Untitled - Lear

RIVISTA DI GRAMMATICA GENERATIVA 

Volume 31, anno 2006 

Direttori: 

Guglielmo Cinque (Università di Venezia) 

Luigi Rizzi (Università di Siena) 

Comitato di lettura: 

Manuela Ambar (Univ. de Lisboa) - Paola Benincà (Univ. di Padova) - 

Adriana Belletti (Univ. di Siena) - Luciana Brandi (Univ. di Firenze) - 

Luigi Burzio (The John Hopkins Univ.) - Noam Chomsky (MIT) - 

Patrizia Cordin (Univ. di Trento) - Violeta Demonte (Univ. Autonoma 

de Madrid) - Alessandra Giorgi (Univ. di Venezia) - Giorgio Graffi 

(Univ. di Verona) - Richard Kayne (New York University) - Michael 

Kenstowicz (MIT) - Giulio Lepschy (Univ. of Reading) - Giuseppe 

Longobardi (Univ. di Trieste) - Lidia Lonzi (Univ. di Milano) - Maria 

Rita Manzini (Univ. di Firenze) - Joan Mascaró (Univ. Autonoma de 

Barcelona) - Marina Nespor (Univ. di Milano-Bicocca) - Jean-Yves 

Pollock (Univ. de Marne-la-Vallée) - Annarita Puglielli (Univ. di Roma 

Tre) - Andrew Radford (Univ. of Essex) - Lorenzo Renzi (Univ. di 

Padova) - Alain Rouveret (Univ. de Paris VIII) - Leonardo Savoia 

(Univ. di Firenze) - Sergio Scalise (Univ. di Bologna) - Laura Vanelli 

(Univ. di Padova) - Jean-Roger Vergnaud (Univ. of Southern 

California) 

Redattore capo: 

Nicola Munaro (Università di Venezia) 

Redattori: 

Cristiano Chesi (Università di Siena) 

Francesco Costantini (Università di Venezia)

Rivista di Grammatica Generativa, 31 – 2006, 3-17 

SUBJECTLESS LANGUAGE: SYNTACTIC ASPECTS 

OF SAMUEL BECKETT’S “ROCKABY” * 

Valentina Bianchi 

Samuel Beckett’s Rockaby is a very intriguing text from a linguist’s point of 

view. In this paper I discuss certain aspects of its syntax for which a linguistic 

analysis proves particularly fruitful. First, I argue that the text shows some syntactic 

phenomena typical of the novelists’ free indirect style (Banfield 1982), and in 

particular, some phenomena belonging in the domain of logophoricity. Secondly, I 

show that Beckett reduces to the minimum the number of finite predicates and uses 

unembedded nonfinite predicates, which are interpreted by means of logophoric 

control. Finally, the most peculiar aspect of the syntax of Rockaby is the omission of 

pronominal subjects with finite verbs. This phenomenon is not found in the ordinary 

use of language, but only in the “abbreviated” written register of diaries (Haegeman 

1990). It is shown that the distribution of overt and omitted subjects in Beckett’s text 

is syntactically conditioned: the omitted subjects are limited to the most prominent 

position of root clauses; more importantly, all the overt subjects except for one 

impersonal subject are found in syntactic environments where subject omission is 

impossible, namely, in clauses introduced by a subordinating conjunction or a whword 

(cf. Haegeman 1990, Rizzi 2000). The linguistic analysis thus shows that 

Beckett avoids overt subjects as much as is syntactically possible. In the final 

section, an interpretation of these facts is attempted on the grounds of the linguistic 

theory of logophoricity. 

1. The syntax of the “stream of consciousness” 

The Woman’s “voice outside the body” in Rockaby seems to be the theatre 

equivalent of the novelists’ stream of consciousness: in fact, the text read by the 

recorded voice presents various syntactic phenomena typical of free indirect style, 

according to the seminal work of Banfield (1982). 

* I wish to thank Francesca Rizzi for having drawn my attention to this text and to the 

problems related to its translation into Italian. All misunderstandings are my own. 

3


Free indirect style – or, in Banfield’s terms, represented speech and thought – is 

syntactically non-embedded, contrary to indirect discourse, and it is shares certain 

features of direct discourse. The source of direct discourse is a speaker; the source of 

represented speech and thought is instead a “centre of consciousness”, a self that is 

referred to by first or third person pronominals. The discourse is a pure expression 

of this self, and not a representation of communication; accordingly, free indirect 

style necessarily lacks the second person – referring to the addressee – and all 

addressee-oriented forms, like e.g. vocatives and imperative verb forms. Furthermore, 

it is not in the present tense, which is anchored to the moment of utterance; however, 

it may contain deictic and demonstrative expressions, which are consistently shifted 

and anchored to the self’s spatial and temporal point of view (see Banfield 1982, 

chapter 3). 

Banfield’s identification of the self (subjective point of view) as a separate entity 

from the speaker has had important consequences for the development of the theory 

of logophoricity (see in particular Sells 1987): also in the ordinary use of language, 

various syntactic phenomena have been discovered which are anchored to a self 

distinct from the speaker. It is impossible here to fully summarize Banfield’s 

analysis, as well as the linguistic literature on logophoricity that followed her seminal 

work. I will only discuss some specific phenomena characteristic of free indirect 

style, which are also found in Beckett’s text. 1 

The text read by the Woman’s recorded voice is in the third person and in the past 

tense (like the free indirect style of e.g. Mrs. Dalloway); the pronominal third person 

refers to the “centre of consciousness” (here, the Woman sitting on the rocker). 

Actually, in Rockaby there is an apparent exception to the lack of second person: the 

final lines contain imperative verb forms, which should in principle be disallowed 

(cf. Banfield 1982, 113 ff.): 

4 

saying to the rocker 

rock her off 

stop her eyes 

fuck life 

stop her eyes 

rock her off 

rock her off 

These imperative forms seem to be contained in a reported speech, even though 

the object pronoun her referring to the Woman should be a first person pronoun in a 

1 In the following discussion I use the term reprise to identify the four parts of the text that are 

separated by a pause and by the word “more” uttered by the Woman sitting on the rocking 

chair.

Subjectless language: syntactic aspects of Samuel Beckett’s “Rockaby” 

reported direct discourse: I will return to this problem in the final section. Anyway, 

note that the imperative forms are addressed to the rocker: this constitutes an 

“internal addressee”, that is, an addressee internal to the Woman’s stream of 

consciousness. 

Another typical feature of free indirect discourse is the use of the anaphoric 

elements him/herself in non-reflexive contexts to refer to the centre of consciousness 

(see Zribi-Hertz 1989). The phenomenon is also found in the ordinary use of 

language (see e.g. Maling (1984) and Sigurdhsson (1990) on Icelandic; Kuno (1987), 

Sells (1987) on English; see also Cole et al. (2001) for a recent cross-linguistic 

overview). In the syntactic literature, these elements are commonly dubbed logophoric 

anaphors. 2 

The logophoric and reflexive uses of the anaphors can be easily distinguished by 

means of the following test. A logophoric anaphor can be substituted for by a 

pronoun with the same reference (indicated by coindexing): 

(1) a. shei looked for another creature like herselfi 

b. shei looked for another creature like heri. 

On the contrary, a reflexive anaphor cannot be replaced by a pronoun with the 

same reference: 

(2) a. shei said to herselfi … 

b. shei said to her k≠ … 

Note also that the logophoric use of the anaphor is subject to a syntactic constraint: 

as discussed by Reinhart & Reuland (1993), an anaphor cannot be logophoric if it is 

the direct argument of a semantic predicate (roughly, a verbal or adjectival predicate 

endowed with a subject), because here it necessarily receives a reflexive interpretation; 

a logophoric anaphor can only be a complement to a noun or a preposition. In the 

syntactic environments where the logophoric anaphor is disallowed, we find the 

pronoun her instead (e.g.: rock her off in the final lines of Rockaby quoted above). 

In Rockaby we find various instances of the logophoric use of the anaphor 

herself rferring to the centre of consciousness, e.g. in the first section: 

for another 

another like herself 

another creature like herself 

[…] 

2 This syntactic feature of free indirect style is difficult to translate in a language like Italian: 

as argued in Bianchi (1999, 115-117), the Italian anaphors se stesso/a do not allow a 

logophoric use, and the only possibility is to use the pronominal forms lui/lei stesso/a, which 

however have a strong emphatic flavour. 

5


6 

one other living soul 

going to and fro 

all eyes like herself 

and in the second section: 

for another 

at her window 

another like herself 

[…] 

one other living soul 

at her window 

gone in like herself 

Third, recall that in free indirect style deictic and demonstrative elements are 

anchored to the physical (spatial and temporal) point of view of the centre of 

consciousness. In Rockaby we find some instances of a demonstrative used in this 

way, in the third section: 

another creature there 

somewhere there 

behind the pane 

and in the fourth one: 

right down 

into the old rocker 

those arms at last 

[…] 

the rocker 


Finally, free indirect style – like direct discourse, and unlike real indirect 

discourse – allows for exclamations and incomplete sentences (Banfield 1982, 71 

ff.). Once again, in Rockaby we find some instances of these phenomena, e.g. in the 

first and second sections: 

when she said to herself 

whom else 

in the third one: 

one blind up 

no more 

never mind a face 

behind the pane


[…] 

no 

a blind up 

like hers 

a little like 

one blind up no more 

and in the fourth one: 

off her head they said 

[…] 

dead one day 

no 

night 

dead one night 

[…] 

saying to herself 

no 

done with that 

In sum, the syntax of the Voice’s text in Rockaby presents various features that 

are characteristic of free indirect style, suggesting that it is a sort of stream of 

consciousness of the protagonist. 

2. Nonfinite predicates 

A striking feature of Beckett’s text is the avoidance of finite verb forms, and the 

frequent use of nonfinite verbs. For instance, in the third section we find an 

exceptionally unembedded gerundive form: 

sitting at her window 

quiet at her window (repeated three times) 

In generative syntax, nonfinite verb forms without an expressed subject are 

assumed to have an implicit subject. The reference of this implicit subject is 

determined by the context in a number of ways (for a thorough discussion, see 

Landau 2000). When the nonfinite form is selected by a matrix predicate, the 

reference of its implicit subject (indicated as PRO) coincides with one of the 

arguments of the matrix predicate; this phenomenon is called obligatory control: 3 

3 Obligatory control also obtains when the nonfinite form is contained in a restricted set of 

adverbials (see Williams 1992, 1994). I leave side here nonfinite verbs used as restrictive 

modifiers. 

7


(3) Shei wanted PROi to stop. 

When the nonfinite form is within an adverbial clause or is an unembedded 

infinitive, the reference of its implicit subject is fixed in a less rigid way. According 

to Landau, Williams and others 4 , in this case the implicit subject is controlled by the 

“centre of consciousness”, namely, there is logophoric control. The point can be 

illustrated by means of the following example (from Williams 1992, 300): 

(4) PRO having travelled all day, the hotel was a vision indeed. 

Clearly, the person who has travelled all day is the person whose subjective point 

of view is reported – that is, the person who perceives the hotel as a vision. By 

default, in ordinary language this is the speaker, but in a narrative it may well be a 

self distinct from the speaker or the author. In the passage of Rockaby quoted above, 

the subject of the unembedded gerundive verb sitting is the Woman on the rocker, 

by logophoric control. The same holds in the following passages of the second 

section: 

8 

so in the end 

close of a long day 

went back in 

in the end went back in 


[…] 



Note that in its first occurrence, the gerundive saying is syntactically adjoined to 

the matrix clause and it is controlled by the unexpressed subject of the a finite verb 

(went back in; I will return to this unexpressed subject below); in the second 

occurrence, the finite verb is not repeated and the gerundive form, unembedded, is 

logophorically controlled. The suppression of the finite verb in the repeated lines 

also occurs in the first section, strenghtening our impression that Beckett studiously 

avoids finite verbs as much as possible. 

A third case of logophoric control is the infinitival form in the following passage 

of the fourth section, which have a purposive flavour: 

she so long all eyes 

famished eyes 

all sides 

high and low 

4 Kuno (1987); Rooryck (2001). Of couse, here I am glossing over several differences in the 

specific analyses offered by these authors.


to and fro 

at her window 

to see 

be seen 

In other cases (fourth again), a nonfinite verb form has an expressed subject; 

note however that the subject is inanimate: 

head fallen 

and the rocker rocking 

rocking away 

Besides nonfinite verbs, in various cases the main predicate is simply a directional 

particle. Consider the following passages from the third section: 

all blinds down 

never one up 

hers alone up 

[…] 

for a blind up 

one blind up 

[…] 

no 

a blind up 

[…] 

one blind up no more 

and from the fourth section: 

let down the blind and down 

right down 


Finally, in the following passage from the third section the main predicates are 

past participles and adjectives, and the subject they are predicated of (the mother) is 

left unexpressed: 5 

off her head they said 

gone off her head 

but harmless 

no harm in her 

5 This implicit subject cannot be easily interpreted as an instance of control, but it resembles 

more closely the phenomenon of “topic drop” discussed by Haegeman (1990), as the mother 

is the topic of the whole passage. 

9


10 

dead one day 

no 

night 

dead one night 

3. The syntax of the subject 

The syntactic devices discussed above give the impression of a “verb-less” and 

“subject-less” language. In fact, in the ordinary use of language every finite verb in 

English has an expressed subject, even when the latter has no actual reference, as in 

the case of weather verbs (5a) and of extraposed subject clauses (5b): 

(5) a. It is raining. 

b. It is clear that you made a mistake. 

c. She is speaking. 

In generative syntax, Modern English is dubbed an “absolutely non-pro-drop 

language”, since it cannot drop a pronominal subject of a finite verb, not even an 

expletive one. Other languages like e.g. Russian allow for partial drop of expletive 

subjects, and fully pro-drop languages like Italian can drop any pronominal subject; 

for a thorough discussion, see among othjìers Rizzi (1986) and Jaeggli & Safir 

(1989). 

At this point we can discuss the most striking aspect of the syntax of Rockaby. 

Besides avoiding finite verbs, Beckett strongly deviates from the ordinary use of 

language in that he often omits the subject of finite verbs. (Note in passing that this 

exceptional character of the syntax of Rockaby cannot be properly rendered when 

translating the text in a pro-drop language like Italian, which has free omission of 

pronominal subjects, as mentioned above.) If I counted correctly, out of 62 finite 

verbs 41 have an expressed subject, and 21 an omitted subject. 6 I report here all the 

unambiguous cases of subject omission with a finite verb: 

6 In the case of two coordinated verbs, the expressed subject precedes the first one, but also 

the second one was obviously counted as having an expressed subject. In some cases, 

however, it is unclear whether a finite verb is really coordinated to the preceding one, e.g. in 

the following passage (repeated several times): 

till in the end 

the day came 

in the end came 

close of a long day… 

I counted the second occurrence of came as endowed with an expressed subject like the first 

one, but actually this is not a straightforward case of coordination. If we consider the second 

occurrence of came subjectless, the number of expressed subjects is 37 out of 62 finite verbs.

second section: 


so in the end 


went back in 


[…] 

so in the end 


in the end went and sat 

went back in and sat 

at her window 

let up the blind and sat 

quiet at her window 

[…] 

fourth section: 

so in the end 


went down 

in the end went down 

down the steep stair 


right down 


[…] 

so in the end 


went down 

in the end went down 



right down 



and rocked 

rocked 

with closed eyes 

[…] 

was her own other 

11


12 

own other living soul 

so in the end 


went down 



right down 


and rocked 

rocked 

saying to herself… 

Given the syntactic characterization of English as a non-pro-drop language, these 

sentences should be utterly ungrammatical. This is not quite correct, however. As 

discussed by Haegeman (1990), Hageman & Ihsane (2001) and Rizzi (2000), in the 

written register of diaries subject drop with a finite verb is allowed. Consider the 

following example, quoted by Haegeman (1990): 

(6) A very sensible day yesterday. ___ saw noone. ___ took the bus to 

Southwark Bridge. 

___ walked along Thames Street…. 

(Virginia Woolf, Diary, vol.5, 1936-41, pp. 203-4) 

As discussed by Haegeman and Rizzi, this type of subject omission with a finite 

verb has structural properties very different from those of full pro-drop in a 

language like Italian: the omitted subject is limited to root clauses, and it must occur 

in the structurally highest position of the clause. Thus, subject omission is impossible 

in a finite clause that is introduced by a wh-phrase or by a subordinating conjunction. 

According to Rizzi’s analysis, these syntactic constraints are due to the nature of 

the understood subject. By hypothesis, subject omission involves an unpronounced 

pronominal category whose content must be syntactically recoverable (in technical 

terms, it must be identified). In a full pro-drop language like Italian, the reference of 

the unpronounced subject can be recovered by means of the “rich” inflection of the 

finite verb, which specifies the values of the person and number features. In Modern 

English, instead, the verbal inflection isn’t “rich” enough to identify a null pronoun. 

Therefore, the unpronounced subject of the written register of diaries is not 

syntactically identified within the clause, but its reference is recovered by its being 

connected to the surrounding discourse. 7 According to Rizzi, this type of discourse 

7 Furthermore, this unpronounced subject differs from that of Italian in that it is not a 

pronominal category but a “null constant”. I will leave aside these technical details here, as 

they are not relevant to my purposes.


identification is only possible when syntactic identification is impossible, namely, 

when the unpronounced subject is in the structurally highest position in the clause, 

so that there isn’t any more prominent category that can in principle act as an 

identifier. This is why subject omission in the written registers of English is limited 

to the highest position of root clauses. 8 

I have dwelled into this somewhat technical discussion because it can help us 

understand the distribution of omitted and explicit subjects in Beckett’s text. In all 

the 21 examples reported above, the finite verb with omitted subject is in a root 

clause; furthermore, the finite verb is only preceded by a coordinating conjunction 

or by the adverbial expression (so) in the end. I will tentatively hypothesize that the 

latter is syntactically peripheral, so that it does not block discourse identification of 

the unpronounced subject. 

The syntactic analysis is even more significant for the distribution of the 41 overt 

subjects of finite verbs in Beckett’s text. Apart from one impersonal pronominal 

subject in a parenthetical clause in the fourth section (off her head they said), 9 all 

overt subjects of finite verbs, both pronominal and nonpronominal, are in a 

syntactically nonprominent position, where discourse identification would be 

impossible: namely, they are preceded by a subordinating conjunction or by a whword. 

I report a few examples below; the subjects and the preceding subordinating 

conjunctions are underlined. 

8 More specifically, on Rizzi's analysis discourse identification is made possible by the fact 

that the topmost part of the syntactic structure of the clause can fail to be realized, leaving the 

subject position as the topmost position in the clause. The same phenomenon is found in a 

specific stage of the acquisition of English, around the age of two years. According to Rizzi, 

the marked possibility of “truncating” the syntactic structure of the clause in child language 

also allows for other phenomena, like the use of nonfinite verbs and nominal predicates in 

matrix clauses. As noted above, this type of verb-less predication is also found in Beckett’s 

text, confirming the clustering of syntactic properties proposed by Rizzi. The correlation is 

linguistically interesting, but I don’t think that the language of Rockaby could possibly be 

considered an imitation of child language (though such an hypothesis may at first sight seem 

appealing, given the regressive character of the Woman’s stream of consciousness). See Rizzi 

(2005) and Haegeman (2007) for a different analysis of the phenomenon in terms of phonetic 

deletion of the highest position of the root clause. 

9 Following Liliane Haegeman (personal communication, Venice, November 21st 2007), this 

exception may be explained away if we think of the phrase “off her head” as a quotation: this 

may either occupy a high structural position in a root clause, or license a phonologically null 

operator in the periphery of a parenthetical clause (see e.g. Collins & Branigan 1997, 10-13); 

in either case, it follows that the subject of the latter does not meet the structural condition for 

subject drop. 

13


First section: 

14 

till in the end 

the day came 

in the end came 


when she said 

to herself 

whom else 

time she stopped 

time she stopped… 

Second section: 

so in the end 


went back in 



whom else 



going to and fro 

time she went and sat 

at her window… 

Fourth section: 


right down 


mother rocker 

where mother sat… 

If we adopt Haegeman’s and Rizzi’s analysis of this exceptional subject omission 

with finite verbs, we come to the conclusion that (apart for maybe one impersonal 

subject), all the expressed subjects in Beckett’s text are syntactically conditioned, 

that is, they occur in a context where an unpronounced subject is impossible. 

4. Some speculations 

In the preceding discussion, I have tried to show that a careful linguistic analysis 

may uncover certain significant aspects of Beckett’s text in a non-impressionistic 

way. The next step is to offer an interpretation of these observations, and this requires


a literary, not an exclusively linguistic competence. Nevertheless, I wish to offer 

some speculations, starting once again from my linguistic background. 

We have noted that the most prominent aspect of the syntax of Rockaby is the 

avoidance of finite verbs and of expressed subjects. These two categories are 

linguistically related: cross-linguistically, expressed subjects are always possible 

(and even required, in non-pro-drop languages) with finite verbs, whereas nonfinite 

verbs usually have implicit subjects. In recent work (Bianchi 2001, 2003, 2006) I 

have proposed a hypothesis to account for the correlation between finiteness and 

overt subjects. The basic insight is that the subject of a finite verb can have any 

value of the person feature, and finite verbs, contrary to nonfinite ones, are endowed 

with person agreement. 10 The person feature is intrinsically deictic because, as 

Jakobson (1971, 134) synthetically puts it, “person characterizes the participants of 

the narrated event with reference to the participants of the speech event”: first 

person refers to the speaker, second person refers to the addressee, and third person 

to anyone and anything else (leaving aside courtesy forms). Holmberg & Platzack 

(1995) have observed that the finite verb has the property of expressing another 

deictic feature, absolute tense, which is interpreted with respect to the time of the 

utterance. Putting these observations together, I have proposed that the finite verb 

syntactically encodes the speech event, which functions as the centre of deixis: the 

tense feature is interpreted w.r.t. the time of the speech event, and the person feature 

is interpreted w.r.t. the participants of the speech event (speaker and addressee). 11 

Going back to Rockaby, let us reconsider the avoidance of finite verbs and the 

marked omission of subjects. These features might be interpreted as the hallmarks of 

an abbreviated register: according to the well known ‘principle of lessness’, Beckett 

strives to obtain a language as synthetic as possible. However, given the linguistic 

considerations offered above, we can attempt a different interpretation: by using 

nonfinite verbs with logophorically controlled implicit subjects and finite verbs with 

discourse-identified implicit subjects, Beckett reduces to the minimum the expression 

of the person feature, and in particular, the opposition between referentially distinct 

10 There are some exceptional cases of nonfinite verbs with inflection for person, like e.g. the 

inflected infinitive of European Portuguese (see Raposo 1987); these are however crosslinguistically 

rare. See Bianchi (2001) for some discussion. 

11 As noted above, Banfield (1982) argues that represented speech and thought is a form of 

pure expression, not of communication: there is no real speech event, no utterance. The centre 

of deixis is the self’s temporal and physical perspective; as for the person feature, the self can 

be referred to by first person (as in Molly Bloom’s stream of consciousness) or third person 

(as in Mrs. Dalloway), but second person is not licensed. In my recent work, I have proposed 

that the speech event is just one instance of a broader notion of logophoric centre: any speech 

or mental event (even a reported one) can function as a centre of deixis, but only the “external” 

utterance event can license a full-fledged person feature. 

15


values of the person feature. Note that in uttering of the words “time she stopped” or 

“rock her off”, the “centre of consciousness” is apparently using a third person 

pronoun rather than a first person one to refer to herself, as if the use of first person 

were linguistically impossible (compare the Mouth’s “vehement refusal to relinquish 

third person” in Not I). This corroborates the impression that the language of the 

disembodied Voice is virtually person-less. 

We may interpret this collapse of the person category by referring to Emile 

Benveniste’s view of the corrélation de personnalité. According to Benveniste 

(1966, 260), the dialogic dimension is constitutive of the linguistic category of 

person: 

16 

C’est cette condition de dialogue qui est constitutive de la personne, car elle 

implique en réciprocité que je deviens tu dans l’allocution de celui qui à son 

tour se désigne par je. […] La polarité des personnes, telle est dans le langage 

la condition fondamentale… 

An obsessive refrain in the Woman’s stream of consciousness is the complete 

absence of another creature like herself, of any potential interlocutor (when she said 

to herself / whom else…), until in the end the mother-rocker is turned into an 

interlocutor (the addressee of the final imperative verbs). In the gradual fade out of 

the Woman’s consciousness, the complete absence of an interlocutor makes the 

deictic category of person virtually vacuous. 

References 

Banfield, A. 1982. Unspeakable sentences.narration and representation in the language of 

fiction. Boston, Routledge & Kegan Paul. 

Bianchi, V. 1999. Consequences of Antisymmetry: Headed Relative Clauses. Berlin, Mouton de 

Gruyter. 

Bianchi, V. 2001. On person agreement. Ms., Scuola Normale Superiore di Pisa. 

Bianchi, V. 2003. On finiteness as logophoric anchoring. In: J. Guéron & L. Tasmovski, eds., 

Proceedings of the Paris Colloquium on Time and Point of View. 

Bianchi, V. 2006. On the syntax of personal arguments. Lingua 116, 2023-2067. 

Cole, P. Hermon, G. & Huang, eds. 2001. Long Distance Reflexives. Syntax and Semantics 33. 

New York, Academic Press. 

Collins, C. & Branigan, P. 1997. Quotative inversion. Natural Language and Linguistic Theory 

15, 1-41. 

Haegeman, L. 1990. Non overt subjects in diary contexts. In: J. Mascaro & M. Nespor, eds., 

Grammar in Progress. Dordrecht, Foris. 

Haegeman, L. & Ihsane, T. 2001. Adult null subjects in the non-pro-drop languages: two diary 

dialects. Language Acquisition 9, 329-346.


Haegeman, L. 2007. Subject omission in present-day written English. Paper presented at the 

Workshop Linguistic Approaches to Narrative Text (University of venice, 20-21 

November 2007). 

Holmberg, A. & C. Platzack. 1995. The Role of Inflection in Scandinavian Syntax. Ofxord, 

Oxford University Press. 

Jaeggli, O. & K. Safir, eds. 1989. The Null Subject Parameter. Dordrecht, Kluwer. 

Kuno, S. 1987. Functional Syntax. Chicago/London, The University of Chicago Press. 

Landau, I. 2000. Elements of Control. Dordrecht, Kluwer. 

Maling, J. 1984. Non-Clause Bounded Reflexives in Icelandic. Linguistics and Philosophy 7, 211- 

241. 

Raposo, E. 1987. Case Theory and Infl-to-Comp: The inflected infinitive in European Portuguese. 

Linguistic Inquiry 18, 85-109. 

Reinhart, T. & E. Reuland. 1993. Reflexivity. Linguistic Inquiry 24, 657-720. 

Rizzi, L. 1986. Null objects in Italian and the theory of pro. Linguistic Inquiry 17, 501-557. 

Rizzi, L. 2000. Early null subjects and root null subjects. In: L. Rizzi & M.A. Friedemann, eds., 

The Acquisition of Syntax. London, Longman. 

Rizzi, L. 2005. Phase Theory and the Privilege of the Root. Ms., University of Siena. 

Rooryck, J. 2000. Configurations of sentential complementation. Perspectives from Romance 

languages. London-New York, Routledge. 

Sells, P. 1987. Aspects of logophoricity. Linguistic Inquiry 18, 445-479. 

Sigurdhsson, H. A. 1990. Long-distance reflexives and moods in Icelandic. In: Modern Icelandic 

Syntax, ed. J. Maling & A. Zaenen, 309-346. New York, Academic Press (Syntax and 

Semantics 24). 

Williams, E. 1992. Adjunct control. Control and Grammar, ed. R. Larson et al., 297-322. 

Dordrecht, Kluwer. 

Williams, E. 1994. Thematic Structure in Syntax. Cambridge, MA, The MIT Press. 

Zribi-Hertz, A. 1989. Anaphor binding and narrative point-of-view:English reflexive pronouns in 

sentence and discourse. Language 65, 695-727. 

17

0. Introduction 


FOCUS IN THE IP: 

THE PARTICLE MA IN FLORENTINE * 

Jacopo Garzonio 

Many works on the cartography of CP and IP (among others Rizzi 1997, Belletti 

2001, Belletti 2004, Benincà and Poletto 2004) have shown that in standard Italian 

there are two types of Focus encoded in the syntax, a structurally high contrastive 

Focus and a structurally low informational Focus. Both types of Focus are 

associated with a dedicated projection: the high Focus is associated with a FocP in 

CP, between the Topic layer and FinP, the low Focus is associated with a FocP 

between IP and vP. In both cases a constituent is moved to the specifier of FocP in 

order to check the [+Focus] feature. 

In Florentine, a variety of Central Italy very similar to standard Italian, both 

Focus types are present, the high contrastive Focus (1a), signaled like in Italian by a 

special intonation indicated here in capitals, and the low informational Focus (1b): 

(1) a. [A MARIO]focus ho dato i libro (non a Giorgio). 1 

to Mario have given the book not to Giorgio 

‘TO MARIO I gave the book (not to Giorgio).’ 

b. l’ho dato [a Mario]focus, i libro. 

it have given to Mario the book 

‘I gave the book to Mario.’ 

* The content of this article was presented at the XIII Giornata di Dialettologia (Padova, 21 

June 2007) and at a postgraduate seminar at the University of Venice (July 2007). I thank 

both the audiences for the helpful discussion. Moreover, for further comments and advice, I 

thank Paola Benincà, Andrea Cattaneo, Guglielmo Cinque, Federico Damonte, Nicoletta 

Penello and Cecilia Poletto. All errors are obviously my own. 

1 In this paper, I do not use the traditional Florentine orthography according to which forms 

that diverge from standard Italian should be written with an apostrophe to indicate the missing 

consonants (Florentine i' ‘the’ vs. Italian il; Florentine 'un ‘not’ vs. Italian non). 

19


However, Florentine also displays a third type of Focus, which is contrastive and 

is marked by the particle ma: 

(2) l’ho dato ma [a Mario]focus, i libro (non a Giorgio). 

it have given ma to Mario the book not to Giorgio 

‘I gave the book to Mario (not to Giorgio).’ 

In this paper I will describe the syntax and the interpretation of this peculiar 

Focus type. The particle ma, homophonous with the conjunction ma ‘but’, precedes 

the focused item, like other focus markers like solo ‘only’, anche ‘too’ and perfino 

‘even’. However, it has a more restrictive syntactic behaviour (for instance, it must 

follow the inflected verb), hence it is possible to assume that it has a precise and 

unique position in the phrase structure. More precisely, I will propose that ma is the 

head of a projection associated with a [+Exhaustive Identification] feature, from 

which derives the contrastive interpretation of the focused item. 

The paper has the following structure: in section 1 I will present some data about 

the syntax of the ma Focus and the position of ma in the clause; in section 2 I will 

take into consideration two specific properties of the syntax of ma, that suggest that 

the focused element is not in its basic position; in section 3 I will briefly describe 

some cases of focusing of CPs and other large constituents; in section 4 I will deal 

with the semantic side of the ma Focus, and show that an operation of exhaustive 

identification is involved; section 5 contains some conclusive considerations. 

1. The position of maP in the clause structure 

The particle ma is a head, since it cannot be modified nor focused itself. This is 

shown in the following examples. In (3a) ma is modified by proprio ‘just, really’, 

while in (3b) it is emphasized by the intonation. Both examples are ungrammatical: 

(3) a. *l’ho dato proprio ma [a Mario]focus, i libro. 

it have given just ma to Mario the book 

b. *l’ho dato [MA]focus a Mario, i libro. 

it have given ma to Mario the book 

I will call the projection headed by this particle maP. It is not possible to insert 

other elements like DPs, PPs or adverbs between ma and the focused item: 

(4) a. *ho dato ma i libro [a Mario]focus (non a Giorgio). 

have given ma the book to Mario not to Giorgio 


b. *ho dato ma a Mario [i libro]focus (non i giornale). 

have given ma to Mario the book not the newspaper 

‘I gave to Mario the book (not the newspaper).’ 

20

Focus in IP: the Particle ma in Florentine 

c. *l’ho dato ma subito [a Mario]focus, i libro (non a Giorgio). 

it have given ma immediately to Mario the book (not to Giorgio) 

‘I gave immediately the book to Mario (not to Giorgio).’ 

These data suggest that there is an adjacency constraint on the formation of this 

Focus structure. From this point of view ma is very similar to quantificational 

elements like solo ‘only’, anche ‘also, too’, perfino ‘even’. Nevertheless, it has 

different syntactic properties. 

First of all, ma can appear only in post-verbal position. It can focus-mark 

subjects, objects or adjunct PPs, but the complex ma+[focused item] cannot precede 

the inflected verb: 

(5) a. Mario vende ma [mobili]focus (non macchine). 

Mario sells ma furniture not cars 

‘Mario sells furniture (not cars).’ 

b. la ruppe ma [i tu fratello]focus, sta finestra (non io). 

it broke ma the your brother this window not I 

‘It was your brother who broke the window (not me).’ 

c. c’andaron ma [le su amiche]focus (non lei). 

there went ma the her friends not she 

‘It was her friends who went there (not her).’ 

d. ci vado ma [con Mario]focus (non con Giorgio). 

there go ma with Mario not with Giorgio 

‘I will go there with Mario (not with Giorgio).’ 

a’. *ma [mobili]focus, Mario vende. 

b’. *ma [i tu fratello]focus ruppe sta finestra. 

c’. *ma [le su amiche]focus c’andarono. 

d’. *ma [con Mario]focus, ci vado. 

Elements like solo ‘only’ or anche ‘too’ do not display similar restrictions: 

(6) a. solo i tu fratello ruppe sta finestra. 

‘Only your brother broke the window.’ 

b. anche le su amiche c’andarono. 

‘Also her friends went there.’ 

Secondly, ma can focus-mark PPs and adjectives only if they are verb complements: 

(7) a. quella casa gl’è ma [di legno]focus (non di mattoni). 

that house clit. is ma of wood not of bricks 

‘That house is made of wood (not of bricks).’ 

21


22 

b. Mario gl’è ma [paziente]focus (non nervoso). 

Mario clit. is ma tolerant not unquiet 

‘Mario is tolerant (not unquiet).’ 

a’. *ho visto una casa ma [di legno]focus 

have seen a house ma of wood 

‘I have seen a wooden house.’ 

b’. *Mario gl’è una persona ma [paziente]focus 

Mario clit. is a person ma tolerant 

‘Mario is a tolerant person.’ 

In similar contexts it is possible to use solo or anche that have scope on a PP or 

an adjective inside a DP: 

(8) a. ho visto una casa solo di legno. 

b. Mario gl’è una persona anche paziente. 

Finally, ma cannot focus-mark a DP inside a PP, while this is marginally possible 

with elements like solo 2 : 

(9) a. *quello gl’era un paese di ma [vecchi]focus 

that clit. was a village of ma old 

b. quello gl’era ma [un paese di vecchi]focus 

a’. quello gl’era un paese di solo vecchi. 

that clit. was a village of only old 

‘That one was a village inhabited only by old people.’ 

All these facts are evidence that ma is a clausal element and that maP has a 

unique position in the clause structure. In the following sections I will analyse the 

relationship between maP and the two FocPs assumed by the cartographic framework. 

1.1 MaP is not in the left periphery 

As I have shown in the preceding section, the complex ma+[focused item] 

cannot appear before the inflected verb. This fact suggests that maP is not in the left 

periphery of the clause and, thus, has nothing to do with FocP in CP. However, 

additional evidence is necessary, since we cannot exclude that the focused item is 

moved to [Spec, Foc] in CP and the rest of the sentence undergoes remnant movement 

2 For similar problems regarding the position of only in English, see Kayne (2000).


to a higher position, for instance the specifier of maP. This explanation is proposed 

by Belletti (2004) for post-verbal contrastive Focus in Italian. 

I present two facts as evidence that maP is lower than CP. Firstly, consider that 

Florentine allows the use of the default subject clitic gl’ and the lack of agreement 

on the verb with a post-verbal plural subject. While it is impossible to have this 

configuration with a pre-verbal focused plural subject, as in (10b) 3 , this option is 

perfectly acceptable with a post-verbal subject focused by ma, as in (10d): 

(10) a. [LE TU SORELLE]focus le son venute (non le mie). 

the your sisters clit.FEM.PL are come.FEM.PL not the mine 

‘YOUR SISTERS have come (not mine).’ 

b. *[LE TU SORELLE]focus gl’è venuto. 

the your sisters clit.DEFAULT is come.MASC.SG 

c. Le son venute ma [le tu sorelle]focus 

d. Gl’è venuto ma [le tu sorelle]focus 

Secondly, if the focused element is a Negative Polarity Item like nulla ‘nothing’, 

the pre-verbal negation is absent with a left periphery Focus, and it is obligatory 

with a ma-Focus; in other words, a Negative Polarity Item focused by ma cannot be 

the only negative element of a clause. Hence, it has not moved through a position ccomanding 

the finite verb (Laka, 1990; Zanuttini, 1991): 

(11) a. [NULLA]focus hanno fatto. 

nothing have done 

b. *(un) hanno fatto ma [nulla]focus 

NEG have done ma nothing 

‘NOTHING they have done.’ 

On the basis of these facts, it can be assumed that maP is not in the left periphery 

of the clause. This is a striking result, as it implies that there is a dedicated projection 

for contrastive Focus in IP. 

1.2 The position of maP in IP 

At this point, the question that has to be addressed is the exact position of maP in 

IP. If we check the relative order of the complex ma+[focused item] and the 

aspectual adverbs, it clearly emerges that ma must follow all these adverbs. In (12) it 

3 Note that this data contrast with those presented by Brandi and Cordin (1981). However, the 

variety I’m taking into consideration here is Urban Florentine, spoken in the city of Florence, 

while Brandi and Cordin have analysed the conservative Rural Florentine of Vaiano, a 

locality 40 kms from Florence. 

23


is shown that ma+[focused item] must follow più ‘any more’, ancora ‘still’, sempre 

‘always’ and digià ‘already’: 

(12) a. Mario un mangia più ma [formaggio]focus (non carne). 

Mario NEG eats any-more ma cheese not meat 

‘Mario does not eat cheese any more (not meat).’ 

b. Mario gl’è ancora ma [in piazza]focus (non a i bar). 

Mario clit. is still ma in square not at the bar 

‘Mario is still in the town square (not in the bar).’ 

c. Mario mangia sempre ma [carne]focus (non formaggio). 

Mario eats always ma meat not cheese 

‘Mario always eats meat (not cheese).’ 

d. Mario gl’è digià ma [in piazza]focus (non a i bar). 

Mario clit. is already ma in square not at the bar 

‘Mario is already in the town square (not in the bar).’ 

24 

a’. *Mario un mangia ma [formaggio]focus più. 

b’. *Mario gl’è ma [in piazza]focus ancora. 

c’. *Mario mangia ma [carne]focus sempre. 

d’. *Mario gl’è ma [in piazza]focus digià. 

Usually ma also follows bene ‘well’, but it should be pointed out that the adverb 

follows ma when is the focused item. This configuration is very frequent when bene 

is argumental, for instance with verbs like sentirsi ‘feel’ or comportarsi ‘behave’. 

On the contrary, if the focused item is different, bene must precede ma. In (13) it is 

shown that with a post-verbal subject focused by ma, only the order bene-ma yields 

grammaticality: 

(13) a. capisce bene ma [Giorgio]focus (non Mario). 

understands well ma Giorgio not Mario 

‘It is Giorgio who understands well (not Mario).’ 

b. *capisce ma [Giorgio]focus bene. 

The data in (12) and (13) are evidence that the position of maP in the IP is very 

low. Ma must follow aspectual adverbs, and even adverbs like bene, which occupy a 

very low position in the clause structure, according to the hierarchy of Cinque 

(1999). Since maP is in a low position in the IP, it is adjacent to the low FocP 

proposed by Belletti (2001; 2004). On the basis of the observation of the data 

presented so far, it is possible to formulate a first hypothesis about the ma Focus: ma 

lexicalises the head of a projection associated with contrastive interpretation; in 

order to receive this contrastive interpretation, an element (DP, PP, adverb) has to


raise to a position adjacent to ma, say the specifier of the low Focus projection. This 

first hypothesis is exemplified by the structure in (14): 

(14) IP 

2 

maP 

2 

ma FocP 

2 

XP Foc’ 

2 

Foc° vP 

This first hypothesis implies that the contrastive interpretation encoded by ma 

requires also the checking of the [Focus] feature. In other words, “contrast” is a 

property given by ma to an element already marked as Focus. 

Before adopting this hypothesis, some other facts must be considered. 

2. More about the syntax of ma 

In this section I will describe more in detail two properties of the syntax of ma 

that can be provided as evidence that the focused item is moved from its basic 

position, i.e. the focused item is not in VP or vP. The first one concerns 

Marginalization (“Emarginazione” in the sense of Antinucci and Cinque, 1977) and 

clitic Right Dislocation, the second one the focusing of the entire VP. 

2.1 Marginalization and Right Dislocation 

So far I have examined mainly sentences with only one internal argument. 

Consider now the case where there are two internal arguments, for instance a direct 

object and a dative PP. In this case, it is possible of course to focus by the particle 

ma one of them. However, the non-focused argument must appear as a clitic Right 

Dislocation. This is shown in (15): 

(15) a. l’ho dato ma [a Mario]focus, i libro (non a Giorgio). 

it have given ma to Mario the book not to Giorgio 


b. gl’ho dato ma [i libro]focus, a Mario (non i giornale). 

to-him have given ma the book to Mario not the newspaper 

‘I gave the book to Mario (not the newspaper).’ 

25


On the other hand, Marginalization is not possible in the same contexts. In other 

words, the resumptive clitic of the sentences in (15) is obligatory: 

(16) a. *ho dato ma [a Mario]focus, i libro (non a Giorgio). 

b. *ho dato ma [il libro]focus, a Mario (non i giornale). 

If there are two or more non focused arguments, they all must be right dislocated. 

Following Cardinaletti (2001) I assume that clitic Right Dislocation and 

Marginalization are structurally different. In the case of Marginalization it is 

possible to demonstrate that what follows the focused element remains in its basic 

position. In the case of Right Dislocation, what follows the focused element is 

moved by some type of process outside its basic position. The main argument in 

favor of this analysis is that after a focused postverbal subject the order of objects is 

free if they are right dislocated, but it is the same as the unmarked order of 

arguments in the case of Marginalization. I report in (17) the relevant Italian 

examples in Cardinaletti (2001): 

(17) a. ce l’ha nascosto [il bambino]focus, il libro, sotto il letto. 

there it has hidden the child the book under the bed 

‘It is the child who has hidden the book under the bed.’ 

b. ce l’ha nascosto [il bambino]focus, sotto il letto, il libro. 

c. ha nascosto [il bambino]focus, il libro, sotto il letto. 

d. *ha nascosto [il bambino]focus, sotto il letto, il libro. 

In Florentine clitic Right Dislocation is obligatory in direct wh questions and 

with post-verbal informational focuses. This is shown in (18) and in (19) respectively: 

(18) a. chi *(l’)ha nascosto, i libro? 

who it has hidden the book 

‘Who has hidden the book?’ 

b. icché tu *(gl’)hai dato, a Mario? 

what you to-him have given to Mario 

‘What have you given to Mario?’ 

(19) a. *(l’)ha nascosto [i bambino]focus, i libro. 

it has hidden the child the book 

‘It is the child who has hidden the book.’ 

b. *(gl’)ho dato [un libro]focus, a Mario. 

to-him have given a book to Mario 

‘I have given a book to Mario.’ 

26


Thus, ma Focus distributes like direct wh questions and Focuses in the IP: one 

element is moved outside its basic position to check some feature ([+Interrogative] 

or [+Focus]), while the other arguments must be right dislocated. This fact supports 

the idea that the focused element in a ma Focus structure is moved outside of VP or 

vP to a position lower than maP, which is very likely the specifier of FocP. 

Why Right Dislocation is obligatory in these structures is not clear. As for what 

concerns the ma Focus, the relevant fact is that it is similar to other constructions 

with feature driven movement. We may assume that obligatory Right Dislocation is 

related to the raising of the verb, which I will consider remnant VP (or vP) 

movement (in the spirit of Hinterhölzl’s, 1997, proposal). Thus, the derivation of a 

ma Focus structure is the following: firstly, one of the arguments is moved outside 

VP to [Spec, Foc]; then all other arguments must be right dislocated; finally the 

“evacuated” VP is raised past maP. This proposal is exemplified in (20): 

(20) [IP glx’ho [VP dato y x]z [maP ma [SpecFoc [i libro]y [VP z ]]]] [a Mario]x (=15b.) 

The DP i libro ‘the book’ is moved to the specifier of the Focus projection, 

enters in a ma Focus configuration and receives contrastive interpretation; the dative 

PP a Mario ‘to Mario’ is right dislocated and the evacuated VP is raised higher than 

maP. 

2.2 Focusing of the VP 

When the verb occurs in an analytic form, the particle ma can appear between 

the auxiliary and the lexical verb. This is shown in (21); in (21a) ma is located 

between the auxiliary avere ‘to have’ and the past participle of the lexical verb; in 

(21b) it is located between the durative auxiliary stare ‘to stay’ and the gerunde of 

the lexical verb: 

(21) a. ho ma letto un libro. 

have ma read a book 

‘I have read a book’ 

b. sto ma leggendo un libro. 

stay ma reading a book 

‘I am reading a book.’ 

These sentences are potential counterexamples for the hypothesis I have 

proposed at the end of section 1, and more in general they challenge the idea that ma 

has a precise and unique position in the clause structure. However, these cases of ma 

higher than the lexical verb (but not higher that the inflected verb) present further 

properties that must be taken into consideration. 

27


Firstly, when ma is between the auxiliary and the lexical verb, and an adverb is 

present, ma must precede the adverb, while the opposite order yields ungrammaticality. 

This is shown in (22): 

(22) a. Mario ha ma sempre mangiato carne. 

Mario has ma always eaten meat 

‘Mario always has eaten meat.’ 

b. Mario sta ma digià andando in piazza. 

Mario stays ma already going in square 

‘Mario is already going to the town square.’ 

28 

a’. *Mario ha sempre ma mangiato carne. 

b’. *Mario sta digià ma andando in piazza. 

Note that this happens also with adverbs that occupy a very high position in 

Cinque’s (1999) hierarchy, like probabilmente ‘probably’. These adverbs can appear 

before the auxiliary or after ma, but not between them 4 : 

(23) a. ?Mario probabilmente ha ma comprato un libro. 

Mario probably has ma bought a book 

‘Probably, Mario has bought a book.’ 

b. ?Mario ha ma probabilmente comprato un libro. 

c. *Mario ha probabilmente ma comprato un libro. 

Secondly, if there are more arguments, with ma between the auxiliary and the 

lexical verb, they must appear in the unmarked order. Direct objects must precede 

datives and adjuncts. Other orders of arguments result in ungrammaticality of the 

sentence. Crucially, in this case no resumptive clitic is present: 

(24) a. Mario (*gl’)ha ma regalato un libro a Giorgio. 

Mario to-him has ma given a book to Giorgio 

‘Mario has given as a gift a book to Giorgio.’ 

b. Mario (*ci) sta ma portando i bambino da i medico. 

4 The sentences (23a) and (23b) are marginal because the contrastive reading conflicts with 

the semantics of probabilmente ‘probably’. However, they are perfectly acceptable when used 

to contradict a previous statement containing probabilmente, like: 

(i) A: Mario ha probabilmente comprato un profumo. 

‘Probably Mario has bought a perfume.’ 

B: No. Mario ha ma probabilmente comprato un libro. 

‘No. Probably Mario has bought a book.’ 

See further for more details about the interpretation of ma Focus with ma between the 

auxiliary and the lexical verb.


Mario there stays ma taking the child to the doctor 

‘Mario is taking the child to the doctor.’ 

c. i bambino (*ci) ha ma nascosto i libro sotto a i letto. 

the child there has ma hidden the book under to the bed 

‘The child has hidden the book under the bed.’ 

a’. *Mario ha ma regalato a Giorgio un libro. 

b’. *Mario sta ma portando da i medico i bambino. 

c’. *i bambino ha ma nascosto sotto a i letto i libro. 

Thirdly, these sentences are appropriate in different contexts. More precisely, a 

sentence like (24a) can be used to contradict all the following assertions: 

(25) a. Mario ha regalato un libro [a suo fratello]. 

‘Mario has given as a gift a book to his brother.’ 

b. Mario ha regalato [un profumo a sua moglie]. 

‘Mario has given as a gift a perfume to his wife.’ 

c. Mario ha [comprato i giornale]. 

‘Mario has bought the newspaper.’ 

Answer: No. Mario ha ma [regalato [un libro [a Giorgio]]]. 

‘No. Mario has given as a gift a book to Giorgio.’ 

It seems that in similar ma Focus structures, it is possible to interpret as the 

contrasted item the most embedded constituent, or larger constituents of the predicate, 

up to the whole VP. However, it is not possible to interpret as the contrasted item a 

constituent of the VP excluding the most embedded ones: 

(26) a. Mario ha regalato [un disco] a Giorgio. 

‘Mario has given as as gift a record to Giorgio.’ 

b. Mario ha [prestato] un libro a Giorgio. 

‘Mario has lent a book to Giorgio. 

Answer: #No. Mario ha ma [regalato [un libro [a Giorgio]]]. 

I will take all the data presented in this section as evidence that, in sentences 

where ma is inserted between an auxiliary and the lexical verb, the whole VP is 

moved to the specifier of FocP. 

Ma, in these cases, must precede any adverb. I assume that these adverbs are 

generated in a lower position, as heads taking the whole VP as complement, and 

then moved with it to the specifier of FocP (see Cinque, 1999, 31): 

29


(27) maP (=22a.) 

2 

ma FocP 

2 

[sempre [mangiato carne]] Foc’ 

2 

Foc° sempreP 

2 

sempre VP 

5 

mangiato carne 

As we have seen, multiple arguments must appear in the unmarked order. I will 

consider this fact a crucial piece of evidence suggesting that the VP is moved to a 

higher position. If it is not moved, it is very hard to explain why its arguments 

cannot be right dislocated, which is obligatory when [Spec, Foc] is occupied by only 

one argumental DP or PP. 

To summarise, we have seen that in a ma Focus construction, the [+Focus] 

feature is checked by the raising of an element to [Spec, Foc] in the IP. If this 

element is an argument, all other eventual arguments must be right dislocated and 

the verb undergoes remnant movement to a position higher than ma. If this element 

is the whole VP, no remnant movement applies, and the arguments appear in the 

unmarked order. 

We can now refine our first hypothesis by adding two details to it: firstly, in a ma 

Focus the verb undergoes remnant movement to a higher position; secondly, the 

[+Focus] feature can be checked not only by DPs or PPs, but also by the whole VP. 

This is not surprising if we consider the data I will present in the next section, where 

I describe some cases of “large” constituents marked as contrastive Focuses by ma. 

3. Focusing of Modal complements and CPs 

In this section I will describe some cases of focusing of large constituents. Ma 

can mark as contrastive Focuses also infinitival complements of modal verbs. Some 

examples are given in (28): 

(28) a. tu devi ma [stare zitto]focus, in questi casi. 

you must ma stay quiet in these cases 

‘You (S) must be quiet, in such cases.’ 

30 

b. vu lo potete ma [comprare]focus

you it can ma buy 

‘You (P) can buy it.’ 


Note that some speakers accept (28b) also if the direct object clitic is reduplicated 

on the infinitive: 

(28) b’. vu lo potete ma [comprarlo]focus 

This example shows that ma somehow blocks the deletion of the lower copy of 

the object clitic in a clitic climbing structure. 

Ma can also focus whole CPs if they are complements of verbs like volere ‘to 

want’ or dire ‘to say’: 

(29) a. Mario vole ma [che noi si vada via subito]focus 

Mario wants ma that we clit. go away immediately 

‘Mario wants that we leave immediately.’ 

b. Giorgio ha detto ma [che un ne sa nulla]focus 

Giorgio has said ma that NEG of-it knows nothing 

‘Giorgio has said that he does not know anything about it.’ 

I leave a complete analysis of these sentences for further research. For now, it is 

sufficient to say that ma can focus-mark very different types of elements, but in any 

case it must follow the inflected verb. This fact cannot receive an adequate explanation 

if we do not assume that ma is IP internal and that there is a FocP in the low IP area. 

4. Deriving the interpretation of the ma Focus 

In the preceding sections I have proposed that ma takes scope on an element 

moved to [Spec, Foc] in order to check a [+Focus] feature. The presence of ma gives 

to this focused element a contrastive interpretation, similar to that of the left periphery 

Focus in Italian and Florentine. This contrastive interpretation is always present in a 

ma Focus structure, as it can be inferred by the fact that a ma Focus cannot be used 

to answer a wh question: 

(30) a. A: a chi tu l’hai dato, i libro? 

to who you it have given the book 

‘Who have you given the book to?’ 

B: #l’ho dato ma [a Mario]focus 

it have given ma to Mario 

‘I have given it to Mario.’ 

b. A: icché tu voi comprare? 

what you want buy 

31


32 

‘What do you want to buy?’ 

B: #voglio comprare ma [un divano]focus 

want buy ma a sofa 

‘I want to buy a sofa.’ 

Note by the way that a ma Focus cannot appear in isolation, even if it is used to 

contrast a wrong supposition of the addressee: 

(31) A: icché tu voi comprare? (una poltrona?) 

‘What do you want to buy? (An armchair?)’ 

B: *ma [un divano]focus 

This fact can be explained in two ways: either by assuming that the ma Focus 

does not allow ellipsis of the background part of the sentence, or by postulating that 

the fragment answer is not the Focus in IP (which is the Focus position in the scope 

of ma). This second solution is similar to some recent proposals, like the one 

advanced by Brunetti (2004), who identifies fragment answers as left periphery 

Focuses followed by sentence ellipsis. 

A ma Focus is acceptable in contexts where an informational post-verbal Focus 

is marginal. As pointed out by Brunetti (2004, 122) among others, in Italian, postverbal 

subjects are more acceptable if the event expressed by the predicate is related 

to an explicit or implicit locative: 

(32) a. ha telefonato Gianni. (from Brunetti, 2004) 

has telephoned Gianni 

b. ??ha dormito il bambino. 

has slept the child 

(32a) is acceptable only if it means ‘Gianni called here’ or ‘Gianni called us’, 

while (32b) is not so good because dormire ‘to sleep’ has no implicit locative 

meaning. The sentence is better with an overt locative: 

(33) in questo letto ha dormito il bambino. 

in this bed has slept the child 

Ma focus does not display a similar restriction. No implicit locative is needed to 

form a ma Focus structure with a post-verbal subject: 

(34) ha dormito ma [i bambino]focus (non io) 

has slept ma the child (not I) 

‘It is the child who has slept (not me).’ 

All the previous examples show that, even if FocP under maP is activated, the 

presence of ma changes the contexts where theese sentences with post-verbal Focus


can be used. In the following section I will try to individuate the precise contribution 

of the particle to the interpretation of the focused element. 

4.1 Ma as a quantificational head 

Consider now the interaction of ma with other focusing elements. In (35)-(37) it 

is checked the compatibility of ma with solo ‘only’, almeno ‘at least’ and perfino 

‘even’: 

(35) a. *Mario ha ma solo dato un esame. 

Mario has ma only given one exam 

b. Mario ha ma dato solo un esame. 

‘Mario has passed only one exam.’ 

c. *Mario ha dato ma solo un esame. 

d. *Mario ha solo dato ma un esame. 

(36) a. *Mario ha ma almeno dato un esame. 

b. ??Mario ha ma dato almeno un esame. 

‘Mario has passed at least one exam.’ 

c. *Mario ha dato ma almeno un esame. 

d. *Mario ha almeno dato ma un esame. 

(37) a. *Mario ha ma perfino dato due esami. 

b. Mario ha ma dato perfino due esami. 

‘Mario has passed even two exams.’ 

c. *Mario ha dato ma perfino due esami. 

d. *Mario ha perfino dato ma due esami. 

As it can be observed only the sentences in (b) are acceptable. In the 

grammatical sentences the focusing element modifies an argument, while ma marks 

the whole VP, since it is between the auxiliary and the lexical verb. All the other 

combinations are excluded: it is not possible to have ma and solo/almeno/perfino 

both as VP modifiers (sentences in (a)), both as argumental modifiers (sentences in 

(c)), and solo/almeno/perfino as VP modifier with ma as argumental modifier. I 

argue that this incompatibility derives from the fact that ma is a quantificational 

head. The elements corresponding to ‘only’, ‘at least’ and ‘even’ are quantificational 

heads, and it is not possible for the same element to undergo similar quantificational 

operations, as those encoded by these elements and by ma. 

The ungrammaticality of the sentences in (d) is crucial: as we have seen, ma has 

a precise position in the clause structure, and its presence prevents the insertion of 

similar quantificational elements in a higher position, even when it focuses only an 

argument and not the whole VP. 

33


Ma Focus is similar from this point of view to the so-called Identificational 

Focus of Hungarian, studied by Brody (1990), É. Kiss (1998), Horvath (1986; 2000; 

2007) among others. This Focus is associated with a precise position in the clause 

structure and diverges from the Informational Focus in expressing exhaustive 

identification: the Focus identifies a particular subset of the contextually relevant set 

of alternatives and excludes all the others. 

In a recent paper, Horvath (2007) has claimed that the syntactic position occupied 

by the Identificational Focus is not a Focus position at all, but a quantificational 

position, and Focus movement is not driven by a [+Focus] feature, but by an 

[+Exhaustive Quantification] feature. While it is not clear whether the CP Focus in 

Italian or Florentine expresses exhaustive identification (see Rizzi, 1997, and 

Brunetti, 2004, for some discussion about this problem), the ma Focus distributes 

exactly like the Identificational Focus of Hungarian. This can be observed in the 

following examples. 

Firstly, É. Kiss (1998) observes that Hungarian Identificational Focus is not 

compatible with universal quantifiers: 

(38) *Mari [minden kalapot]focus nézett ki magának. 

Mari every hat picked out herself.DAT 

*‘It was every hat that Mari picked for herself.’ 

Ma Focus displays a similar restriction. A universal QP cannot be focused by 

ma: 

(39) *Maria ha scelto ma ogni cappello. 

Maria has chosen ma every hat 

Secondly, É. Kiss points out that the Identificational Focus is not compatible 

with even-phrases and also-phrases: 

(40) a. *Mari [egy kalapot is]focus nézett ki magának. 

Mari a hat also picked out herself.DAT 

?‘It was also a hat that Mary picked for herself.’ 

34 

b. *Mari [még egy kalapot is]focus nézett ki magának. 

Mari even a hat also picked out herself.DAT 

*‘It was even a hat that Mary picked for herself.’ 

I have already shown in (37) that ma and perfino ‘even’ cannot modify the same 

constituent. (41) shows that a similar restriction is observable also with anche ‘also, 

too’:

(41) *Maria ha scelto ma anche un cappello. 

Maria has chosen ma also a hat 


Finally, É. Kiss argues that Identificational Focus takes scope, as it is shown by 

the fact that exhaustive identification interacts with other scope-taking elements, like 

univesal quantifiers. Take a sentence like (42): 

(42) minden fiú [Marival]focus akart táncolni. 

every boy Mari-with wanted to-dance 

‘For every boy, it was Mari that he wanted to dance with.’ 

In (42) the universal quantifier takes scope over exhaustive identification, and 

the sentence means that every boy wanted to dance with Mari and not with any other 

girl. If we compare two sentences with a universal quantifier in topic position 5 , the 

first with a standard Informational Focus, the second with a ma Focus, they diverge 

precisely in this respect: only the sentence with the ma Focus means that for every 

boy it was only one specific girl that he liked. The other sentence does not exclude 

that some boys liked also other girls: 

(43) a. a ogni ragazzo piaceva [Maria]focus. 

to every boy liked Maria 

‘Every boy liked Maria.’ 

b. a ogni ragazzo piaceva ma [Maria]focus. 

to every boy liked ma Maria. 

‘For every boy, it was Maria that he liked.’ 

On the basis of these data I propose that maP is associated with an [+Exhaustive 

Identification] feature. At this point it is possible a further refinement of the first 

hypothesis about the ma Focus. 

A ma Focus structure is formed by two syntactic processes: the first one is the 

syntactic movement of a constituent to the specifier of FocP in the IP, the second 

one is the insertion of the particle ma, bearing the [+Exhaustive Identification] 

feature, in the low IP area. The syntactic domain of the particle is the focused 

constituent in [Spec, Foc]. Under this configuration, the element in [Spec, Foc] 

receives the exhaustive identification interpretation, and VP undergoes remnant 

movement to a higher position. 

5 See Cardinaletti (2004) for some discussion about the precise position of “dative subjects” 

of verbs like piacere ‘to like’. 

35


4.2 Ma, ‘only’ and ‘even’ 

From what I have said so far, the only difference between ma and solo ‘only’ 

seems to be that ma has a precise and unique position in the clause structure, while 

solo can appear in different positions, similarly to its English equivalent. However, 

it should be pointed out that an element focused by solo can but does not have to be 

a contrastive Focus, while ma Focus is always contrastive. I think that this distinction 

derives from the fact that the quantification processes operated by solo and ma are 

slightly different. 

Consider the opposition of elements like ‘also’ and ‘even’. From the point of 

view of quantification, they are similar, since they indicate that some property is true 

for an element x, and that the same property is true for elements other than x (Bayer, 

1996, 51). Both the sentences in (44) mean that John invited his sister, and that he 

invited also someone else: 

(44) a. John invited also his sister. 

b. John invited even his sister. 

However, the sentence in (44b) has a further meaning: John’s sister is less likely 

to be invited by John than others are. 

I think that a similar difference exists between ma and solo. More precisely, solo 

indicates that some property is true for an element x and that the same property is 

not true for elements other than x. On the other hand, ma indicates that some 

property is true for an element x, that the same property is not true for elements 

other than x, and that these other elements are less likely to be associated with the 

property. This is why the ma Focus is contrastive and is used to express that some 

explicit or implicit supposition of the addressee is wrong. 

(45) a. Giorgio ha invitato solo la su sorella. 

Giorgio has invited only the his sister 

b. Giorgio ha invitato ma la su sorella (non la su mamma). 

Giorgio has invited ma the his sister not the his mother 

The relation between ma and solo can be observed in diachrony. In some 

northern Italian varieties, the word for ‘solo’ derives from the Latin non magis quam 

‘no more that’: Piedmontese mak, old Paduan nomé, old Lombard nomà. In 

Florentine, like in many other Italian varieties, the word for ‘only’ has a different 

origin. I propose that Florentine ma derives from (non) magis (quam), that has been 

preserved with this particular meaning. 

36

5. Conclusive remarks 


In this paper I have presented some data about the syntax and the interpretation 

of the focusing particle ma in Florentine. This particle is the head of a dedicated 

projection in the low area of IP. Since maP is associated with a [+Exhaustive 

Identification] feature, these data from Florentine confirm that quantificational heads 

can have a precise position in the IP. 

A ma Focus structure requires the movement of the focused constituent to the 

specifier of a lower projection, which is the syntactic domain of the particle, and 

which I have proposed to identify with the low FocP of Belletti (2001) and 

subsequent works. Since maP cannot be located higher than IP, my proposal 

assumes that Belletti’s postulating a Focus position in the IP is correct. This idea has 

been recently confuted by Brunetti (2004), who claims that there is not a Focus 

position in the IP, and that constituents can be focused in situ by entering in an 

Agree relation with the [+Focus] feature in CP. The data presented here cannot be 

directly compared to Belletti’s and Brunetti’s data, which are mainly from standard 

Italian. However, the peculiar configuration required by ma Focus implies that at 

least in Florentine there is a Focus position in IP. 

On the other hand, these data about ma show, contra some of Belletti’s 

conclusions, that a contrastive Focus is not necessarily in CP. 

Some words should be spent on the notion of Contrast. É. Kiss (1998) argues 

that Identificational Focus is characterised by the features [± Exhaustive] and [± 

Contrastive], and that languages vary in the value of these two features. A [+ 

Contrastive] Focus “operates on a closed set of entities whose members are known 

to the participants of the discourse” (É. Kiss, 1998, 267). I think, following Brunetti 

(2004), that “contrast” is not an appropriate syntactic feature. For instance, a left 

periphery Focus in Italian (or in Florentine) is usually contrastive, but in some cases 

it is informational (Benincà and Poletto, 2004). Thus, [+Contrastive] does not seem 

a movement driving feature. In the case of ma Focus, contrast is a pragmatic effect 

derived from the quantificational operation associated with ma. As I have proposed, 

this operation is exhaustive identification (and, thus, ma is similar to solo ‘only’) 

accompanied by a sort of “evaluation” about the non-identified subset of the 

contextually relevant set of alternatives. This type of quantification is worth of 

further research. As it has been noted by Cinque (1999, 180n), perfino ‘even’ does 

not allow its complement to raise past it, while this is possible with solo. As we have 

seen, the same holds for ma, which must precede the focused constituent. In these 

cases a distinctive semantic feature seems to correlate with distinctive syntax. 

References 

Antinucci, F. & G. Cinque. 1977, “Sull’ordine delle parole in italiano: l’emarginazione”. Studi di 

Grammatica Italiana 6, 121-146. 

37


Bayer, J. 1996, Directionality and Logical Form. Dordrecht, Kluwer. 

Belletti, A. 2001, “Inversion as Focalization”. In: A. C. J. Hulk & J.-Y. Pollock (eds.) Subject 

Inversion in Romance and the Theory of Universal Grammar. Oxford - New York, 

Oxford University Press, 60-90. 

Belletti, A. 2004, “Aspects of the Low IP Area”. In: L. Rizzi (ed.) The Structure of CP and IP. 

Cartography of Syntactic Structures 2. Oxford - New York, Oxford University Press, 16- 

51. 

Benincà, P. & C. Poletto. 2004, “Topic, Focus and V2. Defining the CP Sublayers”. In: L. Rizzi 

(ed.) The Structure of CP and IP. Cartography of Syntactic Structures 2. Oxford - New 

York, Oxford University Press, 52-75. 

Brandi, L. & P. Cordin. 1981, “Dialetti e italiano: un confronto sul parametro del soggetto nullo”. 

Rivista di Grammatica Generativa 6, 33-87. 

Brody, M. 1990, “Some Remarks on the Focus Field in Hungarian”. UCL Working Papers in 

Linguistics 2, 201-225. 

Brunetti, L. 2004, A Unification of Focus. Padova, Unipress. 

Cardinaletti, A. 2001, “A Second Thought on Emarginazione: Destressing vs. “Right 

Dislocation” ”. In: G. Cinque & G. Salvi (eds.) Current Studies in Italian Syntax. Essays 

offered to Lorenzo Renzi. Amsterdam, Elsevier, 117-135. 

Cardinaletti, A. 2004, “Toward a Cartography of Subject Positions”.In: L. Rizzi (ed.) The 

Structure of CP and IP. Cartography of Syntactic Structures 2. Oxford - New York, 

Oxford University Press, 115-165. 

Cinque, G. 1999, Adverbs and Functional Heads. Oxford - New York, Oxford University Press. 

É. Kiss, K. 1998, “Identificational Focus versus Information Focus”. Language 74.2, 245-273. 

Hinterhölzl, R. 1997, “A VO-based approach to verb raising”. In: K. Kusumoto (ed.) Proceedings 

of the North East Linguistic Society 27. Amherst, Ma, GLSA, University of 

Massachusetts, 187-202. 

Horvath, J. 1986, FOCUS in the Theory of Grammar and the Syntax of Hungarian. Dordrecht, 

Foris. 

Horvath, J. 2000, “Interfaces vs. the Computational System in the Syntax of Focus”. In: H. 

Bennis, M. Everaert & E. Reuland (eds.) Interface Strategies, Amsterdam, HAG, 183- 

207. 

Horvath, J. 2007, “Separating “Focus Movement” from Focus”. In: S. Karimi, V. Samiian & W. 

K. Wilkins (eds.) Phrasal and Clausal Architecture. Amsterdam, John Benjamins, 108- 

145. 

Kayne, R.S. 2000, Parameters and Universals. Oxford - New York, Oxford University Press. 

Laka, I. 1990, Negation in syntax: On the nature of functional categories and projections. 

Doctoral Dissertation, MIT, Cambridge, Massachusetts. 

Rizzi, L. 1997, “The Fine Structure of the Left Periphery”. In: L. Haegeman (ed.) Elements of 

Grammar. Handbook in Generative Syntax. Dordrecht, Kluwer, 281-337. 

Zanuttini, R. 1991, Syntactic Properties of Sentential Negation: a Comparative Study of Romance 

Languages. Doctoral Dissertation, University of Pennsylvania. 

38


DEVELOPMENTAL PATTERNS IN THE ACQUISITION 


OF COMPLEMENT CLITIC PRONOUNS 

COMPARING DIFFERENT ACQUISITION MODES 

WITH AN EMPHASIS ON FRENCH 

Cornelia Hamann and Adriana Belletti 

Recent research has highlighted the particular status of pronominal clitics in first 

language acquisition of French. Hamann et al. (1996) and Jakubowicz et al. (1997) 

have indicated the delay of complement clitics as opposed to subject pronouns. 

Studies on bilingual and early L2 acquisition have reached a similar conclusion 

(Belletti and Hamann (2004), White (1996), Hulk (1997), Crysmann and Müller 

(2000), Kaiser (1994)) as did studies on SLI children acquiring French (Jakubowicz 

et al. (1998), Hamann et al. (2003), Paradis et al. (2003)). In this paper we reconsider 

the issue, survey data available in the literature, provide new data on French and for 

comparative purposes also on Italian. We undertake a fine grained comparison of the 

developmental patterns in this domain of acquisition which will allow us to uncover 

subtle distinctions hidden under the global term of delay and possibly reveal 

properties of the different modes of acquisition. At the same time, we point out that 

the error types and stages in the different modes of acquisition can reveal properties 

of different grammatical systems, which may remain unnoticed if solely the adult 

system and L1acquisition data are considered. Our comparison can therefore serve 

as a special tool to enhance our general understanding of subtle properties of 

different grammatical systems. 

In this perspective, the analysis of different kinds of clitic placement errors, 

typically found in L2 and sometimes in bilingual acquisition data, but missing in 

monolingual and SLI data constitutes a domain on which we focus our attention. 

Although the brute numbers and percentages of these types of errors are relatively 

limited throughout the literature we review, still we take their existence to be 

meaningful and possibly illuminating for the comparisons undertaken here, concerning 

both the modes of acquisition and the properties of the languages involved. A 

39


comparison of these error types in French, especially in combination with German 

or Italian, and in Italian, especially in combination with German, proves to be 

particularly revealing. 

Our main proposal is that the delay in the acquisition of complement clitics, 

which is found across all acquisition modes, is primarily due to the more complex 

and articulated syntactic derivation that syntactic clitics undergo as compared to 

other classes of pronouns, weak or strong (assuming Cardinaletti and Starke’s (1999) 

typology; see the discussion below). Characteristically, the delay gives rise to stages 

where the clitic complement is omitted (in obligatory contexts), across the different 

modes. 

However, a specific interpretation is required for placement errors in (early and 

adult) L2 and some bilinguals, which are lacking in monolinguals and SLI. We 

propose that these errors are crucially linked to the coexistence of different 

grammatical systems which can occasionally make available a uniform analysis of 

object pronouns. 1 Finally, for those placement errors which do not find a reasonable 

source in the possible interaction or contact between different grammars, we suggest 

a direct active role of UG making different options available in principle. 

1.1. Structure of the Article 

Section 2 is devoted to spelling out our general background assumptions. In 2.1. 

we present properties of the Romance pronominal systems, focusing on the difference 

of complement clitics and subject pronouns in French and its syntactic source. This 

section also introduces some facts about Italian and contrasts the Romance pronoun 

systems with Germanic pronouns, especially the German system. In section 2.2 we 

outline our assumptions about the different modes of acquisition. The theoretical 

assumptions lead to certain expectations, outlined in section 2.3. Section 3 describes 

the methods under review involving spontaneous and elicited production and some 

grammaticality judgements. Section 4 presents the results on complement clitics. In 

section 4.1. we first present data on the delay of complement clitics, discussing, 

monolinguals, bilinguals and L2 learners (early and adult) and the findings for SLI 

children. Different types of placement errors observed in our early L2er will be 

further investigated in section 4.2 with respect to different modes of acquisition; a 

comparison with Italian data in section 4.3 rounds off the result section. A 

summarizing discussion concludes the article in section 5. 

1 Naturally, SLI who are also L2ers may display a different behavior than monolingual SLI in 

this respect. We are not in a position to supply significant data bearing on this subtler 

distinction. 

40

Developmental patterns of complement clitics 

2. General Background: Complement Clitics, Derivation, and Modes of 

Acquisition 

2.1. Assumptions on Deriviation 

Romance complement 2 clitic pronouns are assumed to fill a special functional 

head position in the clause structure. According to different implementations of the 

basic account of their distribution, clitics are assumed to fill a head in the high part 

of the clausal functional structure dedicated to clitic pronouns (Sportiche’s (1996) 

“clitic voice”); they can be taken to move to an Agr-type head in the high part of the 

clause (Kayne (1991), Belletti (1999) and references cited there).The characteristic 

of Romance complement clitic pronouns resides in the fact that they are nominal 

arguments strictly connected to the verbal domain. Clitics are DPs which undergo a 

computation whose crucial step only concerns the head of the DP. Clitic heads and 

the head ultimately hosting the verb are intimately interrelated or actually coincident 

with both the clitic and the verb filling the same functional head position, as in the 

case of cliticization into finite verbs in French, Italian and other Romance languages. 

Whatever the exact implementation of the cliticization process we want to adopt, a 

functional head ultimately hosting the clitic is assumed to be present and active in 

the high part of the clausal functional structure. We can assume that languages differ 

as to whether such a functional head is activated or not: in languages with clitics it 

is, in languages which do not have clitic pronouns it is not. 

In finite clauses, Romance complement clitics are attached to the finite part of 

the verbal construction. If the finite verb is lexical, this results in the order Cl Vfin in 

declaratives. In both French and Italian, in periphrastic complex tenses involving an 

auxiliary and a past participle, the complement clitic pronoun ends up attached onto 

the auxiliary which carries the features related to finite morphology (e.g. person, 

number and tense) and which is the highest verbal form in the clause structure, 

yielding the order Cl Aux Past Participle (Cl Aux PPart). This is a typical ordering 

in Romance. To our knowledge, an order Aux Cl PPart is admitted (only) in 

Brazilian Portuguese (Bianchi and Figueiredo (1994)), where the status of object 

clitics is probably different and closer to that of weak pronouns of the Germanic 

type (Cardinaletti and Starke (1999; 2000) ). For complement clitics the cliticization 

process ultimately involves movement of the clitic as a head (D°) 3 , whereas weak 

pronouns move as maximal projections (DP) from the complement position to some 

2 Under the cover term complement clitics, we here include both direct and indirect object 

clitics (e.g. le, lui and reflexive clitics me te, se… in French) and prepositional clitics (e.g. en, 

y, in French). 

3 In the last step of the derivation, in a movement analysis of cliticization (Belletti (1999) and 

references cited there). 

41


intermediate dedicated position in the clause structure. Hence, the host of a weak 

pronoun is not a verbal head as in the case of typical Romance object cliticization. 

The overall computation affecting syntactic clitics is more complex than that affecting 

weak (and also strong) pronouns as a final further head movement step is included in 

the former but not in the latter. Throughout, we discuss the possible relevance of 

these distinctions in interpreting different outcomes in the acquisition of complement 

clitics. Our central hypothesis is that the complexity of the computation affecting 

complement clitics is precisely at the source of the difficulty manifested in the 

different modes of acquisition. 4 

A peculiar distributional property concerning complement clitics is that in various 

Romance languages, including e.g. Italian, but excluding French, a complement 

clitic can be attached to the finite matrix verb of a complex sentence. In this kind of 

structures, the complement clause is an infinitival and the clitic belongs to the 

embedded clause (e.g. Italian: Lo voglio leggere ‘I it(cl) want to read’). This option, 

often referred to as “clitic climbing”, is characteristically conditioned by the nature 

of the matrix verb typically including modals and aspectuals (e.g. Italian: Lo finisco 

di leggere ‘I it(cl) finish to read’). The size of the verb classes allowing “clitic 

climbing” varies from one Romance language to the other, with French disallowing 

it altogether, Italian allowing it with modals, aspectuals and some raising verbs, with 

differences among speakers and varieties, and, e.g., Spanish allowing it with partly 

different classes of verbs taking an infinitival complement (Cinque (2004) and 

references cited therein for relevant recent discussion). As climbing of the clitic into 

the matrix verb is generally an option, often a much preferred one, the hypothesis 

has been undertaken at least since Rizzi’s (1978) original work, that the option comes 

as a consequence of analyzing an originally biclausal structure as a monoclausal 

one. Whence, the name Restructuring for the process leading to reanalysis. Indeed, 

cliticization is a very local process and a clitic normally cliticizes onto the verb 

4 See Zesiger et al. (2006) for a subtler hypothesis whereby complement clitics may be 

considered problematic since they give rise to a particular instance of crossed chains, typically 

hard at some (initial) stages of acquisition. This factor may be particularly relevant for the 

different profiles observed in elicitation studies (though not necessarily in spontaneous 

production) for the development of the reflexive se and accusative clitics but does not concern 

us here. For concreteness, in this paper we do not elaborate on this alternative and assume the 

hypothesis presented in the text, according to which the complexity of the derivation counts 

as the primary source of difficulty in the acquisition of complement clitics. In the same vein, 

we adopt the working hypothesis that the internal structural make up of the different classes of 

pronominal DPs, with clitics corresponding to a more reduced (deficient in Cardinaletti & 

Starke’s terms) internal structure than weak (and strong) pronouns, does not directly bear on 

the acquisition issues addressed in this paper. We assume that the different external 

computations rather than the possibly different internal make ups are the crucial differential 

factor in acquisition. See also Hamann (2003) for further discussion of this point. 

42


which has it as its complement or onto the aspectual auxiliary, as in the case 

discussed above 5 . It is only in Restructuring contexts that the clitic climbs higher 

than the clause to which it belongs. For the purposes of our discussion here, it 

suffices to have identified the basic descriptive properties of the restructuring 

phenomenology; in this context, it is not crucial to take a stand on whether 

restructuring should be considered a process, as in the original account, or whether a 

clause displaying a climbed clitic should be analyzed as monoclausal altogether as in 

some more recent accounts (Cinque (2004)). 6 In the course of our discussion below 

we will use the term Restructuring to refer to structures where the clitic has climbed 

onto a matrix modal or aspectual verb, the core cases of restructuring in the 

languages displaying the phenomenon. 

Although the main focus of our discussion in this paper will be the acquisition of 

complement clitics in French, we will also provide some comparative considerations 

on the acquisition of French subject clitics. Some preliminary terminological and 

theoretical considerations are in order here. Although French personal pronominal 

subjects (je, tu, il, elle, on, nous, vous, ils, elles ‘I, you, he, she, one, we, you, they 

(m), they (f)’) are often referred to as clitics, their status is not the same as that of 

object complement clitics. Both object and subject clitics are phonological clitics in 

that they do not bear an independent stress and form a phonological word with the 

verb they combine with (directly or in a cluster with other pronominal clitics); 

however, only complement clitics can be properly analyzed as syntactic clitics. If 

complement clitics ultimately count as heads and behave according to this status 

syntactically, subject clitics behave as maximal projections (DP) throughout the 

entire syntactic derivation (Kayne (1991), Cardinaletti and Starke (1999), Laenzlinger 

and Shlonsky (1997)). 7 

According to the distinction mentioned above between weak and clitic pronouns, 

we then assume that French object clitics are syntactic (as well as phonological) 

5 The clitic can also be the complement of the head noun of a DP complement of the verb or 

be an adjunct. We make abstraction of these distinctions which are not directly relevant for 

the point at issue here. 

6 See also the references cited there for a rich bibliographical information on restructuring. 

For different versions of the more traditional account see Rizzi (1978; 1982), Burzio (1986), 

Zubizarreta (1985). 

7 We assume the status of weak pronouns for subject clitics in Standard French, (and also in 

Colloquial French where the facts are less clear; see Friedemann (1995), Hamann (2002) for 

discussion). See Auger (1995), Zrib-Hertz (1994) and references cited there for an alternative 

view according to which also French subject pronouns should be analyzed as filling a head 

inflectional position, hence ultimately as syntactic heads. This analysis essentially assimilates 

French subject pronouns to the subject clitics of Northern Italian dialects; see Brandi & 

Cordin (1989), Poletto (2000) and references cited there for critical discussion. 

43


clitics, while subject clitics are syntactic weak pronouns and clitics just phonologically. 

In the course of our discussion we will occasionally refer to the French pronominal 

subjects as subject clitics following current practice. However, the theoretical and 

distributional distinction concerning their syntactic nature should be kept in mind as 

it directly bears on the different outcomes known from the literature on the 

acquisition of subject vs. object complement clitics. 

Note here that other Romance languages such as Standard Italian do not have 

conspicuous instances of of weak pronouns (neither subjects nor complements). 8 For 

the interpretation of our findings in French we exploit this difference in the 

pronominal systems of the two languages. In this connection, we refer to recent 

results on the acquisition of Italian object clitics by bilingual/L2 speakers and point 

out differences in error patterns in French and Italian, most typically emerging when 

the first or concomitant language is German (Ferrari (2006); Leonini and Belletti 

(2004); Leonini (2006)). 

To round off our discussion of pronominal systems, we summarize here some 

facts about Germanic pronoun systems which we will refer to throughout, using 

German as our main example. In contrast to French/Romance clitics, German personal 

pronouns are ambiguous between strong and weak use and behave differently 

accordingly. As strong pronouns they can be stressed, and coordinated, provided 

they are referring to [+human] arguments. Inanimate subject and object pronouns 

cannot be strong and cannot be coordinated. As weak pronouns they can 

phonologically cliticize to nouns and complementizers. Crucially, they are not verbal 

clitics as the Romance clitics are. 

German also has a series of so called demonstrative pronouns (der, die das, den, 

etc.) which are identical in form to the definite determiners. These d-pronouns are 

again ambiguous between weak and strong use but always show up in the same 

positions as full DPs. (Ich hab’ den schon gesehen ‘ I have this (the, him) already 

seen’/Ich hab’ den Film schon gesehen ‘I have the film already seen’). 3 rd person 

French accusative clitics also coincide in form with the definite determiners. They 

cannot occur in DP-positions, however, but must attach to a functional head, as 

discussed above (Je l’ai déjà vu ‘I it have already seen’/j’ai déjà vu le film ‘I have 

already seen the film). 

2.2. Different modes of acquisition 

Our point of departure is the parametric approach to L1 acquisition according to 

which invariable properties of Universal Grammar (UG) are activated by input data 

which also contain the relevant triggers for the different parametric choices that the 

8 Weak pronouns are limited to the obsolete subject pronoun egli ‘he’ and dative pronoun loro 

(Cardinaletti (1991)) for detailed discussion. 

44


child has to make. Parametric choices are supposed to concern functional heads and 

their feature specifications. 

We also assume that simultaneous bilingual acquisition of more than one 

language from birth follows essentially the same pattern as monolingual acquisition 

for each of the languages involved, thus presupposing early separation of their 

different grammars, as originally proposed by Meisel (1989) (see also Genesee et al. 

(1995), Salustri (2003), and Meisel (2006) for recent discussion). In this view 

interference/contact of the two languages is allowed –if at all- only in the special 

circumstance when input from one of the languages can be reasonably, though 

incorrectly, analyzed by the grammar of the other language (Hulk and Müller 

(2000), and some of the discussion below). In contrast, L2 acquisition, early and 

adult, presupposes by its very nature the existence of an L1 grammar in the language 

learner’s mind. A major issue in L2 acquisition therefore is to make precise the 

impact, if any, of the L1 grammar with its parametric choices on the L2 acquisition 

process. Following proposals on this issue by White (1989; 2000; 2003), Schwartz 

(1998) and related work, we adopt the view that the L2 initial state is the L1 

grammar, that areas of difficulties in L2 acquisition are typically expected in cases 

where the two languages differ in parametric choices leading to misparsing of the L2 

input through the L1 setting. Thus, L2 acquisition characteristically manifests 

Transfer phenomena. However, parametric choices are likely not to be automatically 

transferred, but transfer is characteristically expected in cases of ambiguous input 

prompting an analysis through the L1-grammar. 9 In addition, we also assume an 

active role of UG so that parameters, more specifically the functional feature 

specification they involve, can be reset and functional features not instantiated in the 

L1 can be acquired (Schwartz and Sprouse (1996), Duffield et al. (2002)) in the 

course of L2 acquisition. 10 

We maintain that in L2 acquisition by young children, here referred to as early 

L2, cases of transfer can be overcome quickly, presumably because UG is more 

readily accessible. In fact, it has been argued that early L2 resembles L1 acquisition 

9 In this sense the interference/contact situation in bilingual acquisition and transfer 

phenomena in L2-acquisition may show important similarities. If the input clearly contradicts 

the L1-setting, the transferred setting will typically be very short-lived; see Haznedar (1997) 

for relevant evidence of a short-lived transferred parameter; Haberzettl (2005) for relevant 

discussion reaching partly different conclusions. 

10 The view on adult L2 acquisition referred to above has been labeled the “No impairment” 

hypothesis and stands in contrast to a different view according to which parameters in L2 

acquisition cannot be reset and UG is not operational beyond a critical period (Hawkins 

(2001), Towell and Hawkins (1994), Hawkins and Franceschina (2004), Meisel (2006)). L2 

grammars are thus “impaired” grammars (see White (2000) and Duffield et al. (2002) for 

discussion). 

45


(White (1996)). However, precisely as for some specific transfer phenomena, early 

L2 seems to rather resemble adult L2 acquisition (Belletti and Hamann (2004)). 

Some of the discussion below bears on this issue. 

Specific language impairment (SLI) has been approached from two similar basic 

points of view. On the one hand, it is assumed that language development in 

children with SLI may be substantially delayed but is essentially the same as in 

normal L1 acquisition, which could be interpreted as SLI children having an 

operational UG but needing more trigger experiences to arrive at the valid parameter 

settings. On the other hand, it has been proposed that this type of acquisition is 

deviant from L1 acquisition, which indicates that UG is not fully functional. As it 

has been claimed that error patterns in L2 and SLI resemble each other (see 

Hakansson and Nettelblatt (1996)), we might expect to gain insights into the 

underlying mechanisms of L2 acquisition and SLI by a close comparison of 

development and error patterns in a well investigated area such as clitic use, where 

robust results on L1 acquisition can form the background for investigations of early 

L2, adult L2 and SLI. 

The particular interest of comparing SLI and young L2 learners lies in the 

possibility of teasing apart which phenomena of L2 acquisition are due to transfer 

and which are due to a developmental difficulty with a particular area of grammar, 

as has recently been pointed out by Paradis (2004). Such comparative studies have 

been undertaken only recently, however, and tend to focus on a particular phase of 

development. We maintain, and hope to provide evidence for the claim, that though 

similarities during certain stages of both types of acquisition can be observed 

(Hakanson and Nettelblatt (1996), Paradis (2004)), differences are revealed by a 

consideration of developmental/longitudinal data: SLI ultimately shows a developmental 

pattern closer to that of monolinguals (though slower), whereas early L2ers rapidly 

overcome the difficulties in those cases where they resemble SLI. 

2.3. Main expectations 

Given the above assumptions about the different modes of acquisition and the 

linguistic outline of the pronominal systems in French, Italian/Romance, on the one 

hand, and the Germanic languages on the other hand, we come back to the proposals 

briefly outlined in the introduction. If French object/complement clitics are syntactic 

clitics and are therefore computationally more complex than subject pronouns, 

which are weak pronouns, then we expect a delay and omissions for complement 

clitics in all modes of acquisition. This is so because avoidance and omissions can 

alleviate processing load induced by computational complexity - if only because no 

phonological matrix must be spelled out. Placement errors, however, cannot solely 

be motivated by the greater computational complexity; we propose that in some 

46


instances they are due to the coexistence of grammars leading to misclassifications 

and in other instances to the exploitation of UG options. 

In particular, our assumptions about the role of the L2 system, which may allow 

an analysis in terms of the L1 parameter settings, make us expect that a German 

child learning French may try out an analysis whereby (all) pronouns are treated as 

weak. This analysis would be suggested by the existence of weak pronouns in both 

her L1 and in French (see the status of subject pronouns) and would be aided/ 

strengthened by the similarity in form of object pronouns and determiners, found in 

both languages (2.1.). 

Though such a similarity in form also exists in Italian, the basic absence of weak 

pronouns in the Italian system (footnote 8) should indicate to L2-learners that such 

an analysis is not tenable. Hence it should not be entertained in the typical case. If 

transfer were a stage regardless of properties of the L2, we would expect German(ic) 

learners of Italian and French to manifest the same placement errors. In contrast, we 

propose that the L2 system plays an active role and thus we expect placement errors, 

with pronouns showing up in DP positions, to be more likely when the target/second 

language is French than when it is Italian. 

As for the role played by UG in L2-acquisition, we also expect that 

Restructuring errors may occur during the acquisition of French as a target language, 

Restructuring being a UG option. Furthermore, this option could also be tried out in 

some bilingual settings. 11 

3. The Method 

The core of the French data considered in our discussion is taken from the L1, 

SLI and bilingual/early L2 corpora collected in the framework of the Geneva project 

on “Language and Communication: Acquisition and Pathology”. The Italian data 

(spontaneous and experimental) referred to here were collected in different projects 

within the frame of the research activities undertaken at the Interdepartmental Center 

of Cognitive Studies on Language at the University of Siena. 12 We will also refer to 

examples and, where available, quantitative analyses from the literature in order to 

round off the picture. 

The basis and measure of comparison for this investigation are longitudinal 

studies of the spontaneous production of normally developing monolingual French 

11 To the extent that Restructuring errors are not attested in monolingual L1 acquisition, we 

can speculate that the possible UG option is not equally entertained as it is overwhelmingly 

disfavored on the basis of unambiguous positive evidence, see discussion in section 5. 

12 Many of the results we will be referring to with respect to the Geneva and Siena data have 

been published or have given rise to dissertations. Here they are considered from a fresh 

comparative perspective. 

47


speaking children: Augustin: 10 recordings (2;0- 2;9); Marie: 17 recordings (1;8- 

2;6), Louis: 12 recordings (1;9 - 2;3) as described in Hamann et al. (1996) and 

Rasetti (2003). We also use data from Daniel (1;8-1;11), Nathalie (1;9-2;3), Gregoire 

(1;9-2;3), and Philippe (2;1-2;6) from the Lightbown corpus, and the Leveillé and 

Champaud corpora known from Childes. 

As for impaired language we primarily use the spontaneous productions of 11 

monolingual French children clinically diagnosed as SLI as recorded in Geneva/ 

Lausanne with an age range of 3;10-7;11 at the beginning of recording. For these 

data we make reference to Hamann et al. (2003) and to Baranzini (2003) (see also 

Cronel-Ohayon (2004)). We will also refer to data from Jakubowicz et al. (1998), 

Jakubowicz (2003), Paradis et al. (2003) and Paradis (2004) and others. 13 

Our main source for bilingual/early L2 data is a corpus recording two young 

children with different source languages (exposure from birth), who are communicating 

with each other in French: Elisa, whose first language is German and who was 

recorded between the ages of 4;0-5;5 and Lorenzo, who is of Italian origin and who 

was recorded between the ages of 3;5-4;11. Belletti and Hamann (2004) showed that 

Elisa’s speech exhibits phenomena which are also discussed in the literature on 

(adult) L2, she will therefore be called “early L2er” here. Lorenzo, on the other 

hand, shows a development parallel to that of monolingual French acquisition, so 

that he will henceforth be termed “bilingual”. Systematic exposure to French was 

roughly the same for these two children, and there are 5 recordings, distributed as 

shown in table 1 (see Belletti and Hamann (2004) for further details). 

48 

Table 1: Recordings and ages of Lorenzo and Elisa 

recording Lorenzo age Elisa age 

1 3;5 4;0 

2 3;7 4;2 

3 3;8 4;3 

4 4;4 4;10 

5 4;11 5;5 

13 The typically developing children and the SLI children considered here are in comparable 

developmental stages as they fall roughly into the same MLU range (Hamann (2004)). The 

mean ages of the groups of SLI children discussed in the literature and which we use for 

comparison here are 7;6 (Paradis (2004)), 7;8 and 9;1 (Jakubowicz (2003)), allowing 

comparisons in some cases and projections of further development in others.


The corpora collected in Geneva were transcribed and analyzed by the same 

procedure for all acquisition modes. In particular clitic omissions and placement 

errors were counted in the same way. All judgments have been verified by native 

speakers, which is especially relevant for the identification of omissions. For details 

on the counting and analysis procedures see Hamann et al. (1996), but also White 

(1996), Jakubowicz et al. (1997, 1998), Hulk and Müller (2000) who all identify 

omissions on the basis of the verb’s argument structure and the discourse context. 14 

In addition to our data from the bilingual/early L2 children, we will use data on 

Anouk (Dutch/French bilingual) as discussed in Hulk (1997; 2000), and data on Ivar 

and Caroline (both German/French bilingual) as discussed in Crysmann and Müller 

(2000). For clear cases of early L2 acquisition we also consult the productions of 

Kenny and Greg (L1:English and L2:French) as reported in White (1996) and in 

Prévost and White (2000). Ages and amount of exposure will be provided in the 

context of the relevant discussion. 

As to adult L2-learners of French, we will refer to the literature, especially 

Prévost and White (2000), Herschensohn (2004), and Granfeldt and Schlyter (2004) 

and the speakers they studied. We will occasionally refer to the study of Landow 

(2002) conducted in Geneva with 25 adult L2-learners of French (with Chinese, 

German, English and Romance languages different from French as their L1) to 

strengthen tendencies already observed. 

The Italian data used for comparison are taken from the work by Leonini & 

Belletti (2004), Leonini (2006) for adult L2 acquisition and Ferrari (2006) for 

bilingual acquisition. 

A word of caution is in order before we proceed. Although our comparisons will 

consider data and results obtained in different studies through different procedures, 

including spontaneous production, elicited production and also some grammaticality 

judgments, we believe that the comparisons are revealing and significant as they 

illustrate consistent trends throughout. 

As for placement errors, the different data sources are not specially problematic 

as the errors are clearly manifested. With respect to omissions, the differences in 

data taking may raise the issue of defining what counts as an obligatory clitic 

context. In elicited production all contexts are by definition obligatory, whereas in 

spontaneous production, although for native speakers it may be clear that what is 

missing is a clitic pronoun, there will always remain a doubt that what is being 

omitted could be a lexical complement or a licit omission (see Pirvulescu (2006) for 

discussion). Native speakers’ judgments guided us in the identification of obligatory 

complement clitic pronoun contexts. Although a certain amount of indeterminacy is 

unavoidable in this connection, especially in spontaneous production, many of the 

14 Note especially that omissions which several native speakers judged legitimate are not 

included in our counts. 

49


studies we use for complementation of our data have employed the same procedures 

for the identification of omissions (see references above), so that we work under the 

assumption that the relevant cases of omission from the literature can also be 

classified as pronominal clitic omissions. 

4. Results on Complement Clitics 

4.1. The ‘Delay’ of Complement Clitics and Error types in different modes of 

acquisition 

Complement clitics display a delay in acquisition with respect to subject clitics 

which is rather significant and has given rise to several studies (see the references 

quoted above, and also Schmitz and Müller (in press), Pirvulescu (2006)). In 

monolingual, typical children it concerns a time span of about six months, and in 

SLI children the absence or very rare use of object clitics may persist for several 

years (Jakubowicz et al. (1998), Hamann et al. (2003), and Paradis et al. (2003)). 

Bilinguals seem to show the delay also observed in monolinguals (see Hulk (2000), 

Crysmann and Müller (2000), and Schmitz and Müller (in press)), whereas early and 

adult L2 speakers produce their first object clitics even more than six months later 

than subject pronouns. In comparison to SLI children, the period of non-use of 

object clitics is much shorter in early L2 speakers, however, and they quickly evolve 

in their use as documented also in their error patterns. 

4.1.1. Monolinguals 

The general consensus on subject clitics in monolingual acquisition is that they 

are used from roughly the second birthday, whereas complement clitics are omitted 

till they occur tentatively about 4 months later and more systematically about 6 

months later. Augustin, the child investigated by Hamann et al. (1996) shows this 

clearly in his development between 2;0 and 2;10, see table 2. Louis, one of the 

Geneva children studied by Rasetti (2003) with respect to clitic use, shows the same 

profile. He produces 29.4% subject clitics at the age of 1;9,26, the beginning of 

recording, (Rasetti 2003,155). At this time, he produces no complement clitics. He 

starts using them at a rate of only about 5% from 2;0,8 till 2;1,20 and shows a rise to 

about 11% between 2;2,20 and 2;3,29 (Rasetti 2003,257). Marie, another child 

studied by Rasetti already uses 66.7% subject clitics at the age of 1;8,26, which is a 

rate attained by Augustin at the age of 2;9,30 (p. 155). So it comes as no surprise 

that she uses complement clitics at a rate of 16.7% at that early age already (p. 257). 

Still, even if complement clitics are not radically absent in her early recordings, they 

are much rarer than subject pronouns and her speech also shows a high percentage 

of object omissions at the same time (58.3% at the beginning and 16.7% at 2;5,26). 

Studies on elicited production (Jakubowicz et al. (1996; 1997) and Zesiger et al. 

(2006)) show roughly the same picture, and the recent study on spontaneous 

50


production conducted by Schmitz and Müller (in press) shows an initial absence and 

about the same delay (5 months) for the child Gregoire from the Childes database 

(see also Friedemann (1992) and Rasetti (2003) for the same conclusions regarding 

this child). 

age 

(y;m,d) 

Table 2: Occurrences of subject and complement clitics 

in relevant utterances in the Augustin-corpus 

verbal 

utterances 

subject 

clitics 

% of verbal 

utterances 

complement 

clitics 

% of relevant 

utterances 

2;0,2 57 17 29.8 0 0 

2;0,23 30 4 13.3 0 0 

2;1,15 22 4 18.2 0 0 

2;2,13 55 16 29.1 1 3.8 

2;3,10 45 12 26.6 0 0 

2;4,1 62 10 16.1 0 0 

2;4,22 54 11 20.4 1 5.0 

2;6,16 116 25 21.6 2 3.9 

2;9,2 175 80 45.7 10 14.3 

2;9,30 115 99 63.4 22 33.9 

Total 771 278 36.1 36 10.5 

Moreover, table 3 indicates that if complement clitics are radically absent in 

Augustin’s speech at the beginning of recording, they reach a level of around 30% 

occurrence, the level found for subject clitics at the very beginning, only in the last 

recording where we observe a concomitant decrease in the rate of the occurrence of 

lexical complements as well as in the rate of clitic omissions. The same is true for 

Louis (Rasetti (2003, 257)). 15 

15 See also Wexler, Gavarró, Torrens (2004), Babyonyshev and Marin (2004) for recent 

discussion on the different omission rates in different Romance languages (Spanish, Catalan, 

Romanian in particular) in L1 acquisition. 

51


52 

age 

Table 3: The use of complement clitics in comparison with lexical 

complements and omissions in the Augustin corpus 

comp. 

contexts omissions % 

complement 

clitics % 

lexical 

complements % 

2;0,2 12 4 33.3 0 0 8 66.6 

2;0,23 20 5 25 0 0 15 75 

2;1,15 10 4 40 0 0 6 60 

2;2,13 19 5 26.3 1 3.8 13 69.9 

2;3,10 23 9 39.1 0 0 14 60.9 

2;4,1 20 5 25 0 0 15 75 

2;4,22 21 4 19.0 1 5.0 16 76 

2;6,16 50 10 20 2 3.9 38 76.1 

2;9,2 69 10 14.4 10 14.3 49 71.3 

2;9,30 65 14 21.5 22 33.8 29 44.7 

Total 309 70 22.7 36 11.6 203 65.7 

A clear fact that has emerged from recent research is that object/complement 

clitics are placed correctly from their first occurrences as was pointed out by 

Hamann et al. (1996) for Augustin. This finding was recently corroborated by 

Rasetti (2003, 293) who states that “no placement error is attested in the entire 

Geneva corpus”, i.e. of the monolingual children Augustin, Marie and Louis. 16 

4.1.2 Early L2/bilingual children 

The child called early L2er, Elisa, shows a delay of complement with respect to 

subject clitics which is similar to Augustin’s development, as shown in table 4. Note 

specifically that complement clitics tend to be absent in the early recordings. 

16 As to the order of acquisition of particular complement clitics, it is interesting to note that 

appearance of clitic en generally coincides with the stages where complement clitics begin to 

be systematically produced by the different children analyzed. (Hamann et al. (1996, 324f), 

Rasetti (2003, 293)), see also footnote 21 below.

age 


Table 4: Elisa’s quantitative development of clitic use 

%subjects in finite contexts %complements in complement contexts 

lexical 

subject 

clitic 

subject 

omission total 

lexical 

compl. 

clitic 

compl. 

omission total 

4;0 1/4.2 23/95.8 0 24 4/100 0 0 4 

4;2 2/4.7 41/95.3 0 43 4/50.0 4/50.0 0 8 

4;3 2/9.1 19/86.4 1/4.5 22 4/100 0 0 4 

4;10 0 100/100 0 100 22/55.0 16/40.0 2/5.0 40 

5;5 11/7.3 139/92.7 0 150 38/50.7 33/44.0 4/5.3 75 

total 16/4.7 322/95.0 1/0.3 339 72/55.4 52/40.0 6/4.6 130 

A similar delay has been reported by White (1996) for the early L2 children 

Kenny (5;10-8;1) and Greg (5;6-7;9), both with English as source language (recording 

started after two months of exposure for both children). Kenny produces subject but 

no object clitics between month 7 and month 11 of exposure which is not due to lack 

of complement constructions because he does produce lexical complements. White 

further reports that complement clitics are often omitted in the period when they 

start being produced. 

Paradis (2004) reports similar findings. She investigated 10 Canadian learners of 

French with English as their source language whose mean age is 7.3 years and who 

had been exposed to French for two years in French schools before data taking 

began. She reports that at this stage of acquisition the L2-learners supplied object 

clitics in only 41.5% of the contexts which required pronominalization (i.e. when the 

referent of the clitic had been mentioned in the discourse). Calculating the omission 

rate in all clitic contexts from Paradis’ analysis of particular errors, gives a percentage 

of 37.4% omissions in obligatory contexts. 

In contrast to Augustin and the other monolingual children described so far, 

Elisa’s early recordings show placement errors of a specific kind. We note that 

either complement clitics are absent or that they are found in non-clitic positions, 

notably positions that appear to coincide with those of lexical DPs or strong 

pronouns. This is illustarted in (1) where the clitic appears in isolation (and with 

stress) or, and more systematically so, in the position of lexical DP complements 

(2a, b), an error which has also been reported by White (1996) concerning the early 

L2er Greg (2c). We will call these particular errors the ‘*Cl in isolation’ and the 

53


‘*Cl in object position’ error 17 . Note that Elisa does not use complement clitics in 

the recordings made at the ages of 4;0 and 4;3. In between, at the age of 4;2, we find 

4 uses of object clitics, which all occur in non-clitic positions. At the same time, she 

does not omit complements, but uses lexical DPs, something which has also been 

observed by Paradis (2004) for her early L2-ers of French. 18 

(1) E: c’est à moi, le L: le quoi? Elisa 4;2 in isolation, with stress 

it’s to me, him/the the what 

‘it’s mine, that one the what?’ 

(2a) alors, tu joue avec le Elisa 4;2 after a preposition (2 occurrences) 

so, you play with him 

‘so, you play with it’ 

(2b) non, on laisse le Elisa 4;2 in canonical object position 

no, one leaves him 

‘no, we leave him/it alone’ 

(2c) moi, j’ai trouvé le Greg (month 14 of exposure) 

me, I have found him/it 

‘Me, I’ve found it’ 

Seven months later, at 4;10, she uses complement clitics at a rate of 40% and the 

particular errors exemplified in (1) and (2a,b) have vanished. However, we still find 

placements errors, albeit of a different kind: the clitic is now sometimes located 

between auxiliary and past participle, an error also observed by Hulk (1997) for the 

(Dutch/French) bilingual child Anouk (at age 3;6 and 3;9) 19 . There are 8 contexts 

with auxiliaries in the last recording of Elisa at the age of 5;5, and two uses are 

erroneous: instead of the correct order Cl-Aux-PPart, we find (3a) and (3b) which 

show ‘*Aux-Cl-PPart’, which will henceforth serve as a name for this error 20 . 

17 We keep them distinct for ease of reference although they may actually be instances of the 

same error type. 

18 This tendency also emerged in a clitic elicitation task with adult L2ers of Italian discussed 

in Leonini & Belletti (2004) discussed in 4.2.5. below. The tendency to use full lexical DPs in 

place of a pronominal clitic also appears to be present in monolingual acquisition, during the 

omission stage (Jakubowicz et al (1997), Schaeffer (2000)). 

19 Clitic in object position errors are also signaled by Anouk’s mother, as we point out in 

section 4.2.3. 

20 Note the interesting lack of change of auxiliary (from “have” to “be”) in (3b). Similar data 

are reported in Crysmann & Müller (2000), see section 4.2.3. 

54


(3a) ça a m’ étranglé Elisa 5;5 (repeated) 

that has me strangled 

‘that strangled me’ 

(3b) regarde, là j’ ai m’ étranglé Elisa 5;5 (repeated) 

look, there I have me strangled 

‘look, there I strangled myself’ 

Note that modal/periphrastic contexts are faultless in Elisa’s speech (4a,b): 

(4a) et maintenant tu vas la rattrapper Elisa 4;10 

and now you will her catch 

‘and now you will catch her/it’ 

(4b) je vais les chercher Elisa 5;5 

I will them search 

‘I will go get them’ 

Hulk (2000) reports for Anouk that she sometimes places the clitic before the 

finite verb in these constructions which we will henceforth call ‘Restructuring’ 

errors presented in more detail in 4.2.3. 

We also note that complement omission occurs only in the later recordings 

(though rarely) and seems to surplant the ‘Cl-in object position’ error. It is thus not 

surprising that Paradis (2004) reports clitic omissions but does not find placement 

errors of the kind described here for her older L2 children who had a longer 

exposure before data taking. She observes another tendency, not observed in our 

early L2er, which may be reminiscent of the ‘Cl-in-object position’ error: her L2 

children sometimes inserted a strong pronoun or the demonstrative ça ‘that’ in 

canonical object position (e.g. J’ai vu elle ‘I have seen her’). They thus obey the 

pragmatics of anaphoric reference which leads to the use of a pronominal element 

and also choose the pronoun-type which would in principle be compatible with this 

position. However, the children appear not to have acquired the additional 

constraints which limit the use of strong pronouns in French (see Cardinaletti and 

Starke (1999; 2000) for discussion). This error then seems to show that at this stage 

the children have differentiated between clitic and strong pronouns, but only as far 

as their distribution is concerned. Note that some of the younger L2-children known 

from the literature do not show this differentiation as clitic forms are located in 

complement positions (as in (1) and (2a,b,c) above) or strong forms occur in cliticlike 

positions (see Belletti and Hamann (2004:158, examples (11a,b)) on 

misplacement of ça ‘that’). 

55


Note particularly that with respect to omissions, placement errors and clitic use 

we find strong development for Elisa, our younger, longitudinally followed L2er, as 

discussed in Belletti & Hamann (2004,161; Table 6). 

The child of Italian origin whom we call bilingual, Lorenzo, shows no problems 

with clitic use (see Table 5). Subject pronouns are used correctly and are the 

predominant subjects occurring with finite verbs (see Belletti and Hamann (2004) 

for a discussion of the significance of this finding with respect to the Italian null 

subject option). Complement clitics are present from the beginning in contrast to 

Elisa. Note also that we find no placement errors in Lorenzo’s speech, in particular 

none of the three error types identified for Elisa. 

56 

Table 5: Lorenzo’s quantitative development of clitic use 

% subjects in finite contexts % complements in complement contexts 

lex-s cl-s o-s total lex-o cl-o o-o total 

3;5 1/1.6 56/87.5 7/10.9 64 13/72.2 4/22.2 1/5.5 18 

3;7 2/2.0 97/96.0 2/2.0 101 25/64.1 14/35.9 0 39 

3;8 7/6.1 105/92.1 2/1.8 114 21/84.0 4/16.0 0 25 

4;4 1/1.5 64/98.5 0/0 65 13/56.5 10/43.5 0 23 

4;11 6/3.9 146/94.2 3/1.9 155 24/53.3 21/46.7 0 45 

Total 17/3.4 468/93.8 14/2.8 499 96/64.0 53/35.3 1/0.7 150 

As to the order of acquisition of particular complement clitics, Lorenzo shows a 

similar pattern to the early L2er Elisa (Belletti & Hamann (2004, 165; Table 11). 

Note that for Elisa, Lorenzo as well as monolingual Augustin and Louis the 

appearance of clitic en (‘of that’) seems to coincide with an overall high rate of clitic 

use. 21 

21 En appears late in Lorenzo as well as in Elisa. This data is reminiscent of the remark in 

footnote 16 concerning monolinguals. Moreover, en ‘of that’ seems also prone to omission. It 

is the only clitic which is omitted by Lorenzo (twice, once at the age of 3;7 and once at the 

age of 3;8, not counted in table 7 where only verbal complement clitics are considered). It is 

also omitted by Elisa, at the age of 4;2, though she later also omits le ‘him’, at 4;10 and 5;5. 

We leave open speculations on the seemingly particular status of en in this respect.


4.1.3. Adult L2 

For adult L2ers it has been noted that placement errors and omissions occur 

frequently, (see Towell and Hawkins (1994), Granfeldt and Schlyter (2004) and 

Herschensohn (2004) for a recent discussion). Interestingly, the stages identified by 

these authors, 1. ‘Pronoun in object position’, 2. ‘Object Omission’ and 3. ‘Pronoun 

in intermediate position’ are also observed for Elisa (examples (1), (2) and (3)); the 

three types of placement errors produced by Elisa, ‘* Clitic in isolation’, ‘*Cl in 

object position’, and ‘*Aux-Cl-Ppart’, are the same as described for adult L2. The 

examples in (5) and (6) taken from Granfeldt and Schlyter (2004) illustrate the two 

placement errors occurring in the different phases: 

(5) On prend le gaz et refroidir le Karl, 8 mths exposure *Cl in object position 

one takes the gas and cool him/it 

‘you take the gas and cool it’ 

(6) j’ai le vu Karl, 10 mths exposure *Aux-Cl-Ppart 

I have him/it seen 

‘I have seen him/it’ 

In addition, in adult L2 productions errors of clitic placement in modal contexts 

have been found, where instead of the French order ‘Modal Cl Infinitive’, the order 

‘Cl Modal Infinitive’ with the clitic climbed onto the modal occurs. This is the error 

we call Restructuring error, illustrated in (7) below. In a grammaticality judgement 

experiment conducted by Landow (2002), Restructuring errors as in (7) were also 

found and the ‘Clitic in isolation’ error illustrated in (8) also occurred confirming 

the results obtained from production studies: 

(7) *Il nous peut parler Restructuring 

He us can talk 

‘He can talk to us 

(8) *Qui regardes-tu? LA *Cl in isolation 

Who look at you? HER 

‘Who are you looking at? HER’ 

A closer discussion of these misplacements is taken up in section 4.2.4. 

4.1.4. Children with SLI 

In comparison to the general mastery of subject clitics by children with SLI, 

complement clitics appear extremely delayed. The mastery of subject clitics has 

been observed by Jakubowicz et al. (1998) for elicited production and was 

corroborated in Hamann et al. (2003) with an investigation based on the spontaneous 

speech of the children recorded in Geneva. In the first recordings of these 11 children, 

57


subject clitics occur between 58.9% and 96.0% for the individual children. 22 For 

further analysis the authors group the participants under and up to five years of age 

into the “younger group” (age range 3;10-5;0) and call the children older than five 

years, the “older group” (age range 5;7-7;11, mean age 7.3). 

The younger group has an average omission rate of 16% and an average object 

clitic use of 18%. The older group has fewer omissions, average 8%, but still produces 

complement clitics on average only in 23% of the contexts which require a 

complement (see also Hamann et al. (2003,155, figure 3)). While omissions are 

replaced (in part) by the use of lexical DPs, no real increase in the use of clitics can 

be observed. 

In this first cross-sectional survey, the comparison of the younger and the older 

group indicated that there is no dramatic development in clitic use as found for the 

monolinguals, the bilingual, and the L2 children. 23 Paradis (2004) corroborates the 

rather low percentage of clitic suppliance found for the “older group”. Her 10 SLI 

children are of almost the same age (mean age 7.6 years) and supply clitics in only 

47.3% of obligatory contexts. However, an increase in clitic use has been found for 

older SLI children (mean age 9;1) as discussed in Jakubowicz (2003). 

Placement errors of the types discussed here where not observed in our SLI 

population. Note that Paradis (2004) observes a higher suppliance of lexical objects 

in pronominalization contexts for her SLI children than for her other groups. She 

also reports absence of the ‘Cl-in-object-position’ error and rare occurrence of the 

‘strong pronoun in object position’ error which she found in her L2 speakers. 

4.1.5. The delay of complement clitics in normal development, SLI, L2, and bilingual 

acquisition 

In order to highlight the difference of normal development and SLI on the one 

hand and the similarity of normal and early-L2/bilingual development with respect 

to the described delay, a further comparison can be made. Hamann et al. (1996) 

found that though complement clitics are expected to occur less than subject clitics 

in adult speech, they occur at a ratio of 1:3 on average, i.e. about 75% of all clitics 

used are subject clitics and the remaining 25% are complement clitics. On the basis 

of this adult ratio, the authors showed that Augustin’s clitic use undergoes a dramatic 

22 This holds generally with the exception of 2 children with high infinitive rates. Corentin 

uses only 6.2% and Rafaelle only 15.8% subject clitics whereas their rates of subject omission 

are particularly high. Lexical subjects or strong pronouns without a clitic are not frequently 

used, neither by the younger children (3;10-5;0), nor by the older children (5;7-7;11). 

23 The delay in the production of complement clitics as evidenced by a high omission rate is 

particularly pronounced in the child Rafaelle who was followed in her development till the the 

age of 5;1. At this age she still produces less complement clitics than the monolingual child 

Augustin at the age of 2;10. 

58


development: Whereas only 7.3% of his clitics are complement clitics at the 

beginning of recording, he is close to the adult ratio with 18.2% at the end. The 

same sort of analysis conducted by Hamann et al. (2003) shows that there is no 

development from the younger to the older group of the SLI children they 

investigated (age range 3;10-7;11): the rate of complement clitics remains constant 

and resembles Augustin’s before the ‘clitic-spurt’ as shown in table 6. As for this 

ratio then, SLI children stagnate in the use of complement clitics, at least in the age 

bracket investigated here. 

Table 6: The use of subject and complement clitics in the speech of adults from the 

Augustin corpus, of Augustin and of the younger and older group of SLI children 

adults % Aug 

2;0-2;9 

% Aug 

2;10 

% SLI 

3;10-5;0 

% SLI 

5;7-7;11 

sub-cl 2.332 76.4 179 92.7 99 81.8 333 91.7 681 92.5 

comp-cl 791 23.6 14 7.3 22 18.2 30 8.3 55 7.5 

total 3.123 193 121 363 736 

We performed the same analysis for our two bilingual/early-L2 children (see 

table 7) and found that Elisa’s development resembles Augustin’s in showing a clear 

rise in complement clitic suppliance. Lorenzo does not show much development but 

could be considered proficient from the beginning. A closer analysis of his 

‘complement contexts’ in the last recording revealed that they often involved lexical 

expressions like faire la cuisine ‘do the kitchen – cook’ where pronominalization is 

impossible. Clitics are actually used at a rate of 100% in contexts which require 

pronominalization. 

% 

59


60 

Table 7: Ratio of subject to complement clitic use in Lorenzo and Elisa 

Lorenzo Elisa 

file sub-cl comp-cl total sub-cl comp-cl total 

1 56/93.3 4/6.6 60 23/100 0/0 23 

2 97/87.4 14/12.6 111 45/91.8 4/8.2 49 

3 105/96.3 4/3.7 109 19/100 0 19 

4 64/86.5 10/13.5 74 100/86.2 16/13.8 116 

5 146/87.4 21/12.6 167 139/78.5 38/21.5 177 

total 468/89.8 53/10.2 521 326/84.9 58/15.1 384 

4.1.6. Intermediate Summary 

Summarizing what has been observed so far, L2 development emerges as 

different from impaired language development. We have pointed out that the delay 

of complement clitics can be observed for monolinguals, for bilingual and early L2 

children, for adult L2 and for children with SLI. However, monolinguals, bilinguals 

and early L2 learners show a clear rise in the use of complement clitics, whereas the 

SLI children we analyzed do not show a comparable development (even if their 

omission rates drop) 24 . With respect to development therefore, SLI children exhibit a 

different, slower profile than monolinguals, bilinguals, and early L2 children. 25 

We also observed that early and adult L2 learners show typical error patterns 

which are totally absent in the productions of monolinguals as well as in the 

productions of children with SLI. With respect to the placement of object clitics, we 

thus find that monolinguals pattern with SLI children and both differ from L2 

learners, early or adult. 

24 See also Henry (2006), who found that 10-15 year old adolescents with a childhood 

diagnosis of SLI produced significantly less object clitics than a control group of typical 6year 

olds in an elicited production task; but see Jakubowicz (2003) who observes a rise of 

clitic suppliance in children with SLI between a mean age of 7;8 and a mean age of 9;1. 

25 These observations must be refined and enriched with Paradis’s (2004) findings which 

indicate that, even if a delay is found for all these modes of acquisition, yet omission rates in 

her L2 children pattern with those of SLI children and are significantly different (higher) from 

the omission rates found in monolingual 7-year olds and in monolingual 3-year olds. Thus at a 

certain stage L2 may resemble SLI with respect to clitic omission.


4.2. Placement errors in different modes of acquisition in more detail: analysis 

and discussion 

We now focus more closely on the placement errors identified above by also 

providing further data from the literature and a closer analytical discussion. 

4.2.1. *Cl in isolation (and separation) errors 

Errors of this type are not found in monolinguals nor in SLI children. 

Our early L2 child showed this error type which we repeat below as (9). Even 

though le ‘him’ was clearly stressed in Elisa’s erroneous utterance, note here that 

Lorenzo’s response shows his interpreting the utterance as incomplete, as a DP 

starting with the article but missing the noun. 

(9) Elisa: c’est à moi, LE Lorenzo: le quoi? 

it’s to me, him/the the what 

‘it’s mine, that one the what?’ 

A similar error has not been reported by other authors for early L2 children. It 

did not occur in Kenny’s and Greg’s speech (White (1996, 357)). It also seems to 

have been absent in the bilingual Dutch/French Anouk (Hulk (2000)). 

It occurs in adult L2 learners, however, in a grammaticality judgement task, even 

if their language of origin is a Romance language. Landow (2002) tested 25 adult L2 

learners of French who had different source languages. She found that 5 of 8 

Spanish speakers accepted examples like (9) above, and 4 of 9 speakers from a nonclitic 

language did not reject it. 26 We speculate that in the case of the German child 

Elisa and, possibly, the Spanish speakers the misanalysis could be induced by 

properties of the L1. As pointed out in 2.2. in German the paradigm of articles 

coincides with that of demonstrative pronouns which in colloquial speech tend to 

replace personal pronouns. Articles and third person object clitics also coincide in 

form in French, so that an initial misanalysis is possible. A similar argument can be 

made for the adult L2ers of Spanish origin, considering the shape of the pronominal 

forms in e.g. relative constructions such as lo que (necissito saber ‘What I need to 

know’) where the demonstrative-like head of the relative has the same form as an 

object clitic. 

A related error clearly indicating that the pronoun is not treated as cliticized to 

the verb but is probably treated as a Germanic pronoun occurs in cases where the 

clitic is separated from the verb. Such an error has been reported for the bilingual 

Anouk (Hulk (2000)) as shown in (10a,b) and occurs in the late recordings of the 

early L2er Greg with a subject clitic (White (1996)) as shown in (11a,b). This type 

26 1 of these was German, 1 English, and 2 Chinese speakers were not certain. 

61


of error is rare, however, for early L2ers and bilinguals, and we did not find it 

reported for adult L2ers in the literature. 27 

(10a) Je la aussi mets dans la boite Anouk 3;10,07 

I her/it also put in the box 

‘I also put it in the box’ 

(10b) Tu peux le tres bien faire Anouk 4;06 

You can him/it very good do 

‘You can do it very well’ 

(11a) On juste veut pas Greg month 20 

one just wants not 

‘we just don’t want (that)’ 

(11b) On juste peut voir Greg month 20 

one just can see 

‘we just can see’ 

4.2.2. *Cl in object position 

This error concerns the location of the clitic in a non-clitic, argument like 

position. It is absent from the spontaneous production of monolinguals, and also 

from the speech of SLI children. 

We found it to be the predominant error in the speech of the early L2er Elisa in 

her early recordings (100%) and argued in Belletti and Hamann (2004) that Elisa, 

mislead by the coincidence in the form of articles and clitics described in 2.2. and 

referred to in 4.2. above, is assimilating French clitics to German pronouns which 

can be weak or strong. Analysing French complement clitics as weak pronouns allows 

her to entertain one uniform hypothesis about pronouns in both her languages. This 

error has also been discussed by White (1996) for her two early L2 learners of 

French. Although White did not find a clitic as the complement of a preposition in 

these children’s speech, and Kenny never produces a clitic in the position of a DP 

complement, Greg does produce some such errors as shown in (2c) repeated here as 

(12). 

(12) moi, j’ai trouvè le Greg month 14 

me, I have found him/it 

‘Me, I’ve found it’ 

27 Landow (2002) reports of her L2ers that the sentence *Je lui aussi telephone is judged as 

grammatical by about 50% of the participants, even those of Romance origin. 

62


White’s (1996) table 7 shows that Greg produces more such errors in later 

recordings (at 20, 25, 27 months). At month 14, Greg omits 4 complement clitics, he 

produces 15 correctly placed complement clitics and one case of ‘*Cl in object 

position’. At month 20 he omits 6 clitics, uses 31 correctly and has two errors of the 

type discussed here; at month 25 he omits 12, places 23 correctly and has 1 error, 

and at month 27 he has 15 omissions, 19 correctly placed clitics and 3 such errors. 

Especially the concomitant high omission rate shows that at this stage, he has not 

fully mastered complement clitics so that the occurrence of this error at this stage 

seems to correspond to Elisa’s production of this error after16 months of systematic 

exposure. 

For bilingual children the picture is more articulated. Crysmann and Müller 

(2000) do not find this error in their two French/German bilinguals. However, the 

error has been reported for Anouk (Hulk (2000)), the child with a language 

combination (French/Dutch) very similar to Elisa (French/German) and the children 

discussed by Crysmann and Müller (2000). We find (13a,b,c) as examples, though 

we cannot estimate how important this error is in percentages. Hulk (2000) mentions 

that 10% of Anouk’s complement clitics are placed incorrectly but her list also 

includes the separation error (10) quoted in 4.2.1 and some ‘*Aux Cl PPart’ errors to 

be mentioned in 4.2.3. 

(13) a. Je prends la Anouk 3;03,23 

I take her/it 

‘I take it’ 

b. Je veux la comme ça Anouk 3;03,17 

I want her like that 

‘I want her like that’ 

c. Je couper le pas Anouk 3;04,28 

I cut (inf) him/it not 

‘I don’t cut him/it’ 

This particular error has been discussed often in the literature on adult L2 

acquisition, and we take examples (14a,b,c,d) from Hulk (2000) who quotes Connors 

and Nuckle (1986), Zobl (1980), Gundel and Tarone (1983), Grondin and White 

(1996), van der Linden (1985). Examples (15a,b) and (16a,b) are taken from 

Granfeldt and Schlyter (2004). Percentages of the occurrence of this error are not 

available from the references, but it is noted as being striking and frequent in a 

certain period. 

63


(14) a. Il veut les encore 

He wants them still 

‘He still wants them’ 

64 

b. Le chien a mangé les 

the dog has eaten them 

‘the dog has eaten them’ 

c. Moi j’ai trouvé le 

me I have found him/it 

‘me, I have found it’ 

d. Il ne pas prend le 

he (ne) not takes it 

‘he does not take it’ 

(15) a. Elle demande la Petra, 5 months of exposure 

She asks it/her 

‘she asks for it’ 

b. Elle croit la Petra, 5 months of exposure 

she believes her/it 

‘she believes her/it’ 

(16) a. On prend le gaz et refroidir le Karl, 8 months exposure 

one takes the gas and cool him/it 

‘you take the gas and cool it’ 

b. On refroidir le dedans Karl, 8 months exposure 

one cool him/it therein 

we cool it in there 

4.2.3. *Aux Cl PPart 

As it is generally the case with clitic placement, no error of this type is found in 

monolingual acquisition. See Rasetti (2003) on the monolingual children of the 

Geneva corpus, and see also Zesiger et al. (2006) on elicited production. Neither has 

it been observed in the speech of the 11 SLI children under investigation. 

This error was observed in the early L2er Elisa as noted in 4.1.2., but is not 

mentioned for the two early L2ers of English origin discussed by White (1996). It is 

not observed in our bilingual child Lorenzo, but has been frequently observed for 

bilingual children with Germanic/Romance language combinations.


In Hulk (2000) we find examples (17a,b) and also (18a,b,c,d,e,f) from Anouk’s 

mother’s diary occurring at around 4;06. Comparing Anouk’s age of production of 

the ‘*Cl in object position’ errors cited in (13a,b,c) and of these examples of ‘*Aux 

Cl PPart’, we find that the latter occur later and persist for about a year. (Percentages 

are not available for this error type in Anouk’s speech; the error in (18a,b) can be 

considered of the same type). 

(17) a. T’ as le mis trop chaud Anouk 3;06,25 

you have it put too hot 

‘you have made it too hot’ 

b. Il a le mis à l’ envers Anouk 3;09,01 

He has it put to the wrong side 

‘he put it on wrong’ 

(18) a. On a les tous Anouk around 4;06 

one has them all 

‘we have them all’ 

b. Quand t’ as les tous, tu peux jouer 

when you have them all, you can play 

‘when you have them all, you can play’ 

c. Pourquoi t’ as me reveillé? 

why you have me woken up 

‘why did you wake me up ?’ 

d. Pardon, j’ai pas le vu 

pardon I have not it seen 

‘pardon, I did not see it’ 

e. Toi t’ avais le bu! 

you you have it drunk 

‘you drank it’ 

f. T’ as le pas donné 

you have it not given 

‘you did not give it’ 

Crysmann and Müller (2000) describe problems in this particular area for their 

bilingual French/German children, Ivar (19a,b,c) and Caroline (20a,b). 

65


(19) a. Après il a se réveillé Ivar 3;02,14 

afterwards he has himself woken up 

‘afterwards he woke up’ 

66 

b. Il a se fait mal Ivar 3;02,8 

he has himself done bad 

‘he has hurt himself’ 

c. Il il faut que t’ as te garé sur la rue Ivar 4;00,4 

it it must that you have yourself parked on the street 

‘it’s necessary that you have parked in the street’ 

(20) a. La maman elle a – elle avait se fait mal Caroline 3;08,11 

the mom she has she had herself done bad 

‘the mother has hurt herself’ 

b. Il a- il a se caché Caroline 3;09,22 

he has he has himself hid 

‘he has hid himself’ 

They point out that this kind of placement error concerning the reflexive clitic se 

‘him/her/itself’ always cooccurs with the choice of the wrong auxiliary. Crysmann 

and Müller (2000) show that after age 4;04,4 Ivar stopped making the placement 

mistake and he also stopped using “avoir” with the reflexive clitic. Note here that 

Crysmann and Müller (2000) observe the mistake only arising with reflexive clitics 

and that they assume a principled reason for that. Indeed, they found the same 

mistake only involving reflexives also in an experiment of elicited production run 

with 6 different bilingual children (Crysmann and Müller (2000,227)). However, the 

reason cannot be too principled as the same error is also found with non reflexive 

object clitics as illustrated by the examples (17) and (18) above from bilingual 

Anouk. We also observe that such a misplacement concerning reflexives does not 

occur with monolinguals (Rasetti (2003,298) and Zesiger et al. (2006) on elicited 

production). Our early L2er Elisa produced the error with a reflexive accompanied 

by the wrong choice of auxiliary (3b), but also with an accusative clitic (3a), 

repeated below. 

(3a) ça a m’étranglé, Elisa 5;5 (repeated) 

that has me strangled 

‘that strangled me’


(3b) regarde, là j’ai m’étranglé Elisa 5;5 (repeated) 

look, there I have me strangled 

‘look, there I strangled myself’ 

The error has been discussed for adult L2 by, Granfeldt and Schlyter (2004) as 

mentioned in 4.1.3, example (6). These authors also quote (21) as such an error and 

point out that this type of error occurs later than the ‘*Cl in object position’ error 

illustrated in (15) and (16) for the same speakers. 

(21) il a lui assis (Petra, 7 months of exposure) 

he has him sat down 

‘he sat him down’ 

We also find this error in the adult L2 errors quoted from Connors and Nuckle 

(1986), Zobl (1980), Gundel and Tarone (1983), Grondin and White (1996), van der 

Linden (1985) by Hulk (2000), Towell and Hawkins (1994), and in Herschensohn 

(2004,224) it ranges between 25% and 29% of the relevant configuration, see (22) as 

an example. It also occurred in the grammaticality judgment task administered by 

Landow (2002) particularly in speakers with German as their source language. 

(22) Vous avez la pris Emma II, from Herschensohn (2004) 

you have her taken 

‘you have taken her’ 

In conclusion, we see that this error type does not occur with monolinguals or 

SLIs neither with accusative clitics nor with reflexives. In contrast, it does occur 

with both clitic types for some bilinguals, for others it only occurs with the reflexive, 

and for the bilingual child in two Romance languages it has not been observed. The 

error also occurs for our early L2 child (German/French), and is attested in adult L2, 

particularly in learners of French with German as their source language, and it also 

occurs in speakers with English as L1 (Herschensohn (2004)). This error may stem 

again from a misanalysis of the pronoun allowing for a uniform treatment of 

pronouns in both the learners’ languages, much as in the case of ‘*Cl in object 

position’ discussed in 4.2.2. 

4.2.4. Restructuring errors 

Errors of the type ‘*Cl Mod Infinitive’ have not been reported for monolinguals 

nor for children with SLI. We did not observe such errors in our early L2er either 

(see 4a,b) and can claim that in this respect she conforms to the French pattern from 

the beginning. Such errors have not been reported by White for her early L2 

children, either. 

As to bilinguals, the picture is somewhat mixed. Crysmann and Müller (2000) 

observe the absence of this error type in Ivar and Caroline, but among the errors 

67


quoted by Hulk (2000) for Anouk there are several examples of this kind, one with a 

modal (23a), one with savoir ‘know’ (23b) and one in a periphrastic construction 

with aller ‘go’ (23c). (23a,b) are from Anouk’s mother’s diary and occur at around 

4;06, the periphrastic example occurs earlier in the same file in which some ‘*Aux 

Cl PPart’ errors were found. 

(23) a. Je n’ en veux jamais manger Anouk 4;06 

I ne of that want never eat 

‘I do not ever want to eat that’ 

68 

b. Je le sais pas faire Anouk 4;06 

I it know not do 

‘I can’t do that’ 

c. Tu le va fermer Anouk 3;07,29 

you it will close 

‘you will close it’ 

We noted earlier that our bilingual child, Lorenzo, never misplaces his 

complement clitics. This is also true for Restructuring contexts, where errors could 

be expected through interference of his other language, Italian, which has ‘Clitic 

Climbing’, i.e. Cl-Mod-Inf structures. There is only one context of a clitic with a 

modal, but significantly this occurs early (first recording) and is correct (see (24)), 

thus indicating the same pattern as monolinguals also found, for this child, with 

other aspects of cliticization such as proclisis/enclisis, as discussed in Belletti and 

Hamann (2004,166). 

(24) No, c’est pas moi qui devrais l’amener Lorenzo 3;5 

No, it is not me who should it bring 

‘No, it’s not me who should have brought it’ 

The restructuring error has been found in adult L2 and we point here to the 

findings of Landow (2002) obtained through grammaticality judgments who 

observes that all of her speakers accepted examples (25a,b). This includes Spanish 

speakers for whom this could be interpreted as transfer because Spanish, like Italian, 

has Clitic Climbing: 5 out of 8 Spanish learners of French accepted (25a) and 4 out 

of 8 accepted (25b). But it also includes English speakers, speakers of German 

origin (who find the French order in their language) and Chinese speakers who do 

not have (clitic) pronouns at all. 28 

28 Interestingly, even speakers with the source language Brazilian Portugues accepted (25a,b), 

though Brazilian allows for the French order.

(25) a. *Il nous peut parler 

He us can talk 

‘he can talk to us’ 

b. *Je te veux aider 

I you want help 

‘I want to help you’ 


There are three things to be noted here. First, whereas adult L2ers systematically 

show this error, it is not necessarily documented for early L2. Second, in language 

combinations like Italian/French or Spanish/French we might expect interference or 

transfer in all learner modes. However, our bilingual French/Italian child did not 

produce this error, whereas it is systematically produced by Spanish adult L2 

learners of French, highlighting the difference in these learner modes. Third, adult 

L2 learners who do not have clitics nor Restructuring in their source language but 

also the bilingual child Anouk accept or produce such errors. As the erroneous order 

has neither been in the input nor in the other/source language, we speculate that at 

this point we see UG at work. Restructuring is a valid hypothesis that might be 

entertained once the learner has realized that the pronouns of the new/other language 

are not quite of the same type as in her other language. 29 

4.2.5. Aspects of the acquisition of Italian complement clitics 

In the discussion of error types we have up to now focused on French as the 

target language and have looked for the origin of certain errors in the possibly 

different systems of the source languages. In order to test our assumption about the 

role of the L2 system, we now turn to Italian as the target language, looking at 

learners with source languages considered also for French. We make reference here 

to the similarities and differences discussed in 2.3. which we summarize briefly in 

the following. 

French and Italian have complement clitics of the same nature which share 

morphosyntactic properties of various kinds the most important one being that 

pronominal complement clitcs are verbal clitics which cliticize as heads onto the 

(functional head containing the finite) verb. Different from French, however, Italian 

does not have conspicuous instances of weak pronouns comparable to French 

subject pronouns. 

Interestingly, we find that, despite their similar, possibly identical, nature, the 

acquisition of complement clitics appears to differ in part in Italian from what we 

29 In an elicited production task administered to adult L2 speakers of Italian with English and 

Spanish as the other languages, Bennati and Matteini (2006) found that the Clitic climbing 

option was entertained by both groups. 

69


have seen in the preceding sections on French. 30 We concentrate here on the 

placement errors discussed above. While monolingual and SLI acquisition of object 

clitics in both languages appears to be essentially faultless (as can be deduced from 

the literature; references throughout and, for Italian, Guasti (1992), Schaeffer (2000), 

Bottari et al. (1998)), abstracting away from differences in the delay of acquisition 

and omission rates, bilingual and early/adult L2 acquisition reveal an important 

difference: no placement errors have been so far documented in these modes of 

acquisition of Italian object clitics. Although the literature on the topic is not 

particularly rich, from what is known, no placement errors of the ‘Clitic in isolation’, 

‘Clitic in object position’ and of the ‘Aux CL Ppart’ kind are instantiated in bilingual 

and L2 Italian, neither in elicited production (e.g. Leonini & Belletti (2004)), nor in 

spontaneous production (Ferrari (2006)). 

Leonini and Belletti (2004) investigated 26 adult L2 speakers of Italian where 16 

speakers were of German origin, 3 were French, 2 were Polish, 1 was Dutch, 1 

Russian, 1 Greek, 1 Albanian and 1 was Bosnian. On average, these speakers supplied 

clitics at a rate of 39%, omitted clitics at a rate of 14%, and supplied lexical 

complements at a rate of 40% whereas the control group supplied clitics in 91% of 

the cases, used only 7.7% lexical complements and never omitted the complement. 

These errors strongly resemble our findings for French monolinguals and SLI 

children. They also correspond to some of the error types (omission and suppliance 

of lexical complements) found in our early and adult L2ers. However, Leonini and 

Belletti (2004) observe that no placement errors occurred 31 . 

As to early L2/bilingual speakers of Italian, Ferrari (2006) analyzed two German/ 

Italian bilingual children, Vincenzo (2;5-3;0) and Elisa (2;10-3;5), and found that 

the overuse of lexical complements is the most frequent error. Vincenzo supplied 

clitics in 57% of the cases, omitted clitics in 9% and used lexical complements in 

34% of the cases, her participant Elisa (not to be confused with the early German/ 

French L2 child we were investigating) supplied clitics in 59% of the cases, omitted 

30 See also Schmitz and Müller (in press) who point out a further distinction between the two 

languages showing up in both monolingual and bilingual acquisition, i.e. the fact that Italian 

object clitics seem to appear earlier in Italian than in French. See also results in Leonini 

(2006) on child monolingual acquisition of complement clitics in Italian, which are coherent 

with this conclusion. This is a potentially interesting asymmetry for which some of the 

considerations below and in section 5 may be relevant. 

31 Similar results are discussed in Leonini (2006) for adult L2 and monolingual child Italian. 

The absence of placement errors is confirmed in corpora of spontaneous production put 

together at the University of Siena. 

70


clitics in 12% and supplied lexical complements in 29% of the cases. Placement 

errors of the kind discussed for French did not occur. 32 . 

This difference in the typical error patterns in the acquisition of two closely 

related Romance languages like French and Italian might at first glance be surprising. 

We suggest, however, that it may not be due to hazard but principled once the role 

of L2 is taken into account. We speculated that the origin of the placement errors in 

early and adult L2 French could be the occasional misanalysis of the clitic as a 

weak/strong pronoun. This analysis, however, is not favored by the target language 

Italian which does not instantiate weak pronouns systematically. We will explore 

this idea further also for its theoretical ramifications concerning the comparative 

issue in the following, concluding, section. 

5. Concluding summary and remarks 

We recapitulate here the main findings of our overview, also introducing some 

further element of discussion. For complement clitics we have observed a delay in 

all modes of acquisition, with some variation in the omission rates across groups, 

thus highlighting an area of French grammar which is particularly hard to acquire. 

Adding the developmental perspective, it reveals that impairment gives rise to a 

longer delay. 

Focussing on placement errors, SLI patterns with L1 in not showing placement 

errors in spontaneous data. Placement errors are documented for L2 acquisition, 

early and adult, and for some bilinguals. We particularly discussed four error types: 

‘Clitic in isolation’, ‘Clitic in object position’, ‘Aux CL PPart’ and the Restructuring 

error. 

We propose that the ultimate reason for the detected placement errors of the first 

three kinds in bilingual and L2 French is to be recognized in an occasional 

misanalysis of the object clitic as a “weak” pronoun rather than a syntactic clitic (or 

possibly a “strong” pronoun for the ‘Clitic in isolation’ case). 

We further assume that this misanalysis be favored by the very existence of weak 

pronominal subject pronouns in French, usually referred to as subject clitics (see 

discussion in 2.1.1.), whose nature is more readily compatible with properties of the 

pronominal system of the other language of the learner (e.g. a Germanic language in 

the typical case discussed here). We submit that our L2 Italian data from 4.2.5. 

indirectly support this interpretation. Lack of an analogous misanalysis for Italian 

32 A peculiar misplacement occurred in Restructuring contexts, producing the order ‘*Mod Cl 

Vinf’, discussed in detail in Ferrari (2006), where it occurred at a rate of 63% for Vincenzo 

and a rate of 27% for Elisa. Ferrari interprets this error as stemming from Verb syntax rather 

than from a misanalysis of the Italian clitic pronouns by the two children. Especially 

significant in this connection is the lack of the ‘Aux Cl Ppart’ error in both these children. 

71


complement clitics may be expected as standard Italian lacks conspicuous instances 

of overt weak pronouns which could mislead the learner in the treatment of real 

syntactic (complement) clitics. If this proposal is on the right track, it suggests that 

both, properties of the target second language and of the first/other language may 

influence the analysis adopted by the learner. In such a case, the option shared by 

both languages and covering more data in the target second language will be 

privileged at some initial stage of acquisition. (See also the discussion along similar 

lines in Granfeldt and Schlyter (2004)). The interaction of properties of the 

grammatical systems of the source and the target second language thus plays a 

crucial role. Hence, this leads us to the expectation that placement errors should not 

be found in typical and atypical monolingual acquisition. This is precisely what the 

results from different sources systematically show. 

Our L2 and bilingual data from French and Italian – once taken as suitable 

theoretically relevant empirical evidence - indicate that L2/bilingual results may 

reveal subtle differences between closely related languages in closely related 

domains (see Belletti (2007) for relevant discussion). One such domain discussed 

here is the pronominal system of weak and clitic pronouns in French and Italian. The 

differences between the two pronominal systems are made more visible through the 

peculiar lenses of bilingual and L2 modes of acquisition which thus contribute a 

special means of comparison. Seen from this perspective, the discussed contrasts 

between French and Italian L2-data may supply a further indirect indication for an 

analysis of French subject pronouns as weak pronouns, not as syntactic clitics. 

The “Aux CL PPart” type error as well seems to identify a phase in L2 

acquisition where the hypothesis is being entertained that the complement clitic is a 

weak pronoun. We may further note here that the ‘Aux CL PPart’ which is found in 

the German child Elisa occurs in a phase where subject clitics are clearly acquired. 

Assuming that this means that their nature as weak pronouns is recognized, it could 

again be suggested that the same classification is tentatively extended and tried out 

for complement pronouns. As noted, this would allow Elisa to adopt the same 

analysis for both the French and the German pronoun systems. Arguably, this 

analysis constitutes an overall more economical UG option involving a less complex 

syntactic derivation (see 2.1.) which is directly prompted by properties of both the 

L1 and the L2. The fact that this error is also found in adult L2ers with English as 

L1 corroborates this interpretation as English also has weak pronouns. Again, the 

weak pronoun hypothesis may be entertained given the interaction of the two 

grammatical systems - L1 and L2. 

Clitic Climbing occurring in Restructuring contexts in French is (possibly) attested 

as a transfer phenomenon in adult L2, but interestingly it also occurs in adult L2 

speakers whose L1 does not have clitics, as mentioned in section 4.2.4. As there are 

cases reported in (23a,b,c)) also for the bilingual Dutch/French child, we suggest 

that the child and, for that matter, the adult L2 learners as well, are here trying out an 

72


option which is available through UG. As we are not aware of this type of error in 

L1 (a)typical monolingual acquisition of French, we further suggest that this could 

be a consequence of the fact that the bilingual setting provides an input which is at 

the same time richer and poorer than a monolingual one. On the one hand, there is 

arguably less input data for each single language. On the other hand, more UG 

options manifest themselves through the input data of the two (or more) languages. 

In consequence, different UG hypotheses are likely to be tried out more readily in 

these particular conditions of language acquisition. Note that this is not denying that 

also monolingual acquisition can involve stages corresponding to UG options other 

than the ones implemented in the target (L1) language (see Crain & Thornton 

(1998), and Rizzi (2005), UdDeen (2006), for recent discussion). The suggestion 

here is that the bilingual setting favors the “trying out” yet more. 

We conclude by highlighting the general features of our overview which has 

compared different modes of acquisition in the same empirical domain in French. 

We have suggested that the higher complexity of the syntactic derivation involving 

complement clitics as compared to subject pronouns is responsible for the attested 

delay in the acquisition of complement clitics in all modes of acquisition. We have 

also suggested that complement clitics may be analysed as weak pronouns in some 

modes (early and adult L2), leading to misplacement errors. The latter case is 

interpreted as motivated by two complementary economy considerations: i. the fact 

that the analysis allows for a uniform treatment of pronouns in both the languages 

involved; ii. the fact that the analysis implements a less complex syntactic derivation. 

Furthermore, we have proposed that, although limited in absolute number of instances, 

the very existence of misplacement errors can highlight the role of the different 

grammatical systems involved. 

Finally, the Restructuring error discussed appears to be particularly interesting in 

that it ultimately suggests a direct role of UG. Overall, the study of the various error 

types in the different acquisition modes of complement clitcs in French has offered a 

further novel illustration of the reciprocal contribution that linguistic theory and 

acquisition data can provide to each other to enhance their respective understanding. 

References 

Auger, J. (1995) “Les clitiques pronominaux en francais parlé informel: Un approche 

morphologique”, in Revue Québécoise de Linguistique 24, 21-60. 

Babyonyshev, M. and S. Marin (2004) “The acquisition of object clitic constructions in 

Romanian”, Selected Papers of LSRL 34 Amsterdam/Philadelphia: John Benjamins, 21- 

40. 

Baranzini, L. (2003) Le sujet nul, les infinitives racine et les questions Wh dans la grammaire des 

enfants normaux et des enfants SLI francophones, DEA-thesis, University of Geneva. 

Belletti, A. (1999) “Italian/Romance clitics: Structure and derivation”, in H. van Riemsdijk, ed., 

Clitics in the Languages of Europe, de Gruyter, Berlin, 543-579. 

73


Belletti, A. and C. Hamann (2004) “On the L2/bilingual acquisition of French by two young 

children with different source languages,” in P. Prévost and J. Paradis, eds., The 

Acquisition of French in Different Contexts, John Benjamins, Amsterdam/Philadelphia, 

147-174. 

Belletti (2006) to appear: “Kinds of evidence for linguistic theory”, in the Proceedings of the 17th 

Symposium on Theoretical and Applied Linguistics, Aristotle University, Thessaloniki, 

April 2005, pp. 285-303. 

Bennati, E. and S. Matteini (2006) “Object Clitic Climbing in L2 learners of Italian”, in Language 

Acquisition and Development. Proceedings of GALA 2005, Cambridge Scholars Press, 

35-48 

Bianchi, V. and C. Figueiredo (1994) “On Some Properties of Agreement-Object in Italian and in 

Brazilian Portuguese”, in M. Mazzola, ed., Issues and Theory in Romance Languages 

XXIII, Georgetown University Press. 

Bottari, P., P. Cipriani, A.M. Chilosi and L. Pfanner (1998) “The determiner system in a group of 

Italian children with SLI”, in Language Acquision 7, 285-315. 

Brandi, L and P. Cordin (1989) “Two Italian dialects and the null subject parameter”, in O. 

Jaeggli and K. Safir, eds., The Null Subject Parameter, Kluwer, Dordrecht, 111-142. 

Burzio, L. (1986) Italian Syntax, Reidel, Dordrecht. 

Cardinaletti, A. (1991) “On pronoun movement: The Italian dative loro”, Probus: 3.2, 127-153. 

Cardinalletti, A. and M. Starke (1999) “The typology of structural deficiency. A case study of the 

three classes of pronouns”, in H. van Riemsdijk, ed., Clitics in the Languages of Europe, 

de Gruyter, Berlin, 145-233. 

Cardinaletti, A. and M. Starke (2000) “Overview: the grammar (and acquisition) of clitics”, in S. 

Powers and C. Hamann, eds., The Acquisition of Scrambling and Cliticization, Kluwer, 

Dordrecht, 165-186 

Chillier Zesiger, L., Arabatzi, M., Cronel-Ohayen, S., Franck, J., FrauenFelder, U., Hamann, C., 

Rizzi, L. and Zesiger, P. 2006. The Acquisition and Mastery of French pronouns. 

University of Geneva, Manuscript. 

Cinque, G. (2004) “Restructuring and Functional Structure”, in A. Belletti, ed., Structures and 

Beyond. The Cartography of Syntactic Structure, Vol.3, Oxford University Press, New 

York, 132-191. 

Connors, K. and L. Nuckle (1986) “The morphosyntax of French personal pronouns and the 

acquisition/learning dichotomy”, in O. Jaeggli and C. Silva-Corvalan, Studies in Romance 

linguistics, Foris, Dordrecht, 225-242. 

Crain, S. and R. Thornton (1998): “Investigations in Universal Grammar: A guide to experiments 

on the Acquisition of Syntax and Semantics”, MIT Press, Cambridge 

Cronel-Ohayon, St. (2004) Etude longitudinal d’une population d’enfants dysphasiques, doctoral 

dissertation, University of Geneva. 

Crysmann, B. and N. Müller (2000) “On the Non-parallelism in the Acquisition of Reflexive and 

Non-reflexive Object Clitics”, in S. Powers and C. Hamann, eds., The Acquisition of 

Scrambling and Cliticization. Kluwer, Dortrecht, 207-236. 

74


Duffield, N., L. White, J. Bruh De Gavarito, S. Montrul and P. Prévost (2002) “Clitic Placement 

in L2 French: evidence from sentence matching”, in Journal of Linguistics 38, 487-525. 

Ferrari, I. (2006) “Acquisition of object clitics in Italian by two German/Italian bilingual 

children”, in Language Acquisition and Development. Proceedings of GALA 2005, 

Cambridge Scholars Press, 173-183. 

Friedemann, M.A. (1992) “The underlying position of external arguments in French: A study of 

adult and child grammar”, in GenGenP 1-2 , 115-122. 

Friedemann, M.-A. (1995) Sujets Syntaxiques: position, inversion et pro, doctoral thesis, 

University of Geneva. 

Genesee, F., E. Nicoladis and J. Paradis, (1995) “Language differentiation in early bilingual 

development”, in Journal of Child Language 22, 611-631. 

Granfeldt, J. and S. Schlyter (2004) “Cliticisation in the acquisition of French as L1 and L2”, in P. 

Prévost and J. Paradis, eds., The Acquisition of French in Different Contexts, John 

Benjamins, Amsterdam/Philadelphia, 333-369. 

Grondin, N. and L. White (1996) “Functional categories in child L2 acquisition of French”, in 

Language Acquisition 5(1), 1-34. 

Guasti, M. T. (1992) “Verb syntax in Italian child grammar”, in GenGenP 1-2, 145-162. 

Gundel, J. and E. Tarone (1983) “ ‘Language Transfer’ and the acquisition of pronominal 

anaphora”, in S. Gass and L. Selinker, eds., Language Transfer in language learning, 

Newbury House, Rowley (MA), 281-296. 

Haberzettl, S. (2005) Der Erwerb der Verbstellung in der Zweitsprache Deutsch durch Kinder mit 

russischer und türkischer Muttersprache, Max Niemeyer Verlag, Tübingen. 

Hakanson, G. and U. Nettelblatt (1996) “Similarities between SLI and L2 children. Evidence 

from the acquisition of Swedish word order”, in J. Gilbert and C. Johnson, eds., 

Children’s Language, Vol 9, Lawrence Erlbaum, Mahwah (NJ), 135-151. 

Hamann, C. (2002) From Syntax to Discourse: Pronominal Clitics, Null Subjects and Infinitives 

in Child Language, Kluwer, Dordrecht. 

Hamann, C. (2003) “Phenomena in French normal and impaired language acquisition and their 

implications for hypotheses on language development”, in Probus 15, 91-122. 

Hamann, C. (2004) “Comparing the development of the nominal and the verbal functional 

domain in French language Impairment”, in P. Prévost and J. Paradis, eds., The 

Acquisition of French in Different Contexts, John Benjamins, Amsterdam/Philadelphia, 

109-145. 

Hamann, C., S. Ohayon, S. Dubé, U. Frauenfelder, L. Rizzi, M. Starke and P. Zesiger, (2003) 

“Aspects of grammatical development in young French children with SLI”, in 

Developmental Science 6, 151-158. 

Hamann, C., L. Rizzi and U. Frauenfelder, (1996) “The acquisition of subject and object clitics in 

French”, in H. Clahsen, ed., Generative Approaches to Language Acquisition, 

Benjamin’s, Amsterdam, 309-334. 

Hawkins, R. (2001) Second Language Syntax: A Generative Introduction, Blackwell, Oxford. 

Hawkins, R. and F. Fanceschina, (2004) “Explaining the Acquisition or non-acquisition of 

determiner-noun gender concord in French and Spanish”, in P. Prevost and J. Paradis, 

75


eds., The Acquisition of French in Different Contexts, John Benjamins, Amsterdam/ 

Philadelphia, 175-207. 

Haznedar, B. (1997) “L2 Acquisition by a Turkish-Speaking Child: Evidence for L1-Influence”, 

in E. Hughes, M. Hughes and A. Greenhill, eds., Proceedings of BUCLD 21 st , Cascadilla 

Press, Somerville, Mass. 

Henry, C. (2006) Compensation chez les adolescents dysphasiques: analyse du langage spontané. 

MA thesis Université François-Rabelais, Tours. 

Herschensohn, J. (2004) “Functional categories and the acquisition of object clitics in L2 French”, 

in P. Prevost and J. Paradis, eds., The Acquisition of French in Different Contexts, John 

Benjamins, Amsterdam/Philadelphia, 207-242. 

Hulk, A. (1997) “The acquisition of object pronouns by a Dutch/French bilingual child”, in A. 

Sorace, C. Heycock, and R. Shillcock, eds., Proceedings of the GALA’97 Conference on 

Language Acquisition, 521-526. 

Hulk, A. (2000) “L’acquisition des pronoms clitiques en français par un enfant bilingue françaisnéerlandais”, 

in Canadian Journal of Linguistics 45, 97-117. 

Hulk, A. and N. Müller (2000) “Crosslinguistic influence at the interface between syntax and 

pragmatics”, in Bilingualism: Language and Cognition 3, 227-244. 

Jakubowicz, C. (2003) “Hypthèses psycholinguistiques sur la nature du déficit dysphasique”, in 

Ch. L. Gérard and V. Brun, (eds.), Les dysphasie, Masson, Paris 

Jakubowicz, C., N. Müller, O.-K. Kang, B. Riemer and C. Rigaut (1996) “On the Acquisition of 

the Pronominal System in French and German”, in A. Stringfellow et al., eds., BUCLD 

20, Cascadilla Press, Somerville (MA), 374-385. 

Jakubowicz, C., N. Müller, B. Riemer, and C. Rigaut (1997) “The Case of Subject and Object 

Omission in French and German”, in E. Hughes, M. Hughes and A. Greenhill, eds., 

BUCLD 21, Cascadilla Press, Somerville, MA, 331-342. 

Jakubowicz, C., L. Nash, C. Rigaut, Ch. L. Gérard (1998) “Determiners and Clitic Pronouns in 

French-Speaking Children with SLI”, in Language Acquisition 7, 113-160. 

Kaiser, G. (1994) “More about INFL-ection and Agreement: the Acquisition of Clitic Pronouns in 

French”, in J. M. Meisel, Bilingual First Language Acquisition French and German 

development, John Benjamins, Amsterdam, 131-160. 

Kayne, R. (1991) “Romance Clitics, Verb Movement and PRO”, in Linguistic Inquiry 22, 647- 

686. 

Laenzlinger, C. and U. Shlonsky, (1997) “Weak pronouns and LF clitics: Clustering and 

adjacency effects in pronominal systems of German and Hebrew”, in Studia Linguistica 

51, 2, 154-185. 

Landow, N. (2002) L’acquisition des clititques objets en français langue étrangère, Memoire de 

Licence, University of Geneva. 

Leonini, C. and A. Belletti (2004) “Adult L2 Acquisition of Italian Clitic Pronouns and “Subject 

Inversion”/VS Structures”, in J. van Kampen and S. Baauw, eds., “Proceedings of GALA 

2003”, 293-304, LOT – Occasional Series. 

76


Leonini, C. (2006) Object Clitics and Determiners in the Acquisition of Italian as L1 and L2 in 

Language Acquisition and Development. Proceedings of GALA 2005, Cambridge 

Scholars Press. 

Meisel, J. (1989) “Early differentiation of languages in bilingual children”, in Hyltenstam, K. and 

L. Obler, eds., Bilingualism across the Lifespan: Aspects of Acquisition, Maturity and 

Loss, CUP, Cambridge, 13-40. 

Meisel. J. (2006) “Child second language acquisition or successive first language acquisition”, 

ms., University of Hamburg. 

Paradis, J., M. Crago and F. Genesee (2003) “Object Clitics as a Clinical Marker of SLI in 

French: Evidence from French-English Bilingual Children”, in B. Beachley, A. Brown, 

and F. Conlin, eds., BUCLD 27, Cascadilla Press, Somerville (Mass), 638-649. 

Paradis, J. (2004) “The relevance of specific language impairment in understanding the role of 

transfer in second language acquisition”, in Applied Psycholinguistics 25, 27-82. 

Pirvulescu, M. (2006) “Theoretical implications of object clitic omission in Early French: 

Spontaneous vs. elicited production”, in Catalan Journal of Linguistics 5, 221-236. 

Poletto, C. (2000) The Higher Functional Field, Oxford University Press. 

Prévost, P. and L. White (2000) “Accounting for morphological variation in second language 

acquisition: truncation or missing inflection?”, in Friedemann, M.A. and L. Rizzi, eds., 

The Acquisition of Syntax, Longman, London, 202-235. 

Rasetti, L. (2003) Optional Categories in Early French Syntax: A developmental study of root 

infinitives and null arguments, Doctoral thesis, University of Geneva. 

Rizzi, L. (1978) “A Restructuring Rule in Italian Syntax”, in S. J. Keyser, ed., Recent 

Transformational Studies in European Languages, MIP Press, Cambridge (Mass), 113- 

158. 

Rizzi, L. (1982) Issues in Italian Syntax, Foris, Dordrecht. 

Rizzi, L. (1994) “Some notes on linguistic theory and language development: The case of root 

infinitives”, in Language Acquisition 3, 371-393. 

Rizzi, L. (2005) “Grammatically-Based Target-Inconsistencies in Child Language”, in K. 

UdDeen, J. Nomura, B. Schulz and B.D. Schwartz, eds., The Proceedings of the 

Inaugural Conference on Generative Approaches to Language Acquisition - North 

America (GALANA). UCONN / MIT Working Papers in Linguistics. MIT Press, 

Cambridge (MA). 

Salustri, M. (2003) The Onset of Bilingualism, PhD diss., University of Siena. 

Schaeffer, J. (2000) The Acquisition of Direct Object Scrambling and Clitic Placement, John 

Benjamins, Amsterdam/Philadelphia. 

Schmitz, K., K. and N. Müller (in press) “Strong and clitic pronouns in monolingual and bilingual 

first language acquisition: comparing French and Italian”, in Bilingualism: Language and 

Cognition. 

Schwartz, B. (1998) “The second language instinct”, in Lingua 106, 133-160. 

Schwartz, B. and R. Sprouse (1996) “L2 cognitive states and the full transfer/full access model”, 

in Second Language Research 12, 40-72. 

77


Sportiche, D. (1996) “Clitic constructions”, in J. Rooryck and L. Zaring, eds., Phrase Structure 

and the Lexicon, Kluwer, Dordrecht, 213-276. 

Towell, R. and R. Hawkins (1994) Approaches to Second Language Acquisition, Multilingual 

Matters, Clevedon. 

UdDeen, K. (2006) “It’s gotta be innate: a novel POS argument from Swahili”, in Language 

Acquisition and Development. Proceedings of GALA 2005, Cambridge Scholars Press. 

van der Linden, E. (1985) Toepassing van een regelsystem: de grammatika von het Frans, 

Doctoral thesis, University of Amsterdam. 

Wexler, K., A. Gavarró, V. Torrens (2004), “Feature Checking and Object Clitic Omission in 

Child Catalan and Spanish”, in R., Bok-Bennema, B. Hollebrandse, B. Kampers-Manhe 

and P. Sleeman, eds., Romance Languages and Linguistic Theory 2002, John Benjamins, 

Amsterdam, 253-270. 

White, L. (1989) Universal grammar and second language acquisition, John Benjamins, 

Amsterdam. 

White, L. (1996) “Clitics in L2 French”, in H. Clahsen, ed., Perspectives on Language 

Acquisition, Benjamin’s, Amsterdam, 335-368. 

White, L. (2000) “Second Language Acquisition: From Initial to Final State”, in J. Archibald, ed., 

Second Language Acquisition and Linguistic Theory, Blackwell, Boston, 130-155. 

White, L. (2003) Second Language Acquisition and Universal Grammar. CUP, Cambridge. 

Zobl, H. (1980) “The formal and developmental selectivity of L1 influence on L2 acquisition”, in 

Language Learning 30, 43-57. 

Zribi-Hertz, A. (1994) “Les clitiques nominatifs du francais”. Cahier de Recherche en 

Linguistique, Departement des science du langage, Université de Paris 8. 

Zubizarreta, M. L. (1985) “The relation between morphophonology and morphosyntax: The case 

of Romance causatives”, Linguistic Inquiry 16, 247-289. 

78

Rivista di Grammatica Generativa, 31 – 2006, 79 - 100 

THE TWO FORMS OF THE ADJECTIVE IN KOREAN * 


Soon Haeng Kang 

In Kang (2005) I discussed non-predicative adjectives like ceon ‘former’ in 

Korean arguing against the analysis that denies the existence of adjectives in 

Korean, given that they have been placed under the rubric of “attributive 

determinatives or adnominal modifiers” in the traditional grammar of Korean: 

limited in number, only used in attributive position (cf. Kim 2002). However I’ve 

proposed that these adjectives should be considered as attributive adjectives that 

have a direct modification source in the nominal phrase. In this paper, furthermore, I 

will show that they have other indirect modification counterparts, assuming the two 

typologies of the adjectives proposed by Cinque (2005b). 

The paper is organized as follows: in section 1, I will give a short overview of 

the properties of Korean attributive adjectives, which led some traditional Korean 

linguists to consider them as nouns. Section 2 shows that Korean attributive adjectives 

have their morphological counterparts and Section 3 discusses the distinction between 

adverbial and subject-oriented interpretations, attributive and predicative adjectives, 

assuming the proposal of Cinque (1994). In section 4, I will propose that some 

affixes are responsible for adjectival modification in the noun phrase in Korean. 

Section 5 demonstrates the contrasts between two types of adjectival modification in 

the NP, based on the Cinque (2005b)’s proposal for two origins of adjectives in the 

extended nominal projection. 

2. Some properties of Korean attributive adjectives: the traditional criteria 

Traditionally, attributive adjectives have been distinguished in two groups, 

depending on their ability to take the adverbial suffix, as illustrated in (1) (cf. Sohn 

1999, Mok 2002, among others). 

* I would like to express my gratitude to Guglielmo Cinque for very helpful critical comments 

and discussion. 

79


(1) a. *sun-lo 

pure-ly 

b. cwu-lo 

principal-ly 

c. chak-(u)lo 

book-with 

This shows that cwu ‘principal’ in (1b) can be inflected with the adverbial suffix 

-lo like the noun in (1c), contrary to sun ‘pure’ in (1a). According to this criteria, the 

lexical elements pertaining to the category of non-predicative adjectives as in (1a) 

are say ‘new’, hen ‘old, used’, yeys ‘old, antique’, while hyen ‘present’, ceon 

‘precedent’, cwucencek ‘alleged’ do not belong to this category. In fact, they are 

often classified as nouns, even if both are used only attributively as in (2). 1 

(2) a. ku sun sakykkwun 

That pure swindle 

‘Lit.: that pure swindle’ 

b. *sakykkwun-i sun-i-ta. 

Swindle-NOM pure-COP-DCL 

‘Lit.: a swindle is pure’ 

c. ku cen taytongreyng 

That former president 

‘That former president’ 

d. *ku taytongreyng-i cen-i-ta. 

That president-NOM former-COP-DCL 

‘Lit.: that president is a former’ 

However, this criterion is not fundamental in order to distinguish the two 

categories, since the adjective say ‘new’, considered as a non-predicative adjective, 

can be inflected with the adverbial suffix -lo, while hyen ‘present’ is not able to bear 

this suffix even if considered a noun. 2 

1 Abbreviations used in glosses: NOM ‘nominative’, ACC ‘accusative’, PST ‘past tense’, 

PRES ‘present tense’, PASS ‘passive marker’, AFF ‘affix’, COP ‘copular’, DCL ‘declarative’, 

GEN ‘genitive’, TOP ‘topic maker’. PLU ‘plural’, COM ‘complementizer’, REL ‘relative 

clause marker’. 

2 

The adjective present in English, can appear both in prenominal position and in postnominal 

position with a different meaning: 

(i) a. the present editors 

b. the editors present 

(Cinque 2005b,11) 

80

(3) a. say-lo ciun cip 

new-ly constructed house 

‘A house constructed newly’ 

b. *hyen-lo 

present-ly 

‘presently’ 

The two forms of the adjective in Korean 

Present in prenominal position possesses only a temporal meaning whereas in postnominal 

position it has only a locative value. The corresponding relative clause with present in 

predicate position has only one meaning, that of present in postnominal position, i.e., the 

locative one (see Cinque 2005b,11). The examples corresponding to (i) in Korean are realized 

with two different adjectives; one has a temporal meaning and the other has a locative value, 

both placed in prenominal position: 

(ii) a. hyen peyncipcatul 

present editors 

‘the present editors’ 

b. *peyncipcatul-i hyen-ta. 

editors -NOM present-DCL 

‘Lit.: the editors are present (temporal value)’ 

c. chamsekhan peyncipcatul 

present editors 

‘the editors present’ 

d. peyncipcatul-i chamsekha-ta. 

editors-NOM present-DCL 

‘The editors are present’ 

This is similar to the case of ‘live and alive’ in English, since live can be used in prenominal 

position as an attributive adjective, whereas alive only in postnominal position as a predicative 

one: 

(iii) a. The live/*alive animals 

b. The animals *live/alive (cf. the animals which are *live/alive) 

(Ibid.,11) 

Also the Korean corresponding adjectives nal ‘live’ e salaissnun ‘alive’ show the same pattern 

to the previous case: 

(iv) a. nal cymsungtul 

live animals 

‘The live animals’ 

b. salaissnun cymsungtul 

alive animals 

‘The animals alive’ 

81


Secondly, these adjectives have been considered to be prefixes (cf. Yu 1997, Lee 

2004). 3 Since they cannot be used alone, they are always in adjacent position to the 

nouns that follow them. However, this analysis does not seem to be correct, because 

other elements can be preceded by these attributive adjectives, for example, 

individual-level adjectives: 

3 In fact, Lee (2004, 3-11) inserts some lexical elements to the inside of the non-predicative 

adjectives in Korean such as woy ‘only’, peyng ‘ordinary’, am ‘female’, sus ‘male’. Other 

material cannot be intervened between these elements and the noun: 

(i) a. nay chakhan woy ttal 

my good only daughter 

b. * nay woy chakhan ttal 

my only good daughter 

‘My good only daughter’ 

(ii) a. kwunmeynhan peyng koysu 

diligent ordinary professor 

b. *peyng kwunmeynhan koysu 

ordinary diligent professor 

‘A/the diligent ordinary professor’ 

(iii) a. sanawun sus saca 

ferocious male lion 

b. *sus sanawun saca 

male ferocious lion 

‘A/the ferocious male lion’ 

Kang (2005,7 fn.9) suggests that the attributive adjectives cannot be followed by the nonintersective 

adjectives, considering them like a prefix, and for this reason the phrase (iv)b is 

considered ungrammatical: 

(iv) a. wytayhan cen tatongryung 

great former president 

b. *cen wytayhan tatongryung 

former great president 

‘??A/the great former president’ 

However I think that this analysis is not correct. Indeed, (iv)b is not ungrammatical. Therefore 

the hierarchy (v) for Korean adjectives proposed by Kang (2005) is not correct: 

(v) intersective adjectives > non-intersective adjectives > non-predicative adjectives > N 

One may claim that the basic order is (iv)b and the adjective wytayhan ‘great’ in (iv)a is in 

higher position in the SpecFocus. This would mean that the adjective wytayhan ‘great’ in 

(iv)a is not more non-intersective adjective. Indeed, in (iv)a, wytayhan ‘great’ can be arranged 

with the past morpheme - ess, contrary to that in (iv)b: 

(vi) a. wytayha-ess-ten cen tatongryung 

great-PST-REL former president 

b. cen wytayhan/(*-ess-ten) tatongryung 

former great /(-PST-REL) president 

‘Lit.: A/the former president that has been a great’ 

82

(4) a. ?say ppalgan cha (cf. ppalgan say cha) 

new red car 

‘A/the new red car’ 

b. ?say kun cha (cf. kun say cha) 

new big car 

‘A/the new big car’ 


Therefore, it would seem to be more coherent to consider all adjectives mentioned 

above as attributive ones like in other languages such as Italian and English. 

3. Introductory observations: another form of some attributive adjectives 

Another property of Korean non-predicative adjectives is that some of them have 

another form, depending on their typology of modification. For instance, there are 

two forms of the adjective corresponding to new in English: say and sayroun. These 

two forms can appear prenominally, but the only one, which can be found in 

predicative position, is a say-ro-un, in certain contexts. See the following example 

(the semantic difference between these two forms is shown in the English glosses): 

(5) a. say cha 

new car 

‘A new car; that has just been produced’ 

b. say-rou-n cha 

new car 

‘A new car; that it is a newer model with respect to the previous one’ 

c. i cha-ka design-eyse cen kes pota *say/say-rop-ta 

this car-NOM design-in former one more new-DCL 

‘This car is new with respect to the previous one in the design’ 

In (5b) and (5c), say-ro-u-n and say-rop are identical. It is only for phonetic 

reasons that the consonant -p is transformed into the vowel -u and then -n is added in 

attributive position. The example (5c) shows that only sayroun can appear in 

predicative position. Because of this, sayroun in (5b) could be considered as an 

indirect modification adjective deriving from a relative clause, contrary to say which 

has a direct modification source, since this form cannot appear in predicative 

position. 4 

4 A similar example for the adjective new can be found in Serbian/ Croatian/ Bosnian. These 

languages have two types of adjectives: short-form and long-form adjectives. Long-form 

adjectives have a direct modification origin, but short-form adjectives have a relative clause 

origin (indirect modification) since syntactically both types can appear prenominally as in (i), 

83


However, it is more complicated to determine the origin of an adjective from its 

form, for example, in the case of adjectives like chwucengcek ‘alleged’, sun ‘pure’, 

cwu ‘principal’ which have another form: chwucengcek-i-n, 5 sun-cen-ha-n ‘mere, 

lit.; pure and complete’, 6 dansun-ha-n ‘simple’, lit.: ‘only and pure’, cwuyo-ha-n 

‘principal’, lit.: ‘principal and important’ respectively. 7 

Consider two forms of these adjectives: 

(6) a. chwucengcek/chwucengcek-i-n cangkowan 

alleged/alleged-COP-N minister 

‘A/the alleged minister’ 

b. *ku cangkowan-un chwucengcek/chwcengcek-i—ta 

that minister-TOP alleged/alleged-COP-DCL 

‘Lit.: that minister is alleged’ 

(7) a. sun/dansun-ha-n wuyen 

pure/simple-AFF-N coincidence 

‘A pure coincidence’ 

b. *wuyen-un sun/dansun-ha-ta 

coincidence-TOP pure/simple-AFF-DCL 

‘Lit.: A/the coincidence is pure’ 

but in predicate position, only the short-form adjectives can appear, as in (ii). (in Cinque 

2005b,18): 

(i) a. nov/ novi kaput 

new (short)/new (long) coat 

‘A/the new coat’ 

(ii) a. Njegov kaput je nov/*novi 

his coat is new (short)/*new (long) 

5 This form is formulated by adding the Korean copular –i and the attributive adjective marker 

-n to the basic form. 

6 -ha is an affix and -n is an attributive maker. 

7 According to Alexiadou (2005,5), if a language has two adjectival patterns, such adjectives 

are banned from one of these patterns, as illustrated in (i): 

(i) a. *o ipurgos o proin (Greek) 

the former the minister 

b. *z bylyn rozmawialan slawnyn aktorem (Polish) 

with former talked famous actor 

c. *qian-de zongtong (Chinese) 

former-DE president 

Across languages adjectives such as alleged, former, mere cannot appear in predicative 

position (also see Cinque (2005b)). 

84


Even if these adjectives have different forms, their typologies of modification do 

not vary, since both forms only appear in attributive position. How could the 

differences between the forms of these adjectives be defined? In the next section I 

will continue to characterize their syntactic and semantic properties, introducing the 

distinction between attributive adjectives and predicative adjectives formulated by 

Cinque (1994), and between two sources of adjectival modifications in the extended 

nominal phrase proposed by Cinque (2005b). 

4. Distinction between two types of adjectives 

4.1. Attributive: adverbial vs. subject-oriented 

First of all, let me consider the adjective chwucengcek and chwcengcek-i-n, both 

corresponding to alleged in English 8 . In the same way, also the adjective yamancek 

‘brutal’ can have another form: yamancek-i-n. Both can appear in prenominal position: 

(8) yamancek/yamancek-i-n kongkeyk 

brutal/brutal-COP-N aggression 

‘A/the brutal aggression’ 

In order to explain this phenomenon, we can introduce the adjective brutale 

‘brutal’ in Italian, which can appear before as well as after the noun, but not after the 

complement as in (9). 

8 In fact, Kim (2002, 35 fn.10) notes that modal adjectives (e.g., alleged) in Korean are absent 

and illustrates this with the following example: 

(i) [salinca-lo hyemuy(-lul) pat]-nun salam 

murder-as suspicion (-ACC) received-REL person 

‘A person alleged to be a murder’ (Lit.: ‘A murderer who received suspicion’) 

According to Kim, the adjective corresponding to alleged in (i) is a relative clause with a 

clausal predicate inside. However, we can also say as in (ii). 

(ii) chwucengcek salinca 

alleged murder 

‘A/the alleged murder’ 

The difference between (i) and (ii) can be detected if we compare them with the two uses of 

alleged in (iii). 

(iii) a. The alleged murderer was deported. 

b. The murderer alleged to have killed his own parents was deported. 

(Cinque 2005b,45) 

Alleged in prenominal position in (iii)a has a non-intersective value, but in postnominal 

position with a phrasal complement, it becomes intersective (see Cinque 2005b,45 for the 

further discussion). The same is true for Korean. 

85


(9) a. La loro brutale aggressione all’Albania (subject-oriented) 

b. La loro aggressione brutale all’Albania (manner) 

c. *La loro aggressione all’Albania brutale 

‘Their brutal aggression against Albania’ 

(Cinque 1994, 88-89) 

According to Cinque, the prenominal and postnominal positions of attributive 

adjectives receive two different interpretations: the postnominal position receives a 

strict “manner” interpretation like (9b), whereas the prenominal one has a “subject 

oriented” interpretation as in (9a). Cinque proposes the partial structure containing a 

speaker-oriented adjective as in (10). 

(10) a. La probabile(sp-or) goffa(subj-or) reazione immediata (manner) 

alla tua lettera 

‘The probable clumsy reaction immediate to your letter’ 

b. [ XP APsp-or _ [ YP APsubj-or_[ ZP APmanner/themat_ [ NP N …]]]] 

(Ibid.,92) 

Assuming the distinction between manner adjectives and subject-oriented 

adjectives let me consider yamancek and yamancek-i-n ‘brutal’ in Korean. Both 

forms can appear between the subjective genitive and the noun as in (11). But their 

interpretations are different, in that the one in (11a) receive the manner value, 

whereas the one in (11b) receive subject-oriented value. 

(11) a. kutul-uy yamancek kongkeyk 

they-GEN brutal aggression 

b. kutul-uy yamancek-i-n kongkeyk 

they -GEN brutal-COP-N aggression 

‘Their brutal aggression’ 

In addition, they do not appear in the left peripheral position of NP, since they 

cannot modify the nominal head: 9 

(12) a. *yamancek kutul-uy kongkeyk 

brutal they -GEN aggression 

b. *yamancek-i-n kutul-uy kongkeyk 

brutal-COP-N they -GEN aggression 

‘Their brutal aggression’ 

9 In (11), it is not important that yamancek and yamancek-i-n ‘brutal’ modify kutuluy ‘loro’. 

86


This distinction between subject-oriented and manner interpretation for the two 

forms of the adjective brutal in Korean is supported by the fact that the mutual order 

between these two types of adjectives follows that of Italian. That is, the manner 

adjective yamancek ‘brutal’ can only follow the subject-oriented adjective such as 

erisekun ‘stupid’. The examples (14) are parallel to (13) in Italian: 

(13) a. *L’aggressione stupida brutale/italiana all’Albania 

the aggression stupid brutal/Italian against Albania 

b. La stupida aggressione brutale/italiana all’Albania 

the stupid aggression brutal/Italian against Albania 

‘The stupid brutal Italian aggression against Albania’ 

(Ibid., 91) 

(14) a. Albania-eytayhan kuyul-uy erisekun yamancek kongkeyk 

Albania-against they-GEN stupid brutal aggression 


b. *Albania-eytayhan kutul-uy yamancek erisekun kongkeyk 

Albania-against they-GEN brutal stupid aggression 


This evidence suggests that in Korean, the manner adjectives and the subjectoriented 

adjectives are distinguished morphologically. Moreover the speaker-oriented 

adjectives cannot be preceded by subject-oriented adjectives in the way illustrated in 

example (15). Therefore, the hierarchy between these adjectives would be that of 

Cinque in (10b): 

(15) a. yesangkanunghan eykeyun panwung 

probable clumsy reaction 

‘The probable clumsy reaction’ 

b. * eykeyun yesangkanunghan panwung 

clumsy probable reaction 

‘The probable clumsy reaction’ 

c. ne-uy peynci-ey yesangkanunghan erisekun cwkkakcek-(*-n) pangung 

you-GEN letter-to probable clumsy immediate-(COP-N) reaction 

‘The probable clumsy immediate reaction to your letter’ 

(16) [ XP APsp-or _ [ YP APsubj-or_[ ZP APmanner/themat_ [ NP N …]]]] 

87


4.2. Predicative 

According to Cinque (1994,92), the attributive adjectives in the Romance 

languages precede a noun or appear between the noun and its complement, not after 

the complement (cf. 9c). However, they become grammatical if there is a sharp 

intonational break between the complement and the AP, with the AP bearing stress 

(17a), or coordinated (17b), or modified by a specifier (17c) or a complement (17d): 

(17) a. La loro aggressione all’Albania, BRUTALE 

loro aggressione all’Albania, improvvisa e brutale 

La loro aggressione all’Albania, assai poco brutale 

La loro aggressione all’Albania, brutale nei suoi effetti 

(Ibid.,92) 

Assuming this analysis, let me consider the adjective yamancek, yamancek-in 

which both correspond to brutal in English. The construction (18) is parallel to (17). 

(18) a. AMANCEK-*(I-N), kutul-uy Albania-eytayhan kongkeyk 

brutal-COP-N they-GEN Albania-against aggression 

‘Lit.: Their aggression against Albania, BRUTAL’ 

b. yamncek-*(i)-ko kapcaksure-n kutul-uy Albania-eytayhan kongkeyk 

brutal-COP-and unexpected-N they-GEN Albania-against aggression 

‘Lit.: Their aggression against Albania, brutal and unexpected’ 

c. maywu yamancek-*(i-n) kutul-uy Albania-eytayhan kongkeyk 

very brutal-COP-N they-GEN Albania-against aggression 

‘Lit.: Their aggression against Albania, very little brutal’ 

d. hwoykwoa-eyse yamancek-*(i-n) kutul-uy Albania-eytayhan kongkeyk 

effects -in brutal-COP-N they-GEN Albania-against aggression 

‘Lit.: Their aggression against Albania, brutal in its effects’ 

This demonstrates the contrast between the two forms of the adjective brutal in 

Korean. When it is preceded by a pause and emphasized in the left peripheral 

position (18a), coordinated (18b), or modified by a specifier (18c) or a complement 

(18d), it only becomes grammatical with the adjective containing the copula -i. For 

the attributive adjectives, not only the distinction between adverbial and subjectoriented 

value, but also the distinction between attributive and predicative is of a 

morphological nature. 10 

10 This is limited to the adjectives that have the suffix –cek. Adjectives that do not have the 

suffix -cek, for example yengakhan ‘clever’ and suncynhan ‘naïve’, are subject to interpretations 

determined by their structural positions (ii), as in the Italian phrase (i): 

88


Therefore we would say that yamancek ‘brutal’ can be adverbial or attributive, 

and that yamancek-i-n, which contains the copula -i, can be subject-oriented or 

predicative. 

And this analysis could be applied to other adjectives such as sun ‘pure’ and cwu 

‘principal’ that both have another form added –ha: sun-cen-ha-n, dansun-ha-n, and 

cwuyo-ha-n respectedly. 

In Cinque (2005b,43), some adjectives that are found exclusively in prenominal 

position can appear after a complement in predicative position if they are coordinated 

with other adjectives: 

(19) a. *una coincidenza pura (cf. una pura coincidenza) 

a coincidence pure 

‘a pure coincidence’ 

b. una coincidenza pura e semplice 

a coincidence pure and simple 

‘a coincidence pure and simple’ 

According to Cinque, when coordinated, the adjectives count as “heavy” and can 

thus Access the Spec of a higher FocP, ending up in postnominal position. 

Assuming this analysis of Cinque’s, we can apply it to the adjectives in Korean. 

The adjectives sun ‘pure’ and cwu ‘principal’ for themselves cannot be coordinated 

with other adjectives, similar to the fact that only the form containing the copula -i, 

yamancek-i-n, can be coordinated with other adjectives such as yamancek-i-ko 

kapcakswuren ‘brutal and unexpected’ (cf. 18b). Similarly, the forms of compounds 

dan-sun ‘only and pure’ and cwu-yo ‘principal and important’ can bear a affix as 

–ha, that has the same function of the copula -i, contrary to sun ‘pure’ and cwu 

‘principal’. 

(i) a. L’astuta risposta ingenua di Gianni 

the clever answer naïve of John 

‘John’s clever naive answer’ 

L’ingenua risposta astuta di Gianni 

The naïve answer clever of John 

‘John’s naive clever answer’ 

(Crisma 1996, 65) 

(ii) a. John-uy yengakhan suncynhan taytap 

John-GEN clever naïve answer 

‘John’s clever naive answer’ 

b. John-uy suncynhan yengakhan taytap 

John-GEN naive clever answer 

‘John’s naive clever answer’ 

89


However, if there is an intonational break and an emphasis in left peripheral 

position (20), or a modification by a specifier (21), only the forms with –ha can 

appear: 

(20) a. *SUN/DAN-SUN-HA-N, ku wuyen 

pure/only e pure-HA-N that coincidence 

‘Lit.: That coincidence, only and pure’ 

b. *CWU/CWU-YO-HA-N, ku dongki 

principal/principal e important-HA-N that motive 

‘Lit.: That motive, principal and important’ 

(21) a. nemwu *sun/dan-sun-ha-n wuyen 

very much pure/ only e pure-HA-N coincidence 

‘Lit.: A/the coincidence, very much only and pure’ 

b. nemwu *cwu/cwuyo-ha-n dongki 

very much principal e important-HA-N motive 

‘Lit.: A/the motive, very much principal and important’ 

5. Implication: two types of affix 

5.1. Direct modification: -CEK 

In the previous section, I have shown the presence of the adverbial adjectives in 

Korean: chwucengcek ‘alleged’, yamancek ‘brutal’. These adjectives have the 

morphological similarity, that is, they have the suffix -cek, which is obligatorily 

added to the elements that precede the nouns: 

(22) a. chwuceng-cek/*chwuceng cangkwon 

alleged/allegation minister 

‘A/the alleged minister’ 

b. yaman-cek/*yaman kongkeyk 

brutal/brutality aggression 

‘A/the brutal aggression’ 

90 

In effects, there are some adjectives that share this suffix -cek, as shown in (23): 

(23) a. kwahak-*(cek) selmyung 11 

science-CEK explanation 

11 Without -cek, the noun kwahak ‘science’ would be a internal argument of a head noun 

selmyung ‘explanation’.

‘A/the scientific explanation’ 

b. cwusang-* (cek) sayngkak 

abstraction-CEK idea 

‘A/the abstract idea’ 

c. kensel-*(cek) mannam 

construction-CEK encounter 

‘A/the constructive encounter’ 


The adjectives of the example (23) cannot appear in predicative position, as seen 

in the example (24). 

(24) a. * selmyung-I kwahak-cek-i-ta. 

explanation-NOM science-CEK-COP-DCL 

‘A/the explanation is scientific’ 

b. *sayngkak-i chwsang-cek-i-ta 

idea-NOM abstraction-CEK-COP-DCL 

‘A/the idea is abstract’ 

c. *mannam-i kensel-cek-i-ta 

encounter-NOM construction-CEK-COP-DCL 

‘A/the encounter is constructive’ 

As I have suggested before with respect to the adjective yaman-cek ‘brutal’, the 

adjectives in (24) are adverbial or non-predicative, since they cannot appear in a 

predicative position. Furthermore, they cannot have a FOCUS feature (25a), nor can 

they be found in a left peripheral position with the pause emphasized (25b). 

Moreover they cannot be coordinated with other adjectives (25c), and in case they 

are followed by a specifier (25d) or a complement (25e), they also become 

ungrammatical: 

(25) a. *ku-uy kwahák-cek selmyung 

he-GEN science-CEK explanation 

‘His scientific explanation’ 

b. *KWAHAK-CEK ku-uy selmyung 

science-CEK he-GEN explanation 

‘Lit.: his explanation, SCIENTIFIC’ 

c. *ku-uy kwahakcek-ko sylcaycek selmyung 

he-Gen science-CEK-and logical explanation 

‘Lit.: the explanation, scientific and logical’ 

d. *maywu chwsang-cek sayngkak 

very abstraction-CEK idea 

91


92 

‘Lit.: the idea, very abstract’ 

e. * keylkwoa-meyneyse kenselcek mannam 

effects-in construction-CEK encounter 

‘Lit.: the encounter, constructive in its effects’ 

Therefore, summarising the properties of the adjectives with the suffix -cek: 

1. They have an adverbial interpretation. 

2. They are not able to bear the focus feature. 

3. They cannot appear in the left peripheral position, even if there is an 

intonational break and they bear stress. 

4. They cannot be modified by a specifier or a complement. 

5. They can only be attributive. 

They have a direct modification source. 

5.2. Indirect modification: -I and -HA 

5.2.1. -I 

All the adjectives with -cek can have other forms containing the copula -i in 

attributive position, such as the adjective chwucengcek, chwucengcek-i-n ‘alleged’. 

Therefore also the adjectives in (23) such as kwahakcek ‘scientific’, cwusancek 

‘abstract’ and kenselcek ‘constructive’, have another form with the copula -i in 

attributive position (26). Despite the presence of the copular –i, they do not appear 

predicatively as shown in (27): 

(26) a. kwahakcek-(i-n) selmyung 

scientific-COP-N explanation 

‘A/the scientific explanation’ 

b. cwusangcek-(i-n) sayngkak 

abstract-COP-N idea 

‘A/the abstract idea’ 

c. kenselcek-(i-n) mannam 

construction-COP-N encounter 


(27) a. * selmyung-i kwahakcek-i-ta. 

explanation-NOM scientific-COP-DCL 

‘A/the explanation is scientific’ 

b. *sayngkak-i chwsangcek-i-ta 

idea-NOM abstract-COP-DCL

‘A/the idea is abstract’ 

c. *mannam-i kenselcek-i-ta 

encounter-NOM constructive-COP-DCL 



This would suggest that the adjectives with the copula -i are not derived from 

relative clauses, since their predicative use turns out to be ungrammatical. However, 

they can have a subject-oriented interpretation, contrary to the forms without the 

copular –i as apparent from the semantic contrast of the adjective yamancek, 

yamancek-i-n ‘brutal’ seen in 3.1. In other words, the distinction between adverbial 

adjectives and subject-oriented ones in Korean is morphological. For this reason, the 

adjectives in (27) can appear in predicative position, having a subject-oriented value: 

(28) a. ku-uy selmyung-un kwahakcek-i-ess-ta. 

he-GEN explanation-TOP scientific-COP-PST-DCL 

‘Lit.: His explanation has been scientific’ 

b. Ku-uy sayngkak-un chwsangcek-i-ess-ta. 

he-Gen idea-TOP abstact-COP-PST-DCL 

‘His idea has been abstract’ 

c. Kutul-uy mannam-un kensel-cek-i-ess-ta. 

they-GEN encounter-TOP constructive-COP-PST-DCL 

‘Their encounter has been constructive’ 

Moreover, this type of adjective would be predicative, given their possibility of 

bearing elements that render them heavy: focus (29a), coordination (29b), pause and 

emphasis (29c), and finally modification by a specifier (29d) or a complement (29e): 

(29) a. ku-uy kwahakcek-í-n selmyung 

he-GEN scientific-COP-N explanation 

‘His scientific explanation’ 

b. kwahakcek-i-ko nonrycek-i-n ku-uy selmyung 

scientific-COP-and logical-COP-N he-GEN explanation 

‘Lit.: His explanation, scientific and logical’ 

c. *KWAHAKCEK-I-N ku-uy selmyung 

scientific-COP-N he-GEN explanation 

‘Lit.: His explanation, SCIENTIFIC’ 

d. ku-uy maywu chwsangcek-i-n sayngkak 

he-GEN very abstract-COP-N idea 

‘Lit.: His idea, very abstract’ 

e. keylkwoa-meyneyse kenselcek-i-n mannam 

93


94 

effects-in constructive-COP-N encounter 

‘Lit.: the constructive encounter in effects’ 

This shows the contrast between the adjectives which share the suffix -cek with 

or without the copular –i. In the next section, I would like to characterize the 

properties of the affix –ha. 

5.2.2. -HA 

Let us consider the status of –ha, which is present in adjectives such as dansunha-n/suncen-ha-n 

and cwuyo-ha-n (derived from the adjectives sun ‘pure’ and cwu 

‘principal’ respectively). First of all, the affix -ha is in complementary distribution 

with the copular -i: 

(30) a. dansun-ha/*i-ta 

simple (only and pure)-HA-DCL 

‘Lit.: is simple’ 

b. yamancek-i/*ha-ta 

brutal-COP-DCL 

‘Lit.: is brutal’ 

This suggests that –ha has the same function as the copular affix –i. Indeed, there 

are some adjectives such as kinkup/kinkup-ha-n ‘urgent’, yumyeng/yumyeng-ha-n 

‘famous’. 12 In attributive position, the presence of –ha is optional, as in (31). In 

predicative position, on the other hand, –ha is obligatory, as shown in (32). 13 

(31) a. kinkwup-(ha-n) sanghwang 

urgent-HA-N situation 

‘A/the urgent situation’ 

b. yumyeng-(ha-n) paywu 

famous-HA-N actor 

‘A/the famous actor’ 

(32) a. ku sanghwang-un kinkwup-*(ha)-ta. 

that situation-TOP urgent-HA-DCL 

‘That urgent situation’ 

b. ku paywu-nun yumyeng-*(ha)-ta. 

12 Originally, this idea has been proposed by Mok (2002,15). 

13 This type of adjectives are cwungyo/cwungyo-ha-n ‘important’, ketay/ketatay-ha-n 

‘enormous’, kangryek/kangrye-ha-n ‘strong’… etc.

that actor-TOP famous-HA-DCL 

‘That actor is famous’ 


One characteristic of the affix –ha is that it makes the adjective predicative. This 

means that the coordination, focus, the modification by a specifier and emphasis are 

admitted with adjectives containing –ha: 

(33) a. ku kinkwup-há-n sanghwang 

that urgent-HA-N situation 

‘A/the urgent situation’ 

b. KINKWUP-*(HA-N) ku sanghwang 

urgent-HA-N that situation 

‘Lit.: a/the situation, URGENT’ 

c. ku maywu yumyeng-*(ha-n) paywu 

that very famous-HA-N actor 

‘That very famous actor’ 

This contrast between the adjectives with or without –ha could be supported by 

the fact that the positions of generation of the two forms of the adjective famous in 

Korean could be revealed through their relative ordering with respect to another 

adjective (for example, celmun ‘young’): 

(34) a. ku-nun maywu celmun yumyung-(*ha-n) cakka-i-ta. 

he-TOP very young famous-HA-N writer-COP-DCL 

‘He is a very young famous writer’ 

b. ku-nun maywu yumyung-*(ha-n) celmun cakka-i-ta. 

he-TOP very famous-HA-N young riter-COP-DCL 

‘He is a very famous young writer’ 

This example demonstrates that the adjective with –ha, yumyeng-ha-n ‘famous’ 

can only appear in the higher position relative to the adjective celmun ‘young’, 

contrary to its counterpart without –ha, as in (34a). On the other hand, in the lower 

position relative to the adjective celmun ‘young’, only the form without –ha can 

appear. 

To summarize, in section 4.2, I have argued that the adjectives which occur with 

the copular –i or the affix –ha can be predicative and also have an indirect 

modification origin. 

6. Contrasts between two types of modification 

In the present section, I would like to discuss some evidence supporting the 

conclusion that Korean adjectives are of two types giving rise to two types of 

95


interpretations. I will illustrate the contrasts between stage-level and individuallevel, 

between absolute and relative, between specificity-inducing and nonspecificity-inducing 

with reference to the theory of the two origins of adjectival 

modification proposed in Cinque (2005b). 

6.1. Individual-level vs. stage-level 

In Larson (2000b) and Cinque (2005b), adjectives that have a direct modification 

origin are taken to possess a characteristic or enduring (individual-level) property, 

whereas those that have an indirect modification origin are taken to manifest a 

temporary (stage-level) property. According to Cinque, this difference between the 

two origins of the adjectives is systematic in the Germanic and Romance languages: 

that is, it depends on their pre- and postnominal position, as illustrated in (35) and 

(36). 

(35) a. The visible stars include Aldebaran and Sirius. (ambiguous) 

b. The stars visible include Aldebaran and Sirius. (S-level) 

(36) a. Le invisibili stelle di Andromeda sono molto distanti. (I-level) 

‘A’s stars, which are generally invisible, are very far’ 

b. Le stelle invisibili di Andromeda sono molto distanti. (ambiguous) 

‘A’s stars, which are generally invisible, are very far’ (I-level) 

‘or A’s stars, which are generally visible, but which happen to be 

invisible now…’ (S-level) 

(Cinque 2005b,4-5) 

This shows that in the prenominal position English adjectives introduce 

ambiguity between individual-level and stage-level as in (35a), while in the 

postnominal position they are not ambiguous, they denote only stage-level property 

as in (35b). And in Italian, contrary to English, adjectives in prenominal position can 

have only an individual-level interpretation (36a), while in the postnominal position 

both readings are available (36b). In the case of Korean, the corresponding adjective 

kasicek ‘visible’, which contains –cek, is a direct modification marker for adjectives. 

Therefore, an individual-level reading is available in (37a), while the temporal 

property or the stage-level value can be expressed by another form; kasicek-i-n 

‘visible’ containing the copula -i, as shown in (37). 14 

14 There is a another possibility to express both individual-level and stage-level reading with 

the adjective poinun ‘visible’, containing the present tense marker -nun. This adjective poinun 

‘visible’ would be identical to the reduced relative clause with the temporary property as in 

(ib), while the enduring property expressed with a support of an adverbial hangsang ‘always’ 

as in (ic). 

96


(37) a. Andromeda-nun kasicek pyel-i-ta. (I-evel) 

Aldebaran-TOP visible stras-COP-DCL 

‘The Aldebaran is a visible star’ 

b. Andromeda-nun kasicek-i-n pyel-i-ta. (S-level) 

Aldebaran-TOP visible-COP-N star-COP-DCL 

‘The Aldebaran is a star, which is visible’ 

In addition, these two types of adjective visible in Korean respect the order “Slevel 

> I-level > N” or “indirect modification > direct modification > N” as in (38b). 

And pikasycek-i-n ‘invisibile’, the type of S-level adjective, can only be followed by 

modun ‘every’ through a focus (or emphasis) movement as in (38c): 15 

(38) a. *modun kasicek pikasycek-i-n peyl-tul 

every visible invisible-COP-N stars-PUL 

‘Every visible star invisible’ 

b. modun pikasicek-i-n kasicek peyl-tul 

every invisible-COP-N visibile stars-PLU 

‘every visible stars invisible’ 

c. pikasicek-*(i-n) modun peyl-tul 

invisible-COP-N every stars-PLU 

‘Every visible stars, INVISIBLE’ 

(i) a. Tosekwoan-eyse kongpuha-nun haksayng 

library-in student-PRES student 

‘A/the student who is studying in the library’ 

b. (hanwl-ey) po-i-nun peyl-tul 

sky-in see-PASS-PRES star-PLU 

‘The stars which are seen in the sky now’ 

c. hangsang po-i-nun peyl-tul 

always see-PASS-PRES star-PLU. 

‘The stars which are always seen’ 

15 It is also possible to see the order “S-level > I-level > N” with the adjective poinun 

‘visible’, derived from the reduced relative clauses: 

(i) a. (cikum) an-po-i-nun hangsang po-i-nun peyl-tul 

now not-see-PASS-PRES always see-PASS-PRES star-PLU 

‘The stars that are visible always, invisible’ 

b. * hangsang poinun an-po-i-nun peyl-tul 

always see-PASS-PRES not-see-PASS-PRES star-PLU 

‘Lit.: The stars that are invisible, always invisible’ 

With regard to the order between the relative clauses, see Larson and Takahashi (2004), also 

Cinque (2005b). 

97


6.2. Absolute vs. relative 

In Cinque (Ibid., 16), an adjective like enormous in English is ambiguous between 

relative absolute as in (39): ‘an elephant which was an enormous hing, in absolute 

terms’ (in absolute terms) or ‘an elephant which was enormous with respect to other 

individuals of the same class’: 

(39) (She saw) an elephant which was enormous 

The adjective that corresponds to enormous in Korean is ketay-ha-n and it can 

also introduce ambiguity between these two types of interpretations, as in the 

example (40a), however its counterpart without –ha: ketay ‘enormous’, can have 

only absolute value as in (40b), similarly to the analysis of Cinque. That is, the 

adjectives that have a direct modification source, can have only one absolute value: 

(40) a. kuney-nun ketay-ha-n kokkyri-lul po-ass-ta. (ambiguous) 

She-Top enormous-HA-N elephant-ACC see-PST-DCL 

‘She saw an elephant which was enormous’ 

b. kuney-nun ketay kokkyri-lul po-ass-ta. (absolute) 

She-Top enormous elephant-ACC see-PST-DCL 

‘She saw an elephant which was enormous thing’ 

6.3. Specificity-inducing vs. non-specificity–inducing 

Also in Cinque (Ibid., 8-9), the adjectives with origin of indirect modification 

possess ambiguous interpretations between specific and not-specific, as in the 

example (41), on the other hand, those with origin of direct modification can have 

only a specific interpretation, as in (41b): 16 

(41) a. Mi hanno detto che alla festa interverrà un attore famoso (ambiguous) 

‘They told me that a certain famous actor will come to the party’ 

‘They told me that some famous actor or other will come to the party’ 

b. Mi hanno detto che alla festa interverrà un famoso attore (specific) 

‘They told me that a certain famous actor will come to the party’ 

This property that distinguishes between the origins of the adjectival modifications 

can also be found in Korean in the contrast between the types of adjectives with or 

without –ha, for an adjective yumyeng/yumyeng-ha-n ‘famous’: 17 

16 According to Cinque, the prenominal position of the adjective renders an indefinite DP 

specific, which implies the existence of a particular actor that will come to the party, whether 

or not the speaker knows his identity. 

17 This property can be found also with the adjectives ius/ius-ha-n ‘nearby’: 

98


(42) a. etten yumeng-ha-n paywu-ka party-ey on-ta-ko tul-ess-ta. (ambiguous) 

certain famous-HA-Nactor-Nom party-to come-DCL-Com hear- 

PST-DCL 

‘I heard that a certain famous actor will come to the party’ 

‘I heard that some famous actor or other will come to the party’ 

b. etten yumeng paywu-ka party-ey on-ta-ko tul-ess-ta. (specific) 

certain famoso attore-Nom party-to come-DCL-COM hear-PST-DCL 

‘I heard that a certain famous actor will come to the party’ 

7. Concluding remarks 

Differently from other types adjectives (i.e., the qualitative adjectives), these 

non-predicative adjectives have not been considered to be a part of the adjectives 

that are present in the traditional grammar of Korean. 

In this paper, however, in reviewing the Korean non-predicative adjectives and 

their counterparts containing the copular –i and the affix -ha, I’ve argued that they 

can be distinguished between adverbial and subject-oriented, attributive and 

predicative, origin of direct modification and that of indirect modification with 

regard to the possibility to supply explanations more deepened on the category of 

the adjectives and the adjectival modifications in the nouns phrase in Korean, based 

on the proposals of Cinque (1994) and Cinque (2005b). 

References 

Alexiadou, A. (2005), “Patterns of adjectival modification”, Handout of a paper presented at the 

Second Workshop on Balkan Linguistics, Venice, Italy, 20-21 May 2005. 

Cinque, G. (1994), “On the evidence for partial N movement in the Romance DP”, in Cinque, G, 

J. Koster, J.-Y. Pollock, L. Rizzi & R. Zanuttini (eds.) Paths Towards Universal 

Grammar, Georgetown University Press, Georgetwon. pp. 85-110. 

Cinque, G. (2005a), “Deriving Greenberg’s Universal 20 and Its Exceptions”, Linguistic Inquiry, 

36. 315-332. 

(i) John-un ius-ha-n cyp-ul panghwaha-ko sypehan-ta (ambiguous) 

J. –Top neaby-HA-N house-ACC burn-to want-DCL 

a. ‘John wants to burn some specific house which is near his’ (specific) 

b. ‘John wants to burn some house or other among those which are near his’ 

(non-specific) 

(ii) John-un ius cyp-ul panghwaha-ko sypeha-n-ta (specific) 

J. –Top neaby-HA-N house-ACC burn-to want-DCL 

‘John wants to burn some specific house which is near his’ 

99


Cinque, G. (2005b), “The Dual Source of Adjectives and XP-vs. N-Raising in The Romance DP”, 

ms., University of Venice. 

Crisma, P. (1996), “On the Configurational Nature of Adjectival Modification”, in K. Zagona 

(ed.) Grammatical Theory and Romance Languages (Selected Papers from the 25th 

LSRL), pp. 59-71. 

Kang, S.H. (2005), “On the adjectives in Korean”, University of Venice Working Papers in 

Linguistics, 15. 153-169. 

Kayne, Richard S. (1994), The Antisymmetry of Syntax. Cambridge, MA: MIT Press. 

Kim, M-J. (2002), “Does Korean have adjectives?”, MIT Working Papers, 43. 71-89. 

Kim, Y-K. (1997), “Agreement phrases in DP”. UCL Working Papers in Linguistics, 9. 1-24. 

Larson, Richard K. (2000b), “Temporal Modification in Nominals” Handout of a paper presented 

at the International Round Table ‘The Syntax of Tense and Aspect’, Paris, 15-18 

November 2000. 

Larson, Richard K. and Franc Marušić (2004), “On the indefinite Pronoun Structure with APs: 

Reply to Kishimoto”, Linguistic Inquiry, 35. 268-287. 

Larson, Richard K. and N. Takahashi (in press), “Order and interpretation in pronominal relative 

clauses”, to appear in PROCEEDINGS of WAFL-2, Boğazici University, Istanbul, 

Turkey, 12 October, 2004. 

Lee, Y. H. (2004), “λ-Categorial Grammar for Korean Restrictive Adjectives”, ms., Chosun 

University, Kwangju, Korea. 

Mok, J-S. (2002), “A Study of the Category Kwanhyeongsa [Adnominals] and Hyeongyongsa 

[Adjectives] in Korean Grammar”, Ennehak [Linguistics], 31. 71-99. 

Sohn, H.M (1999), The Korean Language. Cambridge: Cambridge University Press. 

Sproat, R. and C. Shih (1991), “The Cross-Linguistics Distribution of Adjectival Ordering 

Restrictions”, in C. Georgopoulos and R. Ishihara (eds.) Interdisciplinary Approaches to 

Language: Essays in Honor of S-Y. Kuroda, Dordrecht. Kluwer Academic Publishers. pp. 

565-593. 

Yamakido, Hiroko (2000), “Japanese Attributive Adjectives are Not (All) Relative Clauses”, in 

Billerey and Lillehaugen (eds.) Proceedings of WCCFL 19, pp. 588-602. 

Yu, H.K. (1997), A study on Korean Adjectives. Ph.D. Dissertation, Yensei University, Seoul, 

Korea. 

100

Abstract 


ON THE ORDER OF THE PRENOMINAL 

PARTICIPLES IN BULGARIAN 

Vesselina Laskova 

This paper examines the distribution of Bulgarian participles in prenominal 

position. The discussion centers around two main arguments. First, it is argued that 

participial expressions postmodified by adverbs are real verbal participles. We 

provide data that, in Bulgarian, these participles can occur in prenominal position. 

The second goal is to show that when co-occurring in prenominal position, participles 

exhibit certain ordering restrictions, namely, stage-level participles precede individuallevel 

participles. These ordering restrictions conform to what is argued in Larson 

and Takahashi (in press) and Cinque (2005), who suggest that the adnominal 

modification area contains two layers – an individual-level layer, closer to the noun, 

including not only all attributive-only adjectives but also part of the indirect 

modification adjectives, and a stage-level layer, situated higher up and including the 

rest of the indirect modification area. 

Introduction 

In languages like English, participial expressions 1 can occur in prenominal 

position either unmodified or premodified by an adverbial. It has been claimed that 

the English prenominal participial expressions are not verbal participles but 

adjectives. Languages like Bulgarian, however, which do not exclude postmodified 

participial expressions in prenominal position, seem to show, first, that the prenominal 

position is not reserved only for adjectives and, second, that verbal participles 

occurring in prenominal position display certain ordering restrictions, as predicted 

by Larson and Takahashi (in press) and Cinque’s (2005) theory of adnominal 

modification. 

1 We use the term participial expression to refer to all kinds of participle-looking words. We 

reserve the term participle for the real verbal participles. 

101


Unlike the premodified participial expressions, the postmodified ones seem to 

display verbal and not adjectival properties. In English, postmodified participial 

expressions cannot be found in front of the noun but appear only in postnominal 

position. We will provide evidence that postmodified participial expressions, both in 

English and in Bulgarian, exhibit verbal and not adjectival properties. Occurring in 

prenominal position in Bulgarian, postmodified participles can combine with 

transitive participles followed by an object (of which we are sure that they are real 

verbal participles 2 ). We will show that prenominal participial expressions obey 

certain ordering restrictions in front of the noun, namely, the stage-level participle 

must precede the individual-level participle. Neither two stage-level, nor two 

individual-level participles can form a grammatical noun phrase in Bulgarian. 

The paper is organized as follows. In section 1 we introduce the type of participial 

expressions used in prenominal position in Bulgarian. In section 2 we provide 

evidence from Bulgarian and English that the postmodified participial expressions 

are verbal participles and do not have adjectival properties. In section 3 we provide 

examples in support of Cinque’s and Larson and Takahashi’s prediction that if two 

participles occur in prenominal position, it is necessarily the case that the stage-level 

participle precedes the individual-level participle. 

1. Bulgarian prenominal participial expressions 

It has been suggested in the literature (Bresnan 1982, 1995) that all prenominal 

participle-looking words should be considered adjectives. Laczko (2001), however, 

also working within the Lexical-Functional Grammar, as Bresnan, provides data 

from Hungarian showing that verbal participles do occur in front of the noun. We 

are also going to advocate this claim. As far as the English prenominal participial 

expressions are concerned, since only unmodified and premodified participial 

expressions can occur in front of the noun, we will assume that they are not 

adjectival but simply ambiguous between the adjectival and the verbal reading and 

for this reason it is impossible to isolate their verbal characteristics. In Bulgarian, 

however, we can find transitive 3 and postmodified participial expressions 

prenominally. We are going to argue that the last two types of participial expressions 

are real verbal participles and not adjectives, thus showing that the prenominal 

position is not reserved only for adjectival expressions. We will then present data 

from English which show that postmodified participial expressions behave as verbal 

participles also in this language the only difference being that, in English, 

postmodified participles can be found only in postnominal position (for independent 

2 It has been claimed by Wasow (1977) that participles taking a direct object are real verbal 

participles. 

3 We refer to participles taking a direct object as transitive participles. 

102

On the Order of the Prenominal Participles in Bulgarian 

reasons). After having shown that Bulgarian makes use of two syntactic types of 

verbal participles in prenominal position – participles taking a direct object 

complement and postmodified participles – we will concentrate on the order these 

participles exhibit when co-occurring in front of the noun. 

Bulgarian has the following three types of participial expressions occurring in 

prenominal position – passive participles (traditionally called past passive participles), 

past perfect participles (traditionally called past active participles) and what can be 

called progressive participles or present participles (traditionally named present 

active participles). 

(1) otvoreniat vchera magazin (Passive participle) 

opened-the yesterday shop 

“the shop that opened yesterday” 

(2) pristignaliat vchera turgovets (Past perfect participle) 

arrived-the yesterday merchant 

“the merchant who arrived yesterday” 

(3) izuchavashtiat fizika student (Present participle) 

studying-the physics student 

“the student who is studying physics” 

The passive participle form is quite common across languages and is widely 

discussed in the literature. As to the perfect participle, in many languages it has the 

same form as the passive participle (English, Italian, German, etc.). Bulgarian and 

Slovenian, for example, have a separate form for this participle, distinct from the 

form for the passive participle, as reported by Marvin (2002). The progressive 

participle is not uncommon across languages. 

An important peculiarity of the Bulgarian perfect and progressive participles is 

that they can take a direct object also in prenominal position, as shown below. 

(4) zashtitiloto sestra si momche (Perfect) 

defended-the sister his boy 

“the boy who defended his sister” 

(5) chetiashtiat doklada professor (Progressive) 

reading-the report-the professor 

“the professor who is reading the report” 

There is a group of verbs in Bulgarian which obligatorily require a direct object 

complement. 

103


(6) skrivam *(tsennite predmeti) 

hide precious-the objects 

“hide the precious objects” 

(7) nabezhdavam *(priatelkata si) 

accuse (falsely) friend-the my 

“accuse (falsely) my friend” 

The participles deriving from such verbs also require a direct object complement 

(of course we exclude the group of passive participles, which cannot have a direct 

object complement). 

(8) izprazniliat * (kasata) sluzhitel (Perfect) 

emptied-the cash-box-the man 

“the man who emptied the cash box” 

(9) vlacheshtata *(chergata) zhena (Progressive) 

hauling-the rug-the woman 

“the woman hauling the rug” 

There are verbs which, apart from being obligatorily transitive, could also be 

used as intransitive (unaccusative or unergative) verbs: 

(10) a. izkliuchiliat naprezhenieto mehanizum (Perfect - Transitive) 

switched off the tension-the mechanism 

“the mechanism that switched off the tension” 

104 

b. izkliuchiliat mehanizum (Perfect - Unaccusative) 

swiched off the mechanis 

“the mechanism that switched off” 

(11) a. izpulniavashtiat ariata tenor (Progressive - Transitive) 

performing-the area-the tenor 

“the tenor singing this area” 

b. izpulniavashtiat tenor (Progressive - Unergative) 

performing-the tenor 

“the performing tenor” 

We would like to keep apart the cases in which a verb is realized as transitive 

and those in which it is intransitive. We will attribute this phenomenon to the lexical 

ambiguity of the verb. 

Another group of participles are those deriving from verbs which are unambiguously 

intransitive.


(12) padnaliat snoshti sniag (Perfect - Unaccusative) 

fallen-the yesterday night snow 

“the snow that fell down yesterday” 

As was mentioned above, those participial expressions that preserve the direct 

object of the verb will be considered verbal participles. The “bare” or unmodified 

participial forms, we will consider ambiguous between the participial and the 

adjectival reading. We will suggest the same about the premodified participial forms. 

As to the postmodified participial expressions, we will try to show that they exhibit 

verbal and not adjectival properties. 

2. Tests showing the verbal character of the postmodified participial 

expressions 

In this section, we use a number of very well-known tests for distinguishing 

between participles and adjectives, in order to show that the postmodified participial 

expressions share common properties with verbs and not with adjectives. We present 

data from Bulgarian and from English. 

2.1. Bulgarian 

As was stated in the introduction, we will focus mainly on the prenominal use of 

the participial expressions in Bulgarian. Many authors (among them Wasow 1977 

and Bresnan 1982, 1995), analyzing mainly data from English, claim that the 

participle-looking words found in front of the noun are nothing else but adjectives. 

There are also opponents to this idea. Laczkó (2000, 2001) provides data from 

Hungarian showing that verbal participial forms are allowed in prenominal position. 

In this subsection, we are going to provide further evidence in support of this claim. 

Particularly, we will argue in favour of the verbal status of those Bulgarian participial 

expressions which are postmodified by adverbs 4 . 

2.1.1. The degree quantifier 

One of the tests for adjectivality is the compatibility of an expression with the 

degree quantifier. Since the latter combines only with adjectives and never with 

verbs, whatever participle-looking expression is compatible with it, it must be 

considered an adjective (of course, this test applies only to expressions which are 

gradable). 

4 We consider relevant only those adverbs which can never be used with adjectives. Therefore, 

we will restrict ourselves to using only manner adverbs like carefully and politely and their 

Bulgarian analogues. 

105


The example below shows that some unmodified participial expressions are 

compatible with the degree quantifier. 

(13) Nai-nadrastkanata tetradka e tazi na Petia. (Unmodified participial expression) 

most scribbled-the notebook is that of Petia. 

“Petia has the most scribbled notebook.” 

Bulgarian transitive participles (which are verbal participles) are never compatible 

with the degree modifier. 

(14) *Nai-nadraskaloto tetradkata si momche. 

most scribbled-the notebook-the his boy 

As we see below, the same holds true for the post-modified participial expressions. 

Examples (15) – (17) show that unmodified participial expressions can be 

compatible either with the degree quantifier or with a post-modifying adverb, but 

never with both of them at the same time. 

(15) Po-natocheniat nozh rezhe po-dobre. 

more grinded-the knife cuts better. 

“the more grinded knife cuts better” 

(16) Natocheniat vnimatelno nozh se postavia varhu… 

grinded-the carefully knife should be placed upon the… 

“the carefully grinded knife should be placed upon the…” 

(17) *Po-natocheniat vnimatelno nozh se postavia varhu… 

more grinded-the carefully knife should be placed upon the… 

If an expression is compatible both with the degree quantifier and with a 

postmodifying adverb but never with both of them at the same time, there must be a 

difference in the grammatical status of these two combinations. The tests applied 

below seem to further support this conclusion. 

2.1.2. Complements of some verbs 

Another test used in the literature is the possibility of an English adjective to 

appear as a complement of verbs like seem, remain, look, sound, act and become. 

The examples are taken from Wasow (1977). 

(18) John looked eager to win. 

(19) John remained happy. 

It seems that the Bulgarian analogue of the verb remain – ostavam – requires an 

adjectival complement as well. 

106


The examples from Bulgarian show that premodified and unmodified participial 

expressions can occur in this position but participles taking a direct object complement 

and postmodified participial expressions cannot. We see here that, as we suggested 

above, the premodified participial expressions behave like adjectives. 

Unmodified participial expression 

(20) trite ostanali nepochisteni sled partito pomeshtenia 

three-the remained uncleaned after party-the rooms 

“the three rooms that remained uncleaned after the party” 

Premodified participial expression 

(21) Ostanalite vnimatelno podredeni vurhu biuroto dokumenti 5 . 

remained-the carefully ordered on bureau-the documents. 

“the documents that remained carefully ordered on the bureau” 

(22) Ostanaloto vnimatelno razpechatano sled proverkata pismo 6 

remained-the carefully unsealed after examination-the letter 

“the letter that remained carefully unsealed after the examination” 

Transitive participles 

(23) *Ostanaliat podrezhdasht dokumentite sluzhitel. 

remained-the ordering documents-the attendant 

“the attendant that remained ordering the documents” 

Postmodified participial expression 

(24) ?*ostanalite podredeni vnimatelno varhu biuroto dokumenti 

remained-the ordered carefully on bureau-the documents 

“the documents that remained carefully ordered on the bureau” 

(The relevant meaning of the participle ostanalite has to be distinguished from 

the meanings: “remained at that place” and “the rest”) 

5 Exampels (21) and (22) sound a bit odd because of the slight semantic incompatibility of the 

adverb “carefully”, which we use in order to be consistent, and the verb ostavam/remain. We 

aim at showing that, in this environment, an adverb in postposition with respect to the 

participle sounds worse than an adverb in preposition with respect to the participle. We see in 

(27) that, in a predicative use, the same premodified participle sounds better (since it is easier 

to insert the context). 

6 See footnote 5. 

107


(25) *Ostanaloto razpechatano vnimatelno sled proverkata pismo 

remained-the unsealed carefully after examination-the letter 

“The letter that remained carefully unsealed after the examination” 

108 

Predicative use: 


(26) Knigata ostana neprochetena. 

book-the remained unread 

“The book remained unread.” 


(27) Dori sled obiska dokumentite na biuroto i ostanaha vnimatelno podredeni. 

Even after perquisition-the documents-the on bureau-the her remained 

carefully ordered. 

“Even after the perquisition, the documents on her bureau remained carefully 

ordered.” 

Transitive participle 

(28) *Sluzhiteliat ostana podrezhdasht dokumentite. 

Attendant-the remained ordering documents-the 

“The attendant remained ordering the documents.” 

Post-modified participle 

(29) *Dokumentite na biuroto i ostanaha podredeni vnimatelno 

Documents-the on bureau-the her remained ordered carefully. 

“The documents on her bureau remained carefully ordered.” 

2.1.3. Concessional relative phrases with “however” 

Bresnan (1995), claims that only adjectives, and not verbs, can head 

concessional relative phrases beginning with “however”. 

however AP vs. *however VP: however supportive of her daughter she may have 

been vs. however supporting her daughter she may have been… 

(Bresnan, 1995) 

Indeed, neither the Bulgarian analogues of the English concessional phrases with 

however can be headed by a verb. Thus, we can make the prediction that only 

unmodified and premodified participial expressions but not postmodified ones can 

head concessional phrases like kolkoto i…da …/however… The examples below 

show that this expectation seems to be correct.


Unmodified participial expressions 

(30) Kolkoto i nadraskana da e tetradkata, pak shte mi svurshi rabota. 

however and scribbled DA is notebook-the still will to me serve 

“However scribbled the notebook is, it could serve me.” 

Premodified participial expressions 

(31) Kolkoto i vnimatelno podbrani da sa sustavkite … 

however and carefully selected DA are ingredients… 

“however carefully selected the ingredients” 

Post-modified participial expression 

(32) *Kolkoto i nadraskana nevnimatelno da e tetradkata, pak shte mi svurshi 

rabota. 

however and scribbled carelessly DA is notebook-the still will to me serve 

“However carelessly scribbled the notebook is, it could serve me.” 

Transitive participle 

(33) *Kolkoto i podbral sustavkite da e… 

however and selected (masc.) ingredients DA is … 

In prenominal position: 


(34) Kolkoto i nadraskana tetradka da ima Ivan… 

however and scribbled notebook DA has Ivan 

“However scribbled Ivan’s notebook…” 


(35) Kolkoto i vnimatelno podbrani sustavki da izpolzvat… 

however and carefully selected ingredients they use 

“No matter how carefully selected ingredients they use…” 

Postmodified participle 

(36) *Kolkoto i podbrani vnimatelno sustavki da izpolzvat… 

however and selected carefully ingredients they use 

“No matter how carefully selected ingredients they use…” 

The examples above clearly show that the postmodified participial expressions 

cannot fill the slot of the adjectives. The premodified and the unmodified ones, on 

the other hand, qualify as adjectives. 

109


In the next subsection, we will see that Bulgarian is not the only language in 

which the postmodified participles display verbal participles. English seems to pattern 

with Bulgarian in this respect. We present below some tests in support of this view. 

2.2. English 7 

According to the literature, the negative un- prefix can only attach to adjectives. 

It never attaches to verbs. (The negative prefix under consideration here is not to be 

confused with the verbal reversative prefix attaching to verbs as in undo). 

(37) unaccepted * to unaccept 

(38) unquestioned * to unquestion 

Since there are no verbal forms corresponding to the negative participial forms in 

(38) and (39), these forms are considered adjectival. What is of interest for us here is 

that passives like these can be pre-modified but not post-modified by adverbs. (The 

examples were pointed out to me by Megan Linke 8 , p. c.) 

(39) The invitations, politely unaccepted, lay strewn upon the table. 

(40) * The invitations, unaccepted politely, lay strewn upon the table. 

(41) The king’s argument, respectfully unquestioned, rang throughout the room. 

(42) * The king’s argument, unquestioned respectfully, rang throughout the room. 

The fact that adjectival passives cannot be postmodified by adverbs points to the 

conclusion that, also in English, the postmodification of participial expressions is 

characteristic only of the verbal participles and not of the adjectival forms. 

Our claim is further confirmed by the following observation. Only pre-modified 

and not post-modified participial expressions can appear after the verb seem. A 

widespread assumption is that seem can be followed only by adjectives and never by 

verbal expressions. It has not been noticed, however, at least as far as I know, that 

postmodified and premodified participial expressions differ in this respect. We 

present the data below. (The data were pointed out to me by Megan Linke, p. c.) 

(43) The floor has not been waxed and the curtains are still dirty, but the silver, at 

least, seems carefully polished. 

(44) * The floor has not been waxed and the curtains are still dirty, but the silver, 

at least, seems polished carefully. 

7 I thank Steven Franks for the very helpful observations, opinions and suggestions concerning 

the subsection on English participles. 

8 I would like to thank Megan Linke for the various examples she pointed out to me, for the 

pleasant discussions and for her helpful comments on my data. 

110


(45) The red lentils still have pieces of dirt and stone in them, but the green ones 

seem carefully sorted. 

(46) * The red lentils still have pieces of dirt and stone in them, but the green ones 

seem sorted carefully. 

(47) The present seems carefully wrapped up. 

(48) * The present seems wrapped up carefully. 

(49) The room seems carefully cleaned. 

(50) *The room seems cleaned carefully. 

(51) The issue seems carefully explained (in a suitable context it sounds fine) 

(52) * The issue seems explained carefully. 

Another way to test the verbal character of the postmodified participial expressions 

is to see whether they can head concessional phrases with however. Unmodified 

participial expressions clearly can head such a phrase, as we show below. 

(53) However polished the floor was, it didn’t seem completely clean. 

The same holds true for the premodified participial expressions: 

(54) However carefully polished the floor was, it didn’t seem completely clean. 

It is, however, completely impossible to place a postmodified participial 

expression in this environment. 

(55) *However polished carefully the floor was, it didn’t seem completely clean. 

However seems to be compatible with other adjectival participial forms like the 

unmodified ones. This is shown below. (The examples below were pointed out to 

me by Megan Linke p. c.) 

(56) A very promising extension, however untouched, is that of defining 

strategies that decide which presentation forms of the selectors to use, or 

even defining strategies that define such a strategy depending of e. g. the 

speed of the underlying hardware, the size of targeted display, etc., as 

indicated on p. 58. 

http://www.cs.kuleuven.ac.be/publicaties/doctoraten/cw/CW2001_1.pdf 

(57) People have a psychological sense that a used object is worth much less, 

however untouched it is. 

http://www.palminfocenter.com/view_story.asp?ID=4323 

So far, we have provided evidence that postmodified participial expressions 

display verbal properties in behaving differently with respect to the premodified and 

111


the unmodified participial expressions. As to the last two types, we assume that they 

are rather ambiguous between the participial and the adjectival reading. The adverb 

in preposition, unlike the adverb in postposition, is not a signal of the verbal 

character of the participle. 

We have already seen above that premodified participial expressions do not 

pattern with verbal participles. We present below some more examples from Bulgarian 

in support of this claim. 

(58) a. dobre slozhen chovek. 

well-built person (= has a fine physique) 

b. * slozhen dobre chovek. 

built-well person 

(59) a. silno zamursena dreha 

strongly daubed piece of clothing 

b. *zamursena silno dreha 

daubed strongly piece of clothing 

“the strongly daubed piece of clothing” 

What the examples above show is that, with premodified participles, it is 

possible to form fixed expressions. The meaning which emerges in these examples 

is not a real combination of the meaning of the verb and that of the participlelooking 

word. In examples like (59), the participial expression does not convey the 

real meaning of the verb it derives from. What has happened is that the verb has 

been adjectivalized. As we can see, once we place the adverb in postposition, the 

original meaning of the verb reemerges and the example no longer sounds acceptable. 

This observation comes in support of the claim that the premodified participial 

expressions can be adjectival while the postmodified ones are only verbal. We 

mentioned, however, that the premodified participial expressions are actually 

ambiguous, which means that they can also be verbal. In other words, the premodifying 

adverb does not necessarily signal the verbal status of the expression it modifies but 

it does not necessarily signal its adjectival status either. This can be seen in the 

following example, in which the real verbal participles can also be premodified by 

an adverb. 

(60) vnimatelno obrabotiliat dannite sluzhitel 

carefully processed-the data-the attendant 

“the attendant who carefully processed the data” 

112

3. The order of the prenominal participles 


We have seen so far that, apart from the transitive participles, there is another 

group of participial expressions which behave as verbal elements. In this section, we 

will take it for granted that the postmodified participial expressions are verbal 

participles and will try to see how two verbal participles combine in prenominal 

position in Bulgarian 9 . What we will notice is that not all orders between the 

prenominal participles are allowed. We will try to explain this phenomenon in terms 

of Cinque’s (2005) theory of adnominal modification. 

We provide below some examples of two participles occurring in prenominal 

position. 

(61) Vkliucheniat vnimatelno izmervasht naprezhenieto ured pokaza, che … 

swiched on-the carefully measuring voltage-the device showed that… 

“The carefully switched on voltage measuring device showed that…” 

(62) Nanesenoto vnimatelno raztvariashto maznini veshtestvo, ne dopuska 

pronikvaneto na … 

impasted-the carefully solving fat substance not allows penetration-the of… 

“The carefully impasted, fat solving substance does not allow the penetration 

of…” 

Notice that once we change the places of the participles, the examples become 

ungrammatical. 

(63) * Izmervashtiat naprezhenieto vkliuchen vnimatelno ured. 

measuring-the voltage-the swiched on carefully device 

“the device that measures the voltage that has been carefully switched on” 

(64) *Raztvatiashtoto maznini naneseno vnimatelno veshtestvo… 

solving-the fat impasted carefully substance 

“the carefully impasted substance that solves fat” 

Examples like (63) and (64) suggest that not all orders between the participles in 

prenominal position are allowed. How to account for this phenomenon? 

One component of the semantic meaning of these participles is particularly 

relevant to our discussion. Notice that the participles which come first in each of the 

examples express a telic event, an action which has been performed once. The 

participles coming in second position, instead, express either a quality of the entity 

9 It seems impossible to place two transitive participles in prenominal position in Bulgarian. 

The examples sound too heavy and a long pause between the two participles is required. One 

transitive and one postmodified participle, however, can co-occur in front of the noun. 

113


or an activity habitually performed by that entity. We would like to express this 

difference in terms of the stage-level/individual-level distinction. We could assume 

that the participles expressing a telic event are stage-level participles and the ones 

expressing an activity are individual-level participles. With this distinction in mind 

we could turn to the theory developed by Cinque (2005) and Larson and Takahashi 

(in press), about the order of the prenominal reduced relative clauses. 

Discussing data from Japanese, Korean, Chinese and Turkish, Larson and 

Takahashi (in press) suggest that the adnominal modification area is divided into 

two large layers – an individual-level layer, closer to the noun, governed by a 

generic operator and a stage-level layer, higher than the previous one and governed 

by an existential operator. Cinque (2005) develops a theory of adnominal modification, 

adopting this analysis and predicting that if the noun phrase contains more than one 

participles in prenominal position, there would be one participle belonging to the 

stage-level layer and one to the individual-level layer. This is, actually, what the 

Bulgarian data seem to show. Apart from the impossibility to obtain the inverted 

order: individual-level > stage-level, we can notice also the impossibility to place 

two individual-level or two stage-level participles in the same phrase. This is shown 

in the following examples. 

Compare the correct (62) with the not so well-sounding (65). 

(65) ??Izkliucheniat vnimatelno izmeril naprezhenieto ured… 

switched off-the carefully measured (perfect participle) voltage-the device 

“the device that has measured the voltage, that has been switched off” 

In example (65), two telic event participles are used. Therefore the phrase no 

longer sounds good. One could try to save the example by leaving a very long pause 

between the two participles but, still, it does not sound better. 

Equally bad are examples containing two individual-level participles. 

(66) *Izsledvashtiat Jupiter izuchavasht astronomia uchen… 

examinig-the Jupiter studying astronomy scientist… 

“the scientist studying chemistry who studies Jupiter” 

The examples above suggest that participles seem to follow a certain order in 

prenominal position 10 . Higher in the hierarchy are the stage-level participles and the 

individual-level participles follow them. 

10 It is generally not so easy to place two verbal participles in prenominal position in 

Bulgarian. What we claim is that it is possible and easier in the cases in which the higher 

participle displays stage-level properties and the participles following it displays individuallevel 

properties. 

114

4. Conclusion 


In this paper, we have provided data suggesting that verbal participles can really 

appear in prenominal position. We argued against the assumption that all prenominal 

participle-looking words are adjectives and suggested, instead, that the prenominal 

participial expressions in English are rather ambiguous between the participial and 

the adjectival reading. It is not that real participles cannot occur in prenominal 

position in English, as has been argued in the literature. What we suggest is that, 

English, for independent reasons, does not allow for right modification of prenominal 

elements. Since the unmodified and the premodified participial expressions are 

ambiguous the verbal character of these elements is not so visible. Bulgarian, however, 

as well as Hungarian, as shown by Laczkó (2001), do not display any ban on verbal 

participial elements in prenominal position. These languages clearly show that 

verbal participles can occur in front of the noun. 

Another interesting observation was that postmodified participial expressions 

actually do not pattern with the premodified and the unmodified ones. Both in 

Bulgarian and in English, these elements display verbal properties. The two 

languages differ only in terms of the position these participles occupy with respect 

to the noun – in Bulgarian they can be prenominal while in English they can only be 

postnominal. 

The conclusions concerning the verbal status of the postmodified participial 

expressions served as a possibility to explore the co-occurrence of two participles in 

prenominal position. We have seen that it is possible to combine one transitive 

participle and one postmodified participle in front of the noun in Bulgarian. We have 

also observed that two participles can co-occur in prenominal position only if they 

obey the following semantic restriction: the higher participle has to display stagelevel 

properties and the lower participle has to display individual-level properties, a 

restriction discussed in the works of Larson and Takahashi (in press) and Cinque 

(2005). 

References 

Bresnan, J. 1982. “The Passive in Lexical Theory”. In Bresnan, J. (ed.), The Mental Representation 

of Grammatical Relations. The MIT Press. Cambridge, Mass., 3-86. 

Bresnan, J. 1995. “Lexicality and Argument Structure”. In Proceedings of the Paris Syntax and 

Semantics Conference, 1995. 

Cinque, G. 2003a. “The Prenominal Origin of Relative Clauses”. Workshop on Antisymmetry and 

Remnant Movement, New York University. 

Cinque, G. 2005. “The Dual Source of Adjectives and Phrasal Movement in the Romance DP”. 

Ms, University of Venice. 

Embick, David. 2004. “On the Structure of Resultative Participles in English”. Linguistic Inquiry. 

35, 355-392. 

115


Haspelmath, Martin. 1994. “Passive Participles across Languages”. In Fox, Barbara – Paul J. 

Hopper (eds.) Voice: Form and Function. John Benjamins, Amsterdam, pp. 151-177. 

Kayne, Richard. 1994. The Antisymmetry of Syntax. MIT Press, Cambridge, Massachusetts. 

Kratzer, Angelika. 2000. “Building Statives”. In Proceedings of the Twenty-sixth Annual Meeting 

of the Berkley Linguistic Society, ed. by Lisa J. Conathan, Jeff Good, Darya Kavitskaya, 

Alyssa B. Wulf, and Alan C. L. Yu, pp. 385-399. University of California, Berkley 

Linguistics Society. 

Laczkó, Tibor. 2001. “Another Look at Participles and Adjectives in the English DP”. In Butt, M. 

and T. H. King (eds.), Proceedings of the LFG01 Conference. CSLI Publications. 

Stanford. 

Larson, R. and N. Takahashi (in press) “Order and Interpretation in Prenominal Relative Clauses”. 

To appear in PROCEEDINGS OF WAFL-2. Boğazici University, Istanbul, (October 12, 

2004). 

Marvin, Tatjana. 2002. “Past Participles in Reduced Relatives”. ms. University of Lund. 

Wasow, Thomas. 1977. “Transformations and the Lexicon”. In P. W. Culicover, T. Wasow and 

A. Akmajian (eds.), Formal Syntax, pp. 327-360. 

116

Rivista di Grammatica Generativa, 31 – 2006, 119 - 122 

ARE ALL LANGUAGES 

‘NUMERAL CLASSIFIER LANGUAGES’? * 

Guglielmo Cinque 

Greenberg (1975) observes that, “it is generally the case that numeral classifier 

languages will apparently lack a classifier in nouns indicating periods of time, units 

of distance and the word ‘time’ in such phrases as ‘five times’. [In Greenberg 1972] 

it was hypothesized that in these cases the correct interpretation was not that the 

classifier is omitted but that words like ‘day’, ‘mile’ and ‘time’ are themselves 

measures of verbal action so that we have to do with a subtype of the overall 

classifier or measure phrases. In other words, such phrases as ‘five days’ are rather 

to be identified with (Q ↔ Cl) than (Q ↔ N).” (p. 30). 1 

Certain numeral classifier languages provide direct evidence for this conclusion 

as the apparently classifier-less N does not occupy the normal position of the noun 

but that of the “absent” classifier. This is especially evident in Thai, where the noun 

and the numeral classifier are on opposite sides with respect to the numeral: N Num 

CL. 

As Allan (1977, 306f) notes, nouns like ‘year’, in adverbial constructions, 

unexpectedly appear with a numeral without an accompanying numeral classifier: 

(1) nỳŋ pi 

one year 

Num N 

What is even more striking, Allan says, is that they do not appear in the ordinary 

position occupied by the noun (i.e., before the numeral – see mă si tua ‘dog four 

body’ = ‘four dogs’), but after the noun, in the position normally occupied by the 

classifier (see sì tua ‘four body’ = ‘four (of them)’ [animals, coats, etc.]). 

* This squib was originally prepared for a private birthday book in honour of Bernard Comrie. 

I thank Richard Kayne for helpful comments. 

1 This is true of many Tai-Kadai, Sino-Tibetan, Mon-Khmer, and Austronesian languages (see, 

e.g., Thomas 1971, 137; Manley 1972, 126; Goral 1978, 10, 28, 29-30; Kruspe 2004, 209). 

119


Very insightfully he concludes, citing Haas (1942, 204), that in order to 

accommodate these facts pi ‘year’ in (1) “must be interpreted as a classifier, and 

[(1)] must be given a new structural description, [(2)]” (p. 307), adding that “the 

alternative, that the labels [Num] and N […] be swapped, is absurd.” (p. 307). 2 

(2) nỳŋ pi 

Num CL 

It is at this point interesting to note that time units like ‘year’, when used 

adverbially, display properties of numeral classifiers of ‘numeral classifier languages’ 

(rather than those of ordinary nouns) even in ‘non numeral classifier languages’ like 

Italian or English. 3 

For example, it is generally the case that adjectives can modify nouns, and 

mensural “classifiers” (like ‘box’, ‘cup’, ‘kilo’, etc.), but not (sortal) numeral 

classifiers. See the contrast between (3) and (4), observed for Chinese in Cheng and 

Sybesma (1999, 516): 

(3) na yi xiao xiang shu 

that one small CL-box book 

‘that (one) small box of books’ 

(4) a *yi da zhi gou 

one big CL dog 

b *yi da wei laoshi 

one big CL teacher 

Now exactly the same thing is found with the ‘nouns’ anno ‘year’ in Italian, year 

in English, and godina ‘year’ in Bulgarian, when they are used adverbially to 

express a time measure. See (5), (6), and (7): 4 

(5) a Sono rimasto a Londra per tre (*?bellissimi) anni 

I stayed in London for three (beautiful) years 

b Tre (*bellissimi) anni fa ero a Londra 

three (beautiful) years ago I was in London 

2 Also see Simpson (2005, section 7), who suggests that cases like (2) involve raising of the N 

to CL. 

3 The same is true of Bulgarian, which apart from the existence of three (or four) genuine 

numeral classifiers (Greenberg 1972, fn5; Cinque and Krapova 2007) is essentially a ‘non 

numeral classifier’ language. 

4 Although (6)a and b are felt as awkward by some speakers, Richard Kayne tells me that for 

him they are not completely impossible. The Bulgarian examples in (7) are from Cinque and 

Krapova (2007). 

120

(6) a I lived in London (for) three (*beautiful) years 

b Three (*beautiful) years ago I was in London 

(7) a Živjax tri (*prekrasni) godini v London 

I lived three (beautiful) years in London 

b Predi tri (*prekrasni) godini bjax v London 

before three (beautiful) years I was in London 

‘Three (beautiful) years ago I was in London’ 

Are all languages ‘Numeral Classifier Languages’? 

These facts suggest that in this usage Italian anno, English year, and Bulgarian 

godina, are really numeral classifiers in (5), (6), and (7), like Thai pi·is in (2). 

The fact that when they are used as arguments (say as objects of a transitive 

verb), they can be modified by adjectives (see (8), (9), and (10)), further suggests 

that they can also be ordinary nouns; which recalls the case of so-called ‘selfclassifiers’ 

or ‘repeaters’ in many ‘numeral classifier languages’ (see (11), from 

Simpson 2005, 832), except that in Italian, English, or Bulgarian, either the noun or 

the classifier, but not both, can be pronounced: 5 

(8) Ho passato/trascorso tre bellissimi anni a Londra 

I spent three beautiful years in London 

(9) I spent three beautiful years in London 

(10) Prekarax tri prekrasni godini v London 

I spent three beautiful years in London 

(11) a hoong saam hoong (Thai) 

room three CL-room 

‘three rooms’ 

b cun ta cun (Burmese) 

island one CL-island 

‘one island’ 

If this interpretation of the facts is plausible, then the conclusion is that even 

traditional ‘non numeral classifier’ languages are numeral classifier languages, with 

mostly abstract, or non pronounced, classifiers. 6 

5 Perhaps, in the adverbial classifier usage of these nouns, the noun itself raises to the classifier 

head, as Simpson (2005) suggested for Thai. 

6 See Kayne (2003), who argues for the existence of a non pronounced numeral classifier 

‘year’ in English (in expressions like I am seven, at the age of seven, etc.), and Kayne (2005), 

more generally, on the role of non pronounced functional elements in the languages of the 

world. 

121


References 

Allan, Keith (1977) “Classifiers” Language 53. 285-311. 

Cheng, Lisa Lai-Sheng and Rint Sybesma (1999) “Bare and Not-So-Bare Nouns and the Structure 

of NP” Linguistic Inquiry 30. 509-542. 

Cinque, Guglielmo and Iliyana Krapova (2007) “A Note on Bulgarian Numeral Classifiers”, in G. 

Alboiu, A.A. Avram, L. Avram, D. Isac (eds.) Pitar Moş: A Building With a View. 

Papers in Honour of Alexandra Cornilescu. Bucharest: Editura Universităţii din 

Bucureşti, pp. 45-51. 

Goral, Donald N. (1978) “Numerical Classifier Systems: A Southeast Asian Cross-Linguistic 

Analysis” Linguistics of the Tibeto-Burman Area 4 (1) 1-72. 

Greenberg, Joseph H. (1972) “Numeral Classifiers and Substantival Number: Problems in the 

Genesis of a Linguistic Type” Working Papers on Language Universals 9. 1-39. 

Greenberg, Joseph H. (1975) “Dynamic Aspects of Word Order in Numeral Classifier”, in C. Li 

(ed.) Word Order and Word Order Change. Austin, University of Texas Press, pp. 27-46. 

Haas, Mary (1942) “The Use of Numeral Classifiers in Thai” Language 18. 201-205. 

Kayne, Richard S. (2003) “Silent Years, Silent Hours”, in L.-O. Delsing, C. Falk, G. Josefsson, H. 

Sigurdsson (eds.) Grammar in Focus. Festschrift for Christer Platzack 18 November 

2003. Vol. II. Lund, Wallin and Dalholm, pp. 209-226. 

Kayne, Richard S. (2005) “Some Notes on Comparative Syntax, with Special Reference to 

English and French”, in G. Cinque and R.S. Kayne (eds.) The Oxford Handbook of 

Comparative Syntax. New York, Oxford University Press, pp. 3-69. 

Kruspe, Nicole (2004) A Grammar of Semelai. Cambridge, Cambridge University Press. 

Manley, Timothy M. (1972) Outline of Sre Structure. Honolulu, University of Hawaii Press. 

Simpson, Andrew (2005) “Classifiers and DP Structure in Southeast Asia”, in G. Cinque and R.S. 

Kayne (eds.) The Oxford Handbook of Comparative Syntax. New York, Oxford 

University Press, pp. 806-838. 

Thomas, David D. (1971) Chrau Grammar. Honolulu, University of Hawaii Press. 

122

I dattiloscritti devono essere inviati in duplice copia a uno dei componenti la 

direzione, ai seguenti indirizzi: 

Guglielmo CINQUE Dipartimento di Scienze del Linguaggio 

Università di Venezia 

Dorsoduro 1075 - 30100 VENEZIA 

Luigi RIZZI CISCL - Scienze della Comunicazione 

Università di Siena 

Via Roma 56 

53100 Siena 

Prezzi di vendita: 

Area Euro Privati: euro 20,00 

Istituzioni: euro 30,00 

Modalità di pagamento: 

1. Anticipato, senza spese di spedizione, a mezzo versamento su 

conto corrente postale nr. 10644359 intestato a Unipress, via Cesare 

Battisti 231 - 35121 Padova 

2. Contrassegno, con addebito di spese di spedizione, inoltrando l'ordine 

a Unipress, via Cesare Battisti 231 - 35121 Padova (valido solo 

per l’Italia). 

3. Anticipato, senza spese di spedizione, accludendo all'ordine assegno 

bancario o circolare a favore di Unipress (valido solo per 

l’Italia). 

4. Accettiamo VISA e MASTERCARD – comunicate numero carta, 

nome del titolare, codice di sicurezza e data di scadenza 

Subscription price (non-Euro area): 

Foreign Countries Individuals: $ 40.00 [no postage] 

Institutions: $ 50.00 + 3.00 [postage] 

Payment Methods 1. Check enclosed 

2. International Postal Money Order 

3. Post Office Current Account (International Giro Account 

nr 10644359 Unipress, Padova) 

4. We accept VISA and MASTERCARD – please 

send to Unipress: number of Card, authorized signature, 

expiry date and security code. 

All private orders must be prepaid and sent directly to the Publisher and are for personal 

use only

Subjectless language: syntactic 

aspects of Samuel Beckett’s 

“Rockaby” 

Focus in the IP: the Particle ma 

in Florentine 

Developmental patterns in the 

acquisition of complement clitic 

pronouns 

The two forms of the adjective 

in Korean 

On the Order of the Prenominal 

Participles in Bulgarian 

Squib e Discussioni 

Are all languages ‘Numeral 

Classifier Languages’? 

RIVISTA DI GRAMMATICA GENERATIVA 

Volume 31, anno 2006 

Valentina BIANCHI 

Jacopo GARZONIO 

Cornelia HAMANN 

Adriana BELLETTI 

Soon Haeng KANG 

Vesselina LASKOVA 

Guglielmo CINQUE 

Cod. RGG031 ISSN 1122-4428 

Caterina Santinello

Untitled - Lear

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?