Clitic right dislocation in English: cross-linguistic influence in multilingual acquisition

Megan Devlina (devlin-m3@email.ulster.ac.uk),Raffaella Follia (r.folli@ulster.ac.uk),Alison Henrya (AM.henry@ulster.ac.uk),and Christina Sevdalia (c.sevdali@ulster.ac.uk)

aUniversity of Ulster,Jordanstown CampusShore Road,Newtownabbey,Co. Antrim,Northern Ireland.BT37 0QB

Corresponding Author:Email: r.folli@.ulster.ac.ukTel: 00442890366615Postal Add: School of Communication, University of UlsterJordanstown CampusShore RoadNewtownabbeyCO. AntrimBT37 0QB

AbstractThis paper investigates target-deviant constructions produced in English by a child acquiring three languages simultaneously from birth: English, Italian and Scottish Gaelic. These constructions involve the weak pronoun ‘it’ doubling direct object DPs which are dislocated to the right-periphery of the sentence. Additionally, we consider subject dislocations, where the child dislocatesthe subject to the left and the right periphery of the sentence, and doubles it with a variety of pronouns. We argue that both of these constructions are produced as a result of cross-linguistic influence from Italian, where dislocations in general and clitic right dislocations in particular are very frequent in the adult input. We

1

analyse these constructions as involving adjunction. Finally, regarding the wider question of cross-linguisticinfluence and the vulnerability of the C domain, we show that the notion of vulnerability is not necessarily tied to presence of the C domain: while cross-linguistic influence happens with dislocation phenomena exactly because they are syntax-discourse phenomena, it is clear that children can produce them even before they acquire the C-domain fully.

Highlights Cross-linguistic influence as a feature of

multilingual acquisition. CLRD and subject dislocation constructions are

transferred into child’s English from Italian. CLRDs and subject dislocation constructions are

analysed as involving adjunction in both child and adult language.

It has a clitic-like behaviour in English. This data provides further support to the vulnerable

domains hypothesis in the more general sense as the phenomena in question are indeed syntax-discourse phenomena.

KeywordsClitics, dislocation, clitic-right dislocation, multilingual acquisition, cross-linguistic influence, frequency, input, adjunction.

1. IntroductionIn this paper we discuss an unusual doubling constructionwhich appears for a considerable period in the English ofa trilingual child (henceforth S) who has been acquiring English, Italian and Scottish Gaelic from birth. The child uses the pronoun it alongside a DP object which is dislocated to the right edge of the clause.

(1) a. He forget it the teddy (2; 7)b. We will make it bed (2; 7)c. He’s give it back the muffin (2; 9)

2

d. Have to go touch it his tail. (2;10)

This unusual structure is clearly not characteristicof adult English nor is it commonly observed inmonolingual English acquisition1. The construction differsfrom a typical transitive sentence in that a semanticallyand syntactically weak pronoun, it, doubles the object DPwhich is then dislocated to the right. This target-deviant construction has a number of interestingproperties: first, the structure involves a weak element(it) in the canonical object position; second the objectDP is located in the right periphery of the clause;third, it is clearly an information structuring device ofsome kind. As these constructions are not acharacteristic of child English, it appears that they maybe the result of cross-linguistic influence from thechild’s Italian. Specifically, we argue, they resembleItalian Clitic Right Dislocation structures (CLRD),exemplified below in (2) where a clitic is linked to theright dislocated topic ‘il libro’:

(2) L’ho letto, il libro.It-have-1sg read the book. ‘I have read it, the book.’

These constructions occur in the child’s spontaneousdata at a frequency of 6.4 % of all constructionscontaining direct object DPs, over a period of 15 monthsfrom the age of 2;4 – 3;7. Crucially the overallfrequency of the it-doubling constructions is well abovethe accepted 2% where errors are labelled as noise(Platzack, 2001: 365) in the child’s language. Indeedthere is a period of 8 months (between age 2;4 and 3)where the construction occurs with a frequency of 14.6%.

Furthermore, the duration for which theseconstructions occur (a period of 15 months) clearlysuggests that these target-deviant constructions cannot1 Right dislocation is possible in English but it is normally found with subjects and is in fact rare compared to left dislocation. This is indeed true also of some Romance languages but not of Italian where right dislocation is just as frequent as left dislocation (See Hidalgo R. xx for extensive discussion and for a cross-linguistic quantitative overview of right dislocation in English and Spanish).

3

simply be dismissed as performance errors. Rather, theconsistency of the target-deviant constructions in thechild’s output, and the frequency with which theseconstructions occur within this time frame as well as inwithin certain single transcriptions suggest that theyare representative of a grammatical option in the child’sgrammar. Additionally we also consider constructionswhere S dislocates subjects, both to the right and to theleft of the clause with a variety of pronouns (not justit). These constructions although grammatical in English,are not typical of monolingual acquisition. What isrelevant for us is that these constructions also involvea clause internal element doubled with a dislocated DP atthe edge of the clause, somewhat similarly to the itdoubling structures.

The analysis of these phenomena is part of a biggerstudy, Project S (Devlin et al, 2012, Devlin et al, 2013and Devlin, 2014), an extensive longitudinal case studyof the simultaneous trilingual acquisition of English,Italian and Scottish Gaelic from birth investigating theeffect that acquiring three languages simultaneously hason the child’s English, with a particular focus on cross-linguistic influence. Indeed project S has revealedcross-linguistic influence also in the production ofcomplex DPs (Devlin et al, 2013 and Devlin 2014) and theunaccusative/unergative distinction (Devlin at al, 2014).

1.1. Cross-linguistic influence and vulnerable domains

The study of multilingual language development is crucially centred around questions regarding the interaction of the languages as they develop. The interaction between languages in multilingual language development is referred to in the literature with different terms, e.g. interference, transfer, cross-linguistic influence. Interference is nowadays mostly used to refer to code-switching in bilingualism (e.g. seeKoppe and Meisel 1995) while the two latter terms are used interchangeably. One important difference between them is that transfer is more often used in the

4

literature that looks at the cognitive mechanisms that are at play in language selection in bilinguals. This study takes as a starting point the view that the interaction between languages can be viewed as cross-linguistic influence, following Döpke (1998), Hulk and Müller (2000), Müller and Hulk (2001)2, Serratrice et al (2004), Sorace and Serratrice (2009), Sorace (2011) amongothers. The challenge for researchers trying to model this phenomenon is to understand where cross-linguistic influence occurs and why it occurs in some domains ratherthan others as well as why cross-linguistic influence is not always consistently found across bilingual or multilingual populations3.

One popular theory of cross-linguistic influence is the vulnerable domains hypothesis.  Vulnerable domains inacquisition are those domains that are reported to take children longer to acquire.  Müller (2003) observes that cross-linguistic influence does not occur in every domainand that only some grammatical domains are vulnerable to cross-linguistic influence.  Müller and Hulk (2001) follow Platzack (2001) in proposing that the C-domain is the grammatical domain that is vulnerable in acquisition.  Its vulnerability lies in the fact that it is in the C-domain where syntax interfaces with other cognitive domains. Sorace’s (2000, 2004, 2005, among others) research in the field of bilingualism has focusedon the task of accounting for the selective cross-linguistic influence observed in different bilingual populations.  Research spanning over a decade from Soraceet al (Sorace, 2004; 2005; Sorace and Filiaci, 2006; Sorace et al, 2009; Serratrice et al, 2004; Tsimpli et al, 2006; among others) examines the existence of cross-linguistic influence in bilingual populations, including

2 Müller and Hulk (2001) make a distinction between transfer and cross-linguistic influence and argue that transfer refers to direct interference by the dominant language, while cross-linguistic influence can be indirect and purely due to the existence of structural ambiguity in domains that are considered as vulnerable, and is not related to dominance. 3 For a recent evaluation of the cross-linguistic influence hypothesis and a discussion of the phenomena that would be important to consider in relation to it see the issue of Linguistics Approachesto Bilingualism 2011, 1.1.

5

L2 adults, older L2 children and simultaneous bilinguals,across various language combinations. Numerous studies from Sorace et al and from researchers including Tsimpli and Sorace, 2006, 2008; Sorace and Serratrice, 2009; Serratrice et al, 2004; Belletti et al, 2007; Filiaci et al, 2010; (among others) show that the notion of interface is relevant in understanding the nature of thisphenomenon, but also that not all interfaces are equally problematic in acquisition.  The interface hypothesis putforward by Sorace and colleagues is that the syntax-discourse interface is the most vulnerable domain in multilingual acquisition, while internal interfaces such as the syntax-semantic interface and the syntax-lexicon interface should not be affected. In a similar vein, Hulkand Müller’s (2000) study of two bilingual children (a Dutch-French and a German-Italian bilingual) found that cross-linguistic influence occurred in object drop constructions, another phenomenon related to the C-domainand the syntax-discourse interface. They report that as the C-domain is gradually acquired, the rate of target-deviant object omission falls. Müller and Hulk (2001) report that whilst the development path of object drop isthe same for monolingual and bilingual children, target-deviant object omission occurs at a higher rate in bilingual children than monolingual children. Crucially also bilingual children remain in the transitional development phase for much longer than monolingual children.

1.2. It doubling as cross-linguistic influence

The data we discuss in this paper brings further evidenceto support this hypothesis. In the paper we argue that it-doubling structures are a result of cross-linguistic influence at the syntax-discourse interface: in particular they are the result of S transferring CLRD structures from Italian into English. We show that they are due to the high frequency of dislocation structures in one of S’s languages. We follow De Cat (2007) in analysing them as the result of a simple operation of adjunction to the top-most clausal projection, and hence

6

independent at first of the presence of an elaborated clausal structure involving a C layer. One crucial twist is that this happens at a time where S is not using clitics in her Italian, a fact that is unsurprising givenwhat is known about the relatively late acquisition of clitics in the monolingual acquisition of Romance languages (see Hamann et al 1996, Haegeman 1996, Hamann, 2002 and Leonini, 2006 among others)4. If indeed this is acase of cross-linguistic influence, we need to address the following questions:

a. What exactly gets transferred in theseconstructions?

b. Why is cross-linguistic influence observed in thisarea of grammar?

c. How does this relate to prominent views of cross-linguistic influence, such as the “vulnerabledomains hypothesis”?

The paper is organized as follows. In section 2 wepresent information about the subject and the methodologyof the study. Section 3 presents the data, and outlinesthe main proposal. Section 4 discusses and dismissesalternative accounts for the data. In section 5, wepresent our proposal on the ingredients of cross-linguistic influence in more detail; in particular wediscuss the role of it as an English clitic, and we alsopresent and discuss S’s lack of clitics in Italian atthis stage. In section 6, we focus on the cause of cross-linguistic influence and we relate our analysis toprevious work on dislocations, their acquisition and

4 One might wonder whether it is surprising that transfer from one language to the other could occur when a functional element is not available, a clitic, and a construction, here clitic dislocation, is not yet acquired. There is ample evidence in the literature from bothobservational and research data which points to the superiority of comprehension over production in acquisition. Additionally it has been shown that children may omit functors in production which are nevertheless available in the underlying representation (Bloom 1970 and Gerken, Landau and Remez 1990). Finally recent work by Samek-Ludovici (2014) shows that clitic right dislocation may not necessarily involve a clitic element. We thank a reviewer for asking us to make all these possible explanations more explicit.

7

adult CLRD in Italian. In section 7 we discuss theimplications of our analysis for theories of cross-linguistic influence which adopt the notion of vulnerabledomains. Section 8 concludes.

2. Subject and methodology

The data in this paper is part of a larger longitudinalstudy of simultaneous trilingual language acquisition. Sis acquiring English, Italian and Scottish Gaelic frombirth within a “one parent – one language” setting(Ronjat 1913), with the mother interacting with the childin Italian, the father in Scottish Gaelic. English is theshared language of the parents, the language of thechild’s nursery which she has been attending since theage of 0;5, and the language of the wider community. Thechild is growing up in Belfast, while her father speaksCanadian English to the mother who is a second languagespeaker of British English.

The study began when the child was 2;3 and finishedwhen she was 4;9 but the data considered in this paperextends from 2;3 to 3;7. Over this period of 30 months,we captured the vital stages of language developmentthrough weekly recordings of S’s English in a free-playenvironment. The duration of these weekly recordingsessions ranged from 45 minutes to 1 hour 15 minutesdepending on the task at hand that day and the mood ofthe child. The tapes were transcribed using CHAT bytranscribers trained in linguistics and CHATtranscription and 48% of these were checked by a secondtranscriber. All files are now available on the CHILDESdatabase as Project S.

English is S’s most dominant language in terms ofMLU and other characteristics of development. S’s MLUdevelopment is comparable to other monolingual English-speaking children. Figure 2 shows the development of theMLU of Barbara, a monolingual English-speaking child fromthe Belfast corpus (Henry, 1995) and Adam, a monolingualEnglish-speaking child from the Brown Corpus (Brown,1973) and S.

8

Figure 1: A comparison of the MLU development of Barbara,Adam and S

2;4

2;6

2;9

3;2

3;5

3;7

3;11

0

2

4

The MLU development of S, Adam and Barbara

Barbara MLUAdam MLUSofia MLU

Age

As clearly seen in Figure 1, S’s development is entirelyin keeping with the two monolingual children selected.Overall, it appears that S’s development is in the middlerange of her monolingual peers. The choice of comparingS’s MLU development to these two children is based on twoconsiderations: first, Adam’s data is well known andwidely used in the literature; second Barbara’s data is acorpus of Belfast English which is the language varietyused in S’s wider community and that of her nursery .

Devlin (2014) also reports on S’s overallacquisition of English in terms of Brown’s (1973)language development stages (see Appendix B). Her dataanalysis is drawn from a consistent mapping of S’sEnglish to the findings of monolingual English L1acquisition from the literature. This is done by focusingon well-studied linguistic landmarks such as acquisitionof word order, root infinitives, functional morphemes andquestion formation among others. The main overarchingfinding is that S’s language development is, by andlarge, in keeping with monolingual acquisition. SeeAppendix C for a mapping of S’s stages of Englishacquisition to Brown’s original (1973) languagedevelopment stages.

9

3. The data

The child in this study regularly used structuresexemplified further in (3) below: in these structures thepronoun it occurs together with a full DP object whichappears on the right periphery of the sentence. We referto this construction as it-doubling. A list of moreinstances of the it-doubling constructions is given inAppendix 3.

(3) a. You find it my elegetch5 (2;7)

b. He’s give it back the muffin (2;9)c. There have it the big scary dinosaur

(2;10)d. He scrubbing it this (3;4)e. He didn’t eat it the big bunny (3; 7)f. He eat it all up that.(3;7)

At about the same stage of development S alsoproduces other types of dislocations involving thesubject which although grammatical in English are notnormally found in monolingual acquisition at this stageof development (see references and discussion of therelevant literature in section 6.1):

(4) a. He's cross daddy.(2;9)

b. But he's gonna sleep baby. (3;2)c. Margaret he was make it. (3;2)d. The shark they eat you all up. (3;7)

In the next section we analyse in more detail theproperties of it doubling.

3.1 It doubling

5 Elegetch is a lexical loan from Scottish Gaelic’s airgid, the genitiveform of the noun money.

10

It-doubling constructions occur regularly in the child’sdata over a period of 15 months (from 2;3 to 3;7) and area productive feature of the child’s grammar. During thisperiod, 6.4 % of DP objects are doubled with it.

Table 1: Frequency of it-doubling constructions in S’s spontaneous dataTotal no. directobjects

Total no. it-doubling constructions

Total Frequency

671 43 6.4%

Also between the age of 2;4 and 3, hence for a period ofeight months, it-doubling occurs 14.6% of the time. Thisfigure excludes pronominal objects as they cannot bedoubled6. This figure cannot be regarded as “noise” in thedata: according to Platzack (2001: 365) noise iseverything below 2%.

Table 2: Frequency of it-doubling constructions in S’s spontaneous data by months

Age inmonths

Totalno.directobjects

Total no. it-doublingconstructions

TotalFrequency

TotalFrequency

28-30 42 6 12.1%

14.6%

31-33 111 20 15.3%34-36 46 9 16.4%

37-39 188 5 2.6%40-43 241 3 1.2%

6 There is no evidence of pronoun doubling by S or by Adam, and equally none in the adult data of the relevant sessions. Equally pronoun doubling in English seems rather rare:?Her. she doesn’t like cheese?You, you should help himAdmittedly pronouns can be doubled by clitics in Italian. We thank a reviewer for asking us to look into this and to be more explicit.

11

The following table shows the distribution of it-doubling in the data, taking into consideration the fine properties of the doubled element.

Table 3: Distribution of the it-doubling constructionAge of child inMonths

TotalDirect Objects

It-doubling Total it-doubling FrequencyTotal +An

i-Ani

Sg Pl

28-30 42 6 1 5 4 2 12.1%

14.6%

31-33 111 20 4 16 15 5 15.3%

34-36 46 9 5 4 6 3 16.4%

37-39 188 5 1 4 4 1 2.6%40-43 241 3 1 2 3 0 1.2%

These constructions are not representative of L1 Englishacquisition or adult English but they are clearly veryproductive in the child’s grammar.

One important question that needs to be addressedearly on deals with the uniqueness of this construction.In the Brown Corpus (Brown, 1973) we find that Adamproduced similar constructions to those discussed in thisstudy, where an object is doubled by the pronoun ‘it’. Wecan see a couple of examples of this from differenttranscriptions below:

a. CHI: Like it (.) screwdriver? (Adam 01) b. CHI: Mommy get it my ladder. (Adam 01)

c. CHI: Drop it cowboy. (Adam 09)

We scanned the entire Adam corpus and we found that thistype of construction is produced 2.05% of the time, hencevery close to the noise level value discussed above.

Table 4 : Frequency of it-doubling constructions in the Adam corpusTotal no. direct Total no. it-doubling Total

12

objects constructions Frequency2676 55 2.05%

The target-deviant constructions observed in the Adamcorpus closely resemble those under examination in thispaper as they contain an object doubled by the pronoun it.The properties of these constructions in Adam’s data areas follows: an object DP is doubled by the objectpronoun it; the doubled object can be dislocated both tothe right periphery and the left periphery of the clause;the object DP is a bare noun in the majority ofconstructions; the target-deviant constructions occurmost frequently with imperative verbs. In table 5 and 6below we give first a table for Adam’s data which givesthe equivalent breakdown in months. Table 6 compares S’sdata and Adam’s data side by side taking intoconsideration the various types of doubled DP:

Table 5: Distribution of the it-doubling construction in the Adam dataAge of child inMonths

TotalDirect Objects

It-doubling Total it-doubling FrequencyTotal +An

i-Ani

Sg Pl

27-30 935 31 1 30 30 1 3.3%31-33 774 16 1 15 10 6 2.1%34-36 967 8 2 6 8 0 0.8%

Table 6: Detailed comparison of dislocated DPs in theAdam and the S corpus

Det+N Poss+N

Dem+N Bare sg +count

Baresg -count

Plural

Proper N

Dem Pronoun

Adam 0% 8.9% 5.4% 62.5% 1.8% 12.5%

3.6% 5.4%

S 30%

(9.3%‘a’

4.7% 4.7% 16.3% 9.3% 25.6%

2.3% 7%

13

20.9 % ‘the’)

Despite the similarities between the two sets of data,there are also clear differences which are important toconsider. In the case of Adam’s constructions, the objectis predominantly bare, while S uses it-doublingproductively with full DPs of various kinds. Anothercontrasting characteristic of the two target-deviantconstructions is that in Adam’s case the object DP can bedislocated to both the left periphery and the rightperiphery. This contrasts with the it-doublingconstructions produced by S where dislocationconsistently occurs to the right of the clause. Finally,in the Adam corpus these constructions occur mostfrequently in an imperative context with verbs such as getor fix. These fixed expressions, like “get it” or “fixit”, involving one monosyllabic verb and an unstressedpronoun could be arguably analysed as one word, and hencenecessarily not involving a derivation like the onediscussed in this paper. Finally, and crucially, it-doubling constructions in the S corpus differ from thosefound in the Adam corpus data with regards to thefrequency: in the Adam corpus this construction is muchless frequent (2.05%7) and occurs in the data for ashorter period of time (2;3 – 2;10). Given all this, wetreat S’s it-doubling constructions as unique in theacquisition literature, and arguably a feature related toher multilingualism.8

The properties of S’s it-doubling are summarised in thetable below:

7 This frequency was calculated in the same manner as the it-doubling calculations in the S corpus. The frequency is calculated by workingout the total number of direct objects produced by the child and calculating what percentage of these direct objects occurred in it-doubling contexts.8 We have also checked the Belfast corpus and none of the eight children in the Belfast database shows any instance of it-doubling.

14

Table 7: Properties of the it-doubling constructionIt-doubling construction

1 The construction contains a pronoun and an object DPthat co-refer.

2 The pronoun appears in the canonical objectposition.

3 The object DP appears to be dislocated to the rightperiphery of the clause.

4 The dislocated DP can take the form of a bare nounand occasionally a demonstrative9.

5 The pronominal element is always the pronoun it (i.e.it does not appear to agree with the dislocated DPin natural gender, or number).

6 The pronoun co-occurs with both singular and pluralDPs.

7 The pronoun most commonly refers to inanimatereferents or lexical items that can be referred toas it in adult English, i.e. “baby”.

8 In some instances, there is a perceptible pausebetween the pronoun and the dislocated element10.

Let us proceed to discuss some of these propertiesin more detail. One interesting characteristic of theconstruction is that it is being used to double not onlyinanimate DPs, but also singular animate DPs, where onewould expect a doubling element to be him or her, and evenplural DPs, where one would expect the doubling elementto be them.

a. I don’t like it carrots (2; 4)b. Go and get it the toys (2;8)c. You fix it the trousers

(2;11)

9 It-doubling with demonstratives makes up 6% of the it-doubling constructions.10 While our data did not suit an instrumental prosodic analysis because of its acoustic quality, the perceptual analysis was done in conjunction with a trained phonetician and has confirmed the existence of a perceptible pause in a considerable number of cases. We would like to thank Aveen Hassan for her help with this.

15

Note that at this time, S is using feminine, masculine,singular and plural pronouns correctly in other contexts,hence indicating that her inventory of English pronounsdoes not contain only it.

a. INV: Is Emily your friend? (2;7) CHI: Mm-hmm. Don’t like her

b. CHI: I go find them Another interesting property is that the doubled element can be a full DP (5a), a bare noun (5b) or a demonstrative (5c):

(5) a. He forget it the teddy (2;7)

b. Mummy give it back muffin (2;9) c. I want to clean it that (2;9)

As mentioned above, this is different from what is found in the Adam corpus where the object is predominantly barewith this type of construction.

Also the constructions occur with 16 different verbsincluding particle verbs (See Appendix D for a complete list of verbs):

(6) a. I gonna put it away baby.(3;6)

b. He eat it all up that. (3;7)

Evidence that the it-doubling construction is productive in the data also comes from instances of repair where the child repairs an error in an it-doubling construction but keeps the it-doubling component of the construction in the repaired sentence thus showing that this construction is part of the child’s grammar and is the target construction for the child at this time:

(7) CHI: Find it elegetch, mama (2;7)DAN: Oh does your mummy have it?CHI: Have it it a elegetch. CHI: Have it a elegetch.

16

Examining the context in which the target-deviant constructions from this study are produced may help to understand the information structuring function of these constructions in S’s data. For example consider the following examples:

(8) ALI: The duck is eating the banana (2;4) CHI: I don't like it the duck.

These particular examples show clearly that the doubled element encodes information that is already present in the discourse, in other words given information that is the topic of the conversation. Example (12) shows that the child places the old information to the right in order to give emphasis to the preceding information of the clause.

(9) CHI: Ok, you want muffins? (2;9) MEG: Yes.

CHI: You don’t want the muffin?DAN: S[…] , look!CHI: he’s give it back muffin.DAN: who has to give back muffin?CHI: Mummy give it back muffin11.

Above, the child produces two variants of a yes/no question without it-doubling. This is then followed by an utterance containing it-doubling which has a new information element introduced out of the blue, namely the subject of the sentence (he). Here S seems to adopt the it-doubling strategy in an attempt to clearly background the object of the sentence to perhaps emphasise the new information present in the clause. Thisstrategy is used in the final utterance also where the child gives more information regarding who is giving backthe muffin (namely Mummy). Similarly, example (13) also shows the child repeating a construction and placing the old information to the right of the clause and doubling it with it, allowing the preceding information to have emphasis:

11 An anonymous reviewer points out that this could be a repetition effect. 2;9 seems to be late for an explanation based on repetition.

17

(10) DAN: Aw what happened the duck? (2;10)CHI: My broken it. I broke it the duck.

3.2 Subject Dislocations

As we mentioned above, at this stage of development S produces another type of dislocation construction:

(11) a. He's cross daddy (2;9) b. But he's gonna sleep baby (3;2) c. Margaret he was make it (3;2)

d. Animals they go in theres (3;3)e. The shark they eat you all up (3;7)

The table below summarises the properties of these constructions.

Table 8: Properties of dislocation constructionsDislocation constructions

1 Dislocation occurs with subjects2 This construction involves different pronouns and

demonstratives realised in their canonical position,plus a dislocated DP

3 The dislocated DP occurs both to the left and the right periphery of the sentence (unlike in the previous it-doubling construction which involved onlydislocation to the right)

4 The pronoun agrees in gender and number with the dislocated DP

5 These constructions occur at a frequency of 5.6% in the period from 2;6 – 4;3

This set of data involves dislocation structureswhich include a number of different pronouns, unlike withit-doubling. Also, as was not previously seen,dislocations occur both to the right and to the left,even though dislocations to the right are more frequent.S produces dislocation constructions at a higherfrequency than previously reported in the literature forL1 English, namely 5.6%. Crucially for us, these

18

structures are similar to the it-doubling data analysedabove in that they also involve a form of doubling, wherethe element doubled here is the subject, which is eithera pronoun or a full DP. The dislocated element appears onthe left or on the right. As we will see in some detailin section 6.1, dislocation constructions, althoughgrammatical in adult English, are acquired later inEnglish monolingual acquisition (differently for examplefrom French monolingual acquisition, see Van der Lindenand Sleeman 2007).

In summary, the data sets presented here appear toshow two types of dislocation constructions, one non-typical of adult English. Both constructions are also notattested in monolingual English acquisition at thisstage. The first construction involves exclusively thepronoun it doubling DP objects which are dislocated to theright. The frequency and duration of the it-doublingconstructions in the data clearly show that this target-deviant construction is productive in the child’sgrammar. Evidence that the it-doubling construction isproductive in the data also comes from instances ofrepair where the child corrects an error in an utterancehaving used this construction but keeps the it-doublingcomponent of the construction in the repaired sentence.Additionally, we have seen that S produces dislocationsof subjects both to the left and to the right.

3.3 Towards a proposal

In this paper we argue that both of these constructions are the result of cross-linguistic influence from the child’s Italian. Specifically, it-doubling resembles Italian CLRD, exemplified in (2) above and in (12) below where a clitic is linked to the right dislocated topic ‘il libro’12:

12 An anonymous reviewer asks whether it is possible that what is transferred from Italian is not CLRD but hanging topics (HT). An example of a hanging topic in Italian from Fascarelli (2000: 237) is given below:

Io penso che razzismo, ce n’ è in

19

(12) L’ho letto, il libro ‘It-cl have-1sg read the book’‘I have read it, the book’

Like Italian CLRD, the target-deviant it-doubling constructions produced in S’s English contain a right-dislocated constituent, together with a weak element, thepronoun it, which might be considered the most ‘clitic-

I think-1 sg that racism there of-it be-3 sg in

percentualepercentage“As for racism, I think there is a percentage of it”

The main properties of hanging topics (HT) which are still a matter of extensive debate in the literature (see Beninca (1988), Fascarelli(2000), Beninca and Poletto (2004) and Belletti (2008) among many others) are the following: HTs involve a left-peripheral element that is fairly detached from the clause. This peripheral element refers to an argument inside the clause, which is often, but not always, realised as a clitic. Belletti (2008) claims that a topic in a clitic dislocated structure is more connected to the clause, while a HT has a somewhat more separate, independent status. On the theoretical side, it is argued that there can only be one HT per clause, while clitic left dislocation allows multiple topics (see Samek-Ludovici 2014) and only in the left periphery (and not on the right). Moreover, HTs are argued to be some sort of adjunction structure where the HT is base-generated in the left-periphery and isnot moved, and the topic is a (caseless) NP, i.e. there is no real agreement (connectivity) between the left-peripheral topic and the clause-internal (optional) clitic. Finally, HTs are often analysed asinvolving adjunction plus co-reference at LF, rather than a proper syntactic dependency (like CLRDs do). Considering this possibility isclearly important. However, we remain of the opinion that the it-doubling constructions analysed here are transfer from CLRD and not HT in Italian for the following reasons:

i. Crucially HTs are thought to appear exclusively on the left (Fascarelli 2000) unlike S’s doubling constructions which are found on both sides but mostly on the right.

ii. HTs are argued to involve a left-peripheral element intonationally separated by the rest of the clause with a pause, essentially marking a separate, independent element (Belletti 2008). The perceptual analysis carried out on S’s data did reveal a pause in some but not all cases, making the analysis of our data as HT structures quite unlikely.

iii. Finally, as we saw in examples (11), (12) and (13) the discourse context of S’s it-doubling structures is that of

20

like’ of English pronouns13. One difference to note however is that the English pronoun it does not occur in the expected preverbal position for a clitic in Italian, something that clearly suggests that this is not a simplecase of word-for-word surface transfer from one language to another. With respect to the subject dislocations, thehypothesis of cross-linguistic influence from Italian is related to the occurrence of a construction in S’s English which is typically very rare in monolingual English acquisition at this stage of development (as argued explicitly by Van den Linden and Sleeman (2007) and further discussed in section 6.1). Accordingly, here we claim that S is producing a higher rate of dislocations in her English due to the presence of extensive dislocation in the Italian input. The exact nature and reason for this case of cross-linguistic influence lies at the heart of this paper. Before we get into the specifics of our proposal of how this construction arises as the result of cross-linguistic influence from Italian, let us briefly present and subsequently reject possible alternatives to account for our data.

4. Alternative explanations In this section we first briefly review possible analysesof this phenomenon that relate to input (alternatives

topicalization. However, HTs are more akin to ‘aboutness topics’ in the sense of Frascarelli (2000), again dissimilarto the constructions that we are dealing with here.

These reasons make the reviewer’s proposal quite unlikely. However, if the reviewer was right and the source of the cross-linguistic influence observed was indeed HTs, the essence of our argument would in fact remain the same: S’s it doubling would still be the result of cross-linguistic influence of a construction found frequently in Italian. 13 We will come back to the clitic-like nature of it in English in section 5.1.

21

one and two), grammar (alternatives three and four and six) and processing (alternative five).

One plausible suggestion is that the it-doubling structure might occur in the child’s input. As noted above, the child’s mother, an L1 Italian speaker, speaks to S in Italian, but to S’s father in English, and also uses English if there are English-speaking visitors in the household. One could then wonder if the structure under investigation here could be present in the mother’sL2 English; however, when looking into parental reports this is not the case14.

A second alternative explanation of the target-deviant constructions in this study is that it could be a transitivity marker which appears with verbs taking direct objects. Such transitivity markers are found for example in Pacific Creoles, as in the following example from Pijin (Lefebvre 2011: 19):

(13) dig-em siton dig trans stone “dig up a stone”

This would be an alternative explanation for why it can double a plural DP as in the following examples:

(14) I don’t like it carrotsI want it animals

Lefebvre (2011) notes that while transitivity markers tend to occur in Creoles with a substrate language which also has such markers, they also occur in languages such as Australian language-based creoles whose substrate doesnot show these markers. Thus it might appear that transitivity markers are somehow unmarked in Universal

14 One reviewer would like this claim to be supported by a quantitative analysis of the mother’s input. However since the motherspeaks to S exclusively in Italian this request would require the analysis of data from the parents’ interaction in English. This data is not available. However, as we mention in the text, we have explicitly asked the father of S to comment on the production of thistype of structure in the mother’s English and he has confirmed that she does not produce it-doubling.

22

Grammar terms, and so tend to occur when a creolisation situation arises and children form a creole on the basis of an underlying pidgin. However, if transitivity markerswere in some sense natural in UG terms, and UG-unmarked forms were prone to occur in multilingualism, one might expect them to be reported as occurring in many cases of multilingual acquisition. This has not been found to be the case however, making this explanation unlikely.

Third, a possible explanation15 for our data is that the child analyses verbs and the it element together as one word: for example makeit. If this was the case, we presumably would not expect the verb+it combination to occur with many verbs. In contrast, as we saw above, it-doubling is found with 16 different verbs including particle verbs, hence making this hypothesis unlikely (see Appendix D). Additionally, in this case, we also would not expect to find the simple use of the verb and the verb+it combination in free alternation but this is notthe case:

(15) a. I don’t like it carrots (2;3) b. Christina you like this (2;6)

(16) a. I have it some more (2;9) b. No, don’t have chocolate (2;6)

Fourth, a priori it is conceivable that S might be using it as a kind of determiner, so that it teddy might, forher, constitute a DP with it misanalysed as a D. This would be plausible if it was in general immediately followed by a noun or adjective-noun sequence, but this is not the case. There are examples where it precedes a D element, presumably showing that it cannot itself be analysed as a determiner.

(17) a. He forget it the teddy(2;7)b. He broke it the duck (2;10)c. Then I gonna be get it a tree

(3;3)

15 Thanks to Ludovica Serratrice for bringing this to our attention.

23

A further reason why it cannot be analysed as a determineris that it can be separated from the noun by another element such as a particle in a verb-particle construction.

(18) a Mummy he’s give it back muffin (2;9)b. I gonna put it away the baby (3;6)

Fifth, one could wonder if S uses the it-doubling strategy as a device to give herself more time to access the lexical item needed as the object of a sentence. Thus, potentially in saying for example He don’t like it dinosaur S could be using it as a filler while she is searching forthe word dinosaur in her mental lexicon. It has been shown that lexical access is slower in bilinguals than in monolinguals (for example as argued by Soares and Grosjean, 1984 among others) and presumably a third language could make access even slower, so a strategy that gives extra time to find a lexical item would be plausible. However, we can see from the data that this does not seem to be the function of it, as we saw when we discussed the context of these constructions in examples (11) – (13). S uses it-doubling when the word that is doubled has just been mentioned in the previous turn, or indeed in the same turn by S herself, making it very unlikely that what she is experiencing is a word finding difficulty:

(19) CHI: Ok, you want muffins?(2;9)MEG: Yes.CHI: You don’t want the muffin?DAN: S[…] , look…CHI: He’s give it back muffin.DAN: Who has to give back muffin?CHI: Mummy give it back muffin.

(20) DAN: Aw what happened the duck?(2;10)

CHI: My broken it. He broke it the duckc. DAN: A hair bobble?

24

CHI: Hair bobble? I don’t have it hair bobble (2;10)

Finally, an alternative explanation of the target-deviantdoubling construction from this study is that this construction could represent a stage of development that occurs in some or all children acquiring English as an L1(in a way that is reminiscent of Roeper ‘s (1999) defaultgrammar hypothesis). What then would make it more noticeable in this study is that because S is acquiring three languages simultaneously from birth, her development could be argued to be slower in this respect than monolinguals, hence causing her to remain in this stage for a longer period. Above we have mentioned that asuperficially similar construction is found in the Adam corpus (but importantly only in this monolingual corpus).Crucially frequency, length of use of this construction and the fine properties of this construction in the Adam corpus seem to clearly indicate that we are not looking at the same phenomenon. Moreover, Devlin (2014) shows on the one hand that cross-linguistic influence is found in several grammatical domains but that S’s overall languagedevelopment is entirely in line with monolingual children(see Appendix B for details) hence allowing us to excludean analysis along the lines hinted at above.

5. Analysis - part I: the ingredients of cross-linguistic influence

Above we have seen that S uses a structure unattested in monolingual acquisition as an information packaging device to background given information. In thissection we outline our proposal in more detail. As previously mentioned, we argue that it-doubling is the English equivalent of CLRD and is due to cross-linguisticinfluence from Italian. The question that immediately arises from this is how exactly S transfers a strategy that crucially needs (pronominal) clitics in a language that does not have any. This is the role of the followingsection.

5.1 It as an English clitic

25

Arguably English has clitic-like elements including auxiliaries, negation, genitive markers. This limited list of English clitics contains mostly prosodic clitics,namely elements that cannot form a phonological word whenused in isolation, lacking as they do prosodic structure at the level of the word (Anderson 2005). These are also referred to in the literature as simple clitics and are distinguished from pronominal or special clitics (Zwicky 1977). This second type of clitics are typically syntactic clitics, namely elements that take syntactic positions that are special in a number of ways and that cannot be filled by the corresponding free elements of the language (e.g. Romance pronominal clitics). In this paper we are concerned with a case of syntactic clitic but we are using the word clitic in a different, essentially looser way as we explain below.

It is also well known that all third person pronounshave a strong (I like John and her) and a weak variant (I like ‘er), with the notable exception of it which is only weak. However, as we mentioned above, English does not have Romance-type cliticisation where a clitic pronoun can appear in a non-canonical pre-verbal position. Also weak pronouns in English cannot be doubled by a dislocated agreeing element16, so in that sense English lacks a crucial component for a CLRD-like construction. In this paper, we argue that the weak pronoun it is the element that takes the role of the clitic in S’s constructions. The fact that it plays a special functional role (with no agreement) in this type of construction is evidenced by the fact it is the only element used in this fairly wide data set despite S having other pronouns at this stage (as seen above in examples (7a) and (7b). The lack of agreement between it and the dislocated phrase might be surprising and in fact may lead one to think that these constructions could perhaps be cases of hanging topics which crucially lack connectivity. We have discussed above in footnote 12 why we do not think this to be the case.

What about it then and its lack of agreement with thedislocated phrase? One important finding in the literature which seems to be relevant to the issue at 16 We thank an anonymous reviewer for pointing this out to us.

26

hand is the wide-spread cross-linguistic tendency of children to use default pronominal forms (Radford 1990, Clarke 1985, De Cat 2002). Even more importantly for us though, Thavenius (1983) carries out a large corpus studyof pronouns and shows that it is indeed the most common personal pronoun in anaphoric use in child language. Crucially then a child might have acquired other pronominal forms (and we have seen this indeed to be the case with S), but we would still expect to see overuse ofit in anaphoric contexts like in the structures analysed in this paper. 17

Another difference between S’s transferred CLRD and Italian CLRD is the word order of the two constructions: the word order in S’s data is the expected one in a language that does not allow Italian-type pronominal cliticisation, with it occurring post-verbally, never pre-verbally. Constructions like the following are therefore never found in the data.18

(21) a. *He it forget the teddy b. *He it broke the duck

The explanation for the issues above, we would argue, lies in the nature of it in English. The pronoun it is the weakest pronoun in the English system. As a pronoun, it is referentially deficient and cannot introduce a new discourse referent: its interpretation isalways dependent on some referent already present in the discourse. It is also syntactically weak since it cannot befocalised via stress in Standard English (22a), cannot bea direct object pronoun if the indirect object is a full

17 We thank an anonymous reviewer for asking us to be more explicit onthe issue of connectivity. One further (very good) question this reviewer raises is related to the fact that in the subject dislocation briefly analysed in section 3.2 there is connectivity. Wehave no clear explanation for this, other than that there is clearly an asymmetry here between subjects and objects which would be interesting to explore further in future work.18 This is of course not entirely surprising given well-known facts about children generally not making mistakes with basic sentence wordorder. Also, as we argue shortly, S has not yet acquired at this stage the properties of clitics and cliticisation in Italian.

27

DP (22b) and is not grammatical in verb-particle alternations (22 c-g):

(22) a. John gave me THE DOG (not the cat) b. *John gave me IT (the dog, not the cat) c. ?*Mary gave the girl it d. Mary ate her dinner up e. Mary ate up her dinner f. Mary ate it up g. *Mary ate up it

It, as previously suggested by Cardinaletti & Starke (1999), differs from other weak pronouns in English making it most akin to clitics, even though it does not itself show the typical behaviour of a Romance clitic. For example, unlike other English pronouns, it cannot be conjoined:

(23) a. I saw John and him b. *I read the book and it

Nor can it be focused and preposed:

(24) a. Them, I don’t like b.*It, I don’t like

Thus, it, while not occurring pre-verbally like Italian object clitics, and while not agreeing with the dislocated DP like in Italian CLRD construction, is nevertheless the most clitic-like of the English pronouns19. It is therefore not surprising that the childuses it as the equivalent of Italian clitics in a dislocation structure.

5.2 Clitics in Italian at this stage of developmentS was also recorded in her other two languages during thestudy, although not as often (on average one recording a month is available for Italian in the period under 19 Or as an anonymous reviewer puts it it is the orimary default member of the English pronominal paradigm (also in featural terms: cf. 3rd person, inanimate and/or neutral gender, singular number). Wethank the reviewer for this suggestion.

28

consideration here)20. Interestingly, in the first period of language development under consideration here, the child’s Italian shows no use of clitics, even where thesewould be obligatory in adult Italian.

(25) a. M: Si chiama robottinoReflexive clitic call-3sg small robot

S: Come chiama?(2; 7)

How called? (Target: Come si chiama?)

b. M: Come ti chiami?How you-cl call-2sg

S: Chiamo S (2; 7) I am called S… (Target: Mi chiamo S)

Also from when S was very young, her mother has sung to her an Italian lullaby Ninna nanna ninna oh questa bimba a chi la do (Ninna nanna ninna oh this girl to whom her (I) give) which contains a clitic la in each line of the song. In two of the recordings at 2:7 and 2:9 the mother and S both sing the lullaby in turn to a doll and then an animal and S’s reproduction of the song does not contain clitics. This clearly indicates that S has not yet developed clitics at this age. In a recording then at age3;3 we find the following exchange which shows that S hasnow developed the use of clitic pronouns21:

(26) a. S: Ce l’ha big tummy!(3;3) There it has big tummy!b. S: Lei ce l’ha la tail! She there it has the tail

The delay in production of clitics in S is not surprisingas it is well known that the acquisition of object clitics occurs comparatively late in acquisition.

20 The recordings in Italian and Scottish Gaelic have not been fully transcribed and hence are not available in CHILDES.21 In this recording session mother and child are doing puzzles and Suses this construction with clitic right dislocation several times.

29

Research by Hamann (2002) and Leonini (2006) (among others including Hamann et al 1996, Haegeman 1996) into the acquisition of clitics in L1 in Italian report that there is a clear delay in the acquisition of these elements in comparison to determiners and other clitic pronouns such as locative clitics, subject clitics in languages that have them and reflexives22. Hamann et al (1996) report that the first regular use of object clitics in the Augustin corpus occurs at the age of 2;6. S is even further behind this trend, but again this is not surprising given, on the one hand, the complexity of the acquisition task she is facing, and, on the other, the lack of clitic pronouns in two out of three languagesshe is acquiring.

Accordingly, Müller and Hulk (2001) report that object drop and clitic omission is higher in German/Italian, Dutch/French and German/French bilingual children than in monolingual Italian or French children. In fact Müller and Hulk (2001) assume that clitic omission is a discourse strategy for bilinguals. The increased use of object drop and clitic omission in the bilinguals in this study in comparison to their monolingual Italian/French peers is reported to be a result of cross-linguistic influence from the topic-drop languages involved. In particular, Dutch and German use topic-drop as a discourse strategy and they claim that atthis stage of language development the bilingual children22 Related to the issue of acquisition of clitics is the issue of clitic placement. A lot of literature on Italian investigating the acquisition of clitic pronouns shows that by the time children start producing clitics they always place them correctly (Antelmi 1997; Cardinaletti and Starke 2000; Cipriani et al. 1993; Guasti 1993/4; Schaeffer 2000). However, there is another body of literature that shows that this is not always the case, at least not for all languages. For example Petinou and Terzi (2002) report that children overgeneralize enclisis in the initial stages of acquisition of Cypriot Greek. Grohmann (2011) further reports that while children acquiring Cypriot Greek place clitics perfectly by age 3 – 4, they somehow revert to a stage of target deviance by the time they go to school (age 5) where they place clitics in positions that seem to be affected by clitic placement in Standard Modern Greek. While this issue is not directly relevant to us here, it is important to note that acquisition of clitics and their correct placement in the clauseare two separate issues that do not always coincide. We thank an anonymous reviewer for bringing this issue to our attention.

30

have not yet acquired the language-specific morphosyntactic mechanism for full object omission in their French/Italian, i.e., cliticisation. As a way of fulfilling the not yet acquired discourse strategy of object omission in Italian/French, the bilingual childrentransfer the discourse strategy that they use in Dutch/German (topic-drop) into their Italian/French. Whatabout S?Why is she transferring the it-doubling construction from her Italian to her English23? We turn to this question next.

6. Analysis- part II: The why and what of our case ofcross-linguistic influence

23 An anonymous reviewer proposes the following alternative toaccount for the it-doubling construction discussed here: “the use of itmight be the equivalent of the null object in Italian (…)accordingly, the child would use the item that is mostly bleached inEnglish grammar (and which it has acquired), namely it, as thecounterpart of what is null in its Italian grammar”. This proposal isreminiscent of much recent literature on object omission and cliticomission in child language (cf. Schaeffer 2000; Guasti 1993/4;Pirvulescu 2006; Pérez-Leroux, Pirvulescu and Roberge 2008, Costa andLobo 2008 and Tedeschi 2009 among others). Specifically the idea putforward by Pérez-Leroux, Pirvulescu and Roberge (2008) is that allchildren go through an “optional object pronoun” stage, independentlyof whether they speak a language with clitics or not. This wouldindeed be a possible alternative analysis for our data. However,there are two issues with testing this proposal: most of theliterature that investigates this does not look into the frequenciesof object (or clitic) drop produced by children. Instead, theycrucially rely on data from elicitation tasks that test children’sacceptance of sentences with null objects. Unfortunately we cannottest this with S, since the normal age of such tasks in theliterature is between age 3 and 4, and she is past that. Moreover,Tedeschi (2009) having done a task of this sort with 3;9 and 3;11year old Italian L1 children reports that the results of herexperiment do not support the presence of a null object phase in theacquisition of child Italian. If that is true, then by extension wecould assume that S does not go through this stage either. Clearlythen this would not be a case of transfer. For these two reasons wecannot adopt this proposal, although we thank the reviewer for a veryinteresting suggestion.

31

In the sections above we proposed that the it-doubling observed in our data is an instance of cross-linguistic influence from Italian CLRD constructions. An important question in the literature has to do with the cause of cross-linguistic influence in multilingual acquisition. This is still very much an open issue with some researchers arguing that cross-linguistic influence is some sort of “pooling of resources” where the multilingual child uses aspects of grammar already acquired in one language to boost the acquisition of the other language (Gawlitzek-Maiwald and Tracy, 1996). Another prominent view is that cross-linguistic influenceis a result of processing limitations (Serratrice et al 2004), often associated with the vulnerability of one interface which is the locus of cross-linguistic influence, namely the discourse-syntax interface (Sorace 2000). Finally, Müller and Hulk (2001) argue that cross-linguistic influence is due to linguistic complexity. In particular, cross-linguistic influence is found in the domain of grammar where syntax interfaces with discourse,and occurs when the languages involved share some structural characteristics (this is what they call structural overlap). In this section we want to be more specific about the type of cross-linguistic influence observed in our data by answering the following questions: why does S transfer CLRD from Italian into herEnglish, and what makes this transfer possible? In answering this, we will pay attention to the frequency ofdislocation constructions in her Italian input, but also the function of these constructions in grammar more generally. Before we do that, we discuss some prominent literature on the acquisition of dislocations in monolingual and multilingual acquisition.

6.1 Dislocation in acquisition

Dislocations are information packaging devices. Information structure (IS) can be defined as ‘the tailoring of [that] utterance by a sender […] that reflects the sender’s hypotheses about the receiver’s assumption and beliefs and strategies’ (Prince 1981:224).

32

The breadth of this definition reflects the degree to which disparate areas of linguistic knowledge (semantics,syntax, prosody) are at play whenever a sentence is considered in its discourse context. The precise nature and structural correlates of this issue have been widely studied within each area. One other important issue extensively discussed in the literature is the place thiscomponent has within the architecture of grammar (information structure for Lambretch (1994), information packaging for Valduvi (1992), f-structure for Erteschik-Shir (1997)).

Clitic dislocation constructions and dislocation constructions generally are syntactic means to organise asentence into topic/comment. In general, we know that in a language the same truth conditional proposition can be packaged in a different way to express different perspectives: there is clearly a substantial difference between (27b) and (27c).

(27) a. Mary hates crunchy chocolateb. CRUNCHY CHOCOLATES Mary hatesc. Crunchy chocolate, Mary hates (it)

The specific IS phenomena under investigation here have to do with the process of topicalisation.

What about acquisition? There is comparatively little research on the acquisition of dislocation constructions and topicalisation in general and essentially no literature specific to Italian24. Crucially

24 The only previous research on Italian dislocation is from Bates (1976) and Baroni et al (1973). Bates (1976) observes that children acquiring Italian appear to use the comment-topic rule placing new information first in the sentence with the old-information in sentence final position. She also states that this reflects a progression from the one-word stage whereby children only express thecomment. Variation in these constructions is observed when the MLU is3;0. Baroni et al (1973) also show that the children in their study place the new information at the beginning of the clause with the oldinformation following. Both these pieces of research suggest that thedislocation or stranding of extra information to the right is a coping strategy that children use when their processing limits are exceeded. Since there is no more recent literature on the acquisitionof these constructions in Italian, we are using the results of Frenchacquisition to inform our hypothesis given the similarities in the

33

it has been shown that there is extreme variation in the age of acquisition and frequency of use of this type of structure across languages, with English-speaking children using this type of construction much later than for example French-speaking children (see below for a detailed discussion of this literature). Also this variation has been shown to be linked to input since languages vary in the frequency of these constructions inthe adult data. In particular, both Notley et al (2007) and Van der Linden and Sleeman (2007) show that dislocations are rare in the English input, while they are very frequent in French where they are used in 50-70%of the cases (Lambrect 1987, Blasco-Dulbecco 1999 and Notley 2004). In relation to acquisition, Notley et al (2007) and Van der Linden and Sleeman (2007) consider thesame data set from CHILDES and report that the monolingual English-speaking child in their study produces very few left-dislocation constructions, and no right-dislocation constructions between the age of 2;0-3;11 (with a frequency of 2-4% calculated out of all the possible dislocation structures). In contrast, Van der Linden and Sleeman (2007) show that the French-speaking child in their study produces a growing amount of dislocations, mostly right dislocations with a shift later on to left dislocations (again in line with adult input). Notley et al (2007) consider a larger data set for French monolinguals and come to the same conclusion.

Crucially, Notley et al (2007) also compare these results to data from French/English and French/Dutch bilinguals. They show that overall the results for French/English bilinguals show that these children produce more dislocation constructions than English monolinguals (13% right dislocation and 16% left dislocation, compared to 2-4% for the monolinguals). Theyconclude that bilingual children develop differently fromtheir monolingual peers in the domain of topic marking and they argue that this is an indication of a clear effect of cross-linguistic influence. Pérez-Leroux et al’s (2011) study is also a comparison of production of topicalisation in English and French monolinguals. Their study consists of a sentence completion task designed to

frequency and properties of these constructions in the two languages.

34

elicit the continuation of a topicalised structure with aleft dislocated object. Their findings are three fold: first, all children have difficulties with the syntactic demands of this construction; second, they get better at it with age; third, and crucially, French and English monolinguals are reported to develop differently. French-speaking children have an advantage possibly due to theirexposure to CLLD constructions. The non-target-like responses of the English-speaking children are again seento be due to the lower amount of dislocation in the input(following Yang 2002 and 2004).

In a similar vein, De Cat’s (2003) investigation of pragmatic competence in French monolinguals reports that children develop pragmatic competence at a very young age, using topics in more than 90% of obligatory contextsfrom as young as 2;6. De Cat (2000, 2007) shows that dislocations are pervasive in early child French, something which is clearly related to the fact that adultFrench contains frequent dislocations in all dialects forboth subjects and objects. Her proposal is that dislocations reflect a “relative syntactic simplicity” and are effectively a very efficient syntactic tool with a powerful expressive property. Also they are not derivedby movement to the left periphery which is shown not to be projected at this stage of acquisition. De Cat uses eight diagnostics to determine the projection of a C layer and she concludes that the CP is not projected until stage III, but children’s production of dislocationprecedes that. Accordingly, she argues that dislocations are instances of adjunction of dislocated DPs to the topmost maximal projection available to the child at eachstage, either the TP (in stage I and II) or the CP (in stage III). Her proposal is framed within a weak continuity hypothesis of acquisition (Pinker 1984) according to which UG is available to the child from the start, although not accessible in full in the initial stages. In line with this hypothesis, adult dislocations are also derived by adjunction. We return to De Cat’s proposal below.

In summary, from the literature we have learnt that (a) input determines whether children will use dislocation constructions early on; (b) there is

35

variation in languages and English monolingual children produce less dislocations than French monolinguals; (c) French/English bilinguals show a higher frequency of dislocation than English monolinguals due to cross-linguistic influence; (d) dislocation in early stages of development can be derived via adjunction.

6.2 It-doubling is adjunction How does this all relate to the data from our study,

then? S is acquiring two languages (English and Scottish Gaelic25) with a relatively infrequent use of dislocation in the input, and Italian, which on the other hand, uses clitic dislocations extensively. At this stage of development, in line with general findings in the literature, she has not developed the use of clitics in Italian (as we saw in section 5.2). In English, however, she makes extensive use of a structure that resembles CLRD in both its structure and crucially its Information Structure content. The extensive presence of dislocationsin her input, we argue, is the source of her developing it-doubling in English, which is essentially a useful interpretative device to express topics. S is not necessarily transferring the derivation of CLRD from her Italian which arguably involves movement and/or the existence of a CP layer. What she does transfer however, at least in the early stages, is the knowledge that a clause can contain one clause internal and one peripheralelement (referring to the same object) and that this is auseful device to structure information. Following De Cat(2007), we argue that these structures are in the first instance derived via adjunction to the topmost maximal projection acquired by the child. In order to determine which projection the doubling DP is adjoined to we have considered Devlin (2014) through discussion of S languagedevelopment26 and we have considered her data with respect to the C layer diagnostics from the literature (Müller 1994, Müller et al 1996, de Villiers et al 1991,

25 Scottish Gaelic has infrequent right dislocation constructions which are strongly marked (Andrew Carnie, p.c.).26 See Appendix B for a detailed presentation of S’s language development according to Brown’s (1973) stages of acquisition.

36

Hulk 1996, De Cat 2000, 2007). At the start of the it-doubling phase between the age 2;4 – 2;9, S does not display clear evidence for the existence of a C layer: inparticular, yes/no questions are encoded via raising intonation only, she has only few wh-words and most wh questions are repeated in the transcription and thereforeseem to be more like set phrases rather than productive derivation. Also there is only very little subject-auxiliary inversion. However, between 2;10 – 3;7, the CP layer seems now to be in place: wh questions are now productively used, there is extensive use of both finite and non finite embedded clauses, including adjunct and relative clauses, and she also uses several different complementizers. 27

Based on this data then we argue that until 2;10, S is producing the it-doubling construction by adjunction tothe TP on the right, doubling the dislocated DP with the only clitic-like element of English, the weak pronoun it. From age 2;10 onwards and until she stops producing theseconstructions she adjoins them to the CP level in an adult-like manner. 28 29 We adopt this two-step approach following De Cat and assuming with her that children produce dislocations by adjoining dislocated phrases to 27 An anonymous review suggests that rather than right or left TP/CP adjunction, early right dislocations may in fact involve a low, clause internal peripheral position, thus not involving the CP area at all (e.g. Cecchetto 1999, Belletti 2001, Van der Linen and Sleeman2007). According to this analysis, there is a difference between right dislocation and left dislocation where only right dislocation involves this low peripheral position. While this analysis could be used for our main data set, that is the it-doubling phenomenon, it would not easily lend itself to a unified analysis of the other dislocation data we discuss in this paper, namely the subject dislocation constructions in 3.2 above. For this reason whilst we think this is an interesting possibility we think a unified analysis is more appealing. 28 We want to thank an anonymous reviewer for hinting at this possibility in his/her review. 29 One anonymous reviewer asks the interesting and relevant question of how this structure is gradually acquired and then lost in S’s English. We have proposed that it-doubling starts as adjunction to TP,moves on to be adjunction to CP when this layer is available. At thispoint, we think, topicalisation, which is movement to CP in English (Rizzi 1997) becomes a possibility for the developing grammar of S, making the doubling plus adjunction option something superfluous.

37

the topmost projection available to them at each stage, namely TP and CP in turn. Note that this proposal is indeed in accordance with some analyses of CLRDs for adult Italian. There is a debate in the literature regarding the derivation of CLRD, with some researchers arguing for a movement analysis and some others arguing for an adjunction analysis. In the first camp, we find further differentiations: the low topic analysis (Cecchetto 1999; Belletti 2001, 2005) and the left-periphery clause analysis (Cardinaletti, 2002; Samek-Lodovici, 2006). These two approaches differ in their placement of the dislocated element: the low topic analysis keeps the dislocated element internal to the clause, whereas the left-periphery clause analysis placesthe dislocated element in a clause-external position. Crucially however, both these approaches derive CLRD through movement. On the contrary, there is a very different line of analysis that focuses on comparing the derivations of clitic doubling constructions, on one hand, and clitic left dislocations on the other. According to this view, the clitic is base-generated in its surface position as the complement of the verb in clitic doubling constructions, while in clitic dislocation constructions the clitic is an adjunct (Jaeggli (1982, 1986), Borer (1984), Anagnostopoulou (1994, 2005). While these analyses do not specifically discuss CLRDs, they imply that they are the “mirror” of CLLD constructions (cf. Valduvi’s (1992) for an explicit formulation of the mirror hypothesis and Cecchetto (1999)for a critical evaluation of this position). CLRDs are truly the mirror image of CLLDs, and would involve adjunction of the dislocated XP to the right of the node that would host a left dislocated XP. In essence then, assuming the continuity hypothesis, both the child language constructions analysed here and adult dislocations construction should be analysed as derived by adjunction.

This analysis of S’s it-doubling constructions can be naturally extended to the subject dislocation constructions we presented, where she “doubles” the subject, either to right or to the left of the clause. These constructions can be analysed again as adjunctions

38

to the topmost maximal projection produced by S, either to the left or to the right30. Again the Italian input is crucial, as left and right adjunctions are found in adultItalian as well. Finally, the reason behind the overwhelming frequency of right dislocations instead of left dislocations in the data reported here (exclusively with it-doubling constructions, and predominantly with subject dislocations) can be traced back to Van der Linden and Sleeman (2007) who noticed that cross-linguistically children seem to acquire right dislocations before left dislocations. The reason behind this preference is not entirely clear to us, but it is likely to be related to the well-attested finding that children acquire basic clausal word order fairly early, and presumably add any extra, discourse-driven material after but not before the main clause.

Summing up, we have analysed both the it-doubling and subject dislocation constructions found in our data as instances of adjunctions to the topmost projections acquired by the child at each stage. Moreover, we argued that they are the result of cross-linguistic influence from Italian, where dislocations are frequently found. These constructions are transferred as they are a simplerway to topicalise. In English, topicalisation is normallyexpressed through leftward movement to some projection inthe C layer. Here we argue that itdoubling structures involve in the first instance adjunction to the TP (clearly derivation by movement at this point would not be possible while adjuction has been shown to be a default mechanism in developing grammars by Lebeaux 1988,de Villier 1991, Roeper 1991 and De Cat 2007). In this sense they are the simplest type of topicalisation structure in the absence of a C layer31. Finally, arguably this analysis supports the weak continuity hypothesis whereby both child and adult dislocations are instances of the same phenomenon and are best analysed asinvolving adjunction. In the following section we explorethe implications of our analysis for theories that relate

30 De Cat (2000 and 2007) indeed finds both RD and LD in her French acquisition data and she analyses both as adjunction to the top-most projection either to the right or to the left. 31 We thank a reviewer for encouraging us to expand on this issue.

39

cross-linguistic influence to specific domains in grammar, namely the vulnerable domains hypothesis where syntax interfaces with discourse.

7. Analysis- part III: beyond the interface hypothesis.

Following discussion of the relevant literature and analysis of the target-deviant constructions produced in this study, we have proposed that S is transferring CLRD from Italian into her English for information structuringpurposes. The extensive presence of dislocation in her input is the source of her developing and using it-doublingfor an extensive period of time (15 months). The construction is clearly productive and preferred by S at times (remember the cases of repair discussed in section 3.1). This happens despite the fact that it-doubling is a construction that is in a certain sense costly in that itinvolves a numeration involving an extra element. How is it possible then? In the literature there is independent evidence that the notion topic is a primitive (Kiss 1995)and even more crucially that adjunction is one of the earliest structure building operations adopted by the child. It is essentially a default mechanism in developing grammars (Lebeaux 1988, de Villiers 1991, Roeper 1992 and De Cat 2007). All of this can help us understand the phenomenon at hand. What needs to be discussed further is why we see this in S as opposed to monolingual English-speakig children.

We argue that this study broadly speaking provides support for the interface hypothesis as the target-deviant constructions discussed here are a syntax-discourse interface phenomenon: the syntax-discourse interface is vulnerable to cross-linguistic influence dueto the demanding and processing-costly operation of co-ordinating the grammar-internal domain, syntax, with the grammar-external domain, discourse. As discussed at length in section 6, dislocations are information-packaging devices used in order to organise the clause interms of old and new information. As such, they are interface phenomena and it is entirely unsurprising that

40

they are susceptible to cross-linguistic influence. In this sense our findings provide further support to the large body of research discussed above that equates vulnerable domains with interface domains. Two caveats follow. First, the data in question are found before the instantiation of the C-domain in S’s language, in the period between 2;4 – 2;10 where she adjoins the dislocated DPs to the TP, and not to the CP. If we are right in proposing this, the original vulnerable domains proposal that links cross-linguistic influence to vulnerability in the C domain would need to be modified to account for the fact that children display such phenomena even before they acquire the C-domain fully. This is further supported by De Cat’s claim that syntax-discourse ‘[is] not automatically too costly for immatureproduction systems’ (De Cat 2007: 214). Although her study focuses on dislocations in French monolinguals, thedata from our study lend support to her claim and show that multilingual children attempt to relate their utterances to context early on in their language development, even when they do not have all the syntacticmeans at their disposal to do so (for example when the C layer is still not developed). In doing so they are susceptible to cross-linguistic influence between their language systems. Second, the notion of pooling of resources at the heart of the bilingual bootstrapping hypothesis seems to be useful to understand this unusual data set. Notice that this hypothesis rests fundamentallyon the idea that only certain grammatical contexts are subject to this pooling of resources and that in bilinguals two languages do not necessarily develop at the same pace in all areas of grammar: importantly for usthere are well-known differences, which we have discussedat lenght above, in the age of acquisition and frequency of use of dislocation in English versus Romance languages. All of this seems to show that while more workneeds to be done to arrive to a better understanding of this issue, it is clear that an exclusively interface-driven approach to cross-linguistic influence is perhaps too narrow, making more appealing Perez-Leroux’s (2011) more holistic approach to cross-linguistic influence focusing on other important factors such as exposure to

41

input, processing cost of constructions as well as frequency.

8. Conclusion

This paper focuses on two target-deviant constructions that are produced by S, a trilingual child acquiring English, Italian and Scottish Gaelic from birth. The two constructions involve (a) “doubling” of the pronoun it by a full DP object which appears on the right periphery of the sentence, “I like it the carrots”, and (b) subject dislocations to the right and left of the clause, also combined with some form of “doubling”, but this time witha number of different pronouns “He’s cross, daddy”. The interesting thing about these constructions is that whilethe former is not grammatical in monolingual English, thelatter is not typically available at this stage of development in L1 acquisition of English. In this paper, we have analysed both it-doubling and subject dislocation constructions as instances of adjunction to the topmost projections acquired by the child in each stage of language development, TP and then CP. Moreover we argued that they are a result of cross-linguistic influence fromItalian, where dislocations are very frequent in the adult language. These constructions are transferred as they are a simpler way to information-package a sentence.We adopt the weak continuity hypothesis whereby the child’s grammar should use the same rule types and basic elements as adult grammar: accordingly both child and adult dislocations are best analysed as involving adjunction. Finally, regarding the wider question of cross-linguistic influence and the vulnerability of the Cdomain, we have shown that the notion of vulnerability isnot necessarily tied to presence of the C domain. While cross-linguistic influence happens with dislocation phenomena exactly because they are syntax-discourse phenomena, it is clear that children can produce them even before they acquire the C-domain fully.

42

Appendix A

Brown’s (1973) stages of language developmentStage

MLU

1 1.0 – 2.02 2.0 – 2.53 2.5 – 3.04 3.0 – 3.755 3.75 – 4.55+ 4.5+

43

Appendix B

Table 1: A timeline of S’s developmentStage 2 – 3Age: 2;3 – 2;9 (Months: 27 – 33)

MLU: 2.7

Lexical Categories and word order

- Word Order is largely consistent with target with some instances of target deviance (see – adjectives and pronouns inthis section);

- Produces 3-10 word constructions;

- Early recordings show the continued use ofjargon which is untranscribable; these arerepresented in transcriptions by xx / xxx:‘xxx carrots’;

- Constructions consist of declaratives, questions, and imperatives; ‘he’s going to bed’;‘eat it’; ‘

- Co-ordinated constructions are also in use: ‘I put it in my mouth and you stay here with me’;

- Negation is produced in the following forms: ‘no’, ‘not’, ‘can’t’, ‘don’t’: ‘I no want my dinner’. The negative constructions are largely produced in the expected position between subject and verb;

- Yes/no questions are signalled by rising intonation; ‘ok baby, going to bed?’

- Few Wh-questions are in use.

- Early wh-question are possible set-

44

phrases: ‘what’s doggy?’ ‘what’s that?’, ‘where’s teddy’s hair?’, ‘where’s Kanga?’; ‘aw, why you upset?’; ‘where’s your shoes?’

- Complex adjectival and possessive DPs are also present. Adjectival constructions are produced more frequently than possessive constructions as expected

- Adjectival constructions are produced in both the attributive and predicative forms: ‘Pig is scary’, ‘dinosauro gave me sore fingers’, ‘the bumblebee was scary’, ‘they’re wet’;

- Adjectival constructions with target-deviant word order are also produced: ‘They’re all cats beautiful’. These constructionsare limited at this stage of development;

- Target-deviant possessive constructions are also produced but are not overly frequent at this stage: ‘that there flowers mummy’, ‘I’m just the friends pussycat’, ‘where’s baby’s cup of tea?’;

Functional Categories

- As observed in the question constructions above, the child uses the contractible copula at this stage of development. The contractible copula is not just restrictedto question constructions but also declaratives: ‘she’s so pretty’, ‘it’s a lovely cat’, ‘he’s sore’, ‘it’s really hot’;

- The uncontractible copula is produced in this development stage, however not present in 90% of obligatory occasions until later in language development: ‘this isfor you’, ‘elegetch is purple’, ‘this is for you lunch’;

45

- Articles ‘the’ and ‘a’ are used productively: ‘it’s in the cupboard’; ‘we go get a story’;

- The deictic pronouns this, that, are used productively in this stage of development:‘you want some of this?’, ‘what’s that?’;

- S produces all 1st, 2nd and 3rd person pronouns: ‘I want blanket’; ‘you have the socks’; ‘she have socks’; ‘he want to eat it’; ‘we were away to the museum’; ‘they look yummy’;

- It would appear that S has not yet acquired gender with the masculine person pronoun ‘he’ being favoured in production in contexts where the feminine ‘she’ is required: ‘He need to go to bed, doesn’t she? (withboth he and she refering to baby girl doll)’.

- Target-deviant production of the object pronoun it is observed in this stage of development. The object pronoun it is produced in constructions along with the object DP to which it refers: ‘he forget it the teddy’, ‘don’t wake it the baby’, ‘you sing it Winnie the Pooh’.

-- Functional morphemes ‘-ing’, ‘in’, ‘on’ and

plural ‘-s’ all used productively: ‘put the picture in it’; ‘she has to put pyjamas on’; ‘you’re not going to go gym’;

- Present tense auxiliaries are used only inquestion constructions: ‘are you ok?’, ‘how are you?’, ‘am I (dis)gusting?’. Semi-auxiliaries (Valian, 1991) are produced also in this stage: ‘wanna have a torta?’, ‘I gonna get it’;

46

- Auxiliaries omitted widely in this stage: ‘you no crying’,’ you no go to the gym’;

- Regular and irregular past tense is used in this stage: ‘oh no, see you forgot phones’, ‘I found the present book’;

- Over-extension of inflectional morpheme isobserved on both present and past tense verbs: ‘I wants a phones’;

- Target-deviant past tense constructions are produced in the last transcription of this stage ‘yea, he’s fall out (-he fell out)’, ‘he’s fall on legs like this (-he fell on his legs like this)’;

Development Stage 4Age: 2;10 – 3;7 (Months: 34 – 45)

MLU: 3.0

Lexical categories and Word order

- In this stage of development, language production is more complex and jargon-like language is no longer produced;

- Complex DPs – Production of compound nouns with some in target-deviant word order: ‘Hishead got a hair pineapple (pineapple hair)’, ‘that’s not his toy squirrel’;

- Produces comparative adjective: ‘he get all longer’;

- The production of complex sentences/ acquisition of embedding:

Relative clauses: ‘Em’s a girl who was in school’

Infinitival clauses: ‘he don't want to do like that’,

47

‘he doesn’t want to eat wee bit’.

Conjoined clauses: ‘because I like baby’s cow and I like mumma’s cow’;

Subordinate clauses: ‘he don’t really crying cos he don’t like his teddy’,

- Adverbs emerge: ‘he’s too big’, ‘he’s not anymore crying’.

Functional categories

- Possessive ‘s is fully acquired: ‘Mummy’s flowers’;

- Production of possessive pronouns emerge: ‘he don’t like his blue trousers’, ‘it’s only his bed’, ‘put your feet in’;

- Greater awareness of gender-marking: ‘(to baby) You want to go to bed? (to MEG) He wants to go to bed.’ Although arguably, as the baby isfemale and therefore the gender of the pronoun is still not target;

- As observed in stage 1, past tense constructions are produced in a form that resembles the past-continuous tense without–ing on the verb however, the context clearly shows that the child is targeting the simple past: ‘I was go to school’, ‘When at school… I was eat it’, ‘he was hit me’, ‘he was climb up on the tree (-he climbed…)’;

- Prepositions are produced to a greater extent in this stage of development: ‘I’m happy with you’, ‘but he can’t sleep with his hat’, ‘what you doing here on / here on my pond?’;

- Complementizers like because and if are produced: ‘nothing because E. didn’t see me’; ‘I’ll

48

take my shoe off if you’d like to eat it’;

- Both the uncontractible and contractible copula are produced in 90% of obligatory constructions at this stage of development;

- Are is used with plural nouns, i.e. number agreement has been acquired: ‘they are cats’;

- The production of pronouns is largely consistent while the targe- deviant it-doubling constructions are falling out of use at this stage of development.

Development Stage 5Age: 3;8 – 3;10 (Months: 46 – 52)

MLU: 4.0

Lexical categories and Word order

- Negative elements in question constructions are produced: ‘why’s this not working?’, ‘have you not put it in?’, ‘why did he not build a sandcastle?’;

- The production of so emerges: ‘you can sleep and be so cosy’;

- Target-deviant possessive constructions fall out of use at this stage of development.

Functional development

- Our and their are all used consistently: ‘youcan’t see our house’; ‘well they haven’t eaten their cake’;

- The reflexive pronoun ‘myself’ is observed:‘can I eat it by myself?’;

49

- Future is produced with the auxiliary will:‘he will just eat you’;

- Past tense –ed and regular 3rd person functional morphemes are fully acquired inthis development stage as well as the uncontractible auxiliary;

Development Stage 5+Age: 3;11 – 4;4 (Months: 53 - 56)

MLU: 5.2

Lexical development and Word order

- ‘Get’ passive emerges: ‘I don’t remember when he gets stunged’;

- Comparative adjective is produced: ‘and then all the children were making noise really louder’, ‘hmm, bigger?’;

- Target-deviant adjectival constructions fall out of use in this stage of development;

Functional development

- Reflexive pronouns more widely produced inthis stage of development: ‘we can bu [/] build it all by ourself’, ‘they all there by yourself’, ‘because the beetle doesn’t want to stay all by itself’;

- Subject-Auxiliary-Inversion produced regularly in simple wh-questions;

50

Appendix C

The following table provides more data of it-doubling constructions

It-doubling construction Recording (Age)I don’t like it carrots (4) S02 (2;4)I like it carrots S03 (2;4)Do you like it chocolate? S03 (2;4)We will make it bed S10 (2;7)You will sing it Winnie the Pooh (2)

S10 (2;7)

You find it elegetch Mama S10 (2;7)You find it my elegetch S10 (2;7)Have it a elegetch (2) S10 (2;7)He forget it the teddy S10 (2;7)Don’t wake it the baby (2) S10 (2;7)I don't have find it my box S10 (2;7)Find it my box S10 (2;7)He’s making it his bed S10 (2;7)Go and get it the toys S11 (2;8)He was eat it now the apples S11 (2;8)You ma [//] make it dinner S11(2;8)Who is it the dog? S12 (2;8)Well make it some milk S12 (2;8)Close it the door S14 (2;9)You want it cakes? S14 (2;9)Mummy, he’s give it back muffin

S14 (2;9)

Mummy give it back muffin S14 (2;9)The dog eat it some dinner S14 (2;9)I want to clean it that S15 (2;9)There have it a castle S15 (2;9)There have it the big scary dinosaur

S15 (2;9)

No pussycat, leave it the flowers

S16 (2;10)

Have to go touch it his tails

S16 (2;10)

Daddy get it big ice cream S16 (2;10)

51

Watch it my toes S17 (2;10)I want it animals S17 (2;10)He broke it the duck S17 (2;10)No no, don’t eat it paper S18 (3;1)You fix it the trousers S22 (3;2)He don’t like it dinosaur S22 (3;2)Then I gonna be get it a tree

S24 (3;3)

He scrubbing it this S25 (3;4)I gonna put it away the baby S28 (3;6)He eat it all up that S30 (3;7)He didn’t eat it the big bunny

S31 (3;7)

52

Appendix DList of verbs with which it-doubling occurs

- like- Make- Sing- Find- Have- Forget- Wake- Bring- Close- Want- Give- Eat- Want- Leave- Go- Get- Break- Scrubbing

List of particle verbs with it-doubling- Put away- Eat all up- Give back

53

References

Anagnostopoulou, E. 1994. Clitic dependencies in Modern Greek.Ph.D Dissertation, Salzburg University.

Anagnostopoulou, E., 2005. Clitic Doubling. In: Everaet, M., van Riemsdijk, H. (Eds.), The Blackwell Companion to Syntax, vol.1. Blackwell Publishing, Oxford, pp. 519-581.

Antelmi, D. (1997) La Prima Grammatica dell’Italiano.Indagine Longitudinale sull’Acquisizione dellaMorfosintassi Italiana. Bologna: Il Mulino.

Bates, E., 1976. Language and Context: The Acquisition ofPragmatics. Academic Press, New York.

Belletti, A., 2001. ‘Inversion’ as focalization. In: Hulk, A., Pollock, J,Y. (Eds.), Subject Inversion in Romance and the Theory of Universal Grammar. Oxford University Press, Oxford, pp. 60-91.

Belletti, A., 2005. Extended Doubling and the VP Periphery. Probus 17, 1: 1-35.

Belletti, A., Bennati, E., & Sorace, A. 2007.Theoretical and developmental issues in the syntax ofsubjects: Evidence from near-native Italian. NaturalLanguage and Linguistic Theory, 25, 657-689.

Blasco-Dulbecco, M. (1999). Les dislocations en français contemporain: étude syntaxique (Vol. 1). Honoré Champion.

Borer, H., 1984. Parametric Syntax: Case studies in Semitic and Romance languages. Foris, Dordrecht.

Brown, R. 1973. A first language. Cambridge, MA: HarvardUniversity Press.

Cardinaletti, A., 2002. Against optimal and null clitics.Right dislocation vs marginalization. Studia Linguistica 56 (1), 29-57.

Cardinaletti, A., and Starke, M. (2000). Overview: Thegrammar (and acquisition) of clitics, in S. M.Powers and C. Hamann (eds.), The Acquisition ofScrambling and Cliticization. Dordrecht. KluwerAcademic Publishers, 165-186.

Cardinaletti, A., Starke, M., 1999. The typology of structural deficiency: On the three language classes. In: van Riemsdijk, H., (Eds.), Clitics in the Languages of Europe, Mouton de Gruyter, Berlin, pp.

54

145-233.Cecchetto, C., 1999. A comparative analysis of left and

right dislocation in Romance. Studia Linguistica 53 (1), 46-67.

Cipriani, P., Chilosi A., Bottari P., and Pfanner L.(1993). L’acquisizione Della Morfosintassi: Fasi eProcessi. Padova: Unipress.

Costa, J. and Lobo, M. (2008). ‘Clitic omission in theacquisition of European Portuguese: Data fromcomprehension.’ In A. Pires and J. Rothman (eds.),Minimalist Inquiries into Child and Adult LanguageAcquisition: Case Studies across Portuguese. Berlin-New York: Mouton de Gruyter, 1-15.

De Cat, C. (2000). Structure building and the acquisitionof dislocations in child French. In Proceedings of BUCLD (Vol. 24, pp. 242-252).

De Cat, C. (2007). French dislocation. Interpretation, syntax, acquisition. Oxford University Press.

De Cat, C., 2002. French Dislocation. PhD dissertation. University of York.

De Cat, C., 2003. Syntactic manifestations of very early pragmatic competence. In: Beachley, B., Brown, A., Conlin, F., (Eds), Proceedings of BUCLD27. Cascadilla Press, Somerville, MA, pp. 209-219.

De Villiers, J. (1991) Why questions. In T. Maxfield & B.Plunkett (eds.) Papers in the acquisition of wh. Amherst: UMOP. 155 – 173.

DeHouwer, A., 2005. Early bilingual acquisition: focus onmorphosyntax and the Separate Development Hypothesis. In: Kroll, J. and de Groot, A. (Eds), The Handbook of Bilingualism. Oxford University Press, Oxford, pp.30-48.

Devlin, M. (2014) Cross-linguistic influence in theEnglish of a trilingual child: a case study oftrilingual acquisition. PhD dissertation: Universityof Ulster.

Devlin, M., Folli, R., Henry, A., & Sevdali, C. 2012.Clitic Dislocation in the Absence of Clitics: a Studyin Trilingual Acquisition.  University of Pennsylvania WorkingPapers in Linguistics: Vol. 18: Iss. 1, 40-50

Devlin, M., Folli, R., Henry, A., & Sevdali, C.(2013) Vulnerable Domains and Cross-Linguistic

55

Influence: The View from Trilingual Acquisition. InS. Stavrakaki, M. Lalioti, & X. Konstantinopoulou(Eds.), Advances in Language Acquisition, Cambridge ScholarsPublishing, 309-319.

Devlin, M., R. Folli and C. Sevdali (submitted) TheLexicon-Syntax Interface in 3L1 acquisition: AnExperimental Investigation of theUnaccusative/Unergative Distinction. In G. DeAngelis, U. Jessner and M. Kresić (eds.) Specialvolume on “Crosslinguistic influence andmultilingualism”.

Döpke, S. 1998. Competing language structures: Theacquisition of verb placement by bilingual German-English children. Journal of Child Language 25, 555-584.

Erteschik-Shir, N., 1997. The Dynamics of Focus Structure. Cambridge University Press, Cambridge.

Filiaci, F., Sorace, A., & Carreiras, M. 2013.Anaphoric biases of null and overt subjects in Italianand Spanish: A cross-linguistic comparison. Language andCognitive Processes, 1-19.

Gawlitzek-Maiwald, I., Tracy, R., 1996. Bilingual Bootstrapping. Linguistics 34, 901-926

Grohmann, K. (2011) Issues in and for the acquisition ofCypriot Greek. In E. Rinke and T. Kupisch (eds.) Thedevelopment of grammar: language acquisition anddiachronic change. Amsterdam: John Benjamins: 179 –201.

Guasti, M. T. (1993/94) Verb Syntax in Italian ChildGrammar: Finite and Nonfinite Verbs. Languageacquisition 3, 1-40.

Haegemann, L., 1996. Root infinitives, Clitics and Truncated structures. In: Clahsen, H. (Eds), Generative Perspectives on Language Acquisition. John Benjamins Publishing, Amsterdam/Philadelphia.

Hamann, C. 2002. From syntax to discourse: Pronominal clitics, nullsubjects and infinitives in child language. Dordrecht: KluwerAcademic Publisher.

Hamann, C., Rizzi, L., Frauenfelder, U., 1996. On the Acquisition of Subject and Object Clitics in French. In: H. Clahsen, H., (Ed.) Generative Perspectives on Language Acquisition, Benjamins, Amsterdam/Philadelphia.

56

Henry, A. 1995. Belfast English and Standard English: Dialect Variationand Parameter Setting. New York: Oxford University Press.

Hulk, A. (1996) The syntax of wh-questions in child French. In Philip & Winjen (eds.) Connecting children’s language and linguistic theory. 123 – 172.

Hulk, A., & Müller, N. (2000). Bilingual first language acquisition at the interface between syntax and pragmatics. Bilingualism: Language and Cognition, 3(03), 227-244.

Jaeggli, O. 1986. Passive. Linguistic Inquiry, 17, 587-633.Jaeggli, O., 1982. Topics in Romance Syntax. Foris,

Dordrecht.Lambrecht, K., 1987. On the status of SVO sentences in

French discourse. In R. Tomlin (ed.) Coherence and Grounding in discourse. Amsterdam: John Benjamins.

Lambrecht, K., 1994. Information structure and sentence form: Topic, focus and the mental representation of discourse referents. In Cambridge Studies in Linguistics 71. Cambridge University Press, Cambridge.

Lefebvre, C., 2011. Creoles, their Substrates and Language Typology. John Benjamins, Amsterdam.

Leonini, C., 2006. The acquisition of object clitics and definite articles: Evidence from Italian as L2 and L1.PhD dissertation, University of Siena.

Müller, N. (1994). Parameters cannot be reset: Evidence from the development of COMP. Bilingual first language acquisition: French and German grammatical development, 235-269.

Müller, N., 2003. (In)Vulnerable Domains in Multilingualism. John Benjamins, Amsterdam/Philadelphia.

Müller, N., Crysmann B. and Kaiser G. 1996. Interaction between the acquisition of French object drop and the development of the C-system. Language Acquisition 5: 35-63.

Müller, N., Hulk, A., 2001. Crosslinguistic influence in bilingual language acquisition: Italian and French as recipient languages. Bilingualism: Language and Cognition, 4 (1), 1-21.

Notley, A. (2004). The acquisition of topicalisation structures in French-English bilinguals: testing models of cross-linguistic influence. Unpublished ms.,

57

University of Amsterdam.Notley, A., van der Linden, E. H., Hulk, A.C.J., 2007.

Cross-linguistic influence in bilingual children: the case of dislocation. In Baauw, S., Drijkoningen, F., Pinto, M., (Eds.), Romance Language and Linguistic Theory 2005, Selected papers from ‘Going Romance’, John Benjamins, Amsterdam/Philadelphia, 229-259.

Pérez-Leroux, A. T. 2011. What I don’t understand aboutInterfaces. Linguistic Approaches to Bilingualism 1 (1), 71-73.

Pérez-Leroux, A. T., Pirvulescu, M. and Roberge, Y.(2008a). ‘A syntactic transitivity approach to nullobjects in child language.’ Lingua 118, 370- 398.

Pérez-Leroux, A.T., Pirvulescu, M., Roberge, Y., 2011. Topicalization and object omission in child language. First Language 31, 280-299.

Petinou, K. and A. Terzi (2002) Speech patterns inCypriot Greek late-talkers. Applied Psycholinguistics 27:335 – 353.

Pinker, S. 1984. Language learning and language development.Cambridge, MA: Harvard University Press.

Platzack, C. 1999. The Vulnerable C-domain. Brain andLanguage 77, 364-377.

Platzack, C. 2001. The computational system as aminimal feature driven device and the tripartiteTP/VP hypothesis of the universal clause. GLOWNewsletter 46, 46-47.

Prince, E.F., 1981. Topicalization, Focus-Movement and Yiddish-Movement: A pragmatic differentiation. Berkeley Linguistic Society 7: 249-64.

Samek-Lodovici, V., 2006. When right-dislocation meets the left-periphery. Lingua 116, 836-873.

Schaeffer, J. (2000). The Acquisition of Direct ObjectScrambling and Clitic Placement.Amsterdam/Philadelphia. John Benjamins.

Serratrice, L., Sorace, A., Paoli, S., 2004. Crosslinguistic influence at the syntax-pragmatics interface: Subjects and objects in English-Italian bilingual and monolingual acquisition. Bilingualism: Language and Cognition, 183-205.

Soares, C., & Grosjean, F. (1984). Bilinguals in a monolingual and a bilingual speech mode: The effect onlexical access. Memory & Cognition, 12(4), 380-386.

58

Sorace, A. & Filiaci, F. 2006. Anaphora resolution innear-native speakers of Italian. Second LanguageResearch, 22, 339-368.

Sorace, A. 2000. Gradients in auxiliary selection withintransitive verbs. Language, 76 (4), 859-890.

Sorace, A. 2004. Gradience at the lexicon-syntaxinterface: Evidence from auxiliary selection andimplications for unaccusativity. In A. Alexiadou, E.Anagnostopoulou & M. Everaert (Eds.), The unaccusativityPuzzle: Explorations of the syntax-lexicon interface. Oxford: OxfordUniversity Press, 243-268

Sorace, A. 2005. Syntactic optionality at interfaces. InL. Cornips & K. Corrigan (Eds.), Syntax and variation:Reconciling the biological and the social. Amsterdam: JohnBenjamins, 46-111.

Sorace, A. 2011. Pinning down the concept of “interface”in bilingualism. Linguistic Approaches to Bilingualism, 1 (1),1-33.

Sorace, A., & Serratrice, L. 2009. Internal and externalinterfaces in bilingual language development: Beyondstructural overlap. International journal of bilingualism, 13(2), 195-210.

Tedeschi, R. (2009). Acquisition at the interfaces: a case study onobject clitics in Early Italian (Vol. 218). LOT NederlandsGraduate School of Linguistics.

Van der Linden, E., Sleeman, P., 2007. Clitic dislocation: evidence for a low topic position. In: Los, B., van Koppen, M., (Eds.), Linguistics in the Netherlands 2007, John Benjamins, Amsterdam/Philadelphia, 173-187.

Yang, C. (2002). Knowledge and Learning in NaturalLanguage. New York: Oxford University Press.

Yang, C. (2004). Universal Grammar, Statistics, or Both. Trends in Cognitive Sciences, 8 (10), 451-456.

59