Direct Object Relative Clauses. The Acquisition of a Variable Construction Anca Sevcenco, Larisa Avram, Ioana Stoicescui Introduction The present paper investigates how Romanian children deal with variable input during the acquisition of direct object relatives (DOR). In standard Romanian DORs are introduced by the relative pronoun care ‘who/which’, preceded by the preposition pe, traditionally analyzed as an Accusative case marker (DobrovieSorin 1994). We will refer to this type of relatives as +PE relatives. In spoken Romanian, the preposition pe may be omitted in DORs. We will henceforth refer to these relatives as –PE relatives. This “scenario of instability” (DeGraff 1999) is also strengthened by the fact that one and the same individual may use both types of relatives, sometimes during the same conversation. The difference, however, does not seem to be related to the mere optional omission of a preposition. The –PE relatives have been analyzed as introduced not by a relative pronoun, as their +PE counterparts, but by the relative complementizer care, homophonous with the relative pronoun. Importantly, it has been argued that their derivation does not involve movement of the wh-element (Grosu 1994). This suggests that the variation which we find with respect to direct object relatives is only superficially reflected in the presence/absence of the preposition pe. What children actually receive in the input are relative clauses with a different derivational history: movement vs. no movement of the wh-element. If movement creates relatively complex structures (as often assumed, see for example, Roberts & Roussou 2003), the difference between the two types of DORs is also one of degree of computational complexity. In the present study we address the question of how Romanian children cope with the variable input which they are exposed to: + PE (or + movement) and –PE (or – movement) DORs. We rely both on longitudinal data as well as on experimental data. Variable input has been argued to be a possible trigger of changes in grammatical systems over time. The “interaction between chance oscillations in the trigger experience and the biological necessities of the language acquisition device” (Lightfoot 1994: 577) represent the ingredients of language change. Children may opt for one of the available forms, possibly for the less complex one. Frequency of cues in the input has also been shown to play an important part in language growth (see, for example, Yang 2000 and references therein). The Romanian data can therefore shed interesting light on the relationship between variation in the input, acquisition and language change. The remainder of the paper is organized as follows. Section 1 discusses the main properties of the two types of object relatives (+PE and –PE relatives). Section 2 briefly presents the theoretical background with respect to language change and language acquisition. In section 3 we present the longitudinal and the experimental data that we draw upon. The results are discussed in section 4. The paper ends with a set of conclusions in section 5. 1. Romanian Direct Object Relatives In standard Romanian, DORs are introduced by the d-linked relative pronoun care ‘who/which’, preceded by the preposition pe (1): (1) Băiatul pe care îl vezi. boy.the pe who him CL3rd SG M ACC see PRES 2nd P SG ‘The boy whom you see.’ In the spoken language, however, the preposition may be omitted, as in (2): (2) Băiatul care îl vezi. boy.the that him CL3rd SG M ACC see PRES 2nd SG ‘The boy whom you see.’ The use of care without the preposition pe is a common phenomenon in the spoken language. A simplification in the system took place in the 20 th century, when one single Nominative–Accusative form was in use, replacing the earlier rich morphological paradigm (Gheorghe 2009: 420), in which the relative pronoun was inflected for gender and number. In the contemporary language –PE DORsii may be used by both educated and uneducated people, but they are omitted in the written language. The absence of the preposition, though, can create an ambiguous context, in which the same clause can be interpreted both as a subjectiii and as a DOR (3): (3) (i) (ii) Băiatul care îl vede. boy.the that him CL3rd SG M ACC see PRES 3rd P SG ‘The boy whom he sees.’ ‘The boy who sees him.’ If pe is used, the subject relative clause reading is blocked: (4) Băiatul pe care îl vede. boy.the pe that him CL3rd SG M ACC see PRES 3rd P SG ‘The boy who sees him.’ An important property of both +PE and –PE DORs is that, inside the relative, the direct object is obligatorily ‘resumed’ by an Accusative pronominal clitic; in (1) and (2) above, for example, the clitic îl ‘him’ is obligatory and it is co-indexed with the head of the relative. A DOR reading requires the presence of the clitic. However, as seen in (3), in the absence of the preposition, the clitic may not disambiguate the sentence. But there are other clues which can help identify the relative as a direct object one. For example, if a pre-verbal or post-verbal overt subject is present in the relative, the latter can only be interpreted as a DOR. Compare the ambiguous (5a), in which the subject of the relative is null, to the non-ambiguous (5b), where the subject in the relative clause is overt: (5) a. b. Băiatul care îl vede. boy.the that him CL3rd SG M ACC see PRES 3rd P SG ‘The boy that sees him / The boy that he sees.’ Băiatul care (tata) îl vede (tata). boy.the that (father) him PL CL3rd SG M ACC see PRES 3rd SG ‘The boy that father sees.’ Agreement marking on the verb in the DOR can also act as a disambiguator if it is different from the phi-features of the head: it identifies the subject in the relative in the absence of an overtly realized subject as different from the head of the relative (as illustrated in 6). (6) Băiatul care îl desenez. boy.the who him CL 3rd SG M ACC see PRES 1st SG ‘The boy whom I draw.’ The data discussed above show that + PE relatives can only be interpreted as DORs. They are never ambiguous. But –PE relatives may be ambiguous. Their interpretation is facilitated by: (i) an overt subject; (ii) an Accusative pronominal clitic whose phi-feature matrix is identical to the one of the head; (iii) agreement on the verb, with phi-features different from the ones of the head. In terms of acquisition, the +PE relatives do not create ambiguous contexts. Their DOR reading follows straightforwardly from the presence of the preposition. But the interpretation of –PE relatives is dependent on the presence of various disambiguating cues, i.e. on knowledge of those categories associated with the cues. The two types of DOR also evince different syntactic properties. Grosu (1994) proposes that Romanian DORs rely on different syntactic structures, depending on whether the relative connector is a relative pronoun (this is the case of the +PE relatives) or a complementizer (the –PE relatives). According to him, there are two properties which distinguish between the two. When the relative pronoun care is preceded by the preposition pe, the obligatory direct object clitic inside the +PE relative cannot relate to the antecedent of the relative over an island. As can be seen in (7) below, the head of the relative, băiatul ‘the boy’, cannot be co-indexed with the direct object clitic, îl ‘him’, which is embedded in a complex DP island: (7) *Băiatuli pe care [ţiam arătat o fată boy.the pe who you CL2nd SGDAT have shown a girl [care îli simpatizează]]. who him CL3rd SG M ACC likes ‘The boy whom I showed you a girl who likes him.’ (Grosu 1994: 234) Grosu also suggests that relative connectors that are not preceded by the preposition are analyzed as relative complementizers, and not as DPs. The obligatory direct object clitic inside a –PE relative may relate to the antecedent of the relative even over an island. Example (8) contains a direct object clitic embedded in a complex DP island and co-indexed with the head of the relative; yet, the grammaticality of the sentence is in no way compromised: (8) Băiatuli care [ţiam arătat o fată boy.the that you CL2nd SG DAT have shown a girl [care îli simpatizează]]. who him CL3rd SG M ACC likes ‘The boy that I showed you a girl that likes him.’ (Grosu 1994: 234) Grosu argues that –PE relatives represent the outcome of a derivation that does not involve wh-movement. A relative such as the one in (6), for instance, has a null operator base-generated in [Spec, C], as represented in (9). (9) [SPEC,C OPi [C care [TP îli desenez ei]]] The direct object clitic has the status of a resumptive pronoun which is bound by the null operator and binds, in its turn, an empty category in the object position. This proposal explains the grammaticality of (8): resumptive pronouns are expected to occur in islands, configurations from which movement cannot take place. +PE relatives, on the other hand, do not have a null operator base-generated in [Spec, C]. In fact, Grosu offers no proposal about their derivation. Details on this account come from Dobrovie-Sorin (1994). In her view, care occupies the specifier position of a NP which is the direct object of the verb and which undergoes movement to the [Spec, C] of the relative. Since the relative pronoun connector is d-linked and does not have the status of syntactic quantifier (a quantifier that can bind a variable), the trace in direct object position must be doubled by a direct object clitic, which also binds it, as shown in (10): (10) [Spec, C [NP care băiat]i [C [TP îli desenez ti]]] This movement account explains why (7) is ungrammatical: wh-movement of the relative NP crosses the boundary of an island and becomes, therefore, illicit. In terms of input, if Grosu’s (1994) analysis is on the right track, children have to deal not merely with +PE and –PE relatives; they have to deal with DORs of different computational complexity. Although computational complexity is not a notion for which one has a non-controversial definition, several recent studies took the number and the type (internal vs. external) of Merge operations involved in the derivation as a possible measure unit (see, among others, van der Lely 1998, Roberts & Roussou 2003). Complexity effects can be seen both in comprehension and in production. We would therefore expect Romanian children to acquire –PE relatives earlier. This should be seen in better comprehension at an earlier stage, as well as, in terms of production, in a lower avoidance rate and fewer errors. On the other hand, if the presence of the preposition pe is a strong indicator that the sentence should be read as a DOR, we expect the ambiguity of –PE relatives to affect the acquisition process. 2. Theoretical background Roberts & Roussou (2003) propose that language change happens when a population of learners has acquired a grammatical system in which one parameter has a different setting than it used to have in the system of the speakers who provided the input for the respective population of learners. Language changes take place when a cue/trigger that is used as information on what value to set for a certain parameter becomes ʽobscureʼ or ʽambiguousʼ (Roberts & Roussou 2003: 12). If the trigger becomes ʽobscureʼ or ʽambiguousʼ and a language (syntactic) change is thus on the way, learners will choose the structure that comes with the least complex representation. The frequency of the cues can interfere with the acquisition process as well. Yang (2000), for example, argues in favour of frequency effects in parameter setting. What happens, though, when the input provides two structures, with two different derivational histories, whose frequency may vary from one individual to the other? Following Roberts & Roussou (2003), we propose that learners will opt for the less complex representation, at least during the early stages. However, this may happen independently of whether the cue or the trigger is ambiguous or obscure. The cue might become obscure in time if the less complex derivation is a ʽwinnerʼ across more than one generation. Let us return to Romanian DORs now. In the previous section we have shown that the input which children receive contains two types of DORs: +PE and –PE relatives. In the former, relativization from direct object position has been analyzed as the result of wh-movement (in standard Romanian); in the latter, relativization has been argued to be the outcome of a non-movement derivation. In the movement version, DORs are introduced by the relative pronoun care, which, following Dobrovie-Sorin (1994), we assume is a determiner that heads a DP; in the non-movement version, there is no relative DP, but only a relative complementizer, also phonetically realized as care. Following Roberts & Roussou (2003), let us consider movement as a parameter. We could, therefore, say that, in +PE relatives, relativization from direct object position correlates with a positive value for the movement parameter. The language learner may employ the preposition pe as an input trigger which indicates that the movement parameter should get a positive value. The preposition functions as a trigger because it tells the learner that care must be interpreted as a determiner that heads a DP (the DP with the relativized constituent) and we have just seen that relatives introduced by a relative DP rely on movement. Note that pe ends up playing a double part. On one hand, it belongs to the set of properties that distinguish DORs from subject relatives. On the other hand, it takes on the role of an input trigger that associates with the movement derivation of the DOR. In –PE relatives, relativization from direct object position correlates with a negative value for the movement parameter. The absence of the preposition trigger correlates with the reanalysis of the relative pronoun care, formerly a DP, into the complementizer care, a functional head. We can now refine our research question: How do children cope with the variable input they are exposed to when they acquire DORs? Do they show a preference for DORs that result from movement or for those that do not? If the non-movement derivation is a ʽwinnerʼ, we expect children to produce a higher percentage of –PE relatives, at least during the early stages. On the other hand, the preposition is important not only as a cue with respect to the movement/non-movement option, but also for disambiguation. As shown in section 1, –PE relatives may be ambiguous between a SR and a DOR interpretation. Ambiguity is another factor that can interfere with acquisition. From this point of view, the prediction is different; we expect children to deal better with +PE relatives, which are not ambiguous. Do children actually rely on the preposition in their understanding of DORs? If they show better comprehension of DORs with pe, we expect these relatives to be acquired earlier or faster. 3. The Data 3.1. Production data Our discussion of DOR production relies on both longitudinal and experimental data. The longitudinal data come from three corpora of child Romanian, consisting of weekly 60 minute audiotape recordings of natural unstructured conversations of monolingual Romanian children: a girl (B.1;3-3;2) and two boys (A.1;9-3;6 and I. 2;0 – 3;5). The data have been transcribed in Childes format (MacWhinney 2000). For the present analysis, we examined one recording per month for the period 2;02 – 2;11 for child B., 2;02 – 2;11 for child A. and 2;01 – 2;11 for child I. The data are summarized in Table 1 below: Table 1. Longitudinal data Child A. B. I. Age 2;02 – 2;11 2;02 – 2;11 2;01 – 2;11 MLU 2.180 - 3.174 1.819 - 2.878 1.407 - 3.720 Nr of files 10 10 10 DORs are attested relatively early in all the three corpora, but their number is extremely lowiv. In this, the Romanian data do not differ from what has been reported for several other languages. What is directly relevant for the present study, though, is a complete absence of +PE relatives. The results are summarized in Table 2 below: Table 2. +PE vs. –PE relatives in the longitudinal data Child A. B. I. Number of DORs 8 2 4 Number of +PE DORs 0 0 0 Number of –PE DORs 8 2 4 The early DORs uniformly lack both the preposition pe and, in most cases, the obligatory Accusative cliticv. This is illustrated in (11) below, where the missing elements are placed between square brackets: (11) Un brăduţ [pe] care [l-] a adus Moş Crăciun. a fir-tree DIM that has brought Santa Claus ‘A fir tree which Santa Claus has brought.’ [B. 2;07] In the I. corpus, clitics are occasionally used (see 12 below), but their phi-features do not always match those of their antecedent (13): (12) un băiat care-l cheamă tot I. a boy that him CL3rd SG M ACC calls PRES 3rd SG also I. ‘A boy whose name is also I.’ [I. 2;06] (13) o fată mare [pe] care *îl cheamă a girl FEM big FEM that him CL3rd SG M ACC calls PRES 3rd SG ‘A grown up girl whose name is Antonia.’ Antonia. Antonia [I. 2;06] There are two other empirical facts worth mentioning. In the B. corpus, a very short stage when the only C element which is used is the complementizer că ‘that’ has been identified. This complementizer is also attested with the first Subject Relative in this corpus (Avram & Coene 2006). Otherwise, in the three corpora, all the attested relative clauses are introduced by the relative pronoun. No omissions have been found in the examined files. In object relatives, the relative pronoun care is attested as an invariable form, even in contexts where the genitive/dative form is required (see 14 below): (14) asta este un xxx *care i s-a pierdut this is a xxx that him/her CL3rd SG REFL DAT has lost ‘This is a xxx whose ball has been lost.’ mingea ball.the [I. 2;06] These data indicate that during the early stages children use exclusively an invariable element to introduce relative clauses. For DORs, in particular, this shows that they use only –PE relatives during the early stages. One should mention that no similar omission of pe is found in other contexts in which the use of the preposition is required in simple clauses. For example, pe is never omitted with animate direct objects or with definite pronouns outside of relative clauses. In (15) below, the child correctly uses pe with the demonstrative pronoun but omits it with the relative pronoun: (15) pe alea care ai pus tu. pe those that have put you ‘The ones that you put there.’ [A. 2;10] The –PE bias in production is attested with older children as well. Sevcenco, Stoicescu & Avram (2009) report that in an elicited production task with 34 monolingual Romanian speaking children, with ages that range between 5;00 and 7;03 years, 91.16 % of the produced DORs were –PE onesvi. In a task eliciting indirect object relatives with 25 monolingual children, age range 5;00 – 7;04, no inflected care was produced (Sevcenco, Stoicescu & Avram 2009). Since both +PE and –PE relatives are used in the spoken language, child directed speech included, the fact that very young children use exclusively –PE relatives during the early stages may be an effect of the variable input which they receive. In some cases, the same caretaker may use both a +PE and a –PE relative during the same recording session, as seen in (16): (16) a. noua poezie pe care ai învăţat-o la creşă new.the poem pe which have learned clitic CL3rd SG F ACC at nursery ‘The new poem which you have learned at nursery.’ b. poezia de aseară care iai poem.the of yesterday.evening that her CL3rd SG F DAT have SG F ACC spus-o lui mami said it CL3rd SG ACC to Mother ‘The poem of yesterday evening which you told Mother.’ [caretaker in the B. corpus, 2;8.19] However, the input which children receive may be rather different. In the I. corpus, for example, 100% of the DORs used by adults in child-directed speech in the 10 examined files are +PE ones. The analysis of DORs in child-directed speech in the 10 files in the B. corpus reveals that the input which she receives contains both types of DORs with a slight bias for –PE ones (69. 23%)vii. But both I. and B. use –PE relatives during the early stages. The variation in child-directed speech is reinforced by variation in adult-adult interaction, where both types of relatives are used. This is why we believe that the percentage of –PE relatives in the early files of our longitudinal data may not be a real reflection of the input which the children receive. Moreover, the experimental data show that children continue to have a –PE relative bias even after they started going to school, where the input which they receive from teachers is standard Romanian, with +PE relatives. 3.2. Experimental dataviii 3.2.1. Comprehension 1 We first tested the comprehension of direct objects relatives with pe in comparison to the comprehension of subject relatives. The aim of this experiment was to test whether Romanian children had any problems understanding DORs, as reported for various other languages (Sheldon 1974, Arosio, Friedmann, Belletti & Rizzi 2009). This comprehension task was a binary sentence-picture matching one. It included sixteen questions per participant (eight subject relatives and eight DORs). All the test items contained a transitive verb which denoted semantically reversible events. The subject and the direct object DPs were both [+animate] and singular. All the DORs were +PE relatives with an overt post-verbal subject. The subject relatives matched the DOR in terms of length and all of them contained a clitic doubled direct object. Test items are exemplified below: (17) Arată-mi şoarecele care o scoate pe bufniţă din cutie. show.me mouse.the that her CL3rd SG F ACC takes pe owl SG F ACC from box ‘Show me the mouse who takes the owl out of the box.’ (18) Arată-mi pisoiul pe care îl piaptană ariciul. show.me cat.the pe who him CL3rd SG M ACC combs hedgehog.the ‘Show me the cat whom the hedgehog combs.’ The task was administered as a Power Point presentation run on a portable computer in a quiet room at the children’s kindergarten. Each subject was tested individually. The experimenter told the child that she needed some beautiful pictures for a picture book for children but she was not sure whether the pictures she had were good enough. So she asked the child to help her. Two pictures were presented on the computer screen and the child heard a sentence of the type illustrated in (17) and (18) above. (S)he was asked to identify the picture which matched the sentence. For the sentence in (17), for example, in one picture a mouse was taking an owl out of the box and in the other picture the same owl was now the one taking the mouse out of the box. The picture pairs were presented in randomized order so that no more than three identical sentences were targeted. The correct picture of the pair was varied so that the good picture should not be on the right / left side for more than three consecutive trials. 57 monolingual Romanian children, with ages ranging between 35 – 89 months (mean age: 64.70, SD: 11.38) took part in the experiment. We also tested 10 adults. 3.2.2. Comprehension 2 In a second experiment we tested to what extent children can understand –PE relatives. As discussed in section 2, in the absence of the preposition pe, the relative clause may be ambiguous. In this case, there are several possible disambiguators: (i) agreement on the verb; (ii) the overt subject in the relative clause; (iii) the Accusative clitic. This is why we tried to measure which of these cues plays a more important part in the comprehension of –PE relatives. The task was a binary picture-matching task (similar to that used in Arosio, Adani & Guasti 2006, Yatsushiro, Arosio & Forgiarini 2010) and it included eighteen sentences with DORs (6x6x6, see below), all of which were –PE DORs. All the test items contained transitive verbs denoting semantically reversible events. The subject and the direct object were both animate DPs. Test items are illustrated below: (i) –PE relatives disambiguated by the phi-features of the direct object clitic and by agreement on the verb: (19) Arată-mi fata care o stropesc. show me girl.the F SG who her CL3rd F SG ACC splash PRES 3rd PL ‘Show me the girl that they are splashing.’ In (19) above, the head of the relative is feminine singular. The clitic in the relative clause, o ‘her’, matches the features of the head. The verb is 3 rd person plural. (ii) –PE relatives disambiguated by the direct object clitic and an overt postverbal subject: (20) Arată-mi prinţesa care o show me princess.the F SG that her CL3rd F SG ACC fotografiază piticul. photographs PRES 3rd SG dwarf.the M SG ‘Show me the princess that the dwarf photographs.’ (iii) –PE relatives disambiguated only by the direct object clitic: (21) Arată-mi broaştele care le salută. show me frogs.the F PL that them CL3rd F PL ACC greets PRES 3rd SG ‘Show me the frogs that he/she greets.’ In this case, the verb used in the sentence belonged to an inflection class in which the 3rd person singular and the 3rd person plural forms of the present tenses of the indicative are identical. In (21) above, for example, salută could be ‘greet’ or ‘greets’. This task was also administered as a Power Point presentation run on a portable computer in a quiet room at the children’s kindergarten. Each subject was tested individually. Two pictures were presented on the computer screen and the child heard a sentence of the type illustrated in (19), (20) or (21) above. (S)he was asked to identify the picture which matched the sentence. For the sentence in (21), for example, one picture featured a tortoise greeting two frogs; the other picture featured the same two frogs greeting the tortoise. The picture pairs were presented in randomized order so that no more than three identical sentences were targeted. Just like in the previous task, the correct picture of the pair was varied so that the good picture should not be on the right / left side for more than three consecutive trials. 25 monolingual Romanian speaking children, with ages ranging between 48-89 months (mean age: 66.19, SD: 11.96) took part in the experiment. All of them took part in Comprehension task 1 as well. We also tested ten adults as controls. 3.2.3. Results In the first comprehension task, where both the comprehension of direct object relatives (DORs) and subject relatives (SRs) was tested, children understood the +PE direct object relatives 63.59%, while adults, as expected, performed at ceiling. The results are summarized in Figure 1: Figure 1. Comprehension of DORs and of SRs in comprehension task 1 100 90 80 70 60 adults 50 children 40 30 20 10 0 DORs SRs The results in the second task, where the comprehension of DORs without PE and disambiguated by various cues (clitic, clitic and agreement on the verb, clitic and overt subject in post-verbal position) was tested are summarized in Figure 2. Figure 2. Comprehension task 2: –PE relatives 100 90 80 70 60 children adults 50 40 30 20 10 0 CL CL+Agr CL+overt subject Comparing the overall results of the 25 subjects who took part both in task 1 (+PE relatives with an overt subject) and in task 2 (the comprehension of the –PE relatives disambiguated by an overt post-verbal subject) one notices that the difference between the comprehension rate in task 1 is not significantly different from the comprehension rate for the –PE relatives disambiguated by a clitic and an overt post-verbal subject in task 2 (see Table 3). Table 3. Results in comprehension task 1 and 2 Number subjects 25 of Age range (months) 48-89 comprehension of +PE DORs in task 1ix 78.8% comprehension of –PE DORs in task 2 80.29% The data indicate that the presence/absence of pe may not be relevant for the comprehension of DORs. As seen in Figure 1 above, in the absence of the preposition, children rely on other cues, whose contribution is different: (i) clitic + overt subject ranks higher than clitic; DORs disambiguated by a post-verbal overt subject are comprehended 80.29%; (ii) clitic + agrement on the verb ranks higher than clitic; children understood these DORs 67.9 %; (iii) the clitic alone does not seem to play the role of a robust cue; sentences disambiguated only by the clitic were understood 51.23%. In spite of the fact that apparently there is a small difference between (i) and (ii), the difference does not reach significance (see also the discussion in section 5). The results of the longitudinal study data and those from the production experiment show that children have a strong bias towards producing –PE relatives to the detriment of +PE relatives. The overall results in comprehension task 2 indicate that the factors which disambiguate DORs do not have an equal contribution. The overall outcomes between types of responses in Comprehension task 2 are significantly different (H (2) = 16.22, p < 05). The post hoc analysis has revealed that no significant difference holds between the clitic + overt subject and the clitic + Agr cues (U = 229.50, z = - 1.67, ns). However, significant differences arise between (i) the clitic + overt subject and clitic cues (U= 110, z = -4.00, p < 001) and (ii) the clitic + AGR and the clitic cues (U= 195.50, z = -2.31, p < 002). Therefore, these data indicate that the factors with the highest bearing upon the facilitation of comprehension are those that feature the direct object clitic and an overt post verbal subject or the direct object clitic and agreement on verb. 4. Discussion In the present study we focus on how Romanian children cope with the variable input which they are exposed to: + PE (or + movement) and –PE (or – movement) DORs. The longitudinal data as well as experimental production data reveal that children have a strong bias towards –PE relatives. On comparing the results from Comprehension task 1 to those from Comprehension task 2, we noticed that the presence of pe does not facilitate comprehension, i.e. the lack of importance of pe as a disambiguating factor for DORs correlates with its high percentage of omission in production. According to the analysis of relative clauses with/without pe which we have adopted, pe is the trigger associated with a positive setting for the movement parameter in DORs. The fact that it does not affect the comprehension of DORs in the absence of other cues seems to suggest that it may not be a robust cue in DORs. It might be what Roberts & Roussou label an ʽambiguousʼ trigger. The bias towards –PE relatives in production may indicate two things. Firstly, it may point to the fact that children interpret care in DORs as a complementizer. This hypothesis seems to receive support from the fact that in the longitudinal data, all object relatives (direct object, indirect object, prepositional object) are introduced by a non-inflected care or by a care without a preposition. The relative DP in +PE relatives has been reanalyzed in the history of Romanian (Gheorghe 2009). From an inflected DP, it has changed in the 20th century into an invariable DP form. This, we believe, created the right condition for its possible further development into a complementizer. Actually, Grosu (1994), analysing DORs in 20th century Romanian, states that –PE relatives are introduced by a complementizer.x The availability in the contemporary language of two different systems – one with DORs introduced by a DP and one with DORs introduced by a complementizer – creates a variable, unstable input. Children seem to opt for a language in which DORs are introduced by an invariable complementizer. The results from our comprehension tasks reinforce this conclusion. Children do not treat pe as a disambiguator of DORs. Secondly, the two types of relatives, as discussed in section 1, are associated with two different derivational histories: one with wh-movement and one without. Children seem to opt for the non-movement system. That children adopt a nonadult-like non-movement analysis of relatives during the first acquisition stage has also been argued for French, a language where relative clause formation involves movement (see Labelle 1996, as well as arguments against the non-movement analysis of early relatives in Guasti & Shlonsky 1995, Perez-Leroux 1995, Guasti & Cardinaletti 2003). The case of Romanian, however, is different. If Grosu’s (1994) analysis is on the right track and care DORs do not involve movement in the target language, choosing the non-movement construction indicates a bias for less derivationally complex structures when the input is variable; otherwise, children use an available adult-like structure. The phenomenon seems to be similar to the alternation between that and wh-relatives in English. Since the +PE DORs are associated with the ‘norm’ and are taught in schools, one might say that they are acquired as an effect of ‘schooling’. Of course, this may indeed be true. The percentage of +PE relatives is higher with school children and significantly higher with educated people. But even educated people may shift from one type of relative to the other. We believe that we are now in a position to answer the question we started from. When faced with variable input for DORs, Romanian children opt for –PE relatives to the detriment of +PE relatives. The preference for –PE relatives could be motivated in terms of costliness of operations, i.e. not to move seems to be less costly than to movexi. 5. Conclusions This paper has attempted to shed some light on the relationship between variable input and the acquisition of DORs by Romanian children. The analysis of childdirected speech in longitudinal corpora of child Romanian has revealed that the input which children receive features two types of DORs: +PE and –PE, which differ in two important respects. First, the relative connector care may or may not be preceded by the preposition pe. Second, +PE relatives derive from whmovement of the relativized DP headed by care. However, –PE relatives have a different derivational history. The relative connector is not a DP head, but a reanalyzed complementizer. Consequently, –PE relatives do not involve whmovement, but merge of a null operator and merge of the relative complementizer care. The results of our analysis have shown that (i) in production, there is a strong bias towards –PE relatives to the detriment of the +PE ones; (ii) in comprehension, there is no significant difference between +PE and –PE relatives. These results indicate that Romanian children may be treating pe as an obscure trigger in the case of DORs. Children opt for non-movement –PE DORs, i.e. for structures which are derivationally less complex. This preference cannot be accounted for in terms of frequency in the input. The child-directed speech in the examined longitudinal corpora indicated variation among caretakers with respect to the use of +PE/–PE DORs but no variation among children. This shift correlates with the possible change of pe, in time, into an obscure trigger for movement in the derivation of DORs; it might be accounted for in terms of a universal preference for less complex structures that has been shown to play a part in language change phenomena. The acquisition data which we have examined might suggest that, in time, Romanian DORs could be treated uniformly as non-movement structures. References Arosio, F., Adani, F., Guasti, M.T. (2006). Children’s processing of subject and object relatives in Italian. In A. Belletti, E. Bennati, C. Chesi, E. di Domenico & I. Ferrari (Eds.) Language acquisition and development. Proceedings of GALA 2005, pp. 15-27. Newcastle: Cambridge Scholars Press. Avram, L., Coene, M. (2006). The complementizer phrase in child Romanian: An early discourse anchor. In A. Belletti, E. Bennati, C. Chesi, E. di Domenico, I. Ferrari (Eds.) Language acquisition and development. Proceedings of GALA 2005, pp. 28-34. Newcastle: Cambridge Scholars Press. Avram, L., Coene, M. (2007). Object clitics as Last Resort. Implications for language acquisition. In S. Baauw, J. van Kampen & M. Pinto (Eds.) The acquisition of Romance languages. Selected papers from the Romance Turn II 2006, pp. 7-26. LOT, Utrecht. DeGraff, M. (1999). Creolization, language change & language acquisition: An epilogue. In deGraff, V. (Ed.) Language creation & language change: Creolization, diachrony and development, pp. 473-543. Cambridge, Mass.: MIT Press. Dobrovie-Sorin, C. (1994). The syntax of Romanian. Berlin: Mouton de Gruyter. Friedmann, N., Belletti, A. & Rizzi, L. (2009). Relativized relatives: Types of intervention in the acquisition of A-bar dependencies. Lingua 119: 67-88. Grosu, A. (1994). Three studies in locality and case. London: Routledge. Gheorghe, M. (2009). Aspecte de dinamică a construcţiilor cu relative. In G. Pană Dindelegan (Ed.) Dinamica limbii române actuale. Aspecte gramaticale şi discursive, pp. 415- 430. Bucharest: Editura Academiei Române. Guasti, M.T. & Cardinaletti, A. (2003). Relative clause formation in Romance child’s production. Probus 15: 47-89. Guasti, M.T. & Shlonsky, U. (1995). The acquisition of French relative clauses reconsidered. Language Acquisition 4: 257-276. Labelle, M. (1996). The acquisition of relative clauses: movement or no movement? Language Acquisition 1: 95-119. Lightfoot, D. (1994). Review of Ian Roberts’ Verbs and diachronic syntax: A comparative History of English and French. Language 70: 571-578. MacWhinney, B. 2000. The CHILDES project: Tools for analyzing talk. Third Edition. Mahwah, NJ: Lawrence Erlbaum Associates. Perez-Leroux, A.T. (1995). Resumptives in the acquisition of relative clauses. Language Acquisition 4: 105-138. Roberts, I. & Roussou. A. (2003). A minimalist approach to grammaticalization. Cambridge University Press. Ruxăndoiu, L. (2002). Interacţiunea verbală în limba română actuală. Corpus (selectiv). Schiţă de tipologie. Bucharest: Editura Universităţii Bucureşti. Sevcenco, A., Stoicescu, I. & Avram, L. (2009). Subject/Direct object relative Clause asymmetry in child Romanian. Poster presented at RASCAL, Groningen 2009. Sheldon, A. (1974). The role of parallel functions in the acquisition of relative clauses in English. Journal of Verbal Learning and Verbal Behaviour 13: 272281. van der Lely, H.K.J. (1998). SLI in children: movement, economy, and deficits in the computational-syntactic system. Language Acquisition 7: 161–192. Yang, C.D. (2000). Knowledge and learning in natural language. Doctoral dissertation, MIT. Yatsushiro, K., Arosio, F. & Forgiarini, M. (2010). Case and agreement in German children comprehension of Relative Clauses. Poster presented at Let the Children Speak, London, January 22-24, 2010. i Work on this paper was financed by research grant CNCSIS PNII_IDEI 1979. We thank the audience at The Romance Turn 4 and two anonymous reviewers for their extremely helpful comments and suggestions. ii Interestingly, no similar phenomenon is found outside relatives. For example, whquestions with a non-inflected care are, if not totally absent, extremely rare. If educated people occasionally shift from +PE to –PE relatives, no similar shift is seen in the case of wh-questions. iii Subject relatives are always introduced by a “bare” care, both in the standard and in the spoken language. iv Actually, because of the very low number of attested DORs it is difficult to accurately identify the age of emergence when one examines one file per month. With I., for example, the first DOR is attested at 2;03 but the next DORs are attested only at 2;06. No other DOR is attested in the next five files which we examined. v Direct object clitics are not omitted only in early DORs. Non-reflexive 3rd person clitics are omitted in all the clitic obligatory contexts until the age of approximately 3;00 (Avram & Coene 2007). vi An analysis of the Ruxăndoiu corpus ( 2002) of spoken adult Romanian has revealed that, out of a total of 57 DORs uttered, +PE relatives were produced in 73.34% percent of the cases whereas –PE relatives were produced in 26,66% of the cases. The significantly higher percentage of +PE relatives might be due to the fact that some of the subjects knew they were being recorded and may have tried, therefore, to resort to standard Romanian as much as possible. vii For the A. corpus it is not possible to provide the percentage of -PE DORs because the available transcripts contain only some of the child-directed sentences in the recordings. viii We thank Teresa Guasti who generously offered us advice and materials for both comprehension tasks. ix This is the percentage obtained by the 25 participants in task 1 who also took part in task 2. This explains the differences between the data summarized in Figure 1 and Table 3. x One reviewer suggests that the presence of the resumptive pronoun in DORs may play a role in the interpretation of care in the subordinate clause as a complementizer, and not as a relative pronoun. This may be indeed the case. Another factor that may have contributed to the reinterpretation of care as a non-inflecting subordinate clause introducer may be the existence in the language of DORs introduced by the invariable ce, as in Cartea ce am citit ‘The book which I have read’. Such relatives are, however, perceived as outdated. The relatives with invariable care might be filling in the gap left behind by the vanishing ce relatives. But an answer to this question definitely requires further research. xi Obviously, when everything else is “equal”.
© Copyright 2025 Paperzz