1. Introduction

Clitic doubling is an important morpho-syntactic process in many languages (e.g. Spanish, Catalan, Romanian and Hebrew) and functional categories are crucially involved. Clitic-doubling is a construction in which a Full Noun Phrase is doubled by a clitic pronoun that receives the same theta role and checks the same case:

1. Marta te lo dijo a ti

Marta you-DAT it-ACC said to you-DAT

(Marta told you it)

The acquisition of clitic pronouns has been studied in many languages (Italian (Guasti, 1993), French (Friedemann, 1992; Hamann, Rizzi and Frauenfelder, 1995; Kaiser, 1993), German (Jakubowicz et al., 1995), Spanish (Varela, 1988), Catalan (Torrens, 1995)). The only study reported so far on the acquisition of clitic doubling is the PhD thesis by Varela (1988). In this study, however, the author only tested the comprehension of sentences with non-pronominal indirect objects. Therefore, the study does not provide any information about what children know about clitic doubling's obligatoriness or optionality in different contexts.

Due to the lack of evidence on the acquisition of clitic doubling it has been argued that clitic doubling in languages like Spanish should not be found in the early stages (Fox and Grodzinsky, 1994; Fox, Grodzinsky and Crain, 1995). Fox, Grodzinsky and Crain propose that passives in English and clitic doubling in Spanish are the same construction; these authors suggest that the delay on the acquisition of passives is based on the lack of clitic doubling; therefore, they predict the delay of this clitic doubling in the early ages.

The purpose of this paper is to give evidence that shows that very young children know central aspects of clitic doubling in Standard Spanish. The central goal of this study is the investigation of the productivity of this construction; another goal is to study the possible relation between the acquisition of clitic doubling in Spanish and the acquisition of other constructions like floating Quantifiers, dative experiencers, clitic left dislocation, passives in English, and scrambling in Dutch.

2. Theoretical background

2.1. The structure of clitic doubling

There are several accounts to describe the status, base generation and position of clitic pronouns. The base generation hypothesis (Rivas, 1977; Jaeggli, 1982, 1986; Borer, 1984) proposes that the clitic is base generated to the left of the verb, and the movement hypothesis (Kayne, 1975, 1991) proposes that clitics are base generated in the canonical object position and move to the position of the verb by incorporation. Both accounts consider clitics to be arguments, and therefore clitics have to check case and must have a theta role assigned. The movement hypothesis has difficulties in accounting for the cases of clitic doubling, where two arguments (the clitic and the coreferential NP) have to check the same case (accusative or dative) and must have the same theta role assigned² . Our acquisition evidence does not distinguishes between syntactic theories of clitic doubling. The base-generation theory at the moment seems to cover a great deal of syntactic ground; we will assume that theory for concreteness.

To account for clitic doubling, we assume that clitics are heads, and that the object clitics are verbal agreement morphemes (Uriagereka, 1995; Franco, 1992). Following Sportiche (1992) we assume that clitics are heads of their own projection. Following is the structural representation of the clause, including accusative and dative clitics (where Clio = Clitic Indirect Object, and Cldo = Clitic Direct Object):

2. IP

...........

ClioP

Spec Clio'

Clio CldoP

Spec Cldo'

Cldo VP

XP2 Spec V'

V DP

XP1

The specifier of the clitic phrase is a landing site for movement at LF of the coreferent DP, which is XP1 at the base-generated position; XP1 moves to the specifier of the clitic projection, XP2. The specifier position agrees on some features (person, number, gender, Case) with the coindexed clitic. The following example illustrates the structure that we are assuming:

3. Carmen loi prest— XPi1 a Javier

Carmen it-ACCi lent XPi1 to Javier

[CldoP XP2i [[Cldo loi ] [... prest— XP1i [acc] ...]]]

Following Sportiche (1992), we assume that clitics license a feature [F]. This feature may be licensed at LF only in an appropriate agreement relationship. Sportiche states this relationship as follows:

4. Clitic Criterion:

At LF

i. A clitic must be in a spec/head relationship with a [+F] XP

ii. A [+F] XP must be in a spec/head relationship with a clitic

(Sportiche, 1992, pg. 25)

where the argument XP1 has to move in order to satisfy the Clitic Criterion.

Following Sportiche (1992), we consider clitic doubling as the concurrence of a clitic and of an overt XP1, and a covert movement of XP1 to XP2.

2.2. The contexts for clitic doubling

Whether or not clitic doubling is obligatory depends on the status and the case of the doubled object. The contexts described below correspond to Standard Spanish.

For Direct Objects, clitic doubling with a Pronominal Direct Object is obligatory; on the other hand, clitic doubling with a Non-Pronominal Direct Object is impossible.

For Indirect Objects, clitic doubling for Non-Pronominal Indirect Objects is optional with a goal theta role and obligatory with a possessive theta role; however, clitic doubling for Pronominal Indirect Objects is obligatory. For dative experiencers, which have a very different structure, clitic doubling is obligatory. Following are examples for all the contexts described so far:

5. Pronominal Direct Object

*(Los) invitŽ a ellos

(them-ACC) invited-1s to them

(I invited them)

6. Non-Pronominal Direct Object

(*Lo) com’ arroz con frijoles anoche

(it-ACC) ate-1s rice and beans last night

(I ate rice and beans last night)

7. Non-Pronominal Indirect Objects with a goal theta role

Dolores (le) llev— el libro a Carlos

Dolores (him)-DAT brought-3s the book to Carlos

(Dolores brought the book to Carlos)

8. Non-Pronominal Indirect Objects with a possessive theta role

Javier *(le) rompi— la cabeza al boxeador

Javier (him-DAT) smashed-3s the head to the boxer

(Javier smashed the boxer’s head)

9. Pronominal Indirect Objects

Carmen *(le) prest— el ordenador a Žl

Carmen (him-DAT) lent-3s the computer to him

(Carmen lent him the computer)

10. Dative experiencers

A ella *(le) gusta el tŽ

To her-DAT her-DAT likes tea

(She likes tea)

2.3. Other constructions

2.3.1. Clitic-Left dislocation

Clitic-left dislocation is a construction that contains a left-dislocated object and a coindexed clitic. Some authors have suggested that the dislocated object is generated in its surface position (Cinque, 1990). On the other hand, Kayne (1995) and Cecchetto (1995) have proposed that the DP object moves from the argumental position inside VP to the dislocated position (the two-steps movement hypothesis). In this view left dislocation is a type of clitic doubling, where the DP object moves to a dislocated position instead of staying in the argumental position. Evidence supporting this proposal comes from the island sensitivity effect found in clitic left dislocation:

11. ??Beppe, temo la possibilitˆ che lo arrestino

Beppe, fear-1s the possibility that him-ACC arrest-3p

The two-steps movement hypothesis (Ceccheto, 1995) proposes that the object DP moves to check a topic feature; first, the DP has to pass through Spec,CldoP (to check the agreement feature it shares with the clitic head); the second step is A’ movement from Spec,CldoP to the topic position. The clitic incorporates into the verb, and checks case via incorporation into the verb (Belletti, 1994). Although Sportiche does not discuss clitic-left dislocation, it seems to us that this base generation of clitic theory could assume that in clitic-left dislocation forcing to move to topic position on the surface, will force it kids to move to Spec,Cldo also on the surface. If clitic doubling and left dislocation are syntactically similar, then one expects to find clitic doubling and left dislocation at the same stage in development.

2.3.2. Dative experiencers

Dative experiencers have a different structure than common clitic doubling. Dative experiencers occur with the psychological verbs like gustar, and in other non-psych unaccusative predicates like se construction:

12. a Pedro le gusta la bebida

to Pedro him-DAT likes the drink

(Pedro likes drinking)

In this construction the experiencer is the subject (but with a dative case), and the theme is in the object position (but triggers agreement on the verb) (Belleti and Rizzi, 1988). Montrul (in progress) has suggested that dative experiencer clitics are overt agreement in AgrS, manifesting case and agreement features and motivating movement of the dative NP into [spec, AgrS] in the syntax. However, if the dative NP is in Spec, AgrS, we would expect it to agree with the verb, which it does not. Furthermore, an NP in Spec, AgrS should receive nominative case, but the NP receives dative case. Because the argument checks dative case, it seems more plausible to assume that dative experiencers are overt agreement in AgrIO, and that the doubled DP moves to position [Spec,ClioP]. Therefore dative experiencers would be an instance of overt movement of XP1 to position XP2. On this analysis we predict that the dative experiencer has dative case and that it does not agree with the verb. We predict that children produce dative experiencers as they produce clitic doubling at the early ages.

2.3.3. Floating quantifiers

Sportiche (1988) suggests that floating quantifiers are adjacent to a silent DP (trace, pro or PRO); this silent DP is bound by another the quantified DP. Following are some examples of quantifiers in French, taken from Sportiche (1992):

13.

(a) Les enfants on tous mangŽ

Les enfantsi on tous proi mangŽ

(b) Je les ai vu tous

Je lesi ai vu tous proi

(c) Marie les a tous pris

Marie lesi a tous pris proi

where Sportiche suggests an XP1/Cl analysis, under which tous has been floated under syntactic movement of an Accusative XP1 to its associated XP2. In Spanish there are quantifiers of the kind showed in (b).

If the structure of floating quantifiers and clitic doubling is the same, we expect children to produce floating quantifiers and clitic doubling at the same stage in development.

2.3.4. Scrambling

Sportiche (1992) suggests that clitic doubling has the same underlying structure as other constructions like scrambling. Sportiche proposes that the differences between various clitic constructions can be captured by the following set of parameters:

i. Movement of XP1 to XP2 occurs overtly or covertly

ii. The clitic is overt or covert

iii. XP1 is overt or covert

For Sportiche scrambling is the occurrence of an overt XP1 and a covert clitic, and the movement of XP1 to XP2 before LF, whereas clitic doubling involves an overt XP1, an overt clitic, and the movement of XP1 to XP2 at LF.

The following table exemplifies the properties of clitic doubling and scrambling:

Table 1. Properties of clitic doubling and scrambling

The distribution of the different objects is as follows: non-specific accusative objects raise to the [Spec, AGRoP] position, specific accusative objects raise to the [Spec, CldoP] position, and dative objects raise to the [Spec, ClioP].

If clitic doubling and scrambling are syntactically similar and everything else being equal, then we expect to find clitic doubling and left dislocation emerging at the same stage in development.

3. Method

3.1. Goals

This article investigates whether there is evidence of clitic doubling in early language, or whether, on the contrary, this construction does not exist in the early stages - the view suggested by Fox and Grodzinsky (1994) and Fox, Grodzinsky and Crain (1995).

The aim of this research is to find evidence for the productivity of clitic doubling in a range of different possible contexts, and to study the relation between the acquisition of clitic doubling and other constructions like quantifiers, dative experiencers, clitic left dislocation, floating quantifiers, scrambling in Dutch, and passives in English. The right use of the contexts for clitic doubling will be considered as evidence of the knowledge of this construction.

Our proposal predicts that children will produce clitic doubling in the correct contexts, that children will acquire floating quantifiers, dative experiencers, clitic left dislocation and scrambling in Dutch at the same age because they have the same underlying structure; it also predicts that children will acquire passives at a different age than clitic doubling because passives in English and clitic doubling are different constructions.

If children discriminate the contexts where clitic doubling can and must apply, correctly place the clitic and the coreferrential XP1, and use correct agreement for number, person and gender, then there is evidence for the existence of this construction. If children’s grammar does not have the construction of clitic doubling (Fox and Grodzinsky, 1994, Fox, Grodzinsky and Crain, 1995), then children cannot discriminate the contexts of finite and nonfinite verbs.

If children use clitic doubling productively while at the same age children have not acquired passives, then the proposal that these constructions are unrelated is borne out.

3.2. Design

The database employed was collected by L—pez Ornat (1994). The child, Mar’a, speaks Standard Spanish. The design is longitudinal, and the frequency of recordings was once per month. This is a study of spontaneous speech in a natural and representative situation. All data have been transcribed into the CHILDES format (MacWhinney and Snow, 1990). The age ranges between 1;7 and 3;11, and there is a gap between 3;1 and 3;6. The Age range is relevant because Fox, Grodzinsky and Crain postulate that at these ages children have not acquired clitic doubling yet.

The utterances taken into account were sentences containing a verb and a complement, either a clitic pronoun or a full Noun Phrase. The complements taken into account in the study of clitic doubling were non-pronominal direct objects, pronominal direct objects, non-pronominal indirect objects, pronominal indirect objects, dative experiencers, left dislocated direct objects, right dislocated direct objects, and indirect objects.

4. Results

4.1. Clitic doubling

If children know the structure of clitic doubling and the rules governing movement of the NP1, the child will double the NP1 where they should double and will not double where they should not double, or will not produce a high number of errors on the different contexts. Following Fox, Grodzinsky and Crain, if children do not have the structure of clitic doubling, then the child will not double the NP1 or they will commit many errors on the different contexts.

Instances of all possible contexts for clitic doubling occur in Mar’a's transcripts. The data show that the child knows the obligatoriness, optionality and impossibility of use of the different contexts. We have found instances of clitic doubling with Pronominal Direct Objects, where it is obligatory in the adult language:

14.

a. te he rega–ado a ti (M 2;3)

you-DAT have-1s scolded to you

(I have scolded you)

b. t? no me mires a m’ (M 2;9)

you -NOM not me-ACC look-2s at me

(don't look at me)

c. te veo a ti (M 3;11)

you-ACC see-1s to you

(I see you)

Also, we have found instances of clitic doubling with Non-Pronominal Indirect Objects, where it is optional in the adult language:

15.

a. se lo digo a mam‡ (M 1;10)

her-DAT it-ACC tell-1s to mom

(I tell her it)

b. le voy a dar arroz a mi ni–o (M 2;4)

him-DAT go-1s to give rice to my baby

(I'm going to give rice to my baby)

c. Qu’ta/ se /las t? a Irene (M 2;5)

take her-DAT them-ACC you-NOM to Irene

(take them from Irene)

d. le quieren romper las alas a Tobi (M 3;9)

him-DAT want-3p to brake the wing to Tobi

(they want to break Tobi’s wings)

Clitic doubling with Pronominal Indirect Objects also occurs in Mar’a's speech, where it is obligatory in the adult language:

16.

a. te voy a hacer una foto a ti (M 3;6)

you-DAT go-1s to take a picture to you

(I'm going to take a picture of you)

b. no te (lo) digo a ti (M 2;1)

not you-DAT (this) tell-1s to you

(I do not tell you this)

c. me la est‡n haciendo a m’ (3;10)

me-DAT it-ACC are-3p doing to me

(they are doing it to me)

Only one error has been found in Mar’a's transcripts: in a single instance, the child does not double the clitic for a Pronominal Direct Object, where it is obligatory in the adult language. The following is the only doubling error in Mar’a’s transcripts. It shows a failure to double when the adult language requires doubling. There are no instances of doubling where the adult grammar disallows doubling.

17. *voy a peinar a ti (M 2;5)

‘go-1s to comb to you’

ADULT: te voy a peinar a ti

you-ACC go-1s to comb to you

(I'm going to comb you)

The following table shows the number of utterances of doubled and non-doubled clitics for all the contexts and ages. The only error is shown in bold letters:

1;7 - 2;3 2;4 - 3;1 3;6- 3;11

A. INDIRECT OBJECTS

Pronominal (D Obl.)

Clitic-doubled 2 1 7

Omitted 1

Non-pronominal (D Opt.)

Clitic-doubled 6 9 14

Non clitic-doubled 1 1 3

Dative Experiencers (D Obl.)

Clitic-doubled 5 8

Non clitic-doubled

B. DIRECT OBJECTS

Pronominal (D Obl.)

Clitic-doubled 2 1 2

Omitted

Non-pronominal (D Imp.)

Clitic-doubled

Non Clitic-Doubled 182 255 198

Quantifier (D Opt.)

Clitic-doubled 4 1

Non Clitic-Doubled 1 1

Table 2. Productions of clitic doubling for all contexts (where D Obl. = Doubling Obligatory, D Opt. = Doubling Optional, D Imp. = Doubling Impossible).

The following table summarizes the frequency of clitic doubling in impossible, optional and obligatory contexts:

Table 3. Contexts for clitic doubling

Based on this data we can conclude that the child knows that Pronominal Direct Objects and Indirect Objects must double, Direct Object Quantifiers can be doubled, Non-pronominal Indirect Objects (with a goal or possessive theta role) can be doubled, and that Non-pronominal Direct Objects cannot double.

4.2. Other constructions

Mar’a produces dative experiencers, and in all cases she correctly doubles the clitic. Following are some examples:

18.

a. a m’ s’ me gusta (M 2;2)

to me-DAT me likes

(I like it)

b. a ti te gusta el lim—n? (M 2;8)

to you you likes the lemon?

(do you like lemon?)

c. a m’ no me gusta este cafŽ (M 2;8)

to me not me likes this coffee

(I do not like this coffee)

Also, instances of clitic left dislocation occur in Mar’a's speech at a very early age:

19.

a. ese cochecito lo ha comp(r)ado (2;3)

that car it-ACC has bought

(he has bought that car)

b. eso lo sab’as? (2;6)

that it-ACC knew-2s?

(did you know that?)

c. yo a ti te voy a hacer otra (3;9)

I to you-DAT you-DAT going-1s to do another one

(I am going to do another one for you)

Based on the data we can conclude that children have the structure of clitic-left dislocation and dative experiencers, and the rules of movement of the DP object to the position of Spec, FP and to the topic position. The theory we have been assuming claims that clitic-left dislocation, dative experiencers and clitic doubling make use of the same grammatical resources; therefore we expect that children who have clitic-doubling structures will also have the other structures. Data confirm this prediction. Future investigation will have to show whether other doubling theories make the same prediction. Of course it is possible that the structures developed at different but very early times, before the age at which Mar’a started speaking.

Mar’a also produces some instances of floating Quantifiers:

20.

a. me lo voy a comer todo (M 2;5)

myself it-ACC go-1s to eat all

(I’m going to eat it all)

b. lo llevo yo todo (M 2;5)

it-ACC bring-1s I all

(I bring it all)

c. te lo tomas todo (M 2;8)

yourself it-ACC take-2s all

(you take it all)

In the adult language, this is an optional environment for clitic doubling. This also seems to be the case for Mar’a; she also produces floating quantifier structures without clitic doubling:

21.

a. À(los) quieres todos?

them-ACC want-2s all?

(do you want them all?)

b. se (las) quer’a comer a todas (M 2;9)

himself them-ACC wanted-3s to eat to all

'he wanted to eat them all'

Based on the data we can conclude that the child knows that Dative experiencers must double with a quantifier, and that floating quantifiers can optionally double. Therefore, the child produces these constructions at the same time as clitic doubling.

Schaeffer (1995) has found that children scramble direct and indirect objects at a very early stage. However, this movement is optional in children’s grammar, unlike adult grammar. These data show that the construction of scrambling in Dutch exists in the early stages, although it does not show the obligatoriness of the movement.

The proposal made by Sportiche (1992) about the structure of clitic doubling and scrambling in Dutch is borne out by the the occurrence of scrambling on the acquisition of Dutch, but it does not explain the optionality on the early ages.

Note that in Schaeffer’s data, the children tended much more often to scramble pronouns than other NP’s. Perhaps they do not understand the specificity condition on scrambling NP’s, or better, do not understand when particular NP’s are specific. Since obligatory clitic doubling in Spanish mostly involves pronouns, perhaps it is the non-pronominal nature of some Dutch NP’s which causes children to not understand that they are specific. Perhaps the reason that children demonstrate knowledge of obligatory doubling in Spanish but not of obligatory scrambling in Dutch is that the Spanish cases are always pronominal³ .

5. Discussion

As we have already pointed out, Sportiche (1992) suggests that scrambling in Dutch and related languages shows a particular parametric variation on the structures of clitic doubling. Scrambling involves the left-ward movement of an object NP to a position in front of a class of adverbs which start outside the VP. Thus it seems to have in common with clitic doubling, on Sportiche’s account, the property that an NP moves left-ward. In particular, Sportiche suggests that whereas clitic-doubling structures in Spanish involve a visible clitic head and LF movement of the doubled NP, scrambling in Dutch involves an invisible clitic head and surface movement of the doubled NP to the [Spec,Clitic] position. Thus it might be worthwhile to compare what is known about the development of scrambling to what we have established about the development of clitic doubling.

Schaeffer (1995), in a preliminary study, showed that very young children (Niek, 2;7-3;5; Laura, 1;9-3;4) have scrambling in Dutch, in that they, at least sometimes, scramble NP’s which the adult language must scramble. However, Schaeffer’s data seems to show that the scrambling is not obligatory in children even where it is obligatory in adults. This result, if it is substantiated, should be set beside our result which shows that a very young child obligatorily shows clitic doubling when the adult grammar does. The question is: why should clitic doubling be essentially obligatory from the beginning whereas scrambling is not? However, we should take into account the fact that children of Schaeffer's study may be in a previous stage of development, although ages are similar, because developmental stages and ages do not always correlate.

On Sportiche’s account, if the relevant properties are types of movement, then both clitic doubling and scrambling have the same types of movement, except that in clitic doubling the movement is LF movement (i.e. after spell-out) and in Dutch it is surface movement (i.e. before spell-out). Is there any reason to believe that children treat adult surface movement as optional, i.e. that they sometimes do LF movement instead of surface movement? As far as we know, data are contradictory. If we look at a language like French, where there is both in situ and moved wh, Crisma (1992) has found that young children produce structures with moved wh rather than in situ wh; however, Hamann (personal communication) has found that a French-speaking child acquires moved wh much earlier than in situ wh.

The other parameter (on Sportiche’s account) that distinguishes Dutch scrambling from Spanish clitic doubling is whether or not the clitic head is visible or phonetically empty. Could it be that scrambling shows some delay because of a problem with an invisible clitic? Is there some problem with invisible elements in general?

This is possible, although it seems to us not very likely. Consider, for example, null subjects, which are standardly taken to involve an invisible pro. There is good evidence that in pro-drop languages like Italian, young children from the earliest ages know that the language is pro-drop, and have no problems with the invisible subject (for evidence which shows that Italian and English-speaking children know the relevant differences on pro-drop, see Bromberg and Wexler (1995), Wexler (1995)). Thus one can’t say that there is a problem with invisible elements in general. Could there be a problem with invisible heads which are not full phrases, so that the problem is the invisible clitic head? Perhaps, but we know of no evidence.

Thus there seems to be some developmental difference (if Schaeffer’s preliminary data is correct) between scrambling in Dutch and clitic doubling in Spanish, such that the former is only sometimes done and the latter is done correctly, and it seems to be difficult to account for this difference in terms of the parameters proposed in Sportiche’s theory. Of course we do not know why scrambling is only optional, if it is, in young Dutch children. The one point, however, that the two constructions do have in common is that they are both attested very early. If movement of the doubled NP is obligatory at LF in Spanish, it is surprising that this movement is obligatorily done in children in Spanish, but that obligatory surface movement is not carried out in scrambling in young children in Dutch. We leave the question here.

With regard to the acquisition of passives in English Fox and Grodzinsky (1994) and Fox, Grodzinsky and Crain (1995) have suggested that the acquisition of passives is based also on the acquisition of clitic doubling. These authors propose that passives in English and clitic doubling in Spanish are the same construction, and predict that the acquisition of clitic doubling must be delayed as well as the acquisition of passives is.

Following Baker, Johnson, Roberts (1989), Fox, Grodzinsky and Crain assume that the passive morphology is an argument and is 0-marked in Infl. The passive morphology -en, which is a clitic, forms a chain with the NP. After cliticizing on to the verb, the passive morphology transfers the thematic role to the NP.

Fox, Grodzinsky and Crain describe thematic role assignment in passives as following: for actional be-passives (i.e. those with agent or instrument by-phrases) the preposition assigns agenthood to the NP; also there is a syntactic mechanism that transfers the external compositional thematic role to by-phrases. For non-actional be-passives, the non-agentive by-phrase has to receive the external thematic role of the VP. For get-passives, the by-phrase does not receive the external compositional thematic role of the VP.

Fox and Grodzinsky (1994) studied the acquisition of passives in English and they found that children were good at non-truncated actional be-passive, bad at non-truncated non-actional be-passive and good at truncated non-actional be-passive.

Fox and Grodzinsky (1994) propose that the problem with non-actional be-passive by-phrases, occurs because the by-phrase has to receive the external thematic role of the VP; the passive morphology, which is an argument, receives the external thematic role from the verb, and transfers the thematic role to the by-phrase through a process of clitic doubling.

Fox, Grodzinsky and Crain conclude that in passive constructions children do not apply the process that transmits the subject theta role to the subject: this is the reason for the failure in non-agentive passives. For these authors, children have not yet acquired non-actional be-passives because they do not transfer the thematic role of the subject to the by-phrase (this transfer is considered the same structure as clitic doubling), or because they do not transfer the thematic role in the particular passive construction.

Based on the data on the acquisition of clitic doubling that we have analyzed in this study, it seems that either clitic doubling is a different structure than passives; or, assuming that both constructions have the same structure, that bad performance on non-actional be-passive results from some independent problem with this particular passive construction.

6. Concluding remarks

The aim of this article is to study the early acquisition of clitic doubling in Spanish. The contexts, status and productivity of this construction have been tested.

Instances of all contexts have been found: we have found instances of clitic doubling with Pronominal Direct Objects, Pronominal Indirect Objects and dative experiencers, where it is obligatory in the adult language, and Non-pronominal Indirect Objects, where it is optional in the adult language. However, we have not found instances of clitic doubling with Non-Pronominal Direct Objects, which is impossible. Therefore, any explanation on the acquisition of passives in English based on an early lack of the clitic-doubling construction does not apply.

The proposal made by Sportiche (1992) about the structure of clitic doubling is consistent with the data on the acquisition floating quantifiers because instances of this construction have been found in Spanish. Schaeffer (1995) has found that children scramble direct and indirect objects in a very early stage, although this movement is optional in children’s grammar. These data show that the construction of scrambling in Dutch exists in the early stages, although children do not show knowledge of the obligatoriness of this construction. However, we should take into account that they may be in a different stage of development.

We have also found support for our hypothesis from the acquisition of clitic-left dislocation and dative experiencers: instances of both have been found in the speech of Mar’a.

We can conclude that data on the acquisition of clitic constructions, clitic-left dislocation, dative experiencers and floating quantifiers support the hypothesis that these constructions have the same underlying structure, and that the delay on the acquisition of passives is not due to the lack of the construction of clitic doubling. Most important of all, we have shown that the very young child whom we have studied has excellent knowledge of the complex grammatical processes which are responsible for clitic doubling. So far as we know, this is the first demonstration of this knowledge in young children.

7. References

Baker, M., Johnson, K., Roberts, I., (1989) Passive arguments raised. Linguistic Inquiry, 20, 219-251.

Belleti, A. (1994) Case checking and clitic placement. Three issues on Italian/Romance clitics. Manuscript. University of Geneva.

Belletti, A. (to appear) Italian/Romance clitics: structure and derivation. In H. Riemsdijk (ed.) Clitics in the Languages of Europe.

Belleti, A. and Rizzi, L. (1988) Psych verbs and theta theory. Natural Language and Linguistic Theory, 6, 291-352.

Borer, H. (1984) Parametric syntax. Dordrecht: Foris.

Bromberg, H. S. and Wexler, K. (1995) Null subjects in Wh-questions. In C. Sch?tze, J. Ganger, K. Broihier (eds.) Papers on Language Processing and Acquisition. MIT Working papers in Linguistics, 26, 221-247.

Ceccheto, C. (1995) Reconstruction in clitic left dislocation. Unpublished document. DIPSCO, Instituto S. Raffaele, Milano.

Cinque, G. (1990) Types of A'-Dependencies. Cambridge, MA: MIT Press.

Crisma, P. (1992) On the acquisition of Wh-questions in French. Geneva Generative Papers, 1, 115-122.

Fox, D. and Grodzinsky, Y. (1994) Children’s passive: A view from the by-Phrase. Manuscript, M.I.T. and Tel-Aviv University.

Fox, D., Grodzinsky, Y. and Crain, S. (1995) An experimental study of children’s passive. In C. Sch?tze, J. Ganger, K. Broihier (eds.) Papers on Language Processing and Acquisition. MIT Working papers in Linguistics, 26, 249-264.

Franco, J. (1992) On object agreement in Spanish. Ph.D. Dissertation. University of Southern California, Los Angeles.

Friedemann, M.A. (1992) The underlying position of External Arguments in French. Geneva Generative Papers, 1-2, 123-145.

Guasti, T. (1993) Verb syntax in Italian child grammar. Language Acquisition, 3, 1-40.

Hamann, C., Rizzi, L. and Frauenfelder, U. (1995) On the acquisition of the pronominal system in French. Recherches Linguistiques, 24, 83-101.

Jaeggli, O. (1982) Topics in romance syntax. Dordrecht: Foris.

Jaeggli, O. (1986) Three issues in the theory of clitics. In H. Borer (ed.) The syntax of pronominal clitics. Orlando, Fla: Academic Press.

Jakubowicz, C., M?ller, N., Kang, O.K., Rigaut, C., Riemer, B. (1995) On the acquisition of the pronominal system in French and German. Paper presented at the 20th Annual Boston University on Language Development. Boston, Mass.

Kaiser, G.A. (1993) More about INFL-ection and Agreement: The acquisition of Clitic Pronouns in French. In J. Meisel (ed.) Bilingual First Language Acquisition. Amsterdam: Benjamins.

Kayne, R. (1975) French syntax. Cambridge, Mass.: MIT Press.

Kayne, R. (1991) Romance clitics, verb movement, and PRO. Linguistic Inquiry, 22, 647-686.

Kayne, R. (1995) The antisimmetry of the syntax. Cambridge, MA: MIT Press.

MacWhinney, B. and Snow, C. (1990) The CHILDEs project: computational tools for analyzing talk. Carnegie Mellon University: Pittsburgh, PA.

Montrul, S. (in progress) Thematic roles and functional projections in adult SLA. The issue of L1 Influence in Interlanguage. Manuscript. McGill University, Montreal.

Rivas, A. (1977) A theory of clitics. Ph.D. Dissertation. MIT, Cambridge, Mass.

Schaeffer, J. (1995) On the acquisition of scrambling in Dutch. In D. MacLaughlin and S. McEwen (eds.) Proceedings of the 19th Annual Boston University Conference on Language Development, Somerville, Mass: Cascadilla Press, 521-532.

Sportiche, D. (1988) A theory of floating quantifiers and its corollaries for constituent structure. Linguistic Inquiry, 19, 425-449.

Sportiche, D. (1992) Clitic Constructions. Manuscript. UCLA.

Torrens, V. (1995) The acquisition of Syntax in Catalan and Spanish: The Functional Category Inflection. PhD Dissertation. University of Barcelona.

Uriagereka, J. (1995) Aspects of the syntax of clitic placement in western romance. Linguistic Inquiry, 26, 79-123.

Varela, A. (1988) Binding in Spanish: A theoretical and experimental study.. Ph.D. Dissertation, University of Connecticut.

Wexler, K. (1995) Feature interpretability and optionality in child grammar. Paper presented at the “Workhshop on Optionality“. University of Utrecht (The Netherlands).