Académique Documents
Professionnel Documents
Culture Documents
Abstract
Author's Copy
strictions on DCIs in five Romance languages to reveal that clitic-doubling
covaries with the possibility of lexical DCIs and the possibility of combin-
ing the DCI with a bare unergative. We argue that Pylkknens (2002) High
Applicative analysis should be broadened to account for these Romance da-
tives and their variable behaviour so that the DCI clitic is always generated
as the head of a high applicative projection. Doubling and non-doubling lan-
guages diverge based on the type of D-element that merges in the specifier of
ApplP. In clitic-doubling languages (Romanian and Spanish), the DCI essen-
tially matches the Bantu high applicative construction in that lexical datives,
as DPs, are merged in the specifier position of the applicative phrase. The Appl
head in doubling languages may also introduce a pro as an applied argument
whose reference is recovered via the agree relation established with the applied
morpheme. In non-doubling languages (French, Portuguese, Italian), the DCI
functions quite differently. It may only occur in clitic form as the head of the
applied phrase, since lexical indirect objects are PPs and consequently may
not merge in Spec,ApplP. A result of this restriction is that an expletive opera-
tor is merged in the specifier, binding a VP-internal referential DP, accounting
for the restriction on bare unergatives.
1. We would like to thank the following people for their comments and suggestions: Marc
Authier, Emanuel da Silva, Anna-Maria di Sciullo, Jose Ferreira, Nicolas Guillot, Virginia
Hill, Monica Irimia, Michael Lettieri, Stefania Marzano, Diane Massam, Marco Nicolis, Ana
Teresa Prez-Leroux, Mihaela Pirvulescu, Lisa Reed, Milan Rezac, Mario Saltarelli, and two
anonymous Probus reviewers. Funding for this research was provided in part by a Major
Collaborative Research Initiative grant from the Social Sciences and Humanities Research
Council of Canada (412-2003-1003, Di Sciullo).
Author's Copy
Author's Copy
Introduction
Author's Copy
crucial generalization that clitic doubling covaries with the possibility of full
nominal DCIs and with the use of this dative argument in intransitive contexts.
Assuming a high applicative analysis in which the dative clitic is the func-
tional head, we claim that the fundamental difference between DCIs in dou-
bling and non-doubling languages derives from the fact that dative-marked
lexical DPs are available in the former, but not in the latter. That is, in Spanish
and Romanian, a lexical DP may merge as the specifier of the applicative head,
receiving dative case manifested as a bound morpheme, and triggering phi-
feature agreement with the head. There is thus a clear parallel between theses
clitic-doubled DCIs and high applied arguments in the Bantu languages.
On the other hand, no lexical DP may merge with the applied head in French,
Italian, and Portuguese, for in these languages, lexical dative arguments are
not DPs, but prepositional phrases headed by a. Since the Appl head always
selects for an individual in its specifier (Pylkknen 2002), PPs are blocked
from merging in this position. The lack of dative case is therefore claimed to
account for the lack of lexical DCIs in non-doubling languages. In order to
account for the puzzling fact that DCIs in French, Italian, and Portuguese must
co-occur with an otherwise unrelated referential DP in their domain, we claim
that the Appl head still requires the merger of a DP in its specifier and that in
this instance, it is a null operator with scope over VP, binding a referential DP
within it. The operator satisfies the selectional requirement of the Appl head,
whose case and theta properties have been absorbed by the clitic.
In Section 1 we review the properties of high applicative arguments as pro-
posed in Pylkknen (2002). The well-studied Bantu languages, and as it turns
out Latin as well, exemplify the straight-forward case for which the high ap-
Author's Copy
Author's Copy
This section reviews the semantic and syntactic properties of high applicative
arguments as proposed in Pylkknen (2002). While benefactive/malefactive ar-
guments in the Bantu languages provide the classic, straightforward examples
of high applied arguments, we show that the analysis also captures the Latin
Author's Copy
Author's Copy
dativus commodi/incommodi (DCI). In doing so, we emphasize both the con-
tinuation and the divergence of the Romance DCI from its Latin ancestor, a
point to which we return in Section 4.
Appl vP
applicative marker
Author's Copy
Author's Copy
V ApplP
Appl DP
applicative marker
The differences between these two types of applicative constructions make a
number of predictions. First, high applicative heads should be unrestricted in
their ability to combine with the various verb types. In contrast, since low ap-
plicative heads denote transfer of possession, they may not occur with unerga-
tives and make no sense with static verbs. In addition, given that low applied
Author's Copy
Author's Copy
arguments may only occur with the individual which is the internal argument
of the verb, they cannot co-occur with a second internal argument of the verb.
This means that while high applicatives may combine with any so-called di-
transitive verb, low applicatives may not.
Following these diagnostics, the applied arguments shown below, taken from
several Bantu languages, are clearly high; they can occur with unergative verbs
(3a), with static verbs (3b), and with ditransitive verbs (3c).
(3) a. Mukasa o-amb-el-a Katonga
Mukasa.3SG PAST-speak-APPL-FV Katonga
Mukasa spoke for Katonga (Venda: Pylkknen 2002: 10)
b. Nd-o-far-el-a Mukasa khali
1SG-PAST-hold-APPL-FV Mukasa pot
I held the pot for Mukasa (Luganda: Pylkknen 2002: 25)
c. Umugre a-r-mu-he-er-a imbwa ibiryo
woman SP-PR-OP-give-APPL-ASP dog food
The woman is giving food to the dog for him (Kinyarwanda:
adapted from McGinnis 2001: 3)
On the other hand, the diagnostics suggest that a language like English does not
have high applicative arguments. While non-core indirect objects may combine
to form a possession relation with another individual (4a), they may not com-
bine with a transitive verb in the absence of such a relation (4b). Nor can they
combine with unergative (4c), static (4d), or ditransitive verbs (4e); examples
(4ad) from Pylkknen (2002).
Author's Copy
Author's Copy
In what follows, we examine what these diagnostics can tell us about similar
non-core indirect objects in Latin and the Romance languages.
Author's Copy
other individual.3
2. The only reading of this sentence is one in which the indirect object receives the cake.
3. The various functions of the Latin dative have traditionally been described in semantic terms,
and as a result, DCIs tend to encompass both subcategorized and non-subcategorized dative
arguments. For example, Latin grammars generally group the DCIs illustrated in (5) with the
dative argument of so-called intransitive verbs like nocere to harm and invidere to envy.
We use the term DCI only as it pertains to non-subcategorized arguments.
Author's Copy
Author's Copy
Author's Copy
the head in Bantu is a bound morpheme and the applied argument bears no case
marking (6b).
DPDAT Appl
Appl vP
b. Applied benefactive/malefactives in Bantu
ApplP
DP Appl
Appl vP
The present analysis maintains the crucial distinction between the dativus com-
modi/incommodi and the dativus ethicus. They are similar in that neither is sub-
categorized by the verb, but unlike DCIs, the dativus ethicus (DE) is limited to
Author's Copy
Author's Copy
first and second person pronouns. Moreover, rather than denoting the person
who benefits or who is adversely affected by the event, the DE roughly desig-
nates, in emphatic clauses, the person taken as witness among the performers of
the utterance (see Maurel 1982: 79 and Lavency 1997: 164). Woodcock (1959),
for example, translates the first person pronoun of the Latin DE as please! or
pray, and the second person pronoun as for your pleasure I tell you this, you
will be surprised to hear, or lo and behold!. These translations, while slightly
archaic, are noteworthy because they reflect the non-referential reading of the
DE. As Cardinaletti and Stark (1994: 51) assert, the DE is a discourse particle,
and as such, there is no referent to these pronouns, not even derivatively.
Author's Copy
morning
But, lo and behold, Caninius suddenly came to me in the morn-
ing. (Cicero, ad Familiares Epistulae 9,2,1, in Woodcock 1959:
47)
DEs are thus semantically distinct from DCIs in Latin, as they are in Romance.
Moreover, evidence is given in Jouitteau and Rezac (2007) for French, and in
Diaconescu (2004) for Romanian, that the DE is not only semantically distinct
from the DCI, but syntactically distinct as well. We emphasize these distinc-
tions since the term ethical dative is sometimes used in the literature to refer to
DCIs. In order to mitigate terminological confusion surrounding our examina-
tion of DCIs in Romance, we maintain the Latin terms and their definitions as
proposed above. We also avoid a number of other terms that are used more or
less interchangeably in the literature to refer to Romance DCIs, DEs, or both,
some of the more common being ethical, affected, free, or nonlexical
dative, or dative of interest.
Author's Copy
Author's Copy
2.1. French
The DCI in French has been the subject of some consideration and studies have
converged on two points: the DCI is only acceptable in its clitic form, which
is to say that it may not be expressed as an a-phrase, and it must c-command a
referential DP. In this sense, the properties of the French DCI are very different
from its Latin counterpart, yet it has been convincingly argued that the French
DCI is indeed a VP-external argument, as it is in Latin.
Author's Copy
Author's Copy
2.1.1. Constraints on the form. A number of studies have demonstrated that
in French, DCIs are only felicitous as clitics (Rouveret and Vergnaud 1980;
Rooryck 1988; Herschensohn 1992; Authier and Reed 1992; Jones 1996).
4. Examples that are not accompanied by a reference have been provided by our informants.
Author's Copy
Author's Copy
Note that a sentence like (9) might be used to argue that lexical DCIs are actu-
ally permitted in French:
It is important to point out, however, that datives having a possessor reading can
often take on an additional, affected interpretation that resembles that of the
DCI. As a result, possessor datives are often treated as DCIs and vice versa.
We follow Borer and Grodzinsky (1986) in establishing the crucial syntactic
Author's Copy
Author's Copy
distinction between the two types of datives. Like Hebrew, the possessor dative
in French can be questioned but the DCI cannot :
5. The so-called dessus datives (Elle lui a tir dessus) pattern like DCIs in that they do not
have a lexical equivalent and may not be queried (*A qui est-ce quelle a tir dessus?). If
a constituent [dessus pro] is assumed, then it satisfies the syntactic requirement of the DCI,
described further on, to c-command a referential DP.
Author's Copy
Author's Copy
Author's Copy
would with any canonical ditransitive verb.7 Again, Borer and Grodzinskys
Author's Copy
Author's Copy
diagnostic can serve to distinguish these extended datives from DCIs; since
they can be queried, they are not DCIs:
(12) a. qui est-ce que Paul a ouvert cette
to whom is-DEM that Paul AUX.3SG opened this
porte?
door
For whom did Paul open this door?
b. qui est-ce que Paul a fabriqu cette
to whom is-DEM that Paul AUX.3SG fabricated this
table?
table
For whom did Paul fashion this table?
Author's Copy
object, shown in (13a). Second, no possession relation is implied between the
dative and the internal argument of the verb, as demonstrated in example (13b).
(13) a. Je vais te lui crire une lettre.
I go you.DAT DAT.3SG write.INF a letter
Im going to write a letter to him/her for you. (Herschensohn
1992: 126)
b. Jean lui a attrap deux rhumes.
Jean DAT.3SG AUX.3SG caught two colds
Jean caught two colds on her. (Authier and Reed 1992: 295)
On the other hand, unlike Latin and Bantu, a number of studies have pointed
out that DCIs in French are subject to a constraint which prevents them from
occurring in intransitive contexts (Herchensohn 1992, Rooryck 1988, and Au-
thier and Reed 1992). The contrast in (14) illustrates how DCIs are illicit with
unergatives, but when this same verb is used transitively, the DCI is perfectly
acceptable.
(14) a. *Paul lui a bu.
Paul him/her.DAT AUX.3SG drink.PST.PTCP
Paul drank on him. (Authier and Reed 1992: 298)
ministre
minister
I will write it to the minster/ the letter that I wrote to the minster/ the letter has been
written to the minister
Author's Copy
Author's Copy
Author's Copy
(16) Il lui a parl sa fille pendant
he DAT.3SG AUX.3SG spoken to his daughter for
trois heures.
three hours
He spoke to his daughter for him for three hours. (Rouveret and Ver-
naud 1980: 169)
Authier and Reed (1992: 302306) note however, that VP-external adjuncts
denoting cause or time, for example, do not create a licensing environment for
DCIs. Similarly, adverbs do not improve a DCI used with an unergative verb.
(17) a. *Alfred lui a rot pour choquer ses
Alfred DAT.3SG AUX.3SG burped for shock.INF his
invits.
guests
Alfred burped on him to shock his guests.
b. *Ses hommes lui dfileront demain (au
his men DAT.3SG parade.FUT tomorrow (at-the
lieu daujourdhui).
place of-today)
His men will parade tomorrow for him (instead of today).
c. *Marc lui a tudi fort bien cette anne.
Marc DAT.3SG AUX studied very well this year
Marc studied very well this year.
Author's Copy
Author's Copy
Author's Copy
(18) IP
Spec I
Agr-S FP
AgrP
Agr-O VP
lui NP* V
V NP
(Authier and Reed 1992: 307)
Since these datives are taken to be a type of external argument much like the
canonical agentive argument, they are said to depend on case assignment within
the VP (following Burzios generalization). The general constraints on verb se-
lection fall out from this dependency as does the fact that VP-internal adjuncts
also provide a licensing environment for DCIs.
While Authier and Reeds GB analysis captures many of the French facts,
it leaves little room to account for languages that do not require their DCIs to
co-occur with a VP-internal DP, nor does it connect French DCIs with similar
applied arguments in languages like Bantu.
Author's Copy
Author's Copy
2.2. Portuguese
There is a striking parallel between the constraints on French DCIs and those
on European Portuguese DCIs which are also only felicitous as clitics and pro-
hibited in intransitive contexts.
Author's Copy
he connected-me.DAT kindly the light
He kindly switched on the light for me. (Vilela 1986: 146)
b. Ele ligou-lhes amavelmente a luz.
He connected-DAT.3PL kindly the light
He kindly switched on the light for them.
8. In de Andrade Berlincks (1996) survey of Portuguese datives, she claims DCIs may appear
as full nominal arguments preceded by the preposition a:
Recall that a similar example is discussed for French in (11a). We argued that these kinds
of dative arguments are construed as extended datives, or recipients, not DCIs. The same
holds for Portuguese; the dative in (i) is a recipient since recipients and possessor datives can
be questioned, while DCIs cannot:
Author's Copy
Author's Copy
Author's Copy
2.2.2. Constraints on verb selection. Both Vilela (1986) and de Andrade
Berlinck (1996) point out that the Portuguese DCI is not a subcategorized ar-
gument of the verb. According to them, dative clitics of this type always have a
clear benefactive/malefactive interpretation regardless of the verb with which
they appear.
The examples in (21b), (22b), and (23) demonstrate that the DCI need not
be in a possessor relation with the theme, further showing that the Portuguese
DCI is distinct from the possessor dative.
(22) a. O vaso partiu-se-me/lhe.
the vase broke-se-DAT.1/3SG
The vase broke on me/him. (Vilela 1986: 146)
b. O vaso do Emanuel partiu-se-me.
the vase of Emanuel broke-se-me.DAT
Emanuels vase broke on me.
(23) Ele apanhou-me c uma gripe.
he caught-me.DAT here a cold
He caught a cold on me.
De Andrade Berlinck (1996: 133) notes that the DCI in Portuguese is avail-
able for all transitive verbs. Although she does not support this claim in any
detail, our informants have confirmed that this is true, as clearly demonstrated
in examples (19), (20) (21), and (23). The change of state verb in (22) also sup-
ports the claim, since these types of verbs have been argued to be underlyingly
transitive.
Author's Copy
Author's Copy
Author's Copy
(25) a. A mulher dele fugiu anteontem.
the wife of.him fled the day before yesterday
His wife ran off the day before yesterday.
b. *A mulher dele fugiu-lhe anteontem.
the wife of.him fled-him the day before yesterday
His wife ran off on him the day before yesterday.
c. A mulher dele fugiu-lhe com o vizinho.
the wife of.him fled-him with the neighbour
His wife ran off with the neighbour on him.
2.3. Italian
Much like DCIs in French and Portuguese, Standard Italian DCIs are only fe-
licitous as clitics and are illicit with unergative verbs. There are, however, va-
rieties of Italian that accept DCIs as a-phrases and which permit them with
unergative verbs. We do not examine these varieties in detail as it is well be-
yond the scope of this article, but we nonetheless note that they appear to sup-
port the correlation we advance here, between languages that permit dative
clitic-doubling and DCIs that show few restrictions in terms of their form and
Author's Copy
Author's Copy
verb selection. We leave a closer examination of DCIs in the central and south-
ern varieties for future study. In as much as is it possible, we therefore wish to
distinguish Standard Italian from other varieties, recognizing the challenge this
presents given that DCIs are generally associated with informal registers.
2.3.1. Constraints on the form. Standard Italian DCIs are only acceptable
as clitics, as the sentences in (26) show.
(26) a. Gli invitati gli hanno mangiato tutto quello
the guests him.DAT AUX.3PL eaten all that
che rimaneva nel frigo.
which remained in-the fridge
The guests ate everything that was left in the fridge on him.
b. *Gli invitati hanno mangiato tutto ci che
the guests AUX.3PL eaten all that which
rimaneva nel frigo a Marco.
remained in-the fridge to Marco
Author's Copy
Author's Copy
The guests ate everything that was left in the fridge on Marco.
With respect to the a-phrase a Marco in (26b), a DCI reading is unavailable for
all of our informants. Some, however, find it acceptable as a possessor in the
sense that the guests ate everything that Marco had left in his fridge.
In Folli and Harley (2006), Italian benefactives/malefactives are described
as licit both as clitics and a-phrases. They specify, however, that the full nom-
inal form is sharply constrained to a relationship of possession with the theme
and context-specific individuals, as the contrast demonstrates below (examples
taken from Folli and Harley (2006: 124); the translations are our own).9
9. All of our informants strongly reject both of the examples in (27) as DCIs, although some
find them marginally acceptable as possessors (di-possessors being the preferred form). In
informal/non-school central Italian, a DCI reading is possible in (27b), but in this case, the
a-phrase a Marco must also be accompanied by a dative clitic, since this variety of Italian is
a clitic-doubling variety (Mario Saltarelli, p.c.).
This observation appears to support the correlation we establish further on between lexical
DCIs and dative clitic-doubling languages.
Author's Copy
Author's Copy
Author's Copy
to whom the guests AUX.3PL
mangiato tutto ci che rimaneva nel frigo?
eaten all that which remained in-the fridge
On whom did the guests eat everything that was left in the fridge?
Author's Copy
Author's Copy
Like the direct object in (30b), certain VP-internal adjuncts can correct an oth-
erwise illicit DCI:
(31) a. Lorenzo ha ruttato.
Lorenzo AUX.3SG burped
Lorenzo burped.
b. *Lorenzo le ha ruttato.
Lorenzo DAT.3SG AUX.3SG burped
Lorenzo burped on her.
c. Lorenzo le ha ruttato davanti agli
Lorenzo DAT.3SG AUX.3SG burped in front of the
ospiti.
guests
Lorenzo burped on her in front of all the guests.
Note that in Italian, unergative verbs are commonly used with what appears to
be a DCI, as (32a) demonstrates. We point out, however, that in Standard Ital-
ian, forms of the dative clitic other than the first person singular, and a-phrases,
Author's Copy
Author's Copy
are felt to be either ungrammatical or extremely bad, strongly suggesting that
these dative clitics are not DCIs.10
(32) a. Il mio bambino non mi dorme.
the my baby NEG me.DAT sleep
My baby wont sleep for me.
b. ?*Questo bambino non ti/gli dorme proprio.
this baby NEG you/him.DAT sleep properly
This baby wont sleep for you/her.
c. ?*Questo bambino non dorme a Rosa.
this baby NEG sleep to Rosa
This baby wont sleep for Rosa.
2.3.3. Previous analyses. Folli and Harley (2006) propose that Italian DCIs
are independently generated above the VP in the specifier of a high applicative
10. In some varieties of Italian, (32b) is acceptable, and yet the lexical DCI in (32c) is not. Some
speakers have indicated that the lexical DCI is improved, however, when accompanied by the
dative clitic le in a clitic-doubling construction.
Not surprisingly, these speakers also permit DCIs with unergative verbs in general, supporting
the correlation that is established further on between clitic-doubling languages and DCIs that
may occur with unergative verbs.
Author's Copy
Author's Copy
phrase, the head of which is null. In this sense, Italian DCIs are generated in
the same position as applied benefactive/malefactives in Bantu and DCIs in
Latin.11
(33) v
v ApplP
cl/DPDAT Appl
Appl VP
(adapted from Folli and Harley 2006: 133, 136)
To account for the fact that the high applicative head in Italian cannot merge
with an unergative vP structure, a stipulation about the semantics of the head
is proposed such that it must relate the applicative argument in its specifier to
a referential element. Unergatives fail to satisfy this requirement since, follow-
Author's Copy
Author's Copy
ing Hale and Keyser (1987, 2002), these verbs are derived from the incorpo-
ration of a bare noun complement. No analysis is given for the restriction on
a-phrases.
2.4. Romanian
Unlike the formal and selectional constraints that we see in French, Portuguese,
and Italian, Romanian DCIs may appear as clitics, or as clitics doubled with a
dative-marked full nominal argument, and they may appear with both transitive
and intransitive verbs. They thus behave much like their Latin counterparts and
like applied benefactive/malefactive arguments in Bantu.
11. The relative placement of ApplP with respect to vP is inconsequential for the present analysis.
Note that we follow Pylkknen (2002), merging ApplP above the vP and below the external
argument.
Author's Copy
Author's Copy
Author's Copy
thus not generated within the VP as a low applied argument. While DCIs can
occur with transitive verbs, they can also occur with both unergative and static
verbs, shown in (35) and (36), demonstrating that they are neither dependent
on the presence of a direct object nor on an entailment of possession between
it and the direct object.
2.4.3. Previous analyses. Diaconescu argues that Romanian DCIs are gen-
erated in a high applicative phrase, following Pylkknens (2002) proposal for
the syntactic representation of applied benefactive/malefactive arguments in
Bantu.
Author's Copy
Author's Copy
DPSubj Voice
Mihaela
Voice0 ApplP
DP Appl
applied atrenorului/
argument mie, tie
Appl vPACT
applicative marker i
v0 Root
alerg-
Diaconescus proposal for Romanian DCIs parallels our own for Latin (see
Author's Copy
Author's Copy
Section 1.2). The only difference between the two lies in the expression of the
functional head: in Latin it is null while in Romanian it expresses the person
and number features of the applied argument. We adopt Diaconescus account,
with a slight modification, discussed in Section 3.1.
2.5. Spanish
2.5.1. Form. Spanish permits the DCI to appear both as a simple clitic, and
as a clitic doubled by a coreferential full nominal form headed by the dative
case marker a.
Author's Copy
Author's Copy
Author's Copy
that was.1SG caring-for
The neighbors dog, which I was looking after, died on me.
(Roldn 1972: 31)
Judgments vary widely on these cases. Some find only first and second person
clitics to be felicitous while others accept third person clitics but do not accept
full nominal forms like those in (38c), (38d), and (38e).
2.5.2. Verb selection. Unlike French, Portuguese, and Italian, Spanish DCIs
do not appear to have any formal constraints on the type of verb with which
they may occur. They show no dependency on a VP-internal constituent and are
argued to be VP-external arguments in the same way as Romanian and Latin
DCIs and applied benefactive/malefactives in Bantu.
Examples (38c), (38d), and (38e) demonstrate that when Spanish DCIs oc-
cur with (underlying) transitive verbs, a possession relation between the dative
argument and the theme is not required. DCIs are thus distinct from possessor
datives and recipients. Spanish DCIs are also perfectly acceptable with unerga-
tive verbs (39ac)12 and ditransitives (39d).
12. As pointed out by an anonymous reviewer, some verbs alternate between unaccusative and
unergative syntax. We have thus attempted to use relatively uncontroversial unergative verbs
in our examples, following Sorace (2000).
Author's Copy
Author's Copy
Author's Copy
2.5.3. Previous analyses. Conclusions drawn in previous analyses depend
largely on what forms of the DCI the author finds acceptable, but all analyses
appear to converge on the basic idea that the Spanish DCI is a non-core argu-
ment. In fact, Jaeggli (1986) presents them as evidence of syntactic clitics that
are not generated in or associated with any argument position at all. Cuervo
(2003) proposes a high applicative analysis for Spanish DCIs, in which the
dative clitic is the head of the applicative phrase.
(40) Juanita ya le camina.
VoiceP
DP Voice
Juanita
Voice ApplP
Appl vP
le
vdo +Root
camin-
Cuervo (2003: 198)
The absence of a specifier reflects that fact that Cuervo does not accept full
nominal DCIs. She claims that this high applicative head in Spanish is de-
fective in the sense that it does not license an argument DP, comparing it
to other defective heads such as those involved in passive and impersonal se
Author's Copy
Author's Copy
constructions. While we agree with Cuervo that there is indeed something dif-
ferent about this head, it is not clear to us that it is defective. Rather, in the case
of French, Portuguese, and Italian, the pronominal head is super-effective in
that it encodes both the theta role and the case features assigned by the terminal
node.
2.6. Summary
Author's Copy
Where it has been previously observed that the DCI is accompanied by cer-
tain syntactic restrictions, the accounts tend to be stipulative. Authier and Reed
attempt to link the properties of the French DCI to other features of the gram-
mar, but all other accounts remain essentially descriptive. Furthermore, cross-
linguistic variation has been entirely overlooked.
We propose that a comparative approach to DCIs can offer a unified and
a more satisfactory account of their behaviour in Romance. The pattern that
emerges here, through the comparison of five Romance languages, has gone
unnoticed: the DCI in clitic-doubling languages patterns much like high ap-
plied arguments in Latin and Bantu, while the DCI in non-doubling languages
is restricted to its clitic form and it must c-command a referential DP merged
either as a direct object or as part of an adjunct PP. In other words, clitic-
doubling covaries with the possibility of full nominal DCIs and with the use of
Author's Copy
Author's Copy
Based on the evidence assembled in Section 2, we build on the idea that the
non-subcategorized dativus commodi/incommodi in Romance should be an-
alyzed as a high applicative phrase. The generalized ApplP for Romance is
given in (41a), and is compared to its Latin counterpart in (41b), the more
straight-forward case. Note that the functional material above ApplP is omitted
Author's Copy
Author's Copy
for simplicity.
X ApplP
?? Appl
Appl vP
clDAT
v0 VP
b. Latin DCI structure
. . . XP
X ApplP
Appl vP
v0 VP
Author's Copy
Author's Copy
The two constructions differ in two important ways. First, the Romance DCI
construction makes use of the dative clitic as its applicative head whereas, in
Latin, the head is phonologically null. Second, the applied argument in Latin
appears as a lexical DP or a pronoun and, crucially, both are morphologically
marked as datives; see the examples in (5). Romance languages, however, are
not unified in how the Spec,ApplP position is treated, leading to variation in the
realization of the applied argument. Consequently, we must elaborate both on
the nature of the applied argument in modern Romance and on the link between
the cldat and Spec,ApplP. We will see that the variation we observed in Section
2 is to be attributed to the DP/PP status of lexical DCIs and, consequently, also
to the properties of the head.
Author's Copy
Let us first detail the structure in (41a) in clitic-doubling Romance languages
such as Spanish and Romanian. To illustrate, we repeat the structure proposed
by Diaconescu introduced in Section 2.4.3.
DPSubj Voice
Mihaela
Voice0 ApplP
DP Appl
applied atrenorului/
argument mie, tie
Appl vPACT
applicative marker i
v0 Root
alerg-
For sentences involving only the presence of a clitic DCI, like the one in (43)
below, Diaconescu claims that they involve a defective head which does not
project a specifier, essentially extending Cuervos analysis of Spanish DCIs to
Romanian.
Author's Copy
Author's Copy
DPSubj Voice
Mihaela
Voice0 ApplP
Appl vPACT
i
v0 Root
alerg-
(Diaconescu 2004: 212)
This structure is problematic since it suggests that there are two types of high
applicative heads in Romanian with the exact same morphophonology, syntac-
Author's Copy
Author's Copy
tic distribution, and semantic interpretation. Furthermore, if we take the func-
tion of applicative projections seriously which is to introduce a non-core in-
direct object then there should logically be no defective applicative heads.
For these reasons, we propose that the Appl head always projects further. In
the absence of an overt coindexed DP, pro occupies the Spec,ApplP position,
as in (44), assuming the applied argument function.
(44) VoiceP
DPSubj Voice
Mihaela
Voice0 ApplP
DP Appl
pro
Appl vPACT
i/mi
v0 Root
alerg-
Author's Copy
Author's Copy
(45) ApplP
Appli vP
clDAT
The crucial property of this version of the DCI construction is the Case status of
the doubled DCI argument. As Diaconescu and Cuervo argue, the preposition
that marks the DP as a dative la in Romanian, and a in Spanish should
be analyzed as a bound morpheme. As a result, the doubled lexical DCI is
always a morphologically marked dative DP. It thus turns out that the structure
in (41) is nearly identical to a regular clitic projection and differs only in: (1)
the categorial status of the clitic: a regular clitic is a voice projection, in the
sense of Sportiche (1996) not an Appl head; and (2) the fact that its associated
doubled DP has been externally merged: in regular clitic constructions it is
internally merged (i.e., moved) in the specifier of the clitic projection.
There is no reason to believe that this transposition of properties from a regu-
Author's Copy
Author's Copy
lar clitic projection to an applicative projection should not apply to the internal
properties of the dative clitic itself. Indeed, the dative clitic acting as an Appl
head appears to behave exactly like a regular clitic in requiring a DP in a local
agreement relation in order to value its phi features. Assuming that the dative-
marked DP, as a VP-external argument, is linked to a thematic interpretation
and receives Case by virtue of the local relation that it holds with the Appl
head, we obtain the more developed structure in (46).
(46) ApplP
Appl[Case, ] vP
clDAT[u ]
The morphological instantiation of the Appl head is thus essentially a reana-
lyzed dative clitic that still behaves like a quintessential clitic with respect to
feature valuation (i.e. as an agreement marker). The ApplP differs from a reg-
ular dative clitic projection in that it involves a DP merged outside of the VP
domain.
As a result, there are no syntactic restrictions on DCIs in clitic-doubling Ro-
mance languages. The dative clitic as an applicative head continues to behave
as a clitic, but internal merge of a DP to its Spec is blocked by the clash this
would create between the thematic and Case features associated with the Appl
head, thereby forcing external merge of an independent DP (lexical or pro).
Author's Copy
Author's Copy
Author's Copy
As such, all of its features are interpretable. Our approach is thus at odds with
the idea that the dative clitic in its applicative uses corresponds to a defective
head. A defective head is one that is incomplete with respect to some feature(s)
and thus cannot serve as the licensing head for an element in its Spec. In con-
trast, the dative clitic as an applied morpheme in non-doubling languages is
13. For the view that the preposition introducing the indirect object in French is a Case morpheme,
see Jaeggli (1982), Ruwet (1982), Miller (1992).
14. Portuguese -phrases likewise behave as PPs, not DPs, based on the following A/A test orig-
inally proposed to demonstrate the prepositional status of French in Herslund (1988: 46).
A PP can raise out of a DP, as shown in (ii), but the same PP may not raise out of the in-
direct object constituent headed by in (iv), presumably because this in Portuguese is a
preposition.
This diagnostic shows the same results for Italian a. Since the A/A diagnostic is invalid for
Spanish and Romanian, we adopt Cuervos (2003) position that Spanish dative a-phrases are
DPs and that the a in a clitic-doubled dative is therefore not a preposition. We likewise assume,
with Diaconescu (2004), that Romanian la in clitic-doubling contexts is a case marker, not a
preposition.
Author's Copy
Author's Copy
super-effective in being able to encode the Case and theta role properties of
the Appl head and in valuing its phi features in the (extra)linguistic context.
The structure in (47) provides the non-doubling counterpart of (46) above.
(47) ApplP
Appl
Appl[Case, ] vP
clDAT[i , uCase, u ]
Author's Copy
part of the structure is the natural one that Spec,ApplP be filled by a DP or pro
given that the Appl cannot be vacuously merged to vP, i.e., without actually
introducing an applied argument. Interestingly, despite the status of the dative
clitic as the applied argument in non-doubling languages, a restriction on the
DCI projection remains in that the Appl head appears to require the overt pres-
ence of a licensing DP in its c-command domain. This is a strange property
from a theoretical perspective since this dependency relation involves two fully
independent constituents having nothing to share. Logically, this restriction
could be attributed to properties of the dative clitic or to the applicative projec-
tion itself.
Accordingly, one potential approach would be to extend our analysis of dou-
bling languages to non-doubling languages and to conceive of the high applica-
tive phrase as more or less equivalent to a dative clitic phrase in the sense of
Sportiche (1996). However, since the dative clitic receives its theta role from
the position and its reference directly from the (extra)linguistic context (like a
strong pronoun), a dependency on a nominal element in its domain is superflu-
ous. It would be as though the clitic probed for a DP without actually entering
into any kind of agree or checking relation with it. One might liken such a de-
pendency to a remnant property of the clitic, but introducing the possibility of
such vacuous probing into the model does not conform to a minimalist agenda
and as such should be set aside.
We thus turn to the possibility that the dependency relation is a property of
the applicative projection. Does the Appl head itself probe for an overt referen-
tial nominal element in its domain? Motivating such a dependency as a feature
of the projection would in essence be saying that high applicative projections
Author's Copy
Author's Copy
Author's Copy
that the individual is somehow involved in this event. Non-doubling languages
offer a very different picture in that no external merger of a lexical DP occurs.
We have argued above that in non-doubling languages the clitic acts as the
applied argument. This results in the absence of a separate lexical dative DP
which, in turn, could result in a lack of visibility of the commodi/incommodi
semantics. Intuitively, everything seems to function as though the individual in-
troduced by the clitic in Appl could only be affected by the event denoted by the
verb phrase if and only if the event involved another DP. To illustrate, an eat-
ing event would not be sufficient in itself to support the commodi/incommodi
interpretation. To do so it would have to involve, say, something that was eaten
(e.g., everything that was in the refrigerator) or another object (e.g., a cush-
ion or couch on which the eating was done). Under this approach, the Appl
head would thus probe a nominal which would then act both as a signal that
the dative clitic corresponds to a DCI not to a VP-internal argument and
that it is somehow affected by the event. Yet, as tempting as this approach may
be, the introduction of interpretive considerations to assign a functional rai-
son dtre to the restriction is misguided. After all, the commodi/incommodi
interpretation is entirely dependent on the nature of the event of which the DCI
is an argument and there is no logical reason to believe that, for instance, the
addition of an internal argument to an event built on the basis of an unergative
verb should somehow render that event more liable to affect the extra argument
represented by the dative clitic. Furthermore, the lack of such requirement in
Latin and clitic-doubling Romance languages would remain a complete mys-
tery. The restriction on French DCIs cannot reduce to an interpretive require-
ment on the construction. In our view, the apparent interpretive function of the
Author's Copy
Author's Copy
Author's Copy
requirements of the Appl head. Now, for French-type DCIs, even though the
principal function of the applicative head is satisfied semantically by the da-
tive clitic, the syntactic part of the selectional requirement remains unfulfilled;
Spec,ApplP must be filled by an externally-merged DP. Yet, in French, there is
simply no lexical DP candidate that can satisfy this requirement since it would
remain without Case and theta role. We propose that the only candidate in the
lexicon that can satisfy the syntactic selectional requirement of Appl in these
languages is an operator, which incidentally must bind a DP in its domain. The
condition in non-doubling languages on an overt DP merged below the Appl
is thus a result of the merger of an expletive operator in Spec,ApplP. The final
structure we propose for DCIs in non-doubling languages is given in (48).
(48) ApplP
Op[iD] Appl
Appl[Case, , uD] vP
clDAT[i , uCase, u ]
. . . DP
Note that the DP in question must be referential. The following examples are
adapted from Rouveret and Vergnaud (1980: 170). They involve direct objects
that are part of idioms, and as such are non-referential.
(49) a. *Il lui a cass la crote.
he DAT.3SG AUX broken the crust
He had a bite to eat on him/her.
Author's Copy
Author's Copy
Author's Copy
Given the expletive nature of this operator, it may not be surprising that the
DP needs to be referential since a similar link can be found in other expletive
constructions. Furthermore, the reason why unergative verbs are excluded in
DCI constructions in non-doubling languages becomes clearer: the bare N ob-
ject in unergative VPs is non-referential and as such cannot serve for expletive
operator binding.
Finally, the VP-internal DP appears frozen in place and as such cannot be
raised by overt A or A -movement as shown by the following examples with
WH interrogatives (51a vs. 51b, c).15
15. Note that extraction of a DP out of a PP in French cannot be used to support our observation
since preposition stranding is not allowed in Standard French. The sentence in (i) is thus
ungrammatical for independent reasons.
We provide examples in (ii) involving extraction of the whole PP, which are equally ungram-
matical, in contrast with the equivalent sentences without DCI (iii).
Author's Copy
Author's Copy
Author's Copy
commanding DP (his mother), which includes a pronoun coindexed with the
variable.
Author's Copy
Author's Copy
b. La meilleure bouteille a t
the best bottle AUX.3SG AUX
bue.
drink.PST.PTCP FEM
The best bottle has been drunk.
c. Ils lui ont bu la
the him/her.DAT AUX.3PL drink.PST.PTCP the
meilleure bouteille.
best bottle
They drank the best bottle on him.
Author's Copy
The behavior of the DCI construction in French with respect to crossover ef-
fects thus lends support for the structure in (48), although the lack of corefer-
ence between the operator and the VP-internal DP prevent us from concluding
that the DCI construction is in every (relevant) respect identical to the ones that
have traditionally given rise to crossover effects.
Finally, the possibility of internal merger of a DP into Spec,ApplP can be
ruled out based on the fact that this solution would not account for the DPs
which are constituents of oblique objects or circumstantial PPs, in (15), (16),
(25), and (31). Island constraints block movement, and movement of the en-
tire PP is not permitted since only individuals may merge in Spec,ApplP. One
might also consider the possibility that pro and only pro is able to occupy
Spec,ApplP in non-doubling languages, but this would leave no account for
the structural requirement that a referential DP must merge within the VP, a
condition captured by the operator analysis proposed above.
16. Note that in Spoken French, the accusative pronoun is normally not pronounced and this
string surfaces as simply ils lui ont bue. The same holds when the dative clitic is a thematic
argument:
Author's Copy
Author's Copy
3.3. Summary
To sum up, Romance DCI constructions share the same generalized high ApplP
and can be divided into two types according to the nature of the element that
fulfills the function of applied argument: DP or pro in doubling languages or
the dative clitic in non-doubling languages. In the following section we ask
how the Romance languages could have split into these two types.
Author's Copy
Let us first consider, for example, the introduction and expression of the
well-studied applied benefactive/malefactive argument in the Bantu-type lan-
guages. Peterson (2007) shows that the lexical source for the applicative head
is predicative (a verb or an adposition):
Seen from a diachronic perspective, the applicative object was the object of
either a transitive verb or an adposition which governed it as an object. This
syntactic status is simply an aspect of its origin which is not lost when the ap-
plicative objects governing verb or adposition grammaticalizes as an applica-
tive marker. (Peterson 2007: 125)
Mukasa Appl
Appl vP
-el-
Author's Copy
Author's Copy
In contrast, the Appl head in Latin is empty. We assume that the productive
morphological case-marking that existed in Latin made it possible for the head
to be devoid of overt morphological content.
omnibusDAT Appl
Appl vP
How do modern Romance languages fit into this picture considering that the
two structures proposed in (46) and (48) evolved from that in (57b)? We sug-
Author's Copy
Author's Copy
gest that while the lexical items which merge to form the high applicative pro-
jection in the Bantu languages are derived from a predicate and its complement,
those in Romance are derived from a probe and its goal.
The presumed first step in the diachronic change involves the loss of mor-
phological dative case marking on Latin DPs from the High Roman Empire
(Serbat 1996: 435) and the eventual replacement of dative-marked DPs by
PPs (hostibus ad hostes enemies). The consequences for the Latin ApplP
are significant. Since, as we have seen, an Appl head only selects an individual
as its argument, the Latin high ApplP can no longer serve to introduce a regular
DP. However, pronouns remain morphologically case marked and as such can
continue to act as applied arguments in Late Latin. This first step is illustrated
in (58).
Appl vP Appl vP
Assuming that the dative clitic forms in Romance are direct descendants of
the Latin dative pronouns, it is likely that the dative pronoun in the Spec,ApplP
(58b) was at some point reanalyzed as occupying the position of the Appl head,
giving rise to the generalized Romance ApplP proposed in (41a). The process
Author's Copy
Author's Copy
is illustrated in (59).17
Appl vP Appl vP
clDAT
In clitic-doubling languages, PPs were reanalyzed as dative DPs and were con-
sequently able to merge in SpecApplP as an applied argument. In non-doubling
languages, PPs remained PPs, resulting in a significant redistribution of labour
between the applicative morpheme and the element occupying Spec,ApplP.
Given the referential properties of the dative clitic, it was able to absorb the
Case and theta role features of Appl and was also able to acquire reference
from the (extra)linguistic context. The clitic therefore took on the function of
Author's Copy
Author's Copy
both the Appl marker and the applied argument. The syntactically defined se-
lectional requirements of the applicative were satisfied, as they still are, by an
expletive operator.
5. Conclusion
17. This is of course a possible, but not necessary evolution and is likely attributed to the cliti-
cization of the dative pronoun. As pointed out by an anonymous reviewer, when English lost
morphological dative case, the dative pronouns were not reanalyzed as verbal heads.
Author's Copy
Author's Copy
University of Toronto
yves.roberge@utoronto.ca
michelle.troberg@utoronto.ca
References
Author's Copy
Author's Copy
Authier, J.-Marc & Lisa Reed. 1992. Case theory, theta theory, and the distribution of French
affected datives. The Linguistic Review 9:4. 295311.
Borer, Hagit & Yosef Grodzinsky. 1986. Syntactic cliticisation and lexical cliticisation: The case
of Hebrew dative clitics. In Hagit Borer (ed.), Syntax and Semantics 19, 175217. New York:
Academic Press.
Butler, Alastair & ric Mathieu. 2004. Split-DPs, generalized EPP and visibility. In Martha McGin-
nis & Norvin Richards (eds.), MIT Working Papers in Linguistics 49. 4957. Cambridge,
Mass.: MITWPL.
Cardinaletti, Anna & Michal Stark. 1994. The typology of structural deficiency. On the three gram-
matical classes. University of Venice Working Papers in Linguistics 4(2). 41109.
Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michael &
Juan Uriagereka (eds.), Step by step: Essays in honor of Howard Lasnik, 89155. Cambridge,
Mass.: MIT Press.
Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz (ed.), Ken Hale: A life in
language, 152. Cambridge, Mass.: MIT Press.
Cuervo, Mara Cristina. 2003. Datives at large. Doctoral dissertation, Cambridge, Mass., MIT.
de Andrade Berlinck, Rosane. 1996. The Portuguese dative. In William Van Belle & Willy Van
Langendonck (eds.), The dative, vol. 1, 119151. Amsterdam: John Benjamins.
Diaconescu, Rodica. 2004. Romanian applicative constructions. Doctoral dissertation, University
of Ottawa.
Ernout, Alfred & Franois Thomas. 1964. Syntaxe latine. 2nd edn. Paris: Klincksieck.
Folli, Raffaella & Heidi Harley. 2006. Benefactives arent Goals in Italian. In Jenny Doetjes & Paz
Gonzlez (eds.), Romance Languages and Linguistic Theory 2004, 121142. Amsterdam:
John Benjamins.
Garcia, Maryellen. 1979. Spanish preverbal clitics: The affected entity construction. In James P.
Lantolf, Francine Wattman Frank & Jorge M. Guitart (eds.), Colloquium on Spanish and
Luso-Brasilian Linguistics, 4858. Washington, DC: Georgetown University Press.
Hale, Ken & Samuel J. Keyser. 1987. A view from the middle. Lexicon Project Working Papers
10. Cambridge, Mass.: Center for Cognitive Science, MIT.
Author's Copy
Author's Copy
Hale, Ken & Samuel J. Keyser. 2002. Prolegomenon to a theory of argument structure. Cambridge,
Mass: MIT Press.
Herschensohn. 1992. On the economy of Romance non-lexical datives. In Paul Hirschbhler &
Konrad Koerner (eds.), Romance languages and modern linguistic theory, 123134. Amster-
dam: John Benjamins.
Herslund, Michael. 1988. Le datif en franais. Louvain: Peeters.
Jaeggli, Osvaldo. 1982. Topics in Romance syntax. Dordrecht: Foris.
Jones, Michael Allan. 1996. Foundations of French syntax. Cambridge: Cambridge University
Press.
Jouitteau, Mlanie & Milan Rezac. 2007. The French ethical dative, 13 syntactic tests. Bucharest
Working Papers in Linguistics 9(1). 97108.
Lavency, Marius. 1997. Grammaire latine. 2nd edn. Louvain-la-Neuve: Peeters.
Leclre, Christian. 1976. Datifs syntaxiques et datif thique. In Jean-Claude Chevalier & Maurice
Gross (eds.), Mthodes en grammaire franaise, 7396. Paris: Klincksieck.
Maurel, Jean-Pierre. 1982. Datifs libres et datifs lis en latin. Cahiers de Grammaire 4. 5992.
McGinnis, Martha. 2001. Variation in the phrase structure of applicatives. Linguistic. Variation
Yearbook 1. 105146.
Miller, Philip. 1992. Clitics and constituents in Phrase Structure Grammar. London: Garland.
Peterson, David A. 2007. Applicative constructions. Oxford: Oxford University Press.
Pylkknen, Liina. 2002. Introducing arguments. Doctoral thesis, Cambridge, Mass., MIT.
Roberge, Yves & Michelle Troberg. 2007. Thematic indirect objects in French. Journal of French
Language Studies 17. 297322.
Author's Copy
Author's Copy
Roldn, Mercedes. 1972. Concerning Spanish datives and possessives. Language Sciences 21. 27
32.
Rooryck, Johan. 1988. Critres formels pour le datif non lexical en franais. Studia Neophilologica
60. 97107.
Rouveret, Alain & Jean-Roger Vernaud. 1980. Specifying reference to the subject: French
causatives and conditions on representations. Linguistic Inquiry 11. 97202.
Ruwet, Nicolas. 1982. Grammaire des insultes et autres tudes. Paris: Seuil.
Sorace, Antonella. 2000. Gradients in auxiliary selection with intransitive verbs. Language 76:4.
859890.
Serbat, Guy. 1996. Grammaire fondamentale du latin, vol. 6: Lemploi des cas en latin. Louvain-
la-Neuve: Peeters.
Sportiche, Dominique. 1996. Clitic constructions. In Johan Rooryck & Laurie Zaring (eds.),
Phrase structure and the lexicon, 213276. Dordrecht: Kluwer.
Van Hoecke, Willy. 1996. The Latin dative. In William Van Belle & Willy Van Langendoncke
(eds.), The dative, vol.1, 337. Amsterdam: John Benjamins.
Vilela, Mrio. 1986. As categorias de objecto indirecto em portugus. In Morphosyntaxe des
langues romanes. Actes du XVII e congrs international de linguistique et philologie romanes,
vol.4, 139149. Aix-en-Provence: Publications Universit de Provence.
Woodcock, E.C. 1959. A new Latin syntax. London: Methuen.
Author's Copy