence ranked at the top
of all possible grammars. The remaining eight
constraints can be freely re-ranked. Doing so
yields 8! rankings, i.e., 40,320 possible
grammars. The carefulness and thoroughness
with which Hayes & MacEachern deal with this
fact is another of their contributions.
4.3. Metrical vs. natural language grammars
Some recent work in OT includes rather farreaching proposals about structural properties of
grammars of natural language. One example of
such a proposal is the claim that syntax outranks
phonology. This assertion about the structure of
an OT grammar specifically suggests that all
constraints of the syntax are ranked above all
constraints of the phonology. For examples and
argumentation, see Golston (1995) and Rice &
Svenonius (1997).
In the face of this claim, methodological considerations motivated by factorial typologies, as
discussed in 4.2, predict a grammar with the
other ranking, i.e., a grammar in which phonology outranks syntax. Recent work claims that this
is precisely the way to characterize grammars of
metrical poetry, as opposed to grammars of natural language (Rice, to appear). These references
also include discussion of the position of morphology in the grammar; for reasons of space, we
focus on the central argument regarding the
relationship between phonology and syntax.
Data facilitating an argument that phonology outranks syntax will be data in which syntactically ill-formed structures are optimal because
of their phonological properties, while syntactically well-formed structures are non-optimal,
again because of their phonological properties.
Youmans (1983) presents data from Shakespeare
with precisely these properties. Consider the
following tableaux:
Sonnet LVI


Thy edge should blnter be than ppetite

s w s w s ws
Thy edge should be blnter than ppetite
s w s
w s ws

Sonnet LXV



When rocks imprgnable are not so stout

w s
s w sw
s w s
When imprgnable rocks are not so stout
w s w s w s
s w s




In each of these, the first candidate violates the

syntax and the second one violates the metrical
phonology, here limited to the MWR. Hence,
selection of the optimal candidate crucially depends on the ranking of these components. The
first candidate in the first tableau conforms to the
MWR with both blunter and appetite. The word
order of this line, however, is deviant insofar as it
exemplifies a pattern typical of OV structure by
placing the predicative adjective in front of the
verb. The second candidate is a construct in
which the syntax is corrected. This candidate is
metrically ill-formed since the stressed syllable of
blunter appears in a metrically weak position.
The first candidate is the actual line, and hence
we have a well-motivated argument that phonology outranks syntax. Parallel reasoning follows for
the candidates in the second tableau, in which
the optimal candidate displays a marked word


Glot International, Volume 2, Issue 7, September 1996 [1997]

order with a post-nominal adjective, and in which,

again, the syntactic violation is irrelevant. Further
exploration of the extent to which the syntax may
be violated remains a topic for future research.
Shakespeares poetry includes many lines
with marked syntax. This syntactic markedness
is motivated to achieve conformity with the metrical
rules. In this way, Shakespeares poetry reveals the
general nature of metrical grammars: when phonology and syntax conflict, the syntax yields.
In this section, we have seen that an OTbased conception of grammar invites consideration of alternative rankings of the major components of that grammar. Not only does this point
of view give greater insight, for example, into the
syntactic freedom typically ascribed to the poet,

but it also is of theoretical interest, generally in

terms of the factorial typology methodology it
validates, and specifically by highlighting the
central di(erence between grammars of natural
language and grammars of poetry.



The examples given in 34 illustrate selections from the state of the art of research in
generative metrics. Each of these examples is of
interest from both of the perspectives mentioned
in the introduction. We get a greater understanding of the formal structure of poetry, whether it
be variation between Shakespeare and Milton,

Page 6

the inventory of Arabic verse types, restrictions

on folk verse, or the motivations for poetic license. Perhaps even more important, at least for
readers of this journal, are the contributions
these studies make to linguistic theory; i.e., the
importance of metrical poetry as a source of data,
and its analysis as a source of theoretical insight.
The examples in this article o(er a wide variety
of such cases, ranging from evidence for the
prosodic hierarchy (3), to support for the concept
of violable constraints (4.1) and their hierarchical (re)arrangement, both at the level of individual constraints (4.2) and at the level of major
components (or blocks of constraints) (4.3), to
insight into the variation between grammars of
language art and grammars of language (4.3).

A Generative Metrics Bibliography

rnason, K. (1991). The rhythms of Drttkvtt
and other old Icelandic metres. Reykjavk:
Institute of Linguistics, University of Iceland.
Attridge, D. (1982). The rhythms of English poetry. London: Longmans.
Beaver, J. (1971a). Current metrical issues. College English 33, 177197.
Beaver, J. (1971b). The rules of stress in English
verse. Language 47, 586614.
Biggs, H. (1996). A statistical analysis of the
metrics of the classic French decasyllable and
the classic French alexandrine. Ph.D. dissertation, UCLA.
Biggs, H. (1997). A generative metrical analysis
of Ronsards Les Amours. Paper presented at
HILP3, Amsterdam.
Bjorklund, B. (1978). A study in comparative
prosody: English and German iambic pentameter. Ph.D. dissertation, Indiana University.
Burlings, R. (1966) The metrics of childrens
verse: a cross-linguistic study. American
Anthropologist 68, 14181441.
Burzio, L. (1995) The rise of optimality theory.
Glot International 1(6), 37.
Cable, T. (1976). Recent developments in metrics.
Style 10, 313328.
Cable, T. (1991). The English alliterative tradition. Philadelphia: University of Pennsylvania Press.
Cable, T. (1996). Clashing stress in the metres of
Old, Middle, and Renaissance English. In
English historical metrics, C. McCully & J.
Anderson (eds.), 729.
Chen, M. (1979). Metrical structure: evidence
from Chinese poetry. Linguistic Inquiry 10,
Cureton, R. (1992). Rhythmic phrasing in English
verse. London: Longmans.
Devine, A. & L. Stephens (1975). The
abstractness of metrical patterns: generative
metrics and explicit traditional metrics.
Poetics 4, 411430.
Dresher, B.E. (1996). The rise of optimality theory in first century Palestine. Glot International 2(12), 8.
Dresher, B.E. & A. Lahiri (1992). The Germanic
foot: metrical coherence in Old English. Linguistic Inquiry 22, 251286.
Fulk, R. (1992). A history of Old English meter.
Philadelphia: University of Pennsylvania
Golston, C. (1995). Syntax outranks phonology:
evidence from Ancient Greek. Phonology 12,
Golston, C. & T. Riad (1994). Prosodic metrics.
Manuscript, UC Berkeley and Stockholm
Golston, C. & T. Riad (1995). Direct metrics. LSA,
San Diego.
Golston, C. & T. Riad (1996). The phonology of
Classical Arabic meter. Linguistics 35,

Golston, C. & T. Riad (1997). Stress has little to

do with Old English meter. Workshop on
Historical Germanic Linguistics, University
of Troms.
Guthrie, S. (1988). Prosody and the study of
Chaucer: a generative reply to Halle-Keyser.
Chaucer Review 23, 3049.
Halle, M. & S.J. Keyser (1966). Chaucer and the
study of prosody. College English 28,
Halle, M. & S.J. Keyser (1971a). Illustration and
defense of the theory of iambic pentameter.
College English 33, 154176.
Halle, M. & S.J. Keyser (1971b). English stress:
its form, its growth, and its role in verse. New
York: Harper & Row.
Hammond, M. (1991). Poetic meter and the arboreal grid. Language 67, 240259.
Hammond, M. (1992). Review of Rhythm and
meter, P. Kiparsky & G. Youmans (eds.).
Phonology 9, 358362.
Hammond, M. (1993). Review of The English
alliterative tradition, Thomas Cable. Language 69, 400404.
Hanson, K. (1991). Resolution in modern meters.
Ph.D. dissertation, Stanford University.
Hanson, K. (1993). Resolution: evidence from
modern English metrics. NELS 23, 159173.
Hanson, K. (1995). Prosodic constituents in poetic
meter. WCCFL 13, 6277.
Hanson, K. & P. Kiparsky (1996). A parametric
theory of poetic meter. Language 72,
Hayes, B. (1983). A grid-based theory of English
meter. Linguistic Inquiry 14, 35794.
Hayes, B. (1984). The phonology of rhythm in
English. Linguistic Inquiry 15, 3374.
Hayes, B. (1988). Metrics and phonological theory. Linguistics: The Cambridge Survey, Vol.
2: Extensions and Implications, F.
Newmeyer (ed.), 220249. Cambridge University Press.
Hayes, B. (1989). The prosodic hierarchy in meter. In Phonetics and phonology, P. Kiparsky
& G. Youmans (eds.).
Hayes, B. (1995). Metrical stress theory: principles and case studies. Chicago: University of
Chicago Press.
Hayes, B. & M. MacEachern (1997). Are there
lines in folk poetry? UCLA Working Papers
in Linguistics, 125142.
Hayes, B. & M. MacEachern (to appear). Folk
verse form in English. Language. [Rutgers
Optimality Archive #119,
http://ruccs.rutgers.edu /roa.html.]
Helsloot, K. (1995). Metrical prosody: a templateand-constraint approach to phonological
phrasing in Italian. Ph.D. dissertation, University of Amsterdam/HIL. The Hague: Holland Academic Graphics.
Hobsbaum, P. (1996). Metre, rhythm and verse
form. London: Routledge.
Hulst, H. v.d. (1995). Metrical phonology. Glot
International 1(1), 36.

Keyser, S. (1969). Old English prosody. College

English 30, 331356.
Kiparsky, P. (1972). Metrics and morphophonemics in the Rigveda. Contributions to
generative phonology, M. Brame (ed.). Austin: University of Texas Press.
Kiparsky, P. (1973). The role of linguistics in a
theory of poetry. Ddalus: The Journal of the
American Academy of Arts & Sciences 3,
Kiparsky, P. (1975). Stress, syntax, and meter.
Language 51, 576616.
Kiparsky, P. (1977). The rhythmic structure of
English verse. Linguistic Inquiry 8, 189247.
Kiparsky, P. (1989). Sprung rhythm. In Phonetics
and phonology, Kiparsky & Youmans (eds.).
Kiparsky, P. & G. Youmans (eds.) (1989) Phonetics and phonology, vol. 1: Rhythm and
Verse. San Diego: Academic Press.
Klein, W. (1974). Critical remarks on generative
metrics. Poetics 12.
Kooij, J. (1997). Metrical structure and the syntaxphonology interface. Paper presented at
HILP3, Amsterdam.
Leech, G. (1969). A linguistic guide to English
poetry. London: Longmans.
Liberman, M. (1976). The intonational system of
English. Ph.D. dissertation, MIT.
Liberman, M. & A. Prince (1977). Stress and
linguistic rhythm. Linguistic Inquiry 8,
Magnuson, K. & F. Ryder (1970). The study of
English prosody: an alternative proposal.
College English 31, 789820.
Magnuson, K. & F. Ryder (1971). Second
thoughts on English prosody. College English
33, 198216.
Maling, J. (1973). The theory of classical Arabic
metrics. Ph.D. dissertation, MIT.
Manaster-Ramer, A. (1994). Stefan George and
phonological theory. Phonology 11, 317323.
McCarthy, J. & A. Prince (1993). Prosodic morphology I. Unpublished manuscript. University of Massachusetts, Amherst and Rutgers
McCully, C. (1988). The phonology of English
rhythm and metre, with special reference to
Old English. Ph.D. dissertation, University
of Manchester.
McCully, C. & J. Anderson (1996). English historical metrics. Cambridge University Press.
Minkova, D. (1996). Nonprimary stress in early
Middle English accentual-syllabic verse. In
English historical metrics, McCully & Anderson (eds.), 95119.
Neijt, A. (1993). Stress shift in Dutch hexameters. In Linguistics in the Netherlands 1993,
F. Drijkoningen & K. Hengeveld (eds.). Amsterdam: John Benjamins.
Pera, C. (1980). Spanish verse and the theory of
meter. Ph.D. dissertation, UCLA.
Preminger, A. & T. Brogan (eds.) (1993). The new
Princeton encyclopedia of poetry and poetics.
Princeton University Press.

Prince, A. (1989). Metrical forms. In Phonetics and phonology, Kiparsky & Youmans
Prince, A. & P. Smolensky (1993). Optimality
theory: constraint interaction in generative grammar. Unpublished manuscript.
Rutgers University and University of
Colorado, Boulder.
Redford, M. (to appear). Four-beat rhythm
and the metrical hierarchy. In Phonology in progress progress in phonology.
Proceedings of HILP3., G. Booij and J.
van de Weijer (eds.). The Hague: Holland
Academic Graphics.
Rice, C. (to appear). Ranking components:
the grammar of poetry. In Phonology in
progress progress in phonology. Proceedings of HILP3., G. Booij and J. van
de Weijer (eds.). The Hague: Holland
Academic Graphics.
Rice, C. & I. Mseide (1997). Metrically motivated variation in the Chaucerian past
participle. Manuscript, University of
Rice, C. & P. Svenonius (1997). Prosodic V2
in Northern Norwegian. Paper presented
at the Hopkins Optimality Theory Workshop and Maryland Mayfest, Baltimore.
Manuscript, University of Troms.
Russom, G. (1987). Old English meter and
linguistic theory. Cambridge University
Russom, G. (1996). Purely metrical replacements for Kuhns laws. In English historical metrics, McCully & Anderson
(eds.), 3041.
Schlerman, B. (1984). The meters of John
Webster. Ph.D. dissertation, University
of Massachusetts, Amherst.
Standop, R. (1975). Metric theory gone
astray. Language & Society 8, 6077.
Stockwell, R. (1997). On recent theories of
metrics and rhythm in Beowulf. In English historical metrics, McCully & Anderson (eds.), 7394.
Stockwell R. & D. Minkova (to appear). Old
English metrics and the phonology of
resolution. Nowele.
Suzuki, S. (1985). The role of syllable structure in Old English poetry. Lingua 67,
Tarlingskaja, M. (1993). Strict stress-meter in
English poetry. University of Calgary
Yip, M. (1980). The metrical structure of
regulated verse. Journal of Chinese
Linguistics 8, 107125.
Yip, M. (1984). The development of Chinese
verse: a metrical analysis. Language,
Sound, Structure, M. Arono( & R.
Oehrle (eds.), MIT Press.
Youmans, G. (1982). Hamlets testimony on
Kiparskys theory of meter. Neophilologus 66, 490503.
Youmans, G. (1983). Generative tests for
generative meter. Language 59, 6792.
Youmans, G. (1989). Miltons meter. In Phonetics and phonology, Kiparsky &
Youmans (eds.).
Youmans, G. (1996). Reconsidering Chaucers
prosody. In English historical metrics,
McCully & Anderson (eds.), 185209.
Wimsatt, W. (1970). The rule and the norm:
Halle and Keyser on Chaucers meter.
College English 31, 774788.
Zec, D. & S. Inkelas (1990). Prosodically
constrained syntax. In The phonology
syntax connection, S. Inkelas & D. Zec
(eds.), University of Chicago Press.

Glot International, Volume 2, Issue 7, September 1996 [1997]

Page 7


Elan Dresher
The Olde Yankee Grammarian
In his book The Language Instinct, Steven Pinker
observes that the study of grammar provides rare
intellectual pleasure and ought to be of interest
to anyone who is at all curious about the nature
of the human mind. However, he points out that
the highly technical nature of much scholarly
writing in linguistics poses problems for the average reader. And who can blame the grammarphobe, he writes, when a typical passage from
one of Chomskys technical works reads as follows?
He then quotes the following passage from
page 79 of Barriers:
To summarize, we have been led to the following conclusions,
on the assumption that the trace of a zero-level category must
be properly governed:


VP is q-marked by I.
Only lexical categories are L-markers, so that VP is not
L-marked by I.
q-government is restricted to sisterhood without the
qualification (35).
Only the terminus of an X0chain can q-mark or Casemark.
Head-to-head movement forms an A-chain.
SPEC-head agreement and chains involve the same
Chain coindexing holds of the links of an extended
There is no accidental coindexing of I.
I-V coindexing is a form of head-head agreement; if it is
restricted to aspectual verbs, then base-generated structures of the form (174) count as adjunction structures.
Possibly, a verb does not properly govern its q-marked

Let us grant, for the sake of argument, that this

passage is not entirely self-explanatory to someone unfamiliar with the preceding 78 pages.
However it might be paraphrased, is it reasonable to suppose that a person who is curious
about grammar but who has no background in
the subject might be able to pick up a book like
Barriers, turn to page 79 (just before the section
titled Some Further Problems), and expect to
make sense of a set of conclusions that it took the
preceding 78 pages to build up to? Such a person
would be better o(, I think, to start with something a bit more basic: one could try starting with
page 1 of Barriers, for instance. Those with somewhat less background would profit from a textbook, such as Elizabeth Cowpers A concise introduction to syntactic theory. This book presents a
brief but lucid discussion of some of the central
points of the Barriers theory in the last section of
the last chapter. The general reader, however,
will probably profit more from reading Pinkers
book, which does not attempt to explain the Barriers theory at all, but is written with just such a
reader in mind.
Pinkers point, of course, is that more linguists should write for such readers, and I agree
that this would be very desirable and important
to the field. Consider, though, that it takes Pinker twenty pages to explain the barest rudiments
of X-bar theory and the concept of traces. Writing
in this mode, how many pages would it take to
put across the contents of Barriers?
Even more intriguing, to my mind, is this
question: what is it that makes Barriers and
writings on linguistic theory more generally
seem so hard? On a scale of intellectual complexity, is syntax really so much more di+cult than
physics, or neurology, or installing drywall, or

handicapping racehorses, all of which are understood to some extent by millions of amateurs?
I dont think so. I think that the passage
from Barriers quoted above, to continue with this
example, is di+cult for two reasons. The more
trivial one is that the terminology is unfamiliar
to many people. If you dont know what a zerolevel or lexical category is, or what proper government and q-marking are, you cant begin to
make sense of the passage.
Terminology aside, I submit that the main
reason why the concepts of linguistic theory appear di+cult is because they are abstract. If
these ideas were made concrete, they would seem
relatively simple compared to many other things
people deal with every day. Here, then, follows
my proposal for how to make linguistic theory
accessible to a wider audience.
Do you love language but hate grammar? Do you
wonder why you can understand chaos theory but
not the logic of markedness? Has I-language
always been not-for-me-language to you? Help is
on the way! In a unique joint venture, MIT Press
is teaming up with Home Hardware and the
Public Broadcasting System to create a set of
instructional videos designed to complement
their monograph series. Here is a preview of a
portion of the video that will accompany Barriers:
Welcome to another installment of the Olde
Yankee Grammarian. Were outside our grammar
workshop here in Wellfleet on Cape Cod, and
were all set to begin an exciting new project.
Well, Norm, what are you going to show us today?
Hi, Howard. After our programs on Whmovement, when we built chains to derive sentences like (1), a lot of viewers wrote in asking
why (2) is bad:
*Which car do you believe that he stole?
*Which car did you cry when he sold?

Weve got to stop the Wh-word from getting out of

its clause in (2), while still allowing it to escape
in (1). The way were going to do it is by building
some barriers.
Barriers? Sounds intriguing!
Lets head over to the workshop.
(Cut to the interior of the workshop: a workbench,
tools, boxes of supplies and parts.)
Well start by reviewing how to extract the
Wh-phrase in (1). We know that the sentence
starts o( as (3a), and ends up as (3b):

You believe [CP that he stole which car]

Which cari do you believe [CP that he stole ti]

We have (3a) laid out on the workbench. When

we move the Wh-phrase, it will leave behind an
attached trace. So first Ill lift the Wh-phrase out
of its slot and lay it down on the side. While Im
doing that, Howard, could you reach into the
knickknacks box and grab some traces and indices?
Ok, Norm, heres a trace and some is.
Thanks, Howard. It doesnt really matter

Chomsky, N. (1986). Barriers. Cambridge,
Mass.: MIT Press.
Cowper, E. (1992). A concise introduction to
syntactic theory: the government-binding
approach. Chicago, Ill.: The University of
Chicago Press.
Editors of Sunset Books (1995). Basic Woodworking. Menlo Park, Calif.: Sunset
Publishing Corporation.
Pinker, S. (1994). The language instinct: how
the mind creates language. New York:
William Morrow and Company.

e-mail (pdf-only subscription)

air mail
surface mail
surface mail in three batches
(March August December)
postal giro 6261107
cheque (add NLG 17.50/$10.5)
credit card
AMEX Euro/Mastercard
card number
expiry date
zip code + house number

Send my issues by
I will pay by

I am an

until cancellation
one volume
two volumes

My subscription lasts

home page:

current issue
volume 3
volume 2
volume 1


My subscription starts with


individual subscriber


phone: +31 70 4480 203

fax: +31 70 4480 177
e-mail: mail@hag.nl
http: www.hag.nl
P.O. Box 53292
2505 ag The Hague
The Netherlands



In (3a), the CP is the complement of the verb

believe, and the verb assigns it a thematic
role. This is not the case in (2), where the CP
is an adverbial adjunct of the verb.
Anybody who watched our program on qmarking will see where youre going, Norm:
the verb believe directly q-marks a complement that has a thematic role. So Ill get a qmarker out of the box and tie one end to the
verb, and wrap the other around the CP, so
the little q is sitting on the CP. Now Ill pull
the CP snug to the verb, and whoa! Whats
happening here, Norm?
(Chuckles.) Whats happening, Howard,
is that youre discovering what L-marking is
all about. A verb is a lexical category, and

Yes, I subscribe to Glot International!

You believe [CP that he stole which car]

Page 8

recreate it in wood. When the conversation turns

to Minimalism, youll be moving and merging
with the best of them. When they ask you, What
were you doing all that time in your workshop?
youll proudly answer, What was I doing? Almost
At participating hardware stores. Functional
projections sold separately. O(er not valid where
prohibited by law.

Coming soon, by popular demand, to the Olde

Yankee Grammarian: Having trouble following
the talk about Chapter Four down at the general
store? Let Norm and Howard show you how to

order by secure webserver, e-mail, fax or regular mail:


every lexical category has a little L attached to it.

When the verb does not q-mark its complement,
as in (2), the words are far apart and the L doesnt make contact with anything. But when a lexical category directly q-marks its complement, the
verb and the complement are drawn closely together, and the foot of the L pushes against the
barrier of the complement
And forces open the sliding blocker, leaving
enough room for the trace to get through! Well,
thats pretty ingenious, Norm.
Maybe so, Howard, but so simple that any
child can do it. Well, that just about does it for
this edition of the Olde Yankee Grammarian.
Now, heres some information about an upcoming
program weve had lots of requests for.

Holland Academic Graphics

which letter we choose, but experienced grammarians usually start with i, then go on to j and k
in order. It helps keep things straight in complicated constructions, believe me! We nail one
index to the Wh-word, like so. Howard is nailing
another one to a trace. There we go. Now, the
moved constituent has to govern its trace, and it
does that by being attached to it. Youll notice
that every index has a little eyehook to which Im
fastening a lightweight nylon cord. You can use
any sort of string or wire, as long as its sturdy
and wont get frayed. Make sure you cut it long
enough to reach the front of the sentence, and
fasten one end to each of the indices. Now we
move the Wh-phrase to the front and slot it into
place. I like to pull the cord taut by wrapping any
slack around the trace.
Well, thats a nice looking chain, Norm. And
once weve done the necessary work on the auxiliary well be done with sentence (1). But how do
we block (2)?
Good question, Howard. The problem is that
the Wh-word cant govern its trace across the CP.
So were going to build a barrier right there. CP
is a maximal projection. If you look closely, youll
see that every maximal projection has a big
square bracket. Were going to fit a blocking
frame right onto that bracket.
Where do we get blocking frames,
You can make them yourself, or, if you
want to save time, you can get them in the
Barriers section of any participating Home
Hardware dealer. I like to use a hardwood for
the blocking frame, something that wont warp
in the damp New England weather. But you
can use any wood that suits your climate. You
need a three-sided frame, open at the top, with
grooves in the sides. Well screw the bottom of
the frame to the bracket of the maximal projection.
So now we have a wood frame sticking up
from the maximal projection. But the trace
can still go right through the middle of the
frame, Norm.
Thats right, Howard, as long as nothing
is there. But now were going to slide a panel
into the grooves. Plywood will do fine. Plane it
down so that it slides easily. And this is
very important dont glue it, but leave it
free to slide. This is called the blocker.
I see, Norm, when the blocker is inserted
into the blocking frame, it forms a barrier over
the maximal projection, so the trace cant get
through, and thats how we block (2). Of
course, now that weve got barriers on every
maximal projection, Im sure the viewers at
home are wondering how the trace gets
through in (1). We appear to have lost our
account of that sentence. But you viewers who
have followed the show for a number of years
wont be surprised if the Olde Yankee Grammarian has a few more tricks up his sleeve,
and Ill bet its got something to do with that
sliding blocker, Norm!
Right you are, Howard. Lets get (3a)
back on the workbench:

Glot International, Volume 2, Issue 7, September 1996 [1997]




Glot International, Volume 2, Issue 7, September 1996 [1997]



This dissertation examines the relationship between syntactic structure and discourse-contextual meaning of language focusing on the scrambling phenomena in German and Korean. I pursue this issue from the perspective that di(erent
ordering possibilities are motivated and constrained by interactions among syntactic, semantic, and discourse principles of these languages.
In particular, I utilize Optimality Theory (Prince
& Smolensky, in press; Grimshaw, in press) to
demonstrate how these principles interact and
resolve conflicts among one another to yield the
optimal output, i.e., a sentence with a particular word order, in a given context.
Taking the various scrambled variants of a
sentence as competing candidates, I derive each
scrambled structure as the best matching or
optimal output to the given input, which contains discourse-contextual information as well as
syntactic and other grammatical information.
The discourse-contextual information is represented in terms of information structure
(Vallduv 1992; Lambrecht 1994). As a particular
model of information structure, I propose one
based on two crossclassifying discourse features
[New] and [Prom] (short for prominent). To
capture the complex interactions of topic and
focus e(ects on word order in German and Korean, I identify four discourse functions, i.e., topic,
tail, completive focus, and contrastive focus, each
of which is represented with the binary features
[New] and [Prom], as demonstrated in (1).







An example of input is presented in (3) for a

German sentence as in (2). The input includes
predicate-argument information (which leads to
the mapping to grammatical functions) and discourse information among other grammatical
(da) Hans dem Schler
das Buch
gegeben hat
(that Hans the student-DAT the book-ACC given
(that) Hans gave the student the book

y=dem Schler;

z=das Buch;


The outputs are taken to be phrase structural

descriptions or parses, which reflect surface
word orders. All six variants in di(erent word
orders in (4) are the competing candidates.


The current analysis can also account for the

often-problematic contrastive focus e(ect
(Lenerz 1977; Abraham 1986; Moltmann 1990)
with no di+culty. The higher ranking of the constraint PROM over the constraint NEW systematically explains the strong scramblability of a
contrastively focused element ([+New, +Prom]),
e.g., das Buch in (10b).

da Hans
dem Schler das Buch
gegeben hat
da Hans
das Buch
dem Schler gegeben hat
da dem Schler Hans
das Buch
gegeben hat
da das Buch
dem Schler gegeben hat
da dem Schler das Buch
gegeben hat
da das Buch
dem Schler Hans
gegeben hat

Then the candidate outputs are evaluated

against a set of violable constraints given in (5)
through (7), i.e., phrase structural constraints for
canonical order, CN1 and CN2, and information
structuring constraints on new or prominent
elements, NEW and PROM, which are further
interacted with a semantic constraint on indefinite phrases, SPECIfiCITY.
Canonical Structure Constraints:
Subject should c-command non-Subject
Non-Subject elements align reversely
according to the functional hierarchy.
(Subject > Acc.Object > Dat.Object >
Oblique > Adjunct)
Information Structuring Constraints:
A [New] element should precede a
[+New] element.
PROM: A [+Prom] element should precede
a [Prom] element.


(12) gives an example of the constraint interaction in the current OT account. The discourse
context in (10) provides the information structure
such as the one in (11). Given this input, the
scrambled variant (12b), which is equivalent to
(10b), wins over the unscrambled variant (12a)
for example, because the former does not violate
the higher constraint PROM, which the latter
does, although the former violates the lowerranked CN2 and NEW, which the latter does not.
dem Schler

das Buch



a. Hans

dem Schler

das BUCH

b. Hans

das BUCH

dem Schler

c. dem Schler


das BUCH

d. das BUCH


dem Schler

e. dem Schler

das BUCH


f. das BUCH

dem Schler


Potential conflicts among these constraints are

resolved by a particular ranking in each
language. I propose the rankings in (8a) and (8b)
for German and Korean respectively. The di(erence in ranking with respect to CN1 comes from
the fact that scrambling over the subject phrase
is much freer in Korean than in German.


WAS hast du dem Mann
gegeben? Die Zeitung?
the newspaper
what have you the man-DAT given
WHAT did you give the cashier? The newspaper?
Ich habe das BUCH dem Mann
gegeben (nicht die
(not the
I have the book-ACC the man-DAT given
I gave the book to the man (not the newspaper).


Semantic Constraint on Discourse Feature Assignment:
SPECIfiCITY: A nonspecific phrase should not be [New].

a. German
PROM >> CN1 >>

Ich habe dem Kassierer das GELD

I have the cashier-DAT the money-ACC given
*Ich habe das GELD
dem Kassierer gegeben.
I have the money-ACC the cashier-DAT given
I gave the cashier the money.


by Hye-Won Choi
Reviewed by Helen de Hoop
by the author

Page 9


CN1, CN2

This Optimality-Theoretic approach to

scrambling can naturally capture the often observed semantic and pragmatic e(ects associated
with scrambling. First of all, the so-called antifocus e(ect (exhibited in the contrast between
(9a) and (9b)), that only a non-focused element
can scramble (Lenerz 1977; Abraham 1986;
Moltmann 1990; Webelhuth 1992), is easily explained by the interaction between CN2 and NEW.
The current account systematically predicts that
candidate (9a) wins over candidate (9b) because
candidate (9b) violates both CN2 (with the accusative object das Geld preceding the dative object
dem Kassierer) and NEW (with the completively
focused phrase das Geld ([+New]) preceding the
tail element dem Kassierer ([New])), while candidate (9a) violates neither of them.




Moreover, the SPECIFICITY constraint and its

interactions with other constraints account for
the interesting behavior of nonspecific indefinite
phrases: a nonspecific indefinite phrase cannot
scramble unless it is contrastively focused, as
illustrated below (Abraham 1986; Moltmann
meinem Bruder
a. *Ich habe einen Brief
I have a
letter-ACC my
brother-DAT sent
I sent a letter to my brother.
Hans ein BUCH dem Mann
gegeben hat
because Hans a book-ACC the man-DAT given
(nicht eine ZEITUNG)
(not a
because Hans gave a book to the man (not a newspaper)

That an indefinite phrase normally cannot

scramble, i.e., definiteness/specificity e(ect (De
Hoop 1992; Diesing 1992), is explained by the
semantic constraint on the discourse feature
assignment in the input, i.e., the SPECIfiCITY
constraint: since an indefinite phrase cannot be
[New], the NEW constraint would not motivate it
to scramble. However, this constraint does not
prevent an indefinite phrase from being assigned
[+Prom], e.g., by being a contrastive focus. Being
a contrastive focus, the indefinite phrase as in
(13b) can scramble thanks to the highly-ranked
PROM constraint just as the definite phrase in
(10b) does.
This case has been the most troublesome for
previous analyses, which are based either on
specificity (De Hoop 1992; Diesing 1992) or on
focality (Lenerz 1977; Webelhuth 1992); because
the scrambled indefinite phrase ein Buch is nei-

ther specific nor unfocused. This problem disappears in the current OT analysis: the internal
conflict of a contrastively focused indefinite
phrase is systematically dealt with by the ranked
competition between PROM and NEW with the
further restriction by SPECIfiCITY.
As shown above, the analysis of scrambling
proposed in this dissertation provides a way to
simultaneously consider various constraints from
di(erent modules of grammar. This interface
approach not only o(ers a systematic account of
why and how a certain syntactic variation of
structure concurs with a certain semantic or
discoursal variation in interpretation, but also
makes it predictable how the semantic and discoursal e(ects influence each other. Furthermore, this OT approach to scrambling attempts
to explain why variation such as word order
freedom ever happens in language: it is due to
the interest conflict among the constraints in
di(erent components of grammar. Interestingly,
grammar also provides a way of resolving the
conflict. That is why variation is never really free
Abraham, W. (1986). Word order in the middle
field of the German sentence. In Topic, focus,
and configurationality, W. Abraham & S. de
Meij (eds.), 1538.
Bresnan, J. (1995). Lexical-functional syntax.
Ms., Stanford University.
Bresnan, J. (1996). Optimal syntax: Notes on
projection, heads, and optimality. Ms., Stanford University.
De Hoop, H. (1992). Case configuration and noun
phrase interpretation. Ph.D. dissertation,
University of Groningen.
Diesing, M. (1992). Indefinites. Cambridge,
Mass.: MIT Press.
Grimshaw, J. (in press). Projection, heads, and
optimality. To be published in Linguistic
Lambrecht, K. (1994). Information structure and
sentence form. Cambridge: Cambridge University Press.
Lenerz, J. (1977). Zur Abfolge nominaler Satzglieder im Deutschen. Tbingen: Gunter
Moltmann, F. (1990). Scrambling in German and
the specificity e(ect. Ms., MIT.
Prince, A. & P. Smolensky (in press). Optimality
theory: Constraint interaction in generative
grammar. To be published by MIT Press.
Vallduv E. (1992). The informational component.
New York: Garland.
Webelhuth, G. (1992). Principles and parameters
of syntactic saturation. New York: Oxford
University Press.

by Helen de Hoop
Chois dissertation results from an exciting hypothesis, namely that various word orders are
motivated and constrained by interactions among
syntactic, semantic, and discourse principles.
This hypothesis led Choi to develop a clear and
unambiguous theory which is convincingly presented in a well-balanced dissertation. I enjoyed
reading the book very much. In Chois approach,
the unmarked or canonical word order is the
order which is not contextually restricted or constrained. In other words, the unmarked order is
context neutral. The canonically ordered structure is the one which is chosen as the optimal
output over all other possible phrase structural
descriptions, purely based on sentence-internal
information. CN1 and CN2 have the e(ect of favoring the canonical order [SUIODO] over other
scrambled structures, other things being equal. If
CN1 and CN2 were all constraints involved in
phrase structural descriptions and word order,
scrambling would never occur. In chapter 3, Choi
develops a set of information structuring constraints (NEW and PROM), which she argues to be

Glot International, Volume 2, Issue 7, September 1996 [1997]

Page 10

the main motivation for alternative orders. Each

alternative structure denotes a slightly di(erent
interpretation, and thus alternative orders are
not optional in a strict sense.
In this review I will focus on some of the nice
results of the thesis, yet at the same time discuss
some of its problematic aspects, that partly arise
because of the purely syntactic way in which the
theory is organized. Unfortunately, the theory
turns out to make several wrong predictions
about scrambling and this will lead me to reject
what is in fact one of the basic assumptions. All
in all, I am convinced that this dissertation will
serve as a source of inspiration to future research
on scrambling.

In Chois OT approach, optional scrambling of

definites is explained by the interaction of the
constraints NEW and CN2, which are not ranked
with respect to each other. That is, in (2b) CN2 is
satisfied, but NEW is violated, whereas in (2c) it is
the other way around. Because NEW and CN2 are
equally strong, both structures are well-formed.
Note that if this account is correct, we also predict that when the direct object instead of the
indirect object contains the new information,
scrambling must be prohibited (Chois example
(56), p.172):


Scrambling and optionality

Recent approaches to scrambling phenomena
argue that in apparent cases of optional scrambling, there is in fact no optionality. For example,
Diesing & Jelinek (1995) argue that the definite
NP in (1) must scramble in order to escape a
mechanism of existential closure (the
grammaticality judgements are theirs):


a. *?weil ich selten die Katze streichle
since I seldom the cat
weil ich die Katze selten streichle
since I the cat
seldom pet
since I seldom pet the cat

In (3b) both CN2 and NEW are satisfied, whereas

in (3c) both CN2 and NEW are violated. Hence, we
expect a huge di(erence in grammaticality between (3b) and (3c), an expectation that is not
quite borne out, as (3c) is not that bad, in my
opinion. Choi would have an obvious explanation
for this judgement of mine that (3c) is not that
bad, since although NEW is violated here, there is
still a possibility that the focused phrase in (3c) is
contrastive or prominent in which case (3c)
would actually be well-formed (PROM outranks
NEW) (Chois example (60), p.176):

Whereas their analysis definitely predicts (1a) to

be ill-formed, Diesing & Jelinek (1995) are aware
of the fact that the degree of ill-formedness of
(1a) is not exactly in accordance with their prediction. That explains why they use the
grammaticality judgement *?, in their own words
to indicate markedness in the sense that some
contrastive context is required for felicity
(p.129). It is not clear to me how this notion of
markedness is related to the analysis Diesing &
Jelinek propose for scrambling. Yet, the idea
itself that certain word orders are only allowed in
certain contexts is not new, and is in fact explored in several recent studies on scrambling.
For example, both Reinhart (1995) and Choi recognize that scrambling is optional in many cases,
but they argue that there can be no true optionality in the sense that word order variants di(er in
how optimal they are in a certain context. Choi
indeed views scrambling as a process of optimizing structure for a given context. In Chois Optimality Theory approach, using information packaging features ([+/ New] and [+/ Prom]) as a
part of the input representation, each scrambled
variant is the best structural description of a
particular information structure, with respect to
a small number of syntactic and discourse constraints.
The question is whether this implies that a
definite in unscrambled position imposes a restriction on the context in which this word order
variant may be used. If that would be the case,
we would predict that if the unscrambled order is
embedded in a wrong context, the sentence
should become truly ill-formed. Choi correctly
observes, however, that definites do not obligatorily scramble, not even when they bear the feature [New] (it is a fact that (1a) is perfectly wellformed, even in a context where the cat is figuring as a topic in the discourse). Choi illustrates
this, following Lenerz (1977) (Chois example
(35), p.155156):

Wem hat Hans das Buch gegeben?

who has Hans the book given
To whom did Hans give a book?


Ich glaube da Hans dem SCHLER das Buch

I believe that Hans the student
the book
gegeben hat
I believe that Hans gave the book to the student


Ich glaube da Hans das Buch dem SCHLER gegeben

I believe that Hans the book the student
I believe that Hans gave the book to the student





Was hat Hans dem Schler gegeben?

what has Hans the student given
What did Hans give to the student?
Ich glaube da Hans dem Schler das BUCH gegeben
I believe that Hans the student the book given
I believe that Hans gave the book to the student
*Ich glaube da Hans das BUCH dem Schler gegeben
I believe that Hans the book the student given
I believe that Hans gave the book to the student

Was hat Hans dem Schler gegeben? Die Zeitung?

what has Hans the student given
the newspaper
What did Hans give to the student? The newspaper?
Ich glaube da Hans das BUCH dem Schler gegeben
I believe that Hans the book to-the student given
hat (nicht die ZEITUNG)
has (not the newspaper
I believe that Hans gave the book to the student (not the

In other words, since I find (3c) o.k. in the context

of (3a), Choi might reply that this means I assigned the book the feature [+Prom]. Strikingly,
Reinhart (1995) gives an example very similar to
(4b) and claims it to be ungrammatical:
*Ik heb het BOEK nog niet gelezen, maar ik heb de
*I have the book not yet read
but I have the
al wel gelezen.
*newspaper already read

The grammaticality judgements of Choi and

Reinhart for the similar sentences (4b) and (5)
are each others opposite. The di(erence follows
from the fact that Reinharts analysis deals with
unmarked focus and not with Chois contrastive
or prominent focus. Reinharts analysis of scrambling predicts that if a context requires the object
to be in focus, scrambling is not allowed. In example (5), the added context (maar gelezen
but read) should make this clear.
However, as Choi would point out, (5) is wellformed exactly because of this context that yields
the contrastive focus reading. Without this contrast, (5) would be Chois (3c), and Reinhart and
Choi would agree on the grammaticality judgement of the example. I claim, however, that both
Reinharts (5) and Chois (3c) are in fact wellformed, and that scrambling is truly optional for
definites and not necessarily related to information structure nor to contrastiveness (see also De
Hoop 1997).

Scrambling and prominence

Choi (1996) argues that there are in fact two
information structuring notions involved in
scrambling. Elements which are not new scramble as well as elements which are prominent in
the discourse. That prominent elements must

scramble is a stronger condition than that new
elements do not scramble, and in this way Choi
accounts for the scrambling of contrastively focused NPs as in (5) and (4b) above.
Clearly, there should be no di(erence between definite or strong NPs and indefinite or
weak ones in this respect. Weak or indefinite NPs
can also bear contrastive focus, in which case
they scramble (Chois (16a), p.85):
Hans ein BUCH dem Mann gegeben hat (nicht eine
because Hans a book the man given
has (not a
because Hans gave a book to the man (not a newspaper)

This is a very nice result, because now Choi can

elegantly account for the fact that not only weak
NPs on a strong reading (or topics) can scramble,
but also weak, existential NPs when they bear
heavy focus. At the same time, she accounts for
the observation that whereas neutrally stressed
phrases do not scramble, heavily stressed phrases do. Choi argues that the property which distinguishes two types of focus is prominence. She
doesnt really motivate this otherwise attractive
intuition, however. Contrastiveness and prominence are nowhere properly defined.
In Chois theory, contrastively focused elements share this feature prominence with topics.
This is a very interesting result, since notoriously, there is cross-linguistic evidence that topic
and contrastive focus behave alike in certain
syntactic configurations. Choi discusses two cases
where this seems to hold: topicalization in English and the so-called topic-marker nun in Korean
which is better characterized as a prominencemarker.
There is another case where [+Prom] might
be the distinguishing feature, and that is the
standard subject position in Dutch. It has been
observed in the literature that in Dutch weak
subjects in standard subject position only allow
for a strong (topic) reading, as opposed to the
observations made for English (cf. Reuland 1988;
Rullmann 1989; De Hoop 1992) In De Hoop
(1992) I noted one type of exception to this generalization, NPs on a reading that I called the
third reading. This concerns weak NPs that are
interpreted existentially and bear some kind of
heavy focus, as in the examples (De Hoop 1996,
chapter 3, p.182, (107)(108)):
Hij hoopt dat een VROUW hem benadert
he hopes that a woman him approaches
MUURBLOEMEN bloeiden voor
het lage raam
blossomed in.front.of the low window

I wrote on these examples: In (7) een vrouw a

woman does not get a referential or a generic
reading, but rather a contrastive one: he hopes
that a woman will approach him, not a man.
Likewise (8) is certainly not a generic statement
about wall-flowers, but a sentence in which they
have been focused. () Somehow, the property
denoted by the noun is contrastively focused.
Yet, I didnt dwell upon these examples any further and decided to stick to the claim that weak
subjects in standard subject position in Dutch
necessarily have a strong reading. With help of
Chois theory, we might replace this claim by a
better one, namely that subjects in standard subject position in Dutch must be prominent. I would
not say, however, that prominent subjects must
move into the standard subject position in Dutch,
as it is clear that subjects in VP-internal positions can also be strong, or prominent, or
contrastively focused. Compare (9) with (10):
zijn twee vliegtuigen neergeschoten
there are two aircrafts shot.down
Twee vliegtuigen zijn neergeschoten
two aircrafts
are shot.down

Glot International, Volume 2, Issue 7, September 1996 [1997]

In (10) the subject can get either a partitive or
referential reading, or a contrastive focus reading
(Twee VLIEGTUIGEN ), but these readings are
all possible in (9) as well. Apart from these readings, (9) has also the ordinary all-focus reading,
where the stress falls on the noun as well, but it
is not marked or heavy stress. In fact, (10) can
also be all-focus, but the focus on the subject is
more contrastive in a sense, or the fact that those
things that have been shot down, were two AIRCRAFTS, is more prominent. In any case, that
reading is also available in the postverbal subject
Similarly, I am inclined to think that contrastive focus readings are possible in unscrambled position. That is, if Choi would be right,
scrambling would be obligatory for highly prominent (i.e., contrastively focused) phrases such as
ein BUCH in (6). For me, however, ein BUCH can
also get a contrastive focus reading in situ:
Hans dem Mann ein BUCH gegeben hat (nicht eine
because Hans a
book the man given
has (not a
because Hans gave a book to the man (not a newspaper)

Chois (p.c.) explanation for this problematic

example for PROM is that only when the speaker
judges ein BUCH contrastive enough, it is
marked [+Prom]. I dont like this type of explanation as it seems to undermine the core of the
analysis. PROM is a strong constraint, stronger
than NEW, but if speakers may vary as to which
contexts trigger prominence, then we end up with
a constraint that cannot help us predicting which
elements scramble, whilst this is what Chois
dissertation is mainly about. For instance, Choi
points out that stressing the question word may
evoke contrastive focus, as well as providing a
relevant alternative, as well as using a focus
adverb such as nur only.
However, in (12) I make use of all three strategies at once in an attempt to guarantee that
speakers will consider the direct object phrase to
be [+Prom], yet scrambling does by no means
become obligatory; (12b) is still just as felicitous
and well-formed as (12a):
WAS hast du dem Mann gegeben? Die ZEITUNG?!
what have you the man given
the newspaper
WHAT did you give the man? The NEWSPAPER?
Nein, ich habe NUR das BUCH dem Mann gegeben
I have only the book the man given
(nicht die ZEITUNG)
(not the newspaper
No, I gave only the book to the man (not the newspaper)
Nein, ich habe dem Mann NUR das BUCH gegeben
I have the man only the book given
(nicht die ZEITUNG
(not the newspaper)
No, I gave only the book to the man (not the newspaper)

An even stronger case against the obligatory

scrambling of [+Prom] phrases as required by
PROM would be the optional scrambling of topics.
Topics are characterized as [+Prom] and [New],
hence they would violate both information structuring constraints in unscrambled position. In my
opinion, topics can indeed remain in situ as well;
consider the following example:
Wie stehts mit dem Geld?
What about the money?
Ich glaube das Hans dem Kassierer das Geld
I believe that Hans the cashier the money
gegeben hat
given has
Ich glaube das Hans das Geld dem Kassierer
I believe that Hans the money the cashier
gegeben hat
given has
Ich glaube da das Geld Hans dem Kassierer
believe thatthe money Hans the cashier
gegeben hat
I believe that Hans gave the money to the cashier

Chois account cannot explain why das Geld does

not have to scramble all the way into a position

Page 11

before the subject (as in (13c)) if it is considered a

topic. Scrambling over a subject is pretty marked,
in fact, but because it is possible, PROM must be
ranked higher than CN1, and because of this
ranking, scrambling over a subject becomes obligatory in the case of a non-topical subject and a
topical object. This predicted obligatoriness is not
in accordance with the facts, however.
Let us take one step back and see how Chois
constraints can be reformulated in such a way
that this problem of predicted obligatoriness with
respect to scrambling of prominent phrases does
not arise. Even if it is not true that topics obligatorily scramble, I think we can safely maintain
that in scrambled position we often find topics.
More specifically, phrases that occur in scrambled positions are either topics (which captures
the strong readings of weak NPs) or contrastively
focused elements. But neither topics nor contrastively focused constituents obligatorily scramble
(for topics, we might have to make an exception
for unstressed pronouns, but since Choi does not
take pronouns into consideration, I will not do
that either). Therefore, I do not think that we can
use constraints like PROM and NEW which need
as their input arguments that are labeled with
the features [Prom] and [New]. Rather, it seems
that syntactic positions indicate the type of NP in
terms of information packaging that generally
occupy these positions. In unscrambled, canonical
position, all possible readings can arise, but in
scrambled position we find topics (see also Adger
Interpret a constituent in a scrambled position as a topic
(i.e., Vallduvs 1990 link;
see also Vallduv 1993 for a similar
suggestion concerning Dutch scrambling).

However, we also find contrastively focused elements in scrambled positions. I think this can be
explained if we take into consideration the following information packaging universal:
Focused elements ([+new information]) are not topics
and vice versa

One of the advantages of an Optimality Theory

(OT) approach is that (14a) and (14b) can be
universal without the consequence being that
they should always hold. This is because in OT
constraints are soft and potentially conflicting,
and they can always be violated in order to satisfy stronger constraints (see also Burzio 1995).
Therefore, we do not have to immediately worry
about the existence of counterexamples to the
above constraints as long as we can assume that
other constraints may be involved in those cases
(see Vallduv 1990; Krifka 1991; and De Swart
and De Hoop 1995 for some discussion of possible
counterexamples to (14b), for instance).
Now, suppose that in Dutch and German
(14b) outranks (14a): (14b) >> (14a). Suppose
furthermore, following Choi, that scrambling has
a cumulative e(ect on topicality (which boils
down to a more scale-like treatment of
topicfocus structure, compare insights from the
Prague school on communicative dynamism, e.g.,
Hajicov 1983). We then account for the fact that
an NP in scrambled position gets interpreted as a
topic unless it is interpreted as the focus. If we
consider default intonation or unmarked focus to
be strongly related to syntactic structure (cf.
Cinque 1993; Reinhart 1995), then it will follow
automatically that the accent (intonational focus,
stress) in a position where no default focus is
assigned, has to be more marked or heavier in
order for an NP to be interpreted correctly as the
focus (i.e., the new information, cf. Vallduv
1990). This then results in the reading that has
been called contrastive focus reading but that is
di+cult to define properly because of the vagueness of notions like prominence or contrastiveness. In fact, for Rooth (1992) all focus is contrastive; for Vallduv (1990) contrastiveness is just an
epiphenomenon, and apart from contrastively
focused elements, contrastive links (=topics) exist

as well. Assuming the two universal constraints
as in (14a) and (14b) and the ranking between
the two, we account for the observation that intonation seems to be more relevant for the (discourse) interpretation of constituents in Dutch
than syntactic position (see also Zwart 1995). If
there are two well-formed syntactic structures, a
scrambled one and an unscrambled one, then
there are certain tendencies for interpretation
related to default intonational patterns, but actual focus distribution can overrule these tendencies. Furthermore, the cumulative e(ect of topicality will go hand in hand with a cumulative
markedness or heaviness of the intonation that is
needed to let (14b) win from (14a). That seems to
give the correct result, namely that one has to
stress the direct object in (15a)(15d) more and
more in order to get the focus interpretation for
the direct object correct; in (15d) the use of a
focus-attracting determiner like ZULKE seems
to be necessary to save the focus-reading, as
scrambling over subjects is even more marked in
Dutch than it is in German (see also Neeleman
Ik had nooit gedacht dat een student Petra vandaag
I had never thought that a student Petra today the
het boek zou geven
the book would give
Ik had nooit gedacht dat een student Petra HET
I had never thought that a student Petra the
BOEK vandaag zou geven
book today
would give
Ik had nooit gedacht dat een student
I had never thought that a student

HET BOEK Petra vandaag zou geven

the book Petra today would give

Ik had nooit gedacht dat ZULKE BOEKEN

I had never thought that such
een student Petra vandaag zou geven
a student Petra today
would give

In conlusion, the constraints (14a) and (14b) and

their ranking ((14b) >> (14a)) account for the
facts that:
an NP in scrambled position gets interpreted
as a topic unless it is focused;
intonation can overrule syntactic (scrambled,
but also standard subject) position in Dutch in
order to get the interpretation right;
the more a syntactic position is associated
(statistically) with topical fillers in that (relative)
position, the louder one has to shout in order to
get a focus reading;
scrambling is not obligatory for topics, nor for
contrastively focused elements.

Scrambling and specificity

Choi considers information structuring to be
subject to several types of constraints. These
constraints may be morphological, or syntactic, or
semantic. For instance, the specificity e(ect is
one of the cases for which Choi assumes that a
semantic restriction on information feature marking is applied. In this case the semantic constraint SPECIfiCITY prohibits [+New] elements
from receiving a specific feature due to a semantic incompatibility. What is however missing in
this dissertation is a theory about how the discourse feature assignment mechanism of elements actually works.
A connection between specificity and information status is indirectly responsible for the
limited scrambling options of nonspecific NPs,
according to Choi. A nonspecific NP is informationally dependent on the bigger information
unit, which results in a special predicate modifier interpretation. In Chois theory scrambling is
only possible when the phrase has an independent information status. This is a very nice consequence indeed. The indefinite in (16) does not
get an independent [+/new] feature; therefore it
does not have to scramble, and in fact it is not
allowed to scramble because of CN2 that favors
canonical word order.

Glot International, Volume 2, Issue 7, September 1996 [1997]

Ich habe gestern ein Buch gekauft
I have yesterday a book bought
hast du gestern ein Buch gekauft?
at.what.time have you yesterday a book bought
hast du ein Buch gestern gekauft?
b. ??Wann
at.what.time have you a book yesterday bought

In principle, I very much appreciate this idea of

dependent indefinites (see also Van Geenhoven
1996 and Farkas 1996, for suggestions in the
same spirit), but I regret that Choi does not discuss definite NPs here. Intuitively, one would
expect that definite predicate modifiers behave
similarly, but in fact, they do not, as is shown in
Ich habe gestern den Bus verpasst
I have yesterday the bus missed
hast du gestern den Bus verpasst?
at.what.time have you yesterday the bus missed
hast du den Bus gestern verpasst?
at.what.time have you the bus yesterday missed

It will be clear from this example that the scrambled NP in (17b) is not to be characterized as a
topic or as contrastively focused here. Actually, I
do not know whether Choi would agree that it is
surprising that the definite in (17) does not behave like the informationally dependent indefinite in (16) with respect to scrambling. In De
Hoop (1997) I attempt to account for this surprising di(erence between indefinites and definites
and argue that the di(erence cannot be reduced
to general discourse or intonation principles;
instead the di(erence must be sought in a di(erence
in the semantics between the two types of NPs.
Conclusion: scrambling and
On the basis of some potential counterexamples to Chois theory as discussed above I
reject one of the main ideas of the thesis, namely
that scrambling should be viewed as optimizing
structure for a certain context. I have tried to
show that if some word order variants are wellformed, they are well-formed in all possible contexts. In other words, when there is a conflict
between context on the one hand, and word order
on the other, context wins (at least as long as we
are dealing with grammatical word order). This
means that the right intonation must be chosen
that can overrule syntactic position to guarantee
the right interpretation that fits the context. I
have no doubt that an OT-framework can handle
this most elegantly and probably the best. As for
me, I was definitely inspired by Chois dissertation and impressed by the many scrambling data
her theory captures, in particular the link between topicality and contrastive focus cross-linguistically.

Page 12

Hajicov, E. (1983). Topic and focus. Theoretical

Linguistics 10, 268-276.
Hoop, H. de (1992). Case configuration and noun
phrase interpretation. Ph.D. dissertation,
University of Groningen. [1996: Garland
Publishing, New York & London.]
Hoop, H. de (1997). Scrambled definites. Unpublished Ms., OTS, Utrecht University.
Krifka, M. (1991). A compositional semantics for
multiple focus constructions. Proceedings of
SALT 1, 127-158. New York: Ithaca.
Neeleman, A. (1994). Scrambling as a D-structure phenomenon. In Studies on scrambling;
movement and non-movement approaches to
free word order phenomena, N. Corver & H.
van Riemsdijk (eds.). Berlin: Mouton de
Reinhart, T. (1995). Interface economy: focus and
markedness. Unpublished Ms., Tel Aviv/OTS.
Reuland, E. (1988). Indefinite subjects. In Proceedings of NELS 19. Amherst: GLSA.
Rooth, M. (1992). A theory of focus interpretation.
Natural Language Semantics 1, 75-116.
Rullmann, H. (1989). Indefinite subjects in Dutch.
Unpublished Ms., University of Massachusetts, Amherst.
Swart, H. de, & H. de Hoop (1995). Topic and
focus. Glot International 1-7, 3-7.
Vallduv, E. (1990). The informational component. Ph.D. dissertation, University of Pennsylvania, Philadelphia.
Vallduv, E. (1993). Information packaging: a survey. Unpublished Ms., University of Edinburgh.
Zwart, J.-W. (1995). Word order, intonation, and
noun phrase interpretation in Dutch.
WECOL Proceedings.


The research for this review was supported by
the Foundation for Language, Speech, and Logic,
which is funded by the Netherlands Organization
for Scientific Research, NWO (grant 300-75-020).
Adger, D. (1995). Functional heads and interpretation [dissertation summary]. Glot International 1(1), 89.
Burzio, L. (1995). The rise of optimality theory.
Glot International 1(6), 37.
Cinque, G. (1993). A null theory of phrase and
compound stress. Linguistic Inquiry 24,
Diesing, M. and E. Jelinek (1995). Distributing
arguments. Natural Language Semantics 3,
Farkas, D. (1996). Dependent indefinites and
direct scope. Ms., UCSC.
Geenhoven, V. van (1996). Semantic incorporation and indefinite descriptions. Semantic
and syntactic aspects of noun incorporation
in west Greenlandic. Ph.D. dissertation,


Zero Semantics
A Study of the Syntactic Conception
of Quantificational Meaning

Gertjan Postma
diss. Leiden University|Holland Institute of Generative Linguistics|HIL Diss 13

Zero Semantics develops a comprehensive theory of the interface of

morphosyntax and the interpretative module. Taking theories of Heim
and Diesing as a starting point, an interpretative theory is developed
that describes morphosyntactic constraints on both quantificational
interpretation (existential, interrogative, universal, negative quantifications) and lexical interpretation. Especially phenomena on the
borderline of lexical and quantificational interpretation are used to
determine the morphosyntactic mechanisms of interpretation.
Using data from Dutch, Gertjan Postma argues that semantic
assignment does not only occur at the terminal nodes of syntax, but
that syntactic structures are also an independent source of meaning.
Cases exist in which syntactic structure conveys quantificational
meaning rather than the elements that lexicalize it. This phenomenon is
called zero semantics.
At a more technical level, the theory extends generative theories of
Noam Chomsky and Richard Kayne on discontinuous elements (such as
chains, parasitic gap configurations, multiple interrogation) to
configurations with conjoined interpretation. Using contrastive studies
on WH elements and Romance SE-constructions, the author highlights
the structural properties of quantification.
Apart from providing many new observations, this book contributes to
our thinking about the design of natural language. It shows in what
respects natural language differs from the grammar of formal logic.
Furthermore, the theory offers a paradigm for the transition of lexical
elements to functional elements: grammaticalization is not just a
phenomenon that belongs to the realm of diachrony, it is a structural
possibility governed by the synchronic computational component of
The discussion in Zero Semantics has direct relevance for anyone who
is interested in the logical representation of natural language: workers in
syntax, in semantics, and in logic. It is also of interest to workers in
lexicography and in historical grammar.
isbn 90-5569-009-0|x+233pp.|june 1995|NLG 40 (ex VAT, ex P&P)



Holland Academic Graphics

[ s c i e n t i f i c

[ d o c u m e n t ]

P.O. Box 53292

2505 AG The Hague
The Netherlands

p r o c e s s i n g ]

e-mail: mail@hag.nl
http: www.hag.nl
fax: +31 70 4480177


Glot International, Volume 2, Issue 7, September 1996 [1997]

Page 13



by Angeliek van Hout
Reviewed by William Philip
by the author


Verbs and their verb frames are the topic of my

dissertation. The main question I seek to answer
is: what determines that a particular verb appears in an intransitive or a transitive frame, or
in both or even in other verb frames? Investigating adult and child Dutch verbs and verb frames,
my goal is to determine the nature of the lexical
primitives in terms of which verbs are characterized in the lexicon and how these define the way
verbs map onto syntax. The main thesis is that a
verbs event type properties define the mapping
relation, rather than a q-grid, argument structure
or lexical-conceptual structure (LCS). In addition to
presenting a theoretical discussion of this issue, I
also examine it from the perspective of language
acquisition: how does a child acquire the verbs
and their lexical properties of his/her language?
The results contribute to the research program that set out to investigate if the mapping
patterns of verbs can be derived on the basis of
other lexical properties and/or more general properties of the lexicon-syntax interface system (cf.
Gruber 1965; Fillmore 1968; Rappaport & Levin
1988; Jackendo( 1990; Grimshaw 1990; Carrier
& Randall 1993; Levin & Rappaport Hovav 1995).
It is a search for the bare roots of the lexicon.

Verb frame alternations and event

Looking at many di(erent verb frame alternations in Dutch (i.e., transitiveintransitive,
intransitiveoblique, transitiveoblique,
unergativeunaccusative), I show that the meaning di(erences of verbs across di(erent frames
and those between root verbs on the one hand
and prefixed and particle verbs and complex
predicates on the other are systematic. While the
verb essentially refers to the same kind of event
across di(erent frames and constructions, each
expresses a di(erent event type. I conclude that a
verbs flexibility is intrinsically related to event
type-shifting. There are three kinds of event
type-shifts at play in verb frame alternations:
atelictelic, causativization and alternations
with di(erent foci on the same event. I concentrate on the first two and extensively illustrate
and discuss these two kinds of event type-shifts.
In atelictelic event type-shifting, an atelic
event type (i.e., a state or an ongoing process) is
expanded when another state or process is added,
creating a transition between two subevents and
thereby turning it into a telic event type (i.e., an
accomplishment or an achievement). The unergativeunaccusative alternation in (1) and the intransitivetransitive alternation in (2) illustrate this.


Anneke heeft jarenlang / *binnen een jaar bij Raz

Anneke danced with Raz for years / *within a year.
Anneke is *urenlang / binnen een minuut van het podium af gedanst.
Anneke danced o( the stage *for hours / within one

Elena schreef jarenlang / *binnen een jaar.

Elena wrote for years / *within a year.
Elena schreef *jarenlang / binnen een jaar haar proefschrift.
Elena wrote her dissertation *for years / within a year.

Under causativization, another participant (the

causer) is added to the original event, and
sometimes the event type is expanded with an
additional, causing subevent. The causative
inchoative alternation in (3) illustrates causative
event type-shifting without an additional subevent, while the periphrastic causative in (4b)
has an additional participant as well as an additional subevent.


Het ei brak.
The egg broke.
Inge breekt het ei.
Inge broke the egg.

Het ei brak.
The egg broke.
Inge liet het ei breken.
Inge let the egg break.

The event type of a predicate is crucial in determining a particular verb frame. Two generalizations emerge: (i) a telic event type requires an
(underlying) argument in direct object position;
(ii) a causative event type requires two arguments, one in subject and one in direct object

A critique of argument-centered
mapping approaches
In the past three decades, the issue of what
determines in which verb frame(s) a verb occurs,
has focused on the verbs arguments: how many
arguments and of what kind. This approach has
been implemented in various models that refer to
a verbs lexical specification as thematic relations (Gruber 1965), case roles (Fillmore 1968),
q-roles (Chomsky 1981; Baker 1988; Grimshaw
1990), arguments (Williams 1981; Zubizarreta
1987) or LCS-variables (Jackendo( 1990). The
mapping system links an argument of a particular kind to a particular syntactic position (e.g.,
Chomskys 1986 Canonical Structural Realizations rules, Bakers 1988 Uniformity of Theta
Assignment Hypothesis, Jackendo(s 1990 Hierarchical Argument Linking).
Arguing against such argument-centered
mapping approaches, I raise several descriptive
and methodological problems with their lexical
primitives and the lexical rules that account for
verb frame alternations. Moreover, I point out
that these accounts cannot generalize across
di(erent alternations, since they do not take into
account the fact that the event type determines
how an argument is mapped onto syntax. A maximally strong theory should derive the telicity and
the causative mapping generalizations and explain why particle and prefixed variants occur in
di(erent verb frames than their root verbs and,
moreover, why they di(er in the particular ways
they do.

CHESS: checking event-semantic

I conclude that the mapping system needs to
see a verbs basic event type plus aspectual
contributions of other predicates in the VP and
the number of participants that are involved in
the event. It does not need any further semantic
specification of the participants. Developing a
new approach to modeling the lexicon-syntax
interface within the feature checking framework
of the Minimalist Program (Chomsky 1993,
1995), I propose the CHESS model: CHecking
Event-Semantic Structure. Claiming that the
event structure of a predicate must be syntactically identified (cf. Grimshaw 1990; Grimshaw
& Vikner 1993), I define the mapping relation in
terms of checking event-semantic features in
functional configurations. There are two structural argument positions: the specifier positions of
AgrS and AgrO. An argument in either of these
positions identifies an event or subevent by referring to an event participant that is involved in
that (sub)event. Telic event type features must be
checked in AgrOP (cf. Borer 1994). I argue that
the CHESS model accounts for the event-semantic mapping generalizations in a natural way, explaining the phenomenon of lexical-syntactic
flexibility as a derivative of event type-shifting.
This view on mapping also implements a new
version of the Unaccusative Hypothesis
(Perlmutter 1978). Given the condition that telic
features must be checked in AgrOP, the single
argument of a telic one-participant verb (e.g., (1b)
and (3a)) must move through object position
before moving to subject position (i.e., via
SpecAgrO to SpecAgrS). The general mapping
condition on telic event types thus derives NPmovement for one-participant verbs and yields
what is called an unaccusative frame.
Notice that the CHESS mapping conditions
do not make any reference to the semantics of
the event participants; they are semantically
blind. I explicitly argue against mapping systems defined on the semantic properties of a
verbs arguments (i.e., q-roles, LCS-variables).
Instead, I propose some general interpretational
principles defined on properties of the event
structure, the syntactic configuration and particular morphemes which yield an interpretation
(cf. Emonds 1991; Davis & Demirdache 1995).
The (underlying) direct object of a telic predicate
gets interpreted as the participant involved in a
change of state. In a transitive frame, the argument in subject position is interpreted as more
agentive than the object. Passive morphology and
nominalizing su+x -er attribute an agentive
interpretation. Clearly, this is only the beginning
of a more complete theory in which syntax and
semantics co-determine the interpretation of
syntactic configurations.

Learning verbs
After modeling the lexicon-syntax interface
in event-semantic terms, I turn to acquisition
data that confirm some of the claims of the
CHESS model. I present two case studies of verb
learning. One study concerns so-called light
verbs. Assuming that verbs are characterized in
the lexicon with their event types, I argue that
light verbs (e.g., geven give in een kus geven give
a kiss) are easy verbs to acquire, because they
are pure event type denoters without any further
semantics. In a longitudinal study of four Dutch
children, I find support for this prediction. From
early on, these children produce a great variety of
light verb constructions, also including some
overgeneralized ones. Looking at their light
verb usage and evaluating the nature of their
overgeneralized constructions, I conclude that
child Dutch presents evidence for a basic claim of
the CHESS model, namely that verbs are lexically characterized in terms of their event types.
Furthermore, I discuss the acquisition of
intransitive verbs and the split into unergatives
versus unaccusatives. Intransitive verbs pose an
intriguing acquisition problem: how can a child
find out that there are two di(erent subclasses

when they look so similar at the surface? How
does she determine to which subclass a new verb
belongs? The results of an experimental study
aimed at finding out whether two semantic factors, telicity and agentivity, determine unaccusativity in Dutch, show that subjects (4 and 5 yearolds, 7 and 8 year-olds and adults) are able to
divide novel intransitive verbs in two subclasses.
Even the youngest children distinguish two classes, which is evidence in favor of the Strong Continuity hypothesis on language development.
Interestingly, the split is defined by telicity;
agentivity does not play a role. These results
present evidence for another claim made by the
CHESS model, namely that mapping is defined
on the event-semantic properties of predicates.


The research presented in my dissertation
draws on data from adult and child Dutch, but
the claims that are made are suggested to represent more generally the way in which Universal
Grammar organizes the lexicon of the language
and the mapping system that links a verbs lexical features with its syntactic projections.

Baker, M. (1988). Incorporation: A theory of
grammatical function changing. Chicago:
The University of Chicago Press.
Borer, H. (1994). The projection of arguments. In Functional projections, E.
Benedicto & J. Runner (eds.). University
of Massachusetts Occasional Papers 17,
1947. Amherst: GLSA.
Chomsky, N. (1981). Lectures on government
and binding. Dordrecht: Foris.
Chomsky, N. (1986). Knowledge of language,
its nature, origin and use. New York:
Chomsky, N. (1993). A minimalist program
for linguistic theory. In The view from
Building 20: Essays in linguistics in
honor of Sylvain Bromberger, K. Hale &
J. Keyser (eds.), 152. Cambridge, Mass.:
MIT Press.
Chomsky, N. (1995). The minimalist program. Cambridge, Mass.: MIT Press.
Carrier, J. & J. Randall (1993). Lexical mapping. In Knowledge and language, E.
Reuland & W. Abraham (eds.), 119142.
Davis, H. & H. Demirdache (1995). Agents
and events. Paper presented at GLOW
18, University of Troms, Norway.
Emonds, J. (1991). Subcategorization and
syntax-based theta-role assignment.
Natural Language and Linguistic Theory
9, 369429.
Fillmore, C. (1968). The case for case. In
Universals in linguistic theory, E. Bach
& R. Harms (eds.), 188. New York:
Holt, Rinehart and Winston.
Grimshaw, J. (1990). Argument structure.
Cambridge, Mass.: MIT Press.
Grimshaw, J. & S. Vikner (1993). Obligatory
adjuncts and the structure of events. In
Knowledge and language, E. Reuland &
W. Abraham (eds.), 143155.
Gruber, J. (1965). Studies in lexical relations.
Ph.D. dissertation, MIT. Republished in
J. Gruber 1976, Lexical structures in
syntax and semantics. Amsterdam:
Jackendo(, R. (1990). Semantic structures.
Cambridge, Mass.: MIT Press.
Levin, B. & M. Rappaport Hovav (1995).
Unaccusativity: At the syntaxlexical
semantics interface. Cambridge, Mass.:
MIT Press.
Perlmutter, D (1978). Impersonal passives
and the Unaccusative Hypothesis. Proceedings of BLS 4, 157189. Berkeley:
University of California.

Glot International, Volume 2, Issue 7, September 1996 [1997]

Rappaport, M. & B. Levin (1988). What to do
with q-roles. In Thematic relations, W.
Wilkins (ed.), 736. San Diego: Academic
Reuland, E. & W. Abraham (eds.) (1993).
Knowledge and language, Vol. II. Dordrecht: Kluwer.
Zubizarreta, M.-L. (1987). Levels of representation in the lexicon and in the syntax. Dordrecht: Foris.


The principal empirical objective of Van Houts

Event semantics of verb frame alternations is to
formulate a unified semantic account of how it is
possible for verbs to have more than one type of
basic complement structure. Examples of this
phenomenon, which van Hout calls lexical-syntactic flexibility, are given in (1) and (2). Putting
aside optional constituents such as adverbs and
adjunct PPs, the Dutch verb lopen walk may
occur as an unergative intransitive with no complement, as in (1a); or as an unaccusative with an
underlying direct object and an oblique argument, as in (1b) unaccusativity detectable in
Dutch by the presence of auxiliary zijn be or
as an unaccusative intransitive with an underlying direct object but no surface complement, as in
(1c). Similarly, variation in basic complement
structure is found with the Dutch verb schrijven
write, which may occur with no complement, as
in (2a); with a direct object complement, as in
(2b); or with an oblique argument complement, as
in (2c).

Jan heeft gelopen.

Jan walked.
Jan is naar het station gelopen.
Jan walked to the station.
Jan is weggelopen.
Jan walked away.



[SU ]



Jan gaf het boek aan Marie.

Jan gave the book to Marie.
Jan gaf Marie het boek.
Jan gave Marie the book.

Marie las het boek.

Marie read the book.
Het boek werd gelezen.
The book was read.

The dissertation focuses on the two other basic

natural classes of verb frame alternations which
Van Hout identifies. One of these, which includes
the types of alternations exemplified in (6), is the
set of causative alternations, i.e. alternations
determined by the presence or absence of linguistically signalled causation. Due to space limitations, I will not discuss Van Houts analysis of
this sort of lexical-syntactic flexibility.

De foto hangt aan de muur.

The photo hangs on the wall.
Hij hangt de foto aan de muur.
He hangs the photo on the wall.
Hij laat de foto aan de muur hangen.
He lets the photo hang on the wall.

The third class of verb frame alternations, which

receives the most attention in the dissertation,
includes cases such as exemplified in (1) and (2)
above and in (7) and (8) below.

[ DO]
[SU ]

Jan schreef.
Jan wrote.
Jan schreef zijn boekje.
Jan wrote his book.
Jan schreef aan zijn boekje.
Jan worked on his book.


reminds one of Everaert (1986) Van Hout categorizes the major types of lexical-syntactic flexibility which occur in Dutch into three basic semantic natural classes. One of these groupings,
which includes relationships such as those exemplified in (4) and (5), Van Hout characterizes as
alternations with different foci on the
event. This sort of lexical-syntactic flexibility is
only very briefly discussed, its proper analysis set
aside for future research.

by William Philip

Page 14



Rejecting the traditional view that verb frame

alternations such as in (1) and (2) are random
lexical idiosyncrasies traditionally represented
by specifying a non-singleton set of subcategorization features for lexical entries like lopen or
schrijven Van Hout (1996) makes the interesting claim that in general the mapping from the
verbal lexicon to the syntax i.e. the many-tomany relation schematically represented in (3)
can be explained entirely in terms a small set of
principles of semantic interpretation which have
a filtering e(ect on syntactic structure as they
interact with the lexical semantic properties of
the verb and other basic expressions, such as
prepositions and verb particles, that compositionally determine the meaning of the predicate as a
Verbal lexicon

Basic sentence types















After a well-written introduction and very informative chapter 2 the empirical rigor of which


Henk heeft de auto geduwd.

Henk pushed the car.
Hij heeft de auto naar Amsterdam geduwd.
He pushed the car to Amsterdam.

Claartje heeft de hele zommer gezwommen.

Claartje swam all summer.
Ze heeft haar badpak aan flarden gezwommen.
She swam her bathing suit to shreds.

Given the reference to event semantics in the

title, one might expect to find in Van Hout (1996)
an extension of Carlsons (1984) analysis of the
verb frame alternations in (9). Carlson argues
that (9a)s entailment of (9c) and (9d)s nonentailment of (9e) follow for free from world
knowledge when one adopts the Neo-Davidsonian
event-based model-theoretical account of the
semantics of predicates that he proposes.

John ate.
Agent(John) eat
John ate a sandwich.
Agent(John) eat theme(sandwich)
John ate something.
John kicked.
John kicked someone.

Van Hout does not discuss Carlson (1984) at all,

however. Moreover, although she alludes to various event semantic models her account of lexicalsyntactic flexibility actually makes no use of any
kind of model theory. Rather, with a general
hypothesis that is closely related to the work of
Tenny (1987), Grimshaw (1990), Van Valin (1990)
and Borer (1994) and which is actually more of
working hypothesis than a fully thought-out theory Van Hout extends Pustejovskys (1991)
event structure account of predicate aktionsart
in such a way that several of the verb frame alternations she discusses appear to involve a shift


Glot International, Volume 2, Issue 7, September 1996 [1997]

in aspect type (event type). Van Hout calls this

event type-shifting. (The term is incorrectly
attributed to Bach 1986, who, following up on an
observation of Dowty 1972, uses the term packaging to describe the way contextual factors can
cause a process verb to be interpreted eventually e.g. #He finished looking for a girlfriend
versus He finished looking for a book at the library; Bach relates the packaging of a process
verb as an event verb to the packaging of a
mass noun as a count noun e.g. Twee bier,
alsjeblieft Two beers, please.)
A prime example, then, of Van Houts event
type-shifting is the atelic/telic event type shift
which occurs in the contrast between the (a) and
(b) cases of items (1), (2), (7), and (8). (We may
add (9a, b) as well.) All the (a) cases of these
items have an atelic aspectual interpretation,
while all the (b) cases have a telic reading. This is
seen by the fact that for each of the a cases the
predicate can readily be modified by a durative
adverbial such as de hele nacht all night, but not
by a terminative adverbial such as binnen een
uur within an hour. For each of the (b) cases
precisely the converse relation to temporal adverbial modification obtains. Van Hout describes the
semantic e(ect of the atelic/telic type shift in
alternations such as (1a, b) and (2a, b) as follows:
The kind of event a verb refers to remains constant across
The aspect type a verb refers to di(ers across frames.

Referring to the lexical specification of syntactic

adicity as a specification of the number of event
participants read theta roles, understood in
the purely syntactic sense Van Hout also notes
that when the atelic/telic event type shift occurs:
Often, one (or more) of the event participants is mapped onto
an argument position of the verb in each frame. Often, the
exact selection of event participants that are mapped onto
argument positions of the verb and the selection of these
positions di(er [across frames].

Within the extended Pustejovskian analysis of

aspectuality that she proposes, the atelic/telic
type shift in (1a, b) may be represented as in (10).

naar het station lopen




<being at the station>

As a lexical entry, lopen denotes a set of events of

a kind classifiable as having a process event type
(P). This is also the event type of the predicate in
(1a). In the case of (1b), an event type shift has
occurred in the syntax such that the predicate
denotes a set of events of a kind classifiable as
having an internal structure in which one subevent is classifiable as a process and the other as
a (resulting) state. (This is one of four possible
transition event types (T) in Van Houts account.)
The reader will note that it is not entirely
clear what the theoretical status of event typeshifting is intended to be. From its name and the
way it is discussed, it at first seems to be intended as a kind of grammatical operation, comparable, say, to type-shifting in semantic theories
making use of higher-order logic. Understood in
this sense, one is tempted to imagine a model of
language processing based on the formal theoretical principles of event type-shifting according to
which a semantically driven processor e(ects a
change in the event type of a predicate containing
a process verb such as lopen walk by attaching a
directional PP to its VP node thereby realizing
the event type-shift from (1a) to (1b). However,
unlike type-shifting as it is standardly understood, no complete linguistic expression ever
changes its logical type when event type-shifting
occurs. Rather, the event type of a predicate is

determined once and for all by the basic expressions it is made up of. Discussing the type of
alternation in (1a, b), van Hout observes that
the telicity of the predicate is due to an additional goal argument [e.g. naar het station] that
makes the predicate telic or to a telic goal particle such as weg away (p.93). In other words,
Van Hout adopts the standard bottom-up view
of how aspectuality is determined (cf. Pustejovskys (1991) event composition; Verkuyls
(1993) aspect composition), according to which
the aspectual value of a sentence is determined
by a number of di(erent elements and e(ects,
only one of which is the verbs basic event type
(p.91). But now, if the aspectuality of the predicate is compositionally determined in a bottomup fashion, it cannot also be the case that the
contents of the predicate, including its structure,
is determined top-down by a semantic operation
of event type-shifting. The chicken can come
before the egg, or vice versa, but both cannot
come before both. The solution to this apparent
paradox it becomes apparent by chapter 4 is
Van Houts CHESS model of the lexiconsyntax
interface. This is essentially the very interesting
thesis that the syntax of a particular predicate is
licensed just in case its compositionally determined aspect type bears a possible event typeshifting relation to the lexically specified aspect
type of the verb heading this predicate. Thus,
event type-shifting is not really an operation, but
rather a filter or checking mechanism. It is a
semantic constraint on the set of syntactically
distinct complements that a given verbal lexical
entry can be associated with by dint of its ability
to contribute to the determination of di(erent,
related, event types for the predicate as a whole.
In this regard, it will be noted that event typeshifting is quite distinct from Bachs (1986)
packaging. It will also be noted that the question of whether event type-shifting is a part of
grammar is left somewhat open. It could easily be
the case, it would seem, that certain kinds of
events may readily be taken to be subevents of
larger kinds of events simply because of nonlinguistic categories of perception. For example,
the event type-shifting relation between atelic
(1a) and telic (1b) may obtain simply because it is
very easy to imagine a walking event as the subevent of an event of spatial displacement. Thus,
Van Houts notion of the event type-shift may
have the theoretical status of, say, Dowtys (1991)
proto-theta roles, or a Greenburgian linguistic
universal. It may be a meta-grammatical notion
pretending to shed light on why certain types of
alternations frequently occur in natural
language. The same can be said, I believe, of the
other two operations on event structure which
van Hout posits i.e. that underlying causative
alternations and that underlying alternations of
di(erent foci on the event.
Van Houts dissertation is interesting and
well worth reading to those interested in argument structure and argument selection because
of the many questions it raises and the fresh
perspective it o(ers on the organization of lexicon
and on the relation between the lexicon, the computation of syntactic form, and the assignment of
meaning. However, it is hard not to be sceptical
about its principal claim that all forms of lexicalsyntactic flexibility can be reduced to the filtering
e(ect of properties of event structure. Consider
again the alternations in (1) and (2). While the
atelic/telic dichotomy may seem to provide a neat
account of the contrasts in complement structure
observed in (a) and (b) cases, it cannot in principle provide an explanation of the other alternations in (1) and (2). In the case of the syntactic
contrasts between (1b) and (1c) and between (2a)
and (2c), two di(erent types of complement structure have the same aspectual interpretation: (1b)
and (1c) are both telic; (2a) and (2c) are both
atelic. (For Van Hout, the preposition naar in
(1b) heads the predicate, taking the verb as an
argument, so in a trivial sense both verbs map
onto an unaccusative frame (p.94). The fact

Page 15

remains, however, that the syntax of (1b) di(ers

radically from that of (1c) and it is precisely such
alternations that Van Hout pretends to explain.)
Since event type-shifting cannot account for these
alternations, we must turn to one of the other
two basic event structure relations for an explanation. Certainly, these are not causative alternations, so, by the process of elimination we are
left with an operation a(ecting the focus of the
event. But now the problem is that it is not at all
clear how focus structure plays any role in the
contrasts in question. Consider, by way of contrast, the alternations in (4) and (5), which clearly do di(er in focus structure.
Another apparent problem for Van Houts
approach concerns minimal contrast sets such as
in (11). (These examples are from Dowty 1989.)
John dined for an hour.
a. *John dined his lunch.
John ate for an hour.
b. John ate his lunch in an hour.
John devoured his lunch in an hour.
c. *John devoured.

While atelic/telic event type-shifting may explain

how it is possible for eat to occur with the two
complement structures shown in (11b) and (11b),
event type-shifting alone cannot account for why
neither devour nor dine allow the same lexicalsyntactic flexibility. To capture the facts in (11),
Van Hout must make some additional (not implausible) assumptions. The contrast between
(11a) and (11b) would be derived from the lexical
specification of syntactic adicity. While dine is
lexically specified to be a 1place predicate in
Van Houts terms, it has only one event participant eat is lexically specified as a 2place
predicate. (Note that, in terms of meaning, the
kinds of events that dine and eat denote have
exactly the same number of essential participants, i.e. one cannot dine without something
being eaten.) As for the contrast between (11b)
and (11c), under Van Houts analysis, this would
follow from the circumstance that devour is lexically specified as a telic verb (T), while eat is
lexically specified as a process verb (P). Clearly,
this analysis can capture the facts. What is less
clear, however, is that it yields any net theoretic
gain. We seem to have simply replaced a traditional formalism for representing lexical idiosyncracy subcategorization frames with a
pair of di(erent formalisms syntactic adicity
and inherent telicity. While the reader may
feel, as I do, that Van Houts approach does have
greater empirical and/or explanatory adequacy
than the traditional subcategorization approach,
still some argumentation is needed here, it would
Another sort of problem for Van Hout are
contrasts such as in (12), where a sharp contrast
in aspectuality (12a) is atelic; (12b) telic is
found with one and the same complement structure.
Ze hebben de kat gemarteld.
They tortured the cat.
Ze hebben de kat doodgemarteld.
They tortured the cat to death.


As in the case of (11), Van Hout can account for

such facts by positing additional primitive lexical
specifications. But we are still left with a question: how can two di(erent event-types filter in
the same predicate type? If there is a homomorphism between semantic event type and syntactic predicate type, how can event-type have
very much of a filtering e(ect?
A more basic question that arises concerns
the crucial distinction which Van Hout draws
between kind of event and event type. Consider again the contrast in meaning between in (1a)
and (1b). According to Van Hout, in both cases
the same kind of event is denoted, i.e. a walking
event, and the di(erence in meaning is purely
one of event type, i.e. (1a) is atelic while (1b) is

telic. But is this self-evident? Couldnt it alternatively be the case that the fundamental di(erence between (1a) and (1b) is that they denote
two di(erent kinds of events which happen to be
able to co-occur, i.e. a walking event and a
going-somewhere-on-foot event and that this
basic di(erence in meaning results in a di(erence in aktionsart? After all, verbs typically di(er in aktionsart as a consequence of the kind of
events they denote, not because there is any
necessary intensional or extensional relation
between them e.g. consider drink versus die.
This could also be what is going on in the a and b
cases of (1), (2), (7) and (8). If so, it is not the
contrast in aspectuality in the a and b cases of
(1), (2), (7) and (8) that determines the contrasts
in complement structure, as Van Hout claims,
but rather that just the opposite: the contrasts in
aspectuality are indirect e(ects of the contrasts
in kind of event denoted by the respective predicates. This meaning-determines-aspect view of
the situation is not only the standard view, it is
also more consistent with other facts. For example, in support of the meaning-determines-aspect
view, consider the contrast between (13a) and
(13b). On Van Houts account, the only di(erence
between these two sentences is one of aktionsart;
they both denote the same kind of event.
John squirmed. (=moved like a worm on a fishhook)
John squirmed o( the stage. (=left the stage, squirming
the whole way.)
John moved from location A to location B.
d. #John squirmed as fast as his legs would carry him.
John squirmed o( the stage as fast as his legs would
carry him.

Er zijn de hele dag jongetjes naar het station gelopen.
Boys walked/were walking to the station all day long.
Jan schreef boekjes het hele jaar.
(cf. (2b))
Jan wrote books all year long.

Van Hout is well-aware of this phenomenon and

explicitly sets it aside as irrelevant. But is it irrelevant? Borer (1994) cites precisely such facts
as the knock-down empirical argument against
any purely semantic account of the unergative/
unaccusative alternation.
In conclusion, the dissertation is interesting
because of the general conception of the lexiconsyntax interface that it proposes and the many
questions that this generates. No doubt it will

prompt a new and highly productive line of research on this topic. The ideas need considerable
further development, however. In my view, the
most interesting idea the CHESS model
would be much improved if all reference to event
type shifting were dropped, along with the extended Pustejovskian event structure analysis
of aspectuality. In addition, as it stands, it is not
clear that the CHESS model presupposes an
event-based model theory. Its claims could just
as easily be reformulated, it would seem, in a
model-theoretical account of aspectuality that
did not treat events as individuals in the universe of discourse e.g. Dowty (1979) or
Verkuyl (1993). (Those working in event semantics will hope that when Van Hout does tie her
theory more closely to model-theoretic semantics
she will be able to show how an event-based
model theory is preferable.) Finally, I must also
caution the reader that the reading of this dissertation calls for some patience and some charity. The text has a great number of little problems. The argumentation is often quite poor or
incomplete. There are scholarly and factual inaccuracies. There are irrelevant and confusing
digressions that sometimes seem to wander o(
into incoherence. There are misused terms and
unnecessary and poorly defined neologisms.
Semanticists will be appalled, for example, to
hear Van Hout repeatedly characterize the
many-many relation in (3) as a function. Looking past such minor defects, though, the reader
will find the dissertation highly intriguing. The
one word that characterizes Van Hout (1996) is

Page 16

Bach, E. (1986). The algebra of events. Linguistics and Philosophy 9, 516.
Borer, H. (1994). On the projection of arguments. In Functional projections, E.
Benedicto & J. Runner (eds.), 1947.
UMOPS 17, Amherst, Mass.: GLSA.
Carlson, G.N. (1984). Thematic roles and
their role in semantic interpretation.
Linguistics 22, 259279.
Dowty, D. (1972). Studies in the logic of verb
aspect and time reference in English.
Ph.D. dissertation, University of Texas.
Dowty, D. (1979). Word meaning and
Montague grammar. Dordrecht: Reidel.
Dowty, D. (1989). On the semantic content of
the notion of thematic role. In Properties, types and meaning, volume II, G.
Chierchia, B.H.Partee & R. Turner
(eds.), 69129.
Dowty, D. (1991). Thematic proto-roles and
argument selection. Language 67,
Everaert, M. (1986). The syntax of reflexivization. Dordrecht: Foris.
Grimshaw, J. (1990). Argument structure.
Cambridge, Mass.: MIT Press.
Pustejovsky, J. (1991). The syntax of event
structure. Cognition 41, 4781.
Tenny, C. (1987). Grammaticalizing aspect
and a(ectedness. Ph.D. dissertation,
Van Valin, R.D. Jr. (1990). Semantic parameters of split intransitivity. Language 66,
Verkuyl, H. (1993). A theory of aspectuality,
Cambridge: Cambridge University Press.


new titles from john benjamins publishing company

But is this factually accurate? (13b) entails (13c),

but (13a) does not. Moreover, while (13b) may be
modified by the adverbial phrase as fast as his
legs would carry him, as shown in (13e), (13a)
may not, as shown by the ill-formedness of (13d).
These entailment and distributional facts do not
seem to follow directly from any di(erences in
the aspectual properties of the predicates in
(13a) and (13b). Rather, they seem to follow from
di(erences in the kinds of events denoted by
squirm and squirm o( the stage. The latter predicate denotes a set of events of a certain kind of
spatial displacement. When (13b) is true, this set
has a non-null intersection with as fast as his
legs would carry him . In contrast, the predicate
in (13a) does not denote a set of events of any
kind of spatial displacement and therefore always has a null intersection with as fast as his
legs would carry him . This is why (13d) is semantically anomalous. (It is always false and
therefore always infelicitous till accommodation
takes over the job of making sense of it).
Yet another fact that would seem to be problematic for Van Houts general program is
Verkuyls classic observation that the semantic
properties of the nominal constituents in a sentence a(ect aspectuality regardless of their syntactic position (e.g. Verkuyl 1993). Thus, while
(1b) and (2b) are telic, both minimally contrasting sentences in (14) are atelic.

Glot International, Volume 2, Issue 7, September 1996 [1997]

Now in paperback
The Current State of
This state-of-the-art volume presents an outstanding collection of 22 studies on current
issues facing research in second-language
acquisition (SLA). The topics covered range
from the role of universals at various levels of
second-language (L2) knowledge; the way
that linguistic knowledge is represented by
L2 learners; the changing nature of linguistic
theory itself; and the definition of usage phenomena like style shifting and code switching. The introduction to The Current State of
Interlanguage gives a concise yet detailed
overview of research in the field over the past
10 years, and focuses on the present growing
concensus on a number of issues that were
at one point highly controversial.
Pb: 1995. viii, 301 pp.
Europe: 90 272 2165 0
Hfl. 60.

Constructions in Japanese
A frame semantic approach
This study examines the clausal noun-modifying construction (NMC) in Japanese, a muchdiscussed construction that embraces what
have usually been called relative clause and
noun complement constructions. Drawing
upon a broad range of naturally-occurring
NMCs, including types that fall outside the
domains of relative clause and noun complement constructions, Matsumoto argues for
an analysis of NMCs that gives an important
role to semantics and pragmatics. The framework in which this approach is presented
draws from, and further refines, concepts of
frame semantics.
Studies in Language Companion Series, 35
Hb: 1997. viii, 204 pp.
Europe: 90 272 3038 2
Hfl. 130.

The Semantics of Aspect and

Evidence from English and
Biblical Hebrew
The semantics of aspect and modality will
be of interest both to linguists working on
temporality, as a general phenomenon in language, and Hebraists investigating the semantics of the verbal forms in biblical
Hebrew. Tense, aspect and modality are
among the most challenging discussed areas
of language. Similarly, the semantics of the
verbal system in biblical Hebrew has been
investigated since the Middle Ages. Galia
Hatav provides extensive critical overviews of
research in both areas, and suggests a new
approach for analyzing the biblical Hebrew
verb system, showing it to be tenseless.
Studies in Language Companion Series, 34
Hb: 1997. x, 210 pp.
Europe: 90 272 3037 4
Hfl. 150.

Materials on Left Dislocation

Materials on Left Dislocation consists of two
parts. Part I contains a selection of the main
texts on which our present understanding of
the Left Dislocation construction is based. For
various reasons most of these texts had never
been published, or are published in obsolete
places. These articles, by Van Riemsdijk &
Zwarts, Rodman, Hirschbuehler, Vat, Cinque
and Zaenen, contain the first arguments that
pertain to the major questions about Left
Dislocation (for example whether movement
or base-generation is involved), and they
present the rationale for the now standard
distinctions between Hanging Topic LD, Contrastive LD, and Clitic LD.
In Part II a number of recent contributions to
the grammar of Left Dislocation are brought
together. In these articles, by Anagnostopoulou, Demirdache, Escobar, Van Hoof
and Wiltschko, new aspects are being explored such as the relationship between LD

and the grammar of focus and the role of clitic

doubling and its semantic effects in Clitic LD.
Furthermore, the empirical basis is broadened to encompass more languages. Finally,
these articles explore the relationship
between LD and a number of apparently unrelated constructions such as split topicalization.
Linguistik Aktuell/Linguistics Today, 14
Hb: 1997. viii, 349 pp.
90 272 2735 7
Hfl. 145.

Studies on Universal
Grammar and Typological
Edited by A. ALEXIADOU and
The articles of the present volume consist of
generative analyses dealing with several current topics of discussion and debate in syntactic theory, such as clitics, word order,
scrambling, directionality, movement. The
data in the volume are drawn from a number
of typologically diverse languages (e.g. Arabic, Berber, Dutch, Gaelic, Greek, Malagasy).
Linguistik Aktuell/Linguistics Today, 13
Hb: 1997. viii, 252 pp.
Europe: 90 272 2734 9
Hfl. 130.

John Benjamins Publishing Company

Amsteldijk 44 P.O.Box 75577
1070 AN Amsterdam
Tel. +31 20 6738156
Fax +31 20 6739773
John Benjamins North America
P.O.Box 27519
Philadelphia PA 19118-0519 USA
Toll-free ordering: (800) 562-5666
Fax +1 215 836-1204

Book notices

Glot International, Volume 2, Issue 7, September 1996 [1997]

by Rint Sybesma
Shizhe Huang, Quantification and predication
in Mandarin Chinese: A case study of dou.
University of Pennsylvania doctoral dissertation. 1996.
[IRCS report 9636. University of Pennsylvania, 3401
Walnut Street, Suite 400A, Philadelphia, PA
191046228, USA; sdeysher@cis.upenn.edu.] 193 pp.
This is an important book. It makes original
claims and contains many other fresh ideas. It is
mainly concerned with dou but in order to explain
the use and semantics of dou, it also delves deeply
in the semantics of mei every and in the structuring of (Davidsonian) events.
The Mandarin element dou is generally regarded as a universal quantifier, but Huang shows
that that is only part of the story, one of the problems being that dou may also occur in contexts
with no universal quantification. On top of that,
the alternative view that dou is a distributor, i.e.,
more like each in English (see Xu Dings dissertation below), does not cover the whole array of facts
Huangs dissertation starts out from the observation that Mandarin mei every cannot stand
alone in the sense that it only occurs felicitously in
case it cooccurs with an indefinite phrase, a reflexive or dou:
mei-ge chushi zuo yi-ge cai
every-CL cook make one-CL dish
every cook prepares one dish
mei-ge haizi you ziji-de chuang
every-CL child have self-DE bed
every child has his own bed
mei-ge ren
*(dou) kan-le
zhei-ben shu
every-CL person *)DOU read-PERF this-CL book
every person read this book

So what does dou do there and why can it be omitted in case the sentence contains a reflexive or an
indefinite? Phrased di(erently, what do indefinites, reflexives and dou have in common such
that they can somehow license the occurrence of
mei every?
Huang links this question to other distributional facts of dou, which, as noted above, often
expresses universal quantification, while conveying distributivity at other times. But it can also
occur in sentences with no universal quantification at all. In addition, in some cases dou is optional and in specific contexts, it can be replaced
by other elements like ye also and hai still. (And
I repeat the comment I made in the review of the
dissertation of Shyu Shu-ing in Glot International
2/4 that in contexts in which these elements are
interchangeable they must all be unstressed, an
observation which is not taken into account by
Huang either.)
Huangs first step towards an answer lies in
the analysis of mei every. Huang adopts what is
called the skolemized definition of EVERY. A
skolem function links two variables by making the
choice of the value of the one depending on the
choice of the value of the other. If EVERY is a
skolemized phenomenon, EVERY needs a variable
in its scope (it needs to c-command it). For mei
every, being an instantiation of EVERY, this
requirement is met when it has either an indefinite or a reflexive in its scope, both elements, as is
generally assumed, introducing variables. (This
approach also explains the fact why NPs with mei
generally do not occur in object position: from
there it is impossible for them to c-command a

So how does dou fit into this skolemization

picture? The intriguing suggestion Huang makes
in this respect has to do with tense. In earlier
work she had already proposed that in the context
of mei every, dou serves the function which the
tense operator performs in languages like English.
To quote a passage from the introductory chapter
of her dissertation (p.11):
Details aside, our main contention is that whatever unique
function indefinites, reflexives and dou serve in a sentence
with a mei noun phrase, it can be performed by tense in English. Since tense is omnipresent in English, one does not observe similar preference of these items by every. In Chinese,
while systematic tense marking is absent, other devices are
resorted to. In this thesis we have identified the fucntion that
dou and the tense operator in English share to be the role of
an event variable licenser.

So it lexically licenses the event variable for

skolemization. More in particular, as it becomes
clear later, Huang takes dou to be a sum operator
which takes the event variable as its argument.
Viewing dou as a sum operator on events, has
the welcome e(ect that dou can be regarded as
forcing, so to speak, a reading onto the predicate it
is part of to the e(ect that it reads as a plurality of
events. This explains why dou can be used in
contexts with no universal quantification interpretation.
Having linked the (or: one particular) use of
dou to the (or: one particular) use of tense in languages like English, Huang explores what else the
two have in common. In an interesting argument,
involving, among other things, sentence final le,
Huang concludes that one of the characteristics
tense and dou share is that both adduce a partial
order on events.
There is much more of interest in this book.
As I made it clear in the opening line of this notice, I think that this is an interesting book. At
some points Huang may have elaborated on things
more extensively, but that does not reduce the
importance of the work. I always find it inspiring
and stimulating to read something which is full of
both fresh nd viable ideas. It may turn out that
not all of these ideas can be held up, but at least
they constitute a stimulus for further research.

Chaofen Sun, Word-order change and

grammaticalization in the history of Chinese.
Stanford University Press. 1996. ISBN 0804724180.
207 pp. With index. Price: 30.
According to the preface Sun wants to make available a book on the history of the Chinese language
in English, not only for linguists who work on
Chinese but also for others. His fundamental attitude is the following (Preface, p.xii): Instead of
viewing a fragment of the history of a language
from a theoretical point of view, Sun prefers to
investigate a cross section of data from di(erent
historical periods free of theoretical bias to help
develop a more sophisticated understanding of the
history of Chinese. Even so, as we read in chapter
1 (p.2): The aim of this study is to determine, on
the basis of certain linguistic signs, just how far
syntactic changes can depart from complete arbitrariness, and therefore, what types of iconic features those syntactic changes may have involved
in the history of Chinese.
After an introductory chapter mainly on the
type of material and data used in this book, we are
presented with studies of four grammatical devel-

Page 17

opments: the change of word order in prepositional

phrases, the ba-construction, the particle le and
the modal particle de. As far as I can see, these
chapters constitute very adequate descriptions of
what happened and how things changed. In some
cases, Sun also tries to determine what factors
played a role in the changes he describes.
The last chapter summarizes the four case
studies but from a di(erent point of view: Why did
the changes happen the way they did? Why does
language change anyway? In the context of issues
like grammaticalization, Sun critically discusses
di(erent theories of language change, like those
developed by David Lightfoot and Anthony Kroch,
among others. Suns own idea is that concepts like
isomorphism and iconicity play a role in language
change, and that grammatical changes are often
induced by changes at other levels, like semantics
and pragmatics. He argues that the cases he discusses in this book are each a case in point.

C.-T. James Huang & Y.-H. Audrey Li, eds.,

New horizons in Chinese linguistics. Dordrecht,
Kluwer Academic Publishers. 1996. 392 pp. With
index. Hardbound: ISBN 0792338677, price: Hfl.
225; 99; $155. Paperback (North-America only):
0792338685, price: $69.
This book consists of ten chapters and an introduction. Each chapter gives an overview of the
most important developments of the past 1020
years in di(erent fields of Chinese linguistics or
the linguistics of Chinese. The chapters can be
divided into two groups. One group consists of the
chapters which deal with developments in the
study of Chinese in interaction with general linguistics or with the study of other languages. The
second group is made up of chapters that describe
developments in the study of aspects of Chinese
which took place in their own right, which is not to
say, of course, that there is no relevance to the
study of other languages: one could also look at
them as case studies concentrating on one particular language, in this case Chinese. The chapters in
this book are written in such a way that they are
readable and understandable for any linguist; one
does not need to know anything about Chinese
linguistics or even Chinese. The blurb on the back
of the book says it can be used as a textbook, but
the ridiculous pricing of the hardback will make
sure that that is not going to happen outside of
North-America and, yes, that is a pity.
One other feature which makes this a valuable book and research tool is the fact that every
chapter is closed o( with an extensive bibliography, in one case even as long as 19 pages, which
brings together most of the relevant literature on
the topics in this book.
The first group
To start with the first group of chapters (those
which deal with developments in the study of
Chinese in interaction with general linguistics),
chapter 2, written by the editors of this volume
and entitled Recent generative studies in Chinese
syntax (46pp.) presents a very well-organized
overview of a number of issues and proposed analyses which show how the interaction between
general syntactic theory and the study of the Chinese language contributed to both the development of the theory of syntax and the understanding of Chinese grammar. The chapter constitutes
an impressive balancing act in that it seems to
aim at both introducing Chinese linguists to generative grammar and generative grammarians to
Chinese at the same time and I think Huang
and Li succeeded. The issues dealt with certainly
belong to the more complex areas of the field (LF,
wh, donkey sentences, anaphora and binding,
empty categories, and scope, to just mention a
few) but, possibly because these subjects lie very
close to the research interests of the authors, they
seem to have managed to satisfy both the Chinese
linguist and the generative grammarian.
Chapter 3, Recent developments in function-

Book notices
al approaches to Chinese (44pp.) by Yung-O Biq,
James Tai and Sandra Thompson, also belongs to
the first group. It consists of nine sections. After
an introductory section on the ideas behind Functional Grammar (section 1) and two short sections,
mainly consisting of lists of publications (section 2
on the history of functional grammar and Chinese
and section 3, Semantic studies) we get to section 4, one of the two longer sections of this chapter, entitled A cognition-based functional approach to Chinese grammar. It gives an interesting overview of the way a functional approach
throws an explanatory light on how Chinese expresses notions like time and space and deals with
phenomena like noun categorization and iconicity.
Following two other short sections (on syntax and
pragmatics), we find the other main section of the
chapter, section 7 on Discourse approaches to
Chinese which deals with the Chinese sentence
final particles and discourse markers. Section 8
outlines directions for future research and the
last section, section 9, is a short conclusion, pointing out (p.121) that discourse and functional
approaches to the study of Chinese resulted in a
deeper understanding of how Chinese grammar
arises from and is related to the cognitive and
social systems within which it functions.
Thomas Hun-Tak Lees chapter 9, Theoretical issues in language development and Chinese
child language (64pp.) is one of my favourite
chapters. Quite admirably, Lee
manages to summarize the acquisition literature on phonology, syntax and morphology, and
semantics and pragmatics for
both Mandarin and Cantonese,
without, however, glossing over
the details. As to the phonology,
Mandarin consonants, Mandarin
vowels, Cantonese vowels, Mandarin tones, Cantonese tones,
the discrepancies, the things
they have in common when they
are acquired, everything is taken
care of. In the acquisition of semantics and pragmatics section
Lee discusses the subjects binding, quantifier scope, deictic expressions (personal pronouns, but
also expressions of relative time
and space), aspect, classifiers,
questions and sentence final particles. The syntax and morphology
section deals with word order, null
subjects and objects, relativization, complementation and conjunction, and compounding.
Chapter 10, Neurolinguistics: A Chinese perspective
(23pp.) by Daisy L. Hung and Ovid T.-L- Tzeng, is a
very interesting chapter too. It is concerned with
aphasia and the use of cross language comparative
aphasia research. The starting point is quite
straightforward. It is well known that some aphasic
disorders lead to the drop or the incorrect use of
morphology and function words the Indo-European languages come out as Chinese. As the authors
write (p.364), the fact that Chinese grammar has
virtually no flectional morphology like verb conjugation and noun declension and that the word
order is quite flexible (e.g., topicalization)
raise[s] some fascinating questions concerning grammatical impairment in Chinese aphasics: since it is possible to produce
sentences with no grammatical markers of any kind, how can
we identify the symptom patterns that characterize Brocas and
wernickes aphasia in other languages? Thus, by examining the
patterns of language breakdown from a comparative viewpoint,
we can learn a great deal about the relationship between brain
and language.

It turns out that the symptoms displayed by

Chinese aphasics are quite compatible with 100
years of research on aphasia in Indo-European
languages (p.365366). To show this, the authors
discuss four issues: word order problems, nounverb dissociation, morphology (compounding) and
classifiers (arguably the only domain of Chinese
grammar where agreement plays a role).

Glot International, Volume 2, Issue 7, September 1996 [1997]

Arguably, chapter 1, Tonal geometry A
Chinese perspective (28pp.) by Matthew Chen
also belongs to the cross-linguistic group of chapters, but it is much less of an overview article than
most other chapters in the book. Even so, it is a
very clearly written study on the nature and representation of tone, especially the relationship
between register and contour (the primary pitch),
the two major components of tone. Using the tools
of feature geometry, Chen seeks to describe the
anatomy of tone. He deals with both Chinese and
African languages, trying to see how theories
developed on the basis of the study of the one
group of languages can be used to understand
what is going on in the other group. Chen pays a
great deal of attention to tone sandhi phenomena
in several Chinese languages, arguing that they
have much to o(er to further our understanding of
the nature of tones and tonal processes.
The second group
The remaining chapters belong to the second
group and deal with aspects of the study of Chinese with less interaction with, but no less relevance for, the study of other languages. These
chapters address historical issues and issues in
the study of dialects and language diversity.
The first historical chapter, chapter 4, is
Pang-hsin Tings Tonal evolution and tonal reconstruction in Chinese (19pp.). It surveys the

tonal developments from Archaic Chinese

(17001100 B.C.) to the present dialects. It both
looks at the evolution of the tones and the tone
categories (addressing, among others, the issue of
when and how and in what order the tonal categories split between the 5th and 10th century
A.D.) and tries to reconstruct the actual tonal
value of the di(erent tones in di(erent times.
Two things need special mention. First, Ting
argues against the prevalent idea that Archaic
Chinese probably was not a tone language, and
that the four tonal categories were represented,
then, by di(erent final consonants. Instead, he has
reasons to adhere to the more traditional view
that Archaic tones really were pitch tones. (He
does not reject the idea that Chinese was once a
language without tones but it was such a language in an older stage). Secondly, in the part on
tonal reconstruction, Ting makes the interesting
suggestion that, unlike what is commonly assumed, sandhi tone forms preserve the tonal value
of the proto tones much better than the non-sandhi tone forms.
The other historical chapter, chapter 5, is on
syntax, Recent issues in Chinese historical syntax (53pp.). It is written by Alain Peyraube and
next to Thomas Lees, it is my other favourite
chapter in the book. With an admirable pace,
Peyraube guides us along seven research topics,

Page 18

not counting the other problems of section 9.

(The first section briefly addresses general issues
like reasons for grammatical change and issues
more specific to Chinese; section 10 is a short conclusion.) The seven topics are: word order change,
the disposal form (the ba-construction), the passive forms, the di(erent dative constructions,
locative structures, the origin of the perfective
particle le and coordinative conjunctions. All these
sections present a summary of what the issues
are, what the proposals have been and what the
evidence is. In virtually all cases Peyraube makes
clear what he thinks is the most likely scenario for
all these historical developments.
The word order debate is probably the best
known of the topics discussed by Peyraube: in the
70s-80s ideas were put forward that in its older
phases, Chinese had been an SOV language that
had gradually changed to SVO. In the context of
the word order debate, Peyraube quotes the biologist Jean Rostand as saying: Theories come and
go, but the frog is still there (p.162), to which
Peyraube adds: In the case of the long debate on
word order change, the frog unfortunately was
never there, people preferring flowery general
hypotheses to thankless work based on cold facts
(loc. cit.). In other words, as he concludes at the
end of the section, nothing can justify an SOV to
SVO change in Archaic Chinese.
Two chapters address dialect issues. Chapter
6, Dah-an Hos contribution
entitled Stages and strata in
dialectal history case studies
of Heng County, Da County and
Shipo (20pp.), presents three
separate case studies in the
historical phonology (especially
tone) of three dialects of Chinese, spoken in places as geographically apart as Guangxi,
Fujian and Sichuan. One of the
leading ideas in these case studies is that what is called dialectal anomalies, i.e., phenomena
that pose a problem for a
straightforward synchronical
analysis, may give important
clues as to the history of the dialects.
The second chapter on dialects, chapter 8, Chin-chuan
Chengs Quantifying dialect
mutual intelligibility (24pp.) is
of a totally di(erent nature.
Cheng presents a method in
which the mutual intelligibility
between genetically related
languages can be calculated and
quantified in an objective way.
He discusses all the relevant aspects of the method: sources and targets, the weight scale, the
method of calculation. Cheng emphasizes that, for
a start, he is mainly interested in what he calls
systemic mutual intelligibility, i.e., correspondences in patterns in the di(erent dialects. In one
of the examples he gives he compares syllables of
the Jinan dialect to those of the dialect of Peking.
Each element of the syllable (initial, medial, vowel, ending, tone) gets a certain value (0.20). After
comparing a large number of cognate syllables in
both dialects and applying a number of statistical
calculations, we get the figure 0.719 as the value
for the mutual intelligibility between the Jinan
and Peking varieties. Cheng applied these computations to 17 di(erent dialects and it turns out
that the varieties spoken in Wenzhou and Xiamen
are among the least mutually intelligible (0.341).
Probably because of his initial interest in systemics, he does not discuss how non-cognates, which
may a(ect the intelligibility of another dialect to a
large degree, are accounted for in these intelligibility calculations.
In the first part of chapter 7, Linguistic diversity and language relationships (33pp.), William S.-Y. Wang puts (the Chinese) language in a
wider perspective. One of the things he stresses is
that normally when we think of language variation in China, we are only concerned with the

Book notices
Sinitic, or Han, varieties, largely ignoring the, lets
say, more than fifty non-Han languages spoken by
the non-Han ethnic minorities in China. He goes
on to remind us that the present day dialects of
China result from an interaction between the
language of the Han (the Chinese) and the substratum languages of the people living in the areas the Han settled down in. Wang continues
stating that the Han (the Chinese) cannot be defined racially. He quotes from immunological studies, which make it clear that the people who call
themselves Han are often genetically more closely
related to neigboring minorities than to far away
people who also call themselves Han. (Wang concludes that Han Chineseness is an ethnic
concept, based on culture.) It comes as no surprise,
then, that we find words in the Sinitic languages
of a non-Sinitic origin even words as basic as
those for the great rivers of China, jiang [Yangtze,
R.S.] and he [Yellow River, R.S.] are not Han in
origin (p.242).
In the second part of the chapter, Wang concentrates on the Sinitic dialects and the studies
that have been conducted since the beginning of
this century. He discusses several aspects of dialect studies (like, migration) as well as several
research methods, quantifying methods like
Chengs (see above) and parsimony methods which
allow for the incorporation of heterogeneous aspects of language in their computation.

Yilu Zhao, Distributional criteria for verbal

valency in Chinese. Leuven: Peeters. 1996. ISBN
9068317679. 243 pp. Price: BEF 1500.

The aim of this book is (i) to look for criteria for

the valency study of verbal predicates in Chinese
and (ii) to categorize the valents and to attribute
syntactic labels for them.
As to (i) an important part of the book is devoted to the problem that all valency studies,
especially those related to Chinese, have to cope
with, viz., the obligatoriness requirement: the idea
that an element is a valent of a verb if it is obligatory in all contexts is straightforward, but why
would elements that are not obligatory in all contexts not qualify as valents? In Chinese, we have
the additional problem that it is not even always
clear whether an element is obligatory. Zhao proposes and/or develops additional tests (not all are
entirely new): The Interrogative Proform Test
(elements corresponding to shei who, shenme
what and in non-prepositional cases nar
where pass the test), The Formulation Test (to
distinguish non-valents from valents, certain word
order correspondences are argued to be crucial)
and The Preposition Test (based on the observation that valents generally tend to occur without a
preposition). Not all tests always give the right
These tests are all introduced and elaborated
on in chapter 3, and all issues are discussed in
great detail. Zhao presents a wealth of data. Of all
the di(erent patterns she distinguishes, she gives
at least two examples. For instance, with regard to
trivalent verbs, she distinguishes 12 di(erent
formulations as she calls them di(erent word
order types not counting the ungrammatical
orders that she also gives. Her discussion on the
status of the preposition in Chinese, its optionality
in some contexts, its obligatoriness in others is
also very thorough.
The other issue what syntactic labels to
attribute to the di(erent valents is not an easy
matter. In Chinese linguistics, a debate has been
going on for decades on the question as to how to
define the subject and the object of the sentence.
There being no agreement, this is not always easy
to determine. Basically, there have been two
camps: those that said that what agreement is in
European languages is what word order is in Chinese, so everything nominal that occurs preverbally is the subject, and all elements occurring
postverbally is the object. This camp had problems
with sentences like

Glot International, Volume 2, Issue 7, September 1996 [1997]

beijing lai ren
Peking come people
people have come from Peking

where the strict adherence to word order principles to determine what is the subject and what is
the object clashes with certain other intuitions.
The other camp took the intuitions about agenthood and patienthood as their point of departure
and would have pointed at ren in the sentence
above as the subject. But what to do with sentences in which there is no agent not a clear one? To
complicate matters, Chinese has been branded a
topic prominent language by Charles Li and
Sandra Thompson for a good reason, and, again, in
the absence of any type of morphology it is not
always an easy matter to distinguish between the
two. In short, Zhaos task is not simple and I dont
think that her solutions are definitive. But her
discussion of these matters, both in chapter 2
(review of earlier literature) and chapters 4 and 5
is very useful and very thorough. And, as before,
everything is abundantly illustrated with Mandarin data.

Ding Xu, Functional categories in Mandarin

Chinese. HIL/Leiden University doctoral dissertation.
1997. [Distributed by Holland Academic Graphics;
mail@hag.nl.] ISBN 9055690260. 202 pp. Price: Hfl.
Xus dissertation deals with a number of issues
related to the functional superstructure of Mandarin, viz., adverbial placement, CP, negation and
the quantifier dou all (or each).
After a short introduction (3pp.) in chapter 1,
chapter 2 (46pp.) presents the preliminaries: in
the first part it presents a brief introduction into
the minimalist framework within which this study
was written; the second part is a very short, but,
in view of its length, amazingly comprehensive
grammar of Mandarin (Some basic properties of
Mandarin Chinese).
Chapter 3, on adverbial placement (36pp.),
starts out from the X-bar theoretical assumption
that there are only two types of grammatical relation: the relation between the head and its complement and the relation between the head and its
specifier. Pushed to the limit, this has the consequence that there is no such thing as adjunction:
every adjunct heads its own projection. To see
what the consequences of such a claim are for
Mandarin, Xu conducts a meticulous study of the
di(erent types of adverbs that we find in Mandarin (sentential vs. non-sentential; locative, manner, etc), the position they occupy in the sentence,
and the positions they occupy relative to one another. Xu then ends up proposing a tree for Mandarin Chinese which involves, among others,
LocP, InstruP, AttiP and MannerP, and he suggests that the same projections should be postulated in other languages too, but he does not o(er
much discussion on this point. A very nice, be it
short, part of this chapter is about tense and modality in Mandarin.
In chapter 4 (23pp.) Xu tries to find the elements in Mandarin which could be instances of C
in this language. Like others before him, he concludes that the sentence final particles are good
candidates for this as well as the element de in
relative clauses. Towards the end of the chapter,
Xu tries to apply Kaynes antisymmetry analysis
of relative clauses to Mandarin, which leads him
to propose the following structure:
[DP IPj [DP [D0 dei [CP [NP N0] [C0 ti] tj ]]]]

Negation is the subject of the fifth chapter

(32pp.). The goal of this chapter is (a) to find out
where negation is represented in the structure of
Mandarin; and (b) to describe the semantics of the
negation marker bu. As to (a), Xu eventually posits negation as the head of NegP, and he presents
a number of arguments for this position. As to the
semantics of bu, working through many di(erent
types of sentences, Xu comes to the conclusion
that bu is incompatible with the aspect associated

Page 19

with realization or perfectivity. There is an intriguing suggestion in this chapter, but regretfully, Xu has not worked it out in any detail. The
suggestion is that bu will always have some modal
or aspectual element following it and that in some
cases this element may be null. So in sentences
like the following, we have an aspectual/modal
element meaning something like habituality or
Zhang San bu he
Zhang San not drink alcohol
Zhang San does not drink alcohol

Xu also mentions the problem that sentences like

wo bu mai shu
I not buy book
Im not going to buy books

raise (is he forced to say that we have an empty

modal will here?) but he does not really go into
the problem very deeply.
Finally, chapter 6 (47pp.) is about the
quantificational element dou (see Huangs dissertation above). This chapter is worthwhile for several reasons. First, it presents an adequate overview of the literature and the di(erent positions
that have been taken. Second, it pushes the idea
that dou does not really mean all but each
instead. In other words, it is a distributor (yes,
heading DisP) and Xu comes up with several arguments. Thirdly, and probably most importantly,
Xu presents a critical evaluation of the quantificational facts that have been used in the literature
on scope in Chinese. His conclusion is that in
many cases, English scope ambiguity sentences
are translated into Mandarin, and that it is simply
assumed that the English scopally ambiguous
interpretation gets transferred in the translation
process. However, he argues that this is not the
case. This, if true, has consequences for some of
the claims generally thought valid for Chinese.

academic publishers

The Journal of
Germanic Linguistics
Ken Safir
Rutgers University, New Brunswick, NJ, USA
Gert Webelhuth
University of North Carolina, Chapel Hill, USA

The Journal of Comparative Germanic Linguistics provides

a platform for discussion of theoretical linguistic research
into the modern and older languages and dialects of the
Germanic family. Contributions that establish robust
empirical generalizations within a formal theory of grammar
that permit precise discussions are welcome. The empirical
range of the work may either involve cross-linguistic
comparison within the Germanic family or elucidate issues
in Germanic linguistics through the exemplary analysis of
one Germanic language. Papers dealing mostly with
modern English are required to have strong cross-Germanic
implications. Areas of linguistics covered include:
Morphology, Phonology, Pragmatics, Semantics and
Syntax. The journal will be published three times per year
in the English language. Squibs, replies, book and
dissertation reviews will be included.
Subscription Information:
ISSN 1383-4924
1998, Volume 2 (3 issues)
Subscription Rate: NLG 250.00, including postage and

P.O. Box 322, 3300 AH Dordrecht, The Netherlands

P.O. Box 358, Accord Station, Hingham ,MA 02018-0358,U.S.A.


Conference reports

Glot International, Volume 2, Issue 7, September 1996 [1997]

by Joo Costa
GLOW 20, Rabat, 1921/3/97

One of the nice things of conferences is that one

can enjoy a combination of good talks and good
weather! This was true for GLOW 20, which was
held in warm Rabat! For any speaker who might
be nervous about the talk, there was a wonderful
therapy: a ride in one of the petit taxis All your
tension stays in the car and you think: If I survived this, Ill survive the talk!
The following papers were presented:
Henk van Riemsdijk (Tilburg University,
Categorial Feature Magnetism) extended his
work on extended projections, unifying the Categorial Identity Thesis with the apparently contradictory Unlike Feature Constraint in terms of a
Law of Categorial Feature Magnetism, which
makes sure that e.g. Ps may be inserted within an
extended projection of N (in constructions like un
verre de vin) without breaking the projection in
two and without changing the feature specification
of Ps.
Akira Watanabe (Kanda University, Absorption as Feature Checking) proposed that absorption of QPs and Wh-phrases should be reinterpreted as feature checking, making QR and Wh-movement compatible with Chomskys minimalist program.
Sarah Kenelly (Rutgers University, The PFocus Position in Turkish) presented an interesting paper showing that non-specific objects in
Turkish may take wide scope over a subject when
the objects are focused. Her explanation for this
fact hinged on a structural position for contrastively focused elements (right-adjoined to VP),
where the non-specific objects c-command VPinternal material (including the subject).
My own talk (Optimality Theory at the Syntax-Discourse Interface) proposed an OptimalityTheoretic account for situations of conflict between
discourse-based constraints, requiring elements to
stay in situ, and syntactic constraints (Case) requiring elements to move. It was shown that the
di(erences between languages in the way they
optimize these conflicts can be easily formulated
in Optimality terms.
Hans Broekhuis and Joost Dekkers (University of Amsterdam, Minimalism and OT: Derivations and Filters) developed Pesetskys Optimality system, showing that integrating the constraint Obligatory Heads in the system allows for
accounting for relatives in a Flemish dialect and
that a reformulation of Recoverability allows for
accounting for that-t e(ects. Interestingly, their
elaboration of the system led to the conclusion
that the more standard OT-syntax model is su+cient.
Riny Huybregts (Tilburg University, Integrating Indefinite Morpho-Syntax into Fully Interpretable Interface Representations) elaborated on
the theory of Attract F, proposing that features
may be undeleted if necessary for interpretation.
Rajesh Bhatt (UPenn, Obligation and Possession) extended the Freeze-Kayne view on have
and be to the use of these verbs in the expression
of obligation, showing cross-linguistic parallels
between obligation and possession.
Ken Hale (MIT, invited speaker) discussed his
work with Jay Keyser on Argument Structure.
Grant Goodall (Univ. of Texas, El Paso, The
External Argument in Passives and the Syntax/

Argument Structure Interface) presented evidence that by-phrases are not adjuncts, and explored the hypothesis that they are to be represented in the normal position for subjects.
Luis Lpez (Univ. of Missouri at Columbia,
Ellipsis: Logical Form and Discourse) related
ellipsis with D-linking and negation, suggesting
that ellipsis is only licensed within the checking
domain of a discourse-linking functional category.
VP-ellipsis, according to him, is licensed in the
checking domain of SigmaP.
Anders Holmberg (Univ. of Troms, Phonological Feature Movement) analyzed Scandinavian Stylistic Fronting in terms of phonological
feature movement. His proposal was that in these
constructions all that is necessary is that some
phonological matrix occupies the Specifier position
of TopP, making a parallel between this construction and expletives, and explaining why stylistic
fronting only occurs when the subject position is
not filled in.
Maria-Teresa Guasti and Marina Nespor
(Milano and Univ. of Amsterdam, Syntax-Phonology Interactions) looked at certain syntax-phonology interactions, arguing that phonology influences syntax only in the part of the phonological component that may have such an influence (if there
are two options, phonology will decide). They also
argued that the prosodic hierarchy is superior to
recent theories of the influence of syntactic structure on phonology, reinforcing the view on phonology claiming that it is syntax-free.
Angela Ralli (Univ. of Athens, Inflection
Features and the Morphological Component Hypothesis) proposed that certain inflectional features are to be handled within a morphological
component of grammar. She claimed that two
things may happen to features: either they are
checked within morphology or they percolate,
becoming visible to syntax (+interpretable).
Rita Manzini and Anna Roussou (Florence/
UCL and Bangor, Interpretation as Feature Calculus: F-Movement and F-Control) reanalyzed
control in terms of F-movement, proposing that
PRO may be reduced to theta-relations between
Asp(ect) and DP, established in terms of Asp-feature-checking.
Richard Kayne (CUNY, invited speaker) proposed a review of certain constructions, resorting
to VP-preposing as a general mechanism, challenging the traditional view on constituent structure, and suggesting that all movement is overt.
Giuseppe Longobardi (Univ. of Venice, Case
Theory and the Minimalist Program) proposed an
elaboration of Case theory under the minimalist
program, maintaining the traditional distinction
between Spec-head and head-complement and
relating the two types of case-relation to interpretability and locus of checking (PF or LF). His
analysis permitted to avoid certain shortcomings
of Case theory and explain the di(erence between
inherent and abstract case.
K. Scott Ferguson (Univ. of Geneva, Deducing the Invisibility of PP Nodes from Case Checking and Full Interpretation) suggested that the
apparent invisibility of PP nodes for binding could
be derived in a framework assuming that PPs are
dominated by an Agr node which is deleted at LF.
Ian Roberts and Anna Roussou (Univ. of
Stuttgart and Univ. of Wales,Bangor, Interface
Interpretation) proposed a parametrization of
features in terms of +/-p (meaning, whether it

Page 20

needs to be spelled out or not). According to their

proposal +p features could be satisfied by Move (Q
in I-to-C in questions and Agr in V-to-T-to-Agr in
Irish), Merge (Q in clefts in Welsh and Agr in
merged DPs in English), or Move and Merge (Q in
French complex inversion and Agr in French).
This kind of parametrization allowed to solve
certain problems of checking theory based on feature strength.
Elena Benedicto (UMass,Amherst, V-Movement and Its Interpretational E(ects) presented
an analysis of distributional di(erences of determinerless subjects in Catalan and English, arguing that they are the result of independent syntactic operations (V-movement). This led to the conclusion that LF is the result of certain syntactic
operations, and not the cause of these operations.
Murat Kural and George Tsoulas (UCLA and
Univ. of York, Pronouns and the Syntax-Discourse Interface) proposed an integration of discourse aspects in the syntax, by doing indexation
of pronouns via a set of operators (SPEAKER, ADDRESSEE, HERE, NOW and TOPIC) located above the
matrix CP.
Chung-Hye Han (UPenn, Deriving an LF
Interface for Rhetorical Wh-Questions) presented
an analysis of rhetorical wh-questions, showing
that they di(er from regular wh-questions both in
the syntax (making a parallel between rhetorical
questions and neg-inversion) and in the semantics
(where rhetorical questions undergo LF-internal
derivations driven by information structure instructions).
The last talk was presented by Edwin Williams
(Princeton University, Destressing and Ellipsis at the Interface), who elaborated on
his theory that sentence accent and ellipsis
obey the same laws that regulate the distribution of anaphora.
All in all, the conference provided a nice
environment to discuss several views on interface issues and on what people conceive
of as interface problems. A final word for
thanking the organizers, Abdelkader Fassi
Fehri and his crew, for all the work they put
on the conference.
In 1998, there will be two GLOW colloquia: one in Hyderabad, India in January, and
the regular one in Tilburg on Features.

Functional Categories
in Mandarin Chinese
Ding Xu
diss. Leiden University|Holland Institute of Generative Linguistics|HIL Diss 26

Functional Categories in Mandarin Chinese examines

the internal structure of Chinese functional projections,
such as TP, CP, NegP, DisP etc. within the framework
of the Minimalist Program.
It compares some of these categories, as they occur in
Chinese, with their counterparts in other languages.
This comparison aims at providing a principled
explanation for some of the crosslinguistic differences
that one can attest.
The main goal of this study is to show that if a certain
type of functional category is a theoretical requirement
(because there are languages that have been shown to
have it), it will be part of Universal Grammar and thus
be manifested in each individual language.
As shown, linguistic variation can best be explained in
terms of strong versus weak features, in the sense
of Chomsky (1993).
isbn 90-5569-026-0|iii+202pp.|june 1995|NLG 40 (ex VAT, ex P&P)



Holland Academic Graphics

[ s c i e n t i f i c

[ d o c u m e n t ]

P.O. Box 53292

2505 AG The Hague
The Netherlands

p r o c e s s i n g ]

e-mail: mail@hag.nl
http: www.hag.nl
fax: +31 70 4480177