Theoretical Implications of Object Clitic Omission in Early French : spontaneous vs. Elicited Production

Abstract This article examines the phenomenon of object clitic omission in French. Previous research contains contradictory results depending on the source of the data: it seems that in spontaneous production children prefer DPs while in elicited production they prefer omissions. It is proposed that a common methodology be used across different modalities in measuring the rate of omissions, and that the notion of «illicit object omission» be dispensed with. The analysis of the proposed «clitic-contexts» reveals that the strategy favoured by children is omission of all kinds of lexical material in both spontaneous and elicited production. Moreover, it is shown that child behaviour is quantitatively different from the adult one. These findings have consequences on the status of null objects in child grammar: child grammar allows optional object deletion without clitic recoverability, as opposed to adult grammar. Several theoretical approaches are evaluated in the light of the new findings. Key words: object clitic omission, French, spontaneous production, elicited production. 1. Introduction The delay in the acquisition of object clitics with respect to other clitic elements has received much attention in the last decade. The phenomenon has been observed for normally developing children (Friedemann 1992, Hamann et al. 1996, Hamann * I would like to thank Anna Gavarró, Ana Teresa Pérez-Leroux, Yves Roberge and two anonymous reviewers for their comments and suggestions. This research is funded in part by a grant from the Social Sciences and Humanities Research Council of Canada (410-05-0239). Catalan Journal of Linguistics 5, 2006 221-236 Theoretical Implications of Object Clitic Omission in Early French: Spontaneous vs. Elicited Production* Mihaela Pirvulescu University of Toronto at Mississauga 3359 Mississauga Road North, L5L, 1C6 Canada [email protected] Table of Contents 1. Introduction 2. Previous work: elicited vs. spontaneous production 3. Methodology 4. New study 5. Discussion References Cat.Jour.Ling. 5 001-276 3/11/06 16:28 Página 221 2003, Jakubowicz et al. 1996, Jakubowicz and Rigaut 2000, van der Velde et al. 2002, Jakubowicz and Nash to appear), SLI children (Hamann et al. in press, Jakubowicz et al. 1998), bilinguals (Hulk 1997, Hulk and Müller 2000, Kaiser 1994) and early L2 (Prévost, ms., Belletti and Hamann 2004, White 1996). Given the extent of the phenomenon, it is important to investigate what strategy the child is using while object clitics have not been (fully) acquired. Thus far, the results obtained for French appear to be dependent on the source of the data, i.e. according to whether one looks at spontaneous or elicited production. I will show that for French this is the result of a difference in the methodology used and I will argue that the strategy favoured by children is omission of all kinds of lexical material. These findings have consequences on the status of null objects in child grammar while clitics are not yet fully acquired: our data points to the fact that child grammar allows free object deletion without clitic recoverability, as opposed to adult grammar. Moreover, there is a period when this option coexists with the target clitic construction. These important findings will be used to evaluate several theoretical approaches that aim to explain the acquisition of object clitics in French. 2. Previous work: elicited vs. spontaneous production In this section I will review the results from both spontaneous and elicited production of French. The picture that emerges is a seemingly contradictory one: while the low rate of clitic production is associated with a high rate of DP use in spontaneous production, it is associated with a high rate of omissions in elicited production. 2.1. Elicited production Elicited production data contain a high rate of omissions. The studies reported in the literature use an elicited production paradigm (Schaeffer 2000): the experimenter asks a question to a puppet introduced as unreliable, and then asks the child to help correct the puppet’s factual mistake. The goal of the procedure is to establish the introduced object as the definite topic of the discussion, which is a context where a clitic should be used if one wants to refer to this object again. Constructions with null arguments without identification by clitics are frequent; the children favour null objects over clitics (Jakubowicz et al. 1996, Schmitz et al. 2004, PérezLeroux et al. to appear). An example is provided below: (1) Question: Qu’est ce que le père fait avec la balle ? ‘What is the father doing with the ball?’ Expected answer: Il la frappe. ‘He hits it’. Child’s answer: Frappe. ‘Hits’. (Nathaniel, 2;8; Pérez-Leroux et al. 2005) The rate of object omission is high: Jakubowicz et al. (1996, 1997) report a rate of 50% (mean age 2;4); Schmitz et al. (2004) report 44% omissions (mean age 2;6) and 30% omissions (mean age 2;10). Pérez-Leroux et al. (in preparation) 222 CatJL 5, 2006 Mihaela Pirvulescu Cat.Jour.Ling. 5 001-276 3/11/06 16:28 Página 222 report 35% omissions (mean age 3;5). The production of DPs is between 30% and 40% across the mentioned studies (comparable to the adult DP responses in an elicitation setting)1. However, with respect to omissions, the child behaviour contrasts greatly with the adult one: for adults participating in the same elicitation procedure, there were no omissions in the study by Schmitz et al. (2004) which tested four adults, and 4% omissions in the studies by Pérez-Leroux et al. (to appear, in preparation) where nine adults were tested. 2.2. Spontaneous production Data from spontaneous production points towards a delayed acquisition of object clitics. The low production of object clitics in early French is evaluated with respect to the rate of nominative clitic use, determiners and reflexives2. It is shown that during the period where clitics are scarce, children favour DPs: «instead of accusative clitics, the three children produce lexical DPs […]» (Van der Velde et al. 2002: 122; also Jakubowicz et al. 1996, 1997; Jakubowicz and Rigaut 2000). The spontaneous data presented in van der Velde et al. (2002) comes from the interaction of three children. These children were recorded in monthly 45minute sessions. The age of the children ranged between 1;8 and 2;5. The omissions are scarce, varying between 10-15%, while the use of DPs rises as high as 70%. Jakubowicz and Rigaut (2000) obtained similar results. They analyzed the production of 12 children from a 45-minute session of spontaneous interaction with each child. The age of the children ranged between 2;0.13 and 2;5.29. Again, the average rate of omissions was around 10% while the average rate for the lexical nouns was around 70%. The high rate of DP use is surprising given that adults in spontaneous speech seem to use far less: around 34% from the study by Schmitz et al. (2004). A higher rate of omission is reported in Hamann (2002, 2003) who discusses data from one child, Augustin. The age period studied is from 2;0.2 to 2;9.30. The omissions range from 40% to 14.4% in the 10 samples examined. Although the incidence of null objects tends to decline with age, it is still around 20% in three of the last four samples (between 2;4.22 and 2;9.30). Also, one of the children reported in Jakubowicz and Rigaut (2000) has a rate of 28% omissions. Therefore, there seem to be different rates of omission even in data coming from spontaneous production alone. 2.3. The asymmetry To recapitulate, there are two sets of data used in evaluating children’s acquisition of object clitics. Spontaneous production data, which have a low percentage of Theoretical Implications of Object Clitic Omission in Early French CatJL 5, 2006 223 1. Rates of DP production were much lower in other experiments (Schaeffer 2000, Wexler et al. 2004); the source of this contrast remains a topic for future research. 2. «During the development of Augustin’s pronominal system subject clitics appeared systematically from the second birthday whereas complement clitics appeared at least 6 months later, which observation lead to the term ‘delay’ of complement clitics…» (Hamann, 2003:107). Cat.Jour.Ling. 5 001-276 3/11/06 16:28 Página 223 omissions, seem to show that the delay in the acquisition of object clitics is compensated by a massive use of DPs. In contrast, elicited production data, which have a high percentage of omissions, show that when children have not yet fully acquired clitics they favour omissions. The picture that thus emerges is that of an asymmetry in object omission in spontaneous versus elicited production. This is summarized in Table 1. The results obtained in both production procedures show that the child behaviour is strikingly different from the adult one: children use fewer clitics than adults. That being said, the studies differ as to what children use while clitics are not yet produced as in the target. According to the results from spontaneous production, there is an overuse of discourse inappropriate DPs. While under recent theoretical approaches we can expect overlap between the choice of a descriptive DP and a definite pronoun (Farkas 2002), the difference in their rates of use between children and adults indicates a difference at the grammatical level. Similarly, a comparison with the rates of omission found in experimental settings, points towards a grammatical difference between children and adults. In this article I will concentrate on the difference with respect to the rates of omission. Why do we find such a difference between children’s behaviour in spontaneous versus elicited production with respect to the rate of omissions? One possibility, mentioned by Jakubowicz et al. (1996: 384) is that elicited production might have a negative effect on children, such that they tend to omit more objects than during spontaneous interaction: «We believe that the high percentage of object omissions […] is a by-product of the task itself. If these children assumed that French allowed object deletion, they would have omitted objects also in spontaneous interaction data. As shown […] this did not happen.» There is, however, another possibility: the low rate of omissions in spontaneous discourse might be a consequence of the methodology used to count these omissions. Therefore, before firmly concluding that the elicitation task favours object omission (which, if true, is an interesting matter to further investigate), we need to look more clearly at the methodology used to calculate object omissions in spontaneous production. 224 CatJL 5, 2006 Mihaela Pirvulescu Table 1. Percentage References Object omission ~11% Hulk, 1997, Jakubowicz et al. 1997, in spontaneous speech Jakubowicz and Rigaut 2000, van der Velde 1998, van der Velde et al. 2002 Object omission 40%-60% Jakubowicz et al. 1996, 1997; Schmitz et in elicited speech al. 2004; Pérez-Leroux et al. 2005 Cat.Jour.Ling. 5 001-276 3/11/06 16:28 Página 224 3. Methodology 3.1. Omissions in spontaneous production The calculations of omission phenomena are based on the notion of «obligatory context», and the percentage of omissions in spontaneous speech refers to illicit omissions of the direct object. Illicit object omissions are identified with respect to the adult use: they are cases in which adult speech presumably always shows an overt object. Illicit null object constructions in child production seem to be of three kinds (all examples from Müller 2004 and Müller et al. 1996). First, there are those that violate the lexical transitivity of the verb, as in (2). (2) Il met __ dans le bain. ‘He puts in the bathtub.’ The example does not come with a description of the linguistic and extralinguistic contexts. It is to be assumed therefore that the verb mettre should obligatorily appear with a lexically realized object. The second type of illicit construction involves a missing reflexive clitic: (3) Habille dresses ‘He puts his clothes on.’ In this case the verb requires an obligatory reflexive object (and a subject). Finally, the third type of illicit null object construction identified in child speech is the following: (4) Adult: On peut le manger, l’oeuf? ‘Can we eat it, the egg?’ Child: Tu peux manger, oui. ‘You can eat, yes.’ We see here an optionally transitive verb used by the child without an overt object or clitic, although there is a definite linguistic antecedent. This would presumably force the use of the clitic in the answer, in order to recover the null object. Therefore, the omission of the clitic in (4), counts also as an illicit omission. However, as it will be shown in section 4.2, one can find ample evidence that similar examples exist also in adult French. Illicit object omission in spontaneous production therefore seems to encompass a variety of contexts: new information, where a lexical NP would be expected as in (2); reflexives (3); and known information with a definite linguistic antecedent, where a clitic would be expected (4). In elicited production, on the other hand, illicit null objects can be characterized as omissions in the context of known information with a definite linguistic antecedent as in (1) and (4) above. It is therefore necessary to isolate the same type of context in order to be able to comTheoretical Implications of Object Clitic Omission in Early French CatJL 5, 2006 225 Cat.Jour.Ling. 5 001-276 3/11/06 16:28 Página 225 pare children’s production of clitics across modalities. In the absence of this, the notion of «illicit» object omission is of no help. An examination of context was done in Jakubowicz and Rigaut (2000). They found that the omission of a lexical DP was more frequent than the omission of the accusative clitic or the reflexive. Below are examples of each type of omission: lexical DP, reflexive and object clitic (Jakubowicz and Rigaut 2000:137): (5) a. situation: Claire is putting a piece of the puzzle in its place. Claire: pas mis (i.e. la pièce) ‘Not put’ (i.e. the pice of the puzzle) b. Mother: Qu’est-ce qu’ils font? ‘What are they doing?’ (she shows an image of two children brushing their teeth) Child: Ils _ brossent les dents. ‘They are brushing the teeth.’ c. Adult: Le lapin de Pâques, tu la connais cette histoire-là? ‘The Easter bunny, do you know that story?’ Child: Oui. Je _ connais. ‘Yes. I know.’ However, their calculations were based on the total number of constructions with omissions: «In Group 1, among the 12 phrases with a null object, the context indicates that the omitted element is a lexical DP in four cases, a reflexive in 2 sentences and an accusative clitic in one sentence…»3These constructions depend on the discourse situation. Therefore, the same shortcomings apply as in the studies mentioned above. What is needed is a calculation of the omission rate versus overt object rate only in the context that requires/permits the use of the clitic. This is shown in the next section. 3.2. The clitic-context It is well known that clitics are not required by verbal transitivity alone, but by a certain discourse situation. The pronominal clitic must be specific, i.e. it must have an antecedent in the preceding discourse (according to the interpretation of the term «specific» provided by Cardinaletti and Starke (1999) and references therein) and, while it can have a token-reading or a type-reading, it must be an entity that can be presupposed (López 2003). Previous studies on spontaneous production seem to have included contexts where clitics are not required as in (6) below (from Van der Velde et al. 2002: 122): 226 CatJL 5, 2006 Mihaela Pirvulescu 3. «Dans le Group 1, parmi les 12 énoncés à objet nul, le contexte indique que l’élément omis est un DP lexical dans 4 cas, un réfléchi dans 2 énoncés, un clitique accusatif dans 1 énoncé…» (Jakubowicz and Rigaut 2000:137). Cat.Jour.Ling. 5 001-276 3/11/06 16:28 Página 226 (6) a. situation: Hugo’s mother is preparing the animals to construct a farm. Child: je fais une ferme. ‘I’m making a farm’ Hugo, 2;5.5 b. Adult: tu l’aimes cette chanson? ‘Do you like this song?’ Child: j’aime Aladdin. ‘I like Aladdin.’ Victor, 2;5.29 c. Adult: qu’est-ce que tu veux faire? ‘What do you want to do?’ Child: je vais pêcher les poissons. ‘I am going to catch the fish’ Chloé, 2;4.1 Examples such as those in (6) seem to have been included when deciding what the child is using while clitics are not yet fully acquired. However, the examples (6a) and (6c) are not relevant for this question and should not have been included: the child’s behaviour with respect to clitics should only be evaluated in the contexts where the clitic is required or permitted. These are the contexts used in the studies of elicited production. In the rest of this paper I will concentrate on what I call «clitic contexts», which, in adult speech, cover the clitic-identified null object and the clitic drop. For child spontaneous speech, the clitic context is defined as follows (following Pirvulescu and Roberge 2005)4: in the clitic-context the referent is definite; it is the topic of the discussion; and it is contained in the immediately preceding discourse. There are three types of elements, which may appear in the clitic context in children’s spontaneous production: clitics, omissions and DPs. An example of each is provided in (7a-c) (from the York corpus, Mona and Para files, CHILDES database): (7) a. Adult: Elle n’est pas sèche [la pâte à modeler] ‘it is not dry.’ Child: Moi va l’couper. ‘Me cut it.’ Max, 2;2.22 b. Adult: Tu ne veux pas le mettre dans l’eau [le canard]? ‘You don’t want to put it in the water?’ Child: _ mettre ici. ‘put here’ Max, 2;0.14 c. Adult: Voilà la cuillère à Anne. ‘Look there is Anne’s spoon.’ Child: Willy mange la cuillère à Anne. ‘Willy is eating Anne’s spoon.’ Anne, 2;4.20 In the next section I present the new data from spontaneous production available on the CHILDES database as coded and analyzed for the clitic-contexts and the percentage of omissions vs. pronominal clitics. The results from the children are also compared with the results from adult spontaneous child-directed speech. Theoretical Implications of Object Clitic Omission in Early French CatJL 5, 2006 227 4. A similar notion is used in Paradis (to appear): a permissible context for direct object pronominalization is one where the referent for the direct object was mentioned within the previous ten lines in the transcript. See also Tsakali and Wexler (2004). Cat.Jour.Ling. 5 001-276 3/11/06 16:28 Página 227 4. New study In this section I present new data from spontaneous production according to the clitic-context as discussed in the preceding section. The data come from three corpora: two York corpora (the null and clitic counts taken from Pirvulescu, to appear) and the Champaud corpus from the CHILDES database. The three corpora represent the only publicly available recordings that begin from a very early age. The advantage of the York corpora over the Champaud corpus is that in the former the children were recorded more often (every two weeks) and more consistently (the Champaud corpus has time gaps as long as two months). The development of object clitics can therefore be closely followed in the York corpus. 4.1. Results from spontaneous production Tables 2 to 4 show the percentage of null, clitic and lexical objects («Pron» in Table 3 corresponds to ça «this»). The three tables show that omission is constant while clitics are not yet (fully) acquired. Moreover, there seems to be a trade-off between the rate of omissions and the rate of clitics: omissions diminish, as clitics are more and more frequently used. On the other hand, the use of DPs seems to be an option children use in an inconsistent way. The results from the first two children show that while the rate of omission is high, the clitics are used optionally. If we look at particular verbs, we can see that the same verb is used either with a clitic, a lexical noun or an omission. Moreover, there are a variety of verbs that are used in this way. An example is provided in (8) from the Para corpus. 228 CatJL 5, 2006 Mihaela Pirvulescu Table 2. Corpus Para, France; omissions and overt objects in clitic contexts. MLU AGE NULL CL LEXICAL Cl contexts (tokens) 2.58 2;2.30 62.50% 25.00% 12.50% 8 2.83 2;4.2 50.00% 16.66% 33.00% 6 3.01 2;4.20 33.33% 0.00% 66.00% 3 2.61 2;5.4 25.00% 0.00% 75.00% 4 3.25 2;5.18 25.00% 37.50% 12.50% 8 3.38 2;6.2 28.00% 57.14% 14.20% 7 3.21 2;6.18 70.58% 11.76% 17.60% 17 2.87 2;7.1 20.00% 60.00% 20.00% 5 2.85 2;8.3 42.85% 42.85% 14.20% 7 3.66 2;8.20 40.00% 50.00% 10.00% 10 3.56 2;9.15 16.00% 83.00% 0.00% 12 3.59 2;10.18 8.00% 92.00% 0.00% 25 Cat.Jour.Ling. 5 001-276 3/11/06 16:28 Página 228 (8) a. les mettre ‘them put’ b. on met l’oeil ‘we put the eye’ c. je mets __ là ‘I put there’ (Anne, 2;2.30) Examples such as (8) show that resort to omissions vs. overt material is not lexically restricted. 4.2. The nature of omissions with respect to the target We can now return to the notion of «illicit omission» discussed in section 3.1.1, where I concluded that in the absence of a well-identified context this notion is not useful. Can we now, after having identified the omissions in the clitic-context, qualify them as illicit from the point of view of the adult grammar? As shown by several authors, comparable omissions exist in the speech of adults (Fónagy, 1985, Theoretical Implications of Object Clitic Omission in Early French CatJL 5, 2006 229 Table 4. Corpus Champaud, France; omissions and overt objects in clitic contexts. MLU AGE NULL CL LEXICAL Cl contexts (tokens) 1.74 1;10.20 66.60% 0.00% 33.30% 3 1.86 1;11.22 50.00% 0.00% 50.00% 6 2.09 2;0.5 33.30% 0.00% 66.60% 6 2.33 2;1.25 25.00% 0.00% 75.00% 4 2.55 2;3.0 80.00% 0.00% 20.00% 5 3.23 2;5.1 0.00% 100.00% 0.00% 10 3.79 2;5.13 16.60% 66.60% 16.60% 12 4.7 2;5.27 26.30% 52.60% 21.00% 19 Table 3. Corpus Mona, Montréal; omissions and overt objects in clitic contexts. MLU AGE NULL CL LEXICAL PRON Cl contexts (tokens) 1.78 2;0.14 50.00% 25.00% 0.00% 25.00% 8 1.83 2;0.28 40.00% 40.00% 0.00% 20.00% 5 2.11 2;1.16 40.00% 0.00% 0.00% 20.00% 5 2.73 2;2.9 27.27% 36.36% 36.00% 0.00% 11 2.47 2;2.22 40.00% 40.00% 20.00% 0.00% 5 2.51 2;3.20 57.14% 42.85% 0.00% 0.00% 7 2.98 2;4.4 20.00% 80.00% 0.00% 0.00% 5 3.21 2;4.18 16.66% 66.66% 16.00% 0.00% 6 3.73 2;5.1 27.27% 54.54% 0.00% 0.00% 10 3.29 2;5.15 20.00% 80.00% 0.00% 0.00% 10 3.46 2;5.29 0.00% 100.00% 0.00% 0.00% 4 3.68 2;6.12 33.33% 61.11% 5.00% 0.00% 18 Cat.Jour.Ling. 5 001-276 3/11/06 16:28 Página 229 Pérez-Leroux, Ana; Pirvulescu, Mihaela; Roberge, Yves (to appear). «Early object omission in child French and English». Selected Papers from LSRL 35. Pérez-Leroux, Ana; Pirvulescu, Mihaela; Roberge, Yves (in preparation). «A syntactic transitivity approach to null objects in child language». Ms. University of Toronto. Pirvulescu, Mihaela (to appear). «The acquisition of object clitics in French L1: spontaneous vs. elicited production». Proceedings of GALA 2005. Pirvulescu, Mihaela; Roberge, Yves (2005). «Licit and Illicit Null Objects in L1 French». In Theoretical and Experimental Approaches to Romance Linguistics, ed. by R. Gess & E. Rubin. 197-212. John Benjamins: Amsterdam. Schaeffer, Jeannette (2000). The Acquisition of Direct Object Scrambling and Clitic Placement. Amsterdam: John Benjamins. Schmitz, Katrin; Cantone, Katja; Müller, Natascha; Kupisch, Tanja (2004). «Clitic realizations and omissions in early child grammar: a comparison of Italian and French». Paper delivered at the workshop The Romance Turn, Workshop on the Acquisition of Romance Languages. Madrid, Spain. Tsakali, Vina; Wexler, Ken (2004). «Why children omit clitics in some languages but not in others: new evidence from Greek». Proceedings of GALA 2003, vol. 2, 493504. Van der Velde, Marlies; Jakubowicz, Celia; Rigaut, Catherine (2002). «The acquisition of determiner phrase and pronominal clitics by three French-speaking children». Proceedings of the 1999 GALA conference, ed. by I. Lasser, 115-132. Frankfurt am Main: Peter Lang. Wexler, Kenneth (1998). «Very early parameter setting and the unique checking constraint: a new explanation of the optional infinitive stage». Lingua 106, 23-79. Wexler, Kenneth; Gavarró, Anna; Torrens, Vicenç (2004). «Feature checking and object clitic omission in child Catalan». In (eds.). Romance Languages and Linguistic Theory 2002, ed. by Bok-Bennema, R., B. Hollebrandse, B. Kampers-Mahne and P. Sleeman. John Benjamins, Amsterdam/Philadelphia, 253-268. 236 CatJL 5, 2006 Mihaela Pirvulescu Cat.Jour.Ling. 5 001-276 3/11/06 16:28 Página 236