INTRODUCTION
Variation is widely acknowledged in metropolitan French. For example, negation with or without the negative particle ne, and subjects with or without a resumptive clitic represent two emblematic and uncontroversial areas of variation in this language (Ashby, Reference Ashby1981; Lambrecht, Reference Lambrecht1981; Zribi-Hertz, Reference Zribi-Hertz1994; Rowlett, Reference Rowlett1998; Blanche-Benveniste, Reference Blanche-Benveniste2000; Coveney, Reference Coveney2002; Barra-Jover, Reference Barra-Jover2004; Massot, Reference Massot2008, among others). Although research agrees on the existence of variation, it is still a matter of lively debate as to how to formalise it. Two opposing stances are emerging, i.e. variationism and diglossia. The former considers French to be a single grammatical system with internal variation. The language is described as ‘a conjunction of social and linguistic variableness’ where ‘variation [. . .] is better characterized as style-shifting (rather than ‘diglossia’)’ (Gadet, Reference Gadet2003: 5 and Coveney, Reference Coveney, Martineau and Nadasdi2011: 78, respectively).Footnote 1 The latter follows Ferguson's (Reference Ferguson1959) original sociolinguistic formulation for Arabic, Greek, Swiss German and Haitian Creole, and suggests that what is traditionally labelled ‘French’ actually spreads over two distinct grammars (Barra-Jover, Reference Barra-Jover2004, Reference Barra-Jover2010; Rowlett, Reference Rowlett2007; Massot, Reference Massot2008, Reference Massot2010; Zribi-Hertz, Reference Zribi-Hertz2011). Ferguson's approach further implies that the two grammars are complementary. Schematically, one grammar – the ‘high variety’ (H) – is activated in formal situations, often (but not always) when writing, whereas the other grammar – the ‘low variety’ (L) – is used in informal situations (but see Buson and Billiez, this volume on some drawbacks of this dichotomy). Furthermore, variationism and diglossia make very different predictions about variation within each utterance. Indeed, the former ‘predicts that all logically possible combinations of variants are in principle available, and [. . .] that speakers can freely combine marked H and marked L variants’ whereas the latter predicts that ‘each utterance ‘chooses’ one or the other grammar and remains consistent throughout’ (Rowlett, Reference Rowlett2011: 15 and Massot, Reference Massot2010: 98, respectively).Footnote 2
The distinction based on linguistic characteristics and distinct activation situations entails an additional dichotomy with regard to the speakers who are actually diglossic. Indeed, since ‘the actual learning of H is chiefly accomplished by the means of formal education’ (Ferguson, Reference Ferguson1959: 331), the ability to activate H can be attributed to a subgroup of speakers only, i.e. the ‘educated’ ones (Massot, Reference Massot2010: 89). Accordingly, it is further assumed that ‘non-educated’ adults and to an even higher extent preschool children should not be actively diglossic. This exclusion however is too strong since ‘a 5-year-old French-speaking child might have started to actively learn SG [Standard Grammar] without schooling – in a family concerned with the transmission of SG’ (Zribi-Hertz, Reference Zribi-Hertz2011: 241).Footnote 3 Nonetheless, it is generally expected that French-speaking children will initially produce utterances whose characteristics (almost) all belong to L and, only at a second stage closely identified with schooling, should these children generate an increasing number of utterances pertaining to H since ‘adults invariably use L in speaking to children and children use L in speaking to one another’ (Ferguson, Reference Ferguson1959: 331). However, to the best of my knowledge, a clear picture of the emergence of this French diglossia is still missing.
This article therefore proposes to delve into the subject from the point of view of acquisition, by describing and formalising the emergence of diglossia in 37 speakers of metropolitan French between the ages of 2;3 and 4;0. The main aim is to exemplify the existence of two distinct grammars in the mind/brain of these children by highlighting clustering as well as incompatibility effects for each grammar. The outline of the paper is as follows. Section 1 introduces the two sets of naturalistic data underlying this work. Section 2 then examines one of the major characteristics relevant to variation in French, i.e. subjects. The investigation is closely related to the long-standing and still lively debate surrounding the morpho-syntactic status of nominative clitics in colloquial French. I argue that early clitics are generated as verbal prefixes in an initial grammar of French common to all L1 speakers and termed Spontaneous French (G1). Object clitics and past-participle agreement are also examined in this section in order to gain a comprehensive picture of this initial grammar. Section 3 then pinpoints the emergence of two key features incompatible with G1, i.e. subject DPs with no resumptive clitics (henceforth ‘canonical subject DPs’) and the negative particle ne. The emergence of a second grammar termed Normed French (G2) is hence tentatively described and explained, and the G1/G2 distinction as well as the prediction made by diglossia about consistency within every utterance are discussed in the light of the very specific phonological accommodation pattern of the nominative clitic il in the data. Finally, I also briefly discuss learnability, theoretical and open issues relevant to diglossic acquisition.
1 THE CHILD CORPORA
This investigation is primarily based on two sets of spontaneous data collected in two different kindergarten classes of French-speaking children in Southern France. Corpus #1 stems from 5 sessions of audio recordings with 17 children between the ages of 2;3 and 3;1 (mean age 2;8). These data were collected during the first term of these children's first kindergarten year. Corpus #1 consists of 1,072 verbal utterances in total (full details in Palasis, Reference Palasis2005). Corpus #2 comes from audio and video recordings with 20 other children between 2;5 and 4;0 (mean age 3;3). It displays 9,748 verbal utterances which were gathered over the entire first kindergarten year of these children during 13 different sessions (full details in Palasis, Reference Palasis2010a, Reference Palasis2010b). Small groups of 3 to 5 children were constituted on a voluntary basis for both corpora, and the recordings took place whenever possible in a quiet room adjoining the children's usual classroom. One investigator – the same person in both studies – led each group. The data consist of spontaneous child-child and child-investigator interactions. Nevertheless, in order to encourage speech, the investigator suggested activities to each group, i.e. games or books, or asked the children to tell her about their activities in and out of school. The general atmosphere was intended to be informal and friendly.
Corpus #2 was further transcribed and coded along the CHILDES formats.Footnote 4 Each utterance displays three tiers, i.e. a main orthographical transcript, a line of morpho-syntactic codes (%mor), and, whenever relevant, an ‘error’ tier (%err) accounting for the children's non-target utterances. The utterances in (1) exemplify two possibilities with regard to subjects, i.e. a singleton nominative clitic in (1a), and associated strong and clitic forms in (1b). The same 20 children were further recorded and videotaped during their two subsequent kindergarten years, i.e. between the ages of 4;0 and 6;0. However, since these data have not yet been fully processed, they are referred to as Corpus #3 only very sporadically in this article.
(1) Transcripts and codes from Corpus #2:Footnote 5
a. il a fait un rêve (Lucille, 2;9)
%mor: pro:cli:nom|il&3S
b. moi je fais la fourmi (William, 2;9)
%mor: pro:ton:dg:nom|moi&1S + pro:cli:d:nom|je&1S
Let us now examine some of the characteristics of these children's utterances starting with a key aspect of variation in French, i.e. the morpho-syntactic status of nominative clitics.
2 DESCRIBING THE INITIAL GRAMMAR OF FRENCH (G1)
2.1 The morpho-syntactic status of nominative clitics
The morpho-syntactic status of nominative clitics in French has been the matter of much attention for years (e.g. Heap & Roberge, Reference Heap and Roberge2001 for an overview ranging from traditional to generative grammar). As far as generative linguistics is concerned, ever since Kayne's (Reference Kayne1975) seminal work on subject and object clitics, and Rizzi's (Reference Rizzi, Jaeggli and Silva-Corvalan1986) comparison between French and some Northern Italian dialects, the following question has arisen: Should French nominative clitics be analysed as proper syntactic arguments (Kayne, Reference Kayne1975; Rizzi, Reference Rizzi, Jaeggli and Silva-Corvalan1986; De Cat, Reference De Cat2005) or should they be considered as preverbal morphological affixes (Auger, Reference Auger1994; Kaiser, Reference Kaiser and Meisel1994; Zribi-Hertz, Reference Zribi-Hertz1994; Culbertson, Reference Culbertson2010)? The latter option is however envisaged only in colloquial French. Indeed, nominative clitics are analysed as full syntactic arguments in formal French, i.e. as ‘pronouns’ in the etymological sense, and various syntactic properties follow, as illustrated in (2) with Zribi-Hertz’ (Reference Zribi-Hertz1994) examples. Firstly, the syntactic subject can surface as a noun or a clitic pronoun as in (2a). Sentence (2b) then illustrates the standard analysis of left dislocation, i.e. a base-generated topic in a peripheral projection (De Cat, Reference De Cat2007). Topics are ‘preposed elements characteristically set off from the rest of the clause by a ‘comma intonation’ and normally expressing old information’ (Rizzi, Reference Rizzi and Rizzi2000: 245). Accordingly, topics are generally expected to be definite, hence the ungrammaticality with an indefinite or quantified expression, as illustrated in (2c).
(2) Formal French (Zribi-Hertz, Reference Zribi-Hertz1994: 131–4):
a. Jean/Il mangera ici
b. Pierre, il n'a rien dit
c. *Personne, il a rien dit
Consequently, the debate arises in colloquial French due to attested utterances such as personne il a rien dit (Zribi-Hertz, Reference Zribi-Hertz1994: 134), quelqu'un il dansait (Auger, Reference Auger1994: 97), tout le monde il se baignait (Blanche-Benveniste, Reference Blanche-Benveniste2000: 42), or other utterances I have collected randomly over the last few months such as on sait pas pourquoi un truc i marche, personne i veut ça and comme tout le monde i dit. Let us now examine the relevant child data.
2.2 Nominative clitics in early speech
Children's first combinations of words emerge between the ages of 1;6 and 2;0. These early utterances can be either nominal, e.g. la dame là, or verbal with a finite or non-finite verb, e.g. il mange, mangé le chien (De Boysson-Bardies, Reference De Boysson-Bardies1996: 232 and Pierce, Reference Pierce1992: 84, respectively). Nominative clitics represent the most frequent type of subject whenever the utterance displays a finite verb (Prévost, Reference Prévost2009: 131–40 for an overview of the literature). Table 1, drawn from later stages (2;3–3;1 and 2;5–4;0), illustrates this persistent characteristic of child French. Indeed, nominative clitics surface in 73.4% and 95.7% of the children's verbal utterances in Corpus #1 and #2, respectively. Conversely, canonical subject DPs, i.e. DPs with no resumptive clitics, are virtually absent from these early data. These characteristics are also found in adult oral French, as suggested with the 80% clitic and 5.9% to 20% DP rates in Blanche-Benveniste (Reference Blanche-Benveniste and Yaguello1994: 87, 90; Reference Blanche-Benveniste and Merle2003: 83). Nominative clitics are hence pervasive in child and adult oral French.
Different types of subjects between the ages of 2;3 and 4;0
A second characteristic of nominative clitics in child speech is their position with regard to finite verbs. The observed data are clear-cut since Corpus #1 displays absolutely no postverbal clitics, and Corpus #2 contains just 6 given in (3). All the other nominative clitics in these data are hence preverbal, either strictly adjacent to the finite verb, as in Table 1, or separated from the verb by one or several other clitics, e.g. i m l'a piqué. Quite expectedly, all the postverbal clitics surface in interrogative utterances. Indeed, in order to yield clitic inversion, the finite verb, which raises first from V to I to check inflectional features, needs to move further to C to satisfy the Wh Criterion (Rizzi, Reference Rizzi, Belletti and Rizzi1996), e.g. [CPdort j [IPil idortj[VPili dortj]]]. It can be seen from (3d-f) however that 3 of the 6 child utterances displaying this kind of structure are non-target.
(3) Clitic inversions in Corpus #2:
a. que se passe-t-il? (Victor, 3;5)
b. poisson où es-tu? (Dylan, 2;11)
c. où es-tu Maman Ours? (Carla, 3;5)
d. euh s(e) passe-t-il? (Victor, 3;5)
e. qu'est-c(e) qui passe-t-il? (Kelsang, 3;7)
f. qu'est-c(e) que s(e) passe-t-il? (Kelsang, 3;7)
Interrogative utterances hence highlight the fact that nominative clitics are (nearly) exclusively preverbal at this stage. Three different analyses can account for this preverbal position, i.e. no I-to-C movement of the verb, prefixation of the clitic to the verb prior to movement, or a combination thereof.
Let us now consider the different types of DPs young children spontaneously utter with nominative clitics. Corpus #2 overwhelmingly displays definite DPs, as exemplified in (4a). However, in very occasional cases, the associate can be indefinite, as in (4b-d).Footnote 6 The latter are hence reminiscent of the adult examples in Section 2.1, and comply with the morphological analysis of the clitic since un garcon, personne, quelqu'un and tout le monde, being indefinite or quantified, cannot be dislocated topics.
(4) Different types of associates in child speech:
a. l'escargot i dort (Dylan, 2;8)
b. un garcon il attrape un œuf (Mathilde, 3;0)
c. personne i m l'a dit (Matteo, 4;5, Corpus #3)
d. oh quelqu'un i l'a déchiré! (Dylan, 3;10, Corpus #3)
As far as nominative clitics are concerned, the observed child data show the following characteristics so far: (i) overwhelming co-occurrence with finite verbs only, (ii) preverbal position, (iii) possible co-occurrence with indefinite and quantified DPs. This set of properties suggests a morphological handling of these clitics in the child system. Arguments against the morphological analysis exist however. De Cat (Reference De Cat2005) for instance critically discusses some of the consequences of this analysis, i.e. (i) possible redundancy of the pre- and post-verbal agreement markers, (ii) optionality of the preverbal clitics in some cases, (iii) unavailability of these clitics for syntactic operations due to their pre-syntactic insertion, (iv) morphological analysis of the preverbal clitics that appear between the subject clitic and the finite verb, and (v) the existence of doubling in spoken French, i.e. the co-occurrence of a morphological clitic and a DP in subject position. Nevertheless, it has also been claimed that these arguments do not invalidate the morphological analysis. Firstly, most of these arguments do not apply to early French, as illustrated in the near total absence of optionality and clitic inversion, and the possible co-occurrence with indefinite and quantified DPs. Secondly, whenever these arguments do apply, i.e. redundancy and morphological analysis of other clitics, they are construed as unproblematic (see Palasis, Reference Palasis, Kaiser and Remberger2009, Reference Palasis2010b for a thorough discussion). The morphological status of nominative clitics is hence deemed to be a major characteristic of an initial grammar of French termed Spontaneous French (G1) in this work. Let us now test two additional phenomena relevant to the morpho-syntactic debate, i.e. object enclisis and past-participle agreement with object proclitics.Footnote 7
2.3 Accusative clitics in early speech
The morphological status assumed for nominative clitics in Spontaneous French implies that any constituent intervening between a nominative clitic and a finite verb is also morphological (in line with ‘clitics can attach to material already containing clitics, but affixes cannot’, Zwicky and Pullum, Reference Zwicky and Pullum1983: 504). This implication entails that: (i) accusative clitics are verbal prefixes too, and (ii) the co-occurrence of an accusative clitic and an associate DP instantiates clitic doubling, not clitic dislocation. Let us hence firstly turn to occurrences that could question the morphological analysis of clitics, i.e. object enclitics.
2.3.1 Object enclitics
Object enclitics occur in positive imperatives, e.g. prends-le. Enclisis takes place when the finite verb moves to the left periphery (Rizzi, Reference Rizzi and Rizzi2000; Han, Reference Han1998). Thus, the order v-cl is possible only if the accusative clitic is syntactically independent from the verb. Indeed, if the clitic were morphological, viz. prefixed, the syntactic operation would move the verb together with its prefix to the left periphery yielding *le-prends. Consequently, object enclisis can only surface if the child manipulates the clitic as a proper syntactic argument. Although in complementary distribution with nominative proclitics (e.g. imperative prends-le vs. indicative je le prends), object enclitics could hence cast doubt on the morphological analysis of nominative and accusative proclitics in the rest of the data.
On the one hand, young children regularly resort to imperative forms in their speech (Corpus #1: 15.9% of finite verbs; Corpus #2: 14.8%). On the other hand, both spontaneous and elicited data usually illustrate a lag of several months between the emergence of nominative and accusative clitics since the latter are uttered on a regular basis in obligatory contexts at circa 5;0 (Prévost, Reference Prévost2009: 140–143). Before this period, children usually resort to strong pronouns, full DPs, or object omission. Imperative forms in Corpus #2 comply with these early preferences since imperatives are primarily uttered intransitively, as in (5a), and when the verb is transitive the object is overwhelmingly strong, as shown in (5b). Thus, only 7 different children utter a total of 15 object enclitics, as exemplified in (5c). Although always correctly positioned with regard to the finite verb, object enclisis is hence not productive at this stage. Consequently, except for the 15 above-mentioned exceptions which could belong to an emerging G2, no overt evidence invalidates the morphological handling assumed so far for G1. Moreover, Haverkort and Weissenborn (Reference Haverkort and Weissenborn1991) noticed that younger children (2;0–2;4) initially produce positive imperatives with cl-v sequences, as illustrated in (5d).
(5) Positive imperatives in child speech:
a. regarde! (Matteo, 2;11)
b. eh Lucille touche ça! (William, 2;10)
c. mets le sur la table! (Nina, 3;3)
d. le mets là-dedans! (2;0, Haverkort and Weissenborn, Reference Haverkort and Weissenborn1991: 3)
Haverkort and Weissenborn (Reference Haverkort and Weissenborn1991) analysed this non-target order as an absence of I-to-C movement. However, in line with the morphological hypothesis and further to what was said with regard to nominative clitic inversion in Section 2.2, it could also be envisaged that utterance (5d) displays a verb with a prefixed clitic in either I or C.
2.3.2 The Doubling/Agreement Correlation
The ‘Doubling/Agreement Correlation’ proposed by Tsakali and Anagnostopoulou (Reference Tsakali, Anagnostopoulou, Kallulli and Tasmowski2008) is also of particular relevance to our morpho-syntactic debate. Indeed, according to this correlation object-clitic doubling and participle agreement with direct objects are in complementary distribution, i.e. ‘if a language has participle agreement it lacks clitic doubling’, and vice versa (Tsakali and Anagnostopoulou, Reference Tsakali, Anagnostopoulou, Kallulli and Tasmowski2008: 322). Formal French and Modern Greek illustrate these two possibilities, as shown in (6).
(6) The Doubling/Agreement Correlation:
a. [–] Doubling, [+] Agreement (e.g. Formal French):
je li’ai achetée (*la tartei)
b. [+] Doubling, [–] Agreement (e.g. Modern Greek):
tini eho agorasi tin turtai
it-acc-fem-sg have-1sg bought-pp-masc the tart-acc-fem-sg
‘I have bought the tart’
It is generally acknowledged however that ‘for many speakers, this agreement [past-participle agreement with direct-object clitics] is not made in spoken French’ (Kayne, Reference Kayne and Kayne2000: 10). Following Tsakali and Anagnostopoulou's correlation and Kayne's statement, we might suggest that early French displays the [+] Doubling/[–] Agreement option, i.e. has morphological accusative clitics, and we can examine Corpus #2 in order to test child data for this correlation. As indicated earlier, however, accusative clitics are still scarce at this stage. Consequently, this investigation merely represents a first attempt to add the correlation to the debate, and Corpus #3 should be more informative on this matter. Indeed, utterances displaying an accusative clitic together with an associate DP and a past participle are very rare in Corpus #2 with only two utterances meeting these requirements, as in (7a-b).
(7) Past-participle agreement in Corpus #2:
a. eh tu l’i as pris le chien-chieni? (William, 3;0)
b. tu l’i as pas mis bien çai (Matteo, 3;3)
c. y'a Matteo qui me l'a pris! (Nina, 3;4)
(7a-b) display a masculine object which does not yield a modification of the default masculine past-participle form. Consequently, no overt evidence emerges in these two utterances with regard to [+] Agreement, and the morphological analysis of clitics, corresponding to the [+] Doubling/[–] Agreement option, can be maintained for G1. (7c) points in the same direction since it illustrates the [–] Agreement option with a previously-mentioned feminine object (la pomme).
2.4 From G1 to G2
Drawing on the observation of naturalistic child data, Section 2 aimed to describe the grammar of French that emerges spontaneously with young children in the initial stages of their linguistic development, hence its label Spontaneous French (G1). One pivotal characteristic of this grammar is the morphological status of its nominative and accusative clitics. In order to buttress the hypothesis on diglossia, we now need to describe the emergence of a second grammar (G2), by identifying properties incompatible with G1, and test the prediction made by diglossia about grammatical consistency within utterances.
3 WHY ADD G2?
Diglossia posits that, at a particular point in time, speakers move from their initial single grammar to a choice between two cognate grammars. The immediate questions are why and when. Actually, we have already mentioned child utterances non-ascribable to G1. This section firstly details these occurrences, and then turns to a further crucial topic, viz. negation.
3.1 Canonical subject DPs
Canonical subject DPs, e.g. le train est là (see Table 1), represent the most emblematic incompatibility with G1 due to the absence of the nominative prefix in this configuration.Footnote 8 It is hence suggested that the corresponding 39 sentences in Corpus #2 pertain to a different grammar, i.e. G2. Two types of utterances emerge within the relevant series. Indeed, only 29 sound genuinely spontaneous since the other 10 correspond to previously heard nursery rhymes or stories, as in (8a-b) and (8c), respectively.
(8) Previously heard utterances in Corpus #2:
a. le facteur n'est pas passé (Victor, 3;7)
b. vole mon petit chagrin (Lucille, 3;4)
c. pas moi dit le chat (Nina, 3;2)
Interestingly, these utterances display three main characteristics with respect to children, development and constituents. Firstly, as far as children are concerned, only 9 out of 20 utter canonical subject DPs, as detailed in Table 2. Furthermore, the vast majority are produced by one child only, Lucille, who utters 46.1% of the series. Secondly, the progression between the first and the last recordings is also of interest since 9 of the 29 spontaneous utterances belong to the last session in contrast with the other sessions that display a maximum of 3 DPs each.
The emergence of canonical subject DPs in Corpus #2
Finally, these utterances stand out because of the presence of quantified elements which are scarce in the rest of the data. Personne for instance surfaces a total of 5 times in verbal utterances and only once as a subject, as in (9b).
(9) Quantified subjects in Corpus #2:
a. et quelqu'un regarde la main (Sara, 3;11)
b. personne me le retourne comme ça (Lucille, 3;4)
Drawing on these facts, it is surmised that these utterances illustrate the emergence of a second grammar, G2, and that contrary to G1, G2 displays syntactic clitics. These are hence not obligatory, not limited to proclisis, and not compatible with indefinite and quantified DPs, as illustrated in (2), (3) and (9), respectively. Few children show signs of G2 at this stage. The most obvious utterances come from Lucille as will be confirmed in the next section on the emergence of discontinuous negation.
3.2 The emergence of the negative particle ne
Sentential negation represents another uncontroversial area of variation in French. Formal French displays discontinuous negation, i.e. Neg1 (ne) + Verb + Neg2 (pas, plus, etc.), whereas colloquial French shows simple negation, viz. the absence of the negative particle ne (Ashby, Reference Ashby1981, Reference Ashby2001; Lambrecht, Reference Lambrecht1981; Bernini & Ramat, Reference Bernini and Ramat1996; Rowlett, Reference Rowlett1998; Coveney, Reference Coveney2002; Barra-Jover, Reference Barra-Jover2004; Massot, Reference Massot2008, among others). Seminal work on early acquisition (1;8–2;2) shows the correct placement of pas with regard to finite and non-finite verbs, e.g. veux pas lolo vs. pas manger la poupée (Déprez & Pierce, Reference Déprez and Pierce1993: 40). These facts are interpreted as evidence that young children project and master functional categories and syntactic movements pertaining to verbal inflection. No explicit comments are made however on markers other than pas and the preverbal particle ne. Let us hence turn to Corpus #2 in order to delve into these matters.
Children negate 13.5% of their verbal utterances in Corpus #2. This rate is consistent with previous analyses (e.g. Friedemann, Reference Friedemann1993: 233). Table 3 shows that child negation displays three different markers at this stage, i.e. pas, plus and jamais, and that pas is by far the most frequent one. The table also highlights the rarity of discontinuous negation in these data since the negative particle ne emerges in 16 utterances only, i.e. 1.2% of the negated sentences.
Negative markers in Corpus #2
The 16 occurrences of discontinuous negation are given in (10). They illustrate that ne is uttered by 4 children only (out of 20). Moreover, the type of utterance is of considerable significance: 11 of these occurrences are either songs (Mathilde) or nursery rhymes (Victor). Consequently, as for clitic inversion and canonical subject DPs, none of the utterances seems to pertain to genuine spontaneous speech.
(10) Discontinuous negation in Corpus #2:
a. Mathilde (3;2), 10 occurrences in a row, same pattern, singing, e.g.:
il ne chante plus/il ne saute plus/il ne danse plus
b. Victor (3;7), 1 occurrence, nursery rhyme:
le facteur n'est pas passé
c. Lucille (2;10–2;11–3;2), 3 occurrences:
i. ta photo n'est pas là?
ii. non non ce n'est pas moi
iii. maman elle n'est pas là
d. Sara (3;10), 2 occurrences:
i. il n'a pas gagné pour les œufs
ii. ne te cache pas avec la photo!
The 13 recording sessions over a period of 7 months exhibit precise information on the moment ne emerges in the corpus. Table 4 indicates the relevant ages and corresponding sessions. Lucille is the first child to utter ne at 2;10, Sara is the fourth at 3;10, and the 16 remaining children – who are between 3;0 and 4;0 by the end of the investigation – only use simple negation. Unfortunately, the exact lag between the emergence of simple and discontinuous negation cannot be calculated within this corpus since the relevant 4 children were already producing simple negation from the outset of the study (in accordance with their ages reported in the last column of Table 4).
The emergence of discontinuous negation in Corpus #2
3.3 A possible developmental trajectory
The data reported in (10) present an additional peculiarity in comparison with the rest of the corpus. Thus, (10b) and (10ci) exemplify the co-occurrence of two rare phenomena, viz. the emergence of ne together with the absence of the nominative clitic (1.2% and 0.4% of the utterances, respectively).Footnote 9 A comparison of the names and ages in Tables 2 and 4 shows that 4 children produce canonical subject DPs and discontinuous negation in Corpus #2, and that both phenomena emerge at the same age for Lucille (2;10), Victor (3;7) and Sara (3;10). The fourth child, Mathilde, shows a lag, and hence a sequence, between the emergence of her first canonical DP (2;9) and discontinuous negation (3;2). This order is interestingly also displayed by the remaining children in Table 2 since William, Dylan, Nina, Kelsang and Enzo rarely produce canonical subject DPs in this corpus and no discontinuous negation. Canonical subject DPs hence seem to emerge prior to or no later than discontinuous negation within the course of acquisition.
Since discontinuous negation and canonical subject DPs are traditionally assumed to belong to the ‘high variety’ (H) of French, it is hypothesized that we might be starting to gain a picture of the actual emergence of G2 through a second set of possibly correlated properties. The two distinct series of characteristics for G1 and G2 are recapitulated in Table 5.Footnote 10
Two distinct series of properties (provisional version)
Table 5 sketches the possible existence of two cognitively distinct grammars in the mind/brain of French speakers. Drawing on adult data, Massot (Reference Massot2010: 103) suggests a clausal domain of application for each grammar. Applied to Corpus #2, this proposal predicts for instance that a morphological clitic will not co-occur with the negative particle ne, and that a syntactic clitic will not co-occur with simple negation. The following section illustrates this particular point relying on the very specific behaviour of clitic il in the child data.
3.4 The strict accommodation pattern of il
A typical characteristic of the nominative clitic il in Corpus #2 is its very strict accommodation pattern according to the phonological environment. Indeed, the analysis of the 1,688 il clitics shows that this element surfaces as [il] before a vowel and [i] before a consonant, as illustrated in (11).
(11) Strict accommodation pattern of il in Corpus #2 (provisional version):
a. il/_V:
il est à moi (William, 2;9)
b. i/_C:
i veut manger (Lucille, 2;9)
On the one hand, these facts do not seem particularly enlightening since French-speaking children favour this CVCV pattern from their babbling (De Boysson-Bardies, Reference De Boysson-Bardies1996: 80). On the other hand, the pattern described in (11) admits of 11 exceptions in the corpus, and their analysis seems to point to an interesting phenomenon with regard to the G1/G2 distinction and the prediction made by diglossia about grammatical consistency within utterances. Indeed, as can be seen from the relevant sentences reported in (12), the exceptions are definitely not random. Thus, 10 of these utterances display the negative particle ne which we ascribed to G2. Following Massot's (Reference Massot2010) proposal on consistency, it is hence further assumed that: (i) the non-elided il clitics in these utterances are also generated by G2, and (ii) these clitics are syntactic arguments, contrary to the 1,677 other il clitics that are generated by G1 as verbal prefixes. Corpus #2 also exemplifies consistency within G1 since elided clitics co-occur with simple negation only, as illustrated in (13).Footnote 11 Rizzi's (Reference Rizzi, Jaeggli and Silva-Corvalan1986: 401) rejection of *personne il ne mange could then be reinterpreted in terms of G1 and G2 grammars as belonging to neither G1 (which yields personne i mange) nor G2 (due to the constraint on Topics).
(12) The 11 non-elided forms of il/_C (G2):
a. 9 occurrences, all on the same pattern (Mathilde, 3;2, singing):
il ne chante/saute/danse plus
b. il n'a pas gagné pour les œufs (Sara, 3;10)
c. et après oh il dit qui est là (Lucille, 3;2)
(13) Elided clitics co-occur with simple negation only (G1):
a. i (*ne) colle pas (Chloé, 2;7)
b. comme ça i (*ne) nous mangera pas (Lucille, 3;1)
The correlations exemplified in (12a-b) and (13) between ne and syntactic clitics on the one hand, and simple negation and morphological clitics on the other hand call for the following additional comments. Firstly and similarly to negation, a lag appears between the emergence of the two different types of clitics. Section 2 reported the early appearance of morphological clitics at circa 1;8. The utterances in (12) now point to an emergence of syntactic clitics at 3;2 for Mathilde and Lucille, and various subsequent ages for the other children (still to be confirmed in Corpus #3). G2 hence definitely emerges at different moments of the linguistic development, but always after G1.Footnote 12 The second comment relates to the morpho-syntactic status of ne. Since clusters such as i-ne-chante pas (cl-neg-v) and ne-i-chante pas (neg-cl-v) are not attested, the data suggest that ne is not available as a verbal affix in Spontaneous French, contrary to some Northern Italian dialects (Rizzi, Reference Rizzi, Jaeggli and Silva-Corvalan1986; Zanuttini, Reference Zanuttini1997). Thus, ne is available only in Normed French as the head of an independent negative projection, viz. NegP (following Pollock, Reference Pollock1989 and Rowlett, Reference Rowlett1998, among others). Finally, as ne never co-occurs with a nominative prefix, a blocking effect is presumed between ne and morphological clitics. This blocking effect is further interpreted as a trigger that forces a speaker to move from G1 to G2, since ne is totally unavailable in G1 whether as a verbal affix or as an independent syntactic head. The very existence of this blocking effect should also be accounted for. At this stage of the investigation, I merely tentatively question the status of pas in both grammars, suggesting possible different underlying structures altogether in line with the various existing analyses of pas in the literature, e.g. specifier of NegP (Pollock, Reference Pollock1989), initial VP adjunction (Rowlett, Reference Rowlett1993, Reference Rowlett1998), multiple specifiers in VP (Péters, Reference Péters1999), or surmise a possible connection with the overall development of quantification, further to Schapansky's (Reference Rowlett2002) proposal that negation with ne requires negation and quantifying features.
4 CONCLUSION
This article aims to lend support to the French diglossia hypothesis (Barra-Jover, Reference Barra-Jover2004, Reference Barra-Jover2010; Rowlett, Reference Rowlett2007; Massot, Reference Massot2008, Reference Massot2010; Zribi-Hertz, Reference Zribi-Hertz2011). Drawing on naturalistic data from native French children under 4, two cognate but distinct grammars are hence described by hinging upon two crucial distinctions, i.e. the morpho-syntactic status of nominative clitics and the emergence of the negative particle ne. Thus, the initial Spontaneous French grammar (G1) displays morphological proclitics and simple negation, whereas the subsequent Normed French grammar (G2) features syntactic clitics and discontinuous negation. The detail of these characteristics is recapitulated in Table 6. G1 is common to all native French speakers, whereas G2 gradually emerges in one fifth of the observed children. It is also proposed that the existence of discontinuous negation in G2 represents one of the triggers forcing a speaker to move from G1 to G2 due to the total unavailability of the particle ne in G1. Finally, the data on canonical subject DPs, discontinuous negation, and the elision pattern of clitic il seem to support the prediction made by diglossia and argued for by Massot (Reference Massot2010) about grammatical consistency within each utterance. These observations call for further comments on learnability and theoretical issues, and provide directions for future research.
Two distinct series of properties (final version to date)
The early acquisition of two grammars inevitably raises questions concerning learnability. Firstly, I follow the overall generative rationale for language acquisition implying ‘[. . .] that the human language faculty predisposes the individual to become bilingual and that adequate theories of language and of grammar need to reflect this fact’ (Meisel, Reference Meisel, Cenoz and Genesee2001: 12). Accordingly, diglossia does not challenge our cognitive capabilities. Its formalisation however remains a matter of lively debate, especially within the Principles and Parameters framework (from Chomsky, Reference Chomsky, Hornstein and Lightfoot1981 to Yang and Roeper, Reference Yang, Roeper and Boeckx2011). This article highlights clustering and blocking effects. Should these be construed as mere tendencies, macro-parameters, micro-parameters, learning biases (Newmeyer, Reference Newmeyer2005; Baker, Reference Baker and Biberauer2008; Palasis, Reference Palasis2010b; Boeckx, Reference Boeckx, Di Sciullo and Boeckx2011, respectively), or does Rowlett's ‘bolt-on approach’ account for ‘the non-randomness of the location/nature of the variation’ (Rowlett, this volume)?
Additional acquisition data are always welcome in order to further our understanding of linguistic competence. When fully processed, Corpus #3, i.e. naturalistic data from the same children between the ages of 4;0 and 6;0, should give us a wealth of detail about the emergence and development of French diglossia. Furthermore, elicitation tasks with children and Event-Related brain Potential measures with adults have already lent support to diglossia as far as Arabic is concerned (Khamis-Dakwar, Froud & Gordon, Reference Khamis-Dakwar, Froud and Gordon2011 and Khamis-Dakwar & Froud, Reference Khamis-Dakwar, Froud and Mughazy2007, respectively). These paradigms could therefore inspire further research on French.
Last but not least, several important domains of acquisition had to be left aside in this article, notably interrogative and relative clauses. Nevertheless, diglossia could also well account for the ‘degraded status’ of the wh-in-situ question Jean ne mange pas quoi? reported in Boskovic (Reference Boskovic, Boskovic and Lasnik1998: 246) since the author interestingly highlights a blocking effect of neg on wh-movement in covert syntax.Footnote 13 As far as relative clauses are concerned, Guasti's (Reference Guasti2004: 240) statement that ‘children's deviations are likely due to their ignorance of relative pronouns, a lacuna that one must assume is filled in during the school years through explicit teaching [and] lexical learning’ lends direct support to the developing aspect of G2. Both these matters are, of course, already on the research agenda.
