1 INTRODUCTION
Much has changed in recent years in the study of language variation and change. In the context of French it was once a commonplace to claim that there had been no significant syntactic change since the end of the seventeenth century, and that the label ‘Modern French’ reflected a three-century-long period of grammatical stability (Rowlett, Reference Rowlett2007: 9). According to Gadet's (Reference Gadet, Beeching, Armstrong and Gadet2009) survey of work in the area, this is at least in part due to the ‘ideology of the standard’ (Milroy and Milroy, Reference Milroy and Milroy1985) which surrounds the French (of France), and, relatedly, to the fact that the use of linguistic corpora, especially in relation to the spoken language, developed later for French than, for example, English or Italian.Footnote 2 The relatively recent interest in syntactic variation within specifically spoken French is very closely associated with the work of the late Claire Blanche-Benveniste's Groupe aixois de recherches en syntaxe (GARS) at the University of Provence which has looked at VP-related valency variation (micro-syntax) and CP-related discourse/pragmatic variation (macro-syntax),Footnote 3 and has even placed a question mark over the status of the sentence as the fundamental unit of syntactic description. Work on syntactic variation has now broadened out beyond narrow normative written French and takes in: (i) social/stylistic variation within France (Armstrong, Reference Armstrong2001; Blanche-Benveniste with Martin, Reference Blanche-Benveniste and Martin2010); (ii) French in contact situations (Spaëtt ed. Reference Spaëtt2010); (iii) diatopic variation outside France, including comparison with ‘le français de référence’ and across ‘non-standard’ varieties (Gadet and Jones, Reference Gadet and Jones2008); and, more recently still, (iv) French as used in online/computer-mediated environments (van Compernolle Reference van Compernolle2008; Damar, Reference Damar2008; Doehler, Reference Doehler2011; Stark, Reference Stark2011).
The article is organised as follows: section two illustrates syntactic variation in French and the ‘variationist’ approach to it; section three sets out an alternative account based on the notion of diglossia; section four demonstrates how surface forms can have two distinct sets of properties in the two putative grammars; section five explores the nature of the relationship between the two grammars that Massot proposes, and suggests an alternative to the two-grammar hypothesis; section six contains some concluding remarks.
2 SYNTACTIC VARIATION IN FRENCH
Syntactic variation in the French of contemporary France is the subject of a 2008 Paris 8 PhD thesis by Benjamin Massot. Massot catalogues a number of familiar areas of syntactic variation in French. First, the preverbal negative marker ne co-exists with a null variant Ø:
(1)
Footnote 4 Second, topical subjects can, but do not have to, be clitic left dislocated:
(2)
Third, non-presuppositional yes–no interrogatives either display ‘inversion’Footnote 5 of the finite verb or the use of the interrogative marker est-ce que ‘is it that’:
(3)
Fourth, exclamatives can be marked by (Ah) que . . .! or Qu'est-ce que . . .!:
(4)
Fifth, the first-person plural subject proform nous ‘we’ alternates with on ‘one’:
(5)
Finally, in a number of specific (non-adjectival) contexts there is variation as to whether or not past participles show overt agreement for gender and number with a preceding direct object:
(6)
See Zribi-Hertz (Reference Zribi-Hertz2011: 236–7) for details of a wider range of areas of variation across different levels of grammar.
Uncontroversially, Massot (Reference Massot2008) notes that, in terms of sociolinguistic status, the variation illustrated in (1)–(6) is not free. Rather, in each case the (a) and (b) variants have more or less clearly marked and contrasting stylistic profiles: the (a) examples are high status (H), while the (b) ones are low status (L). Further, given the absence of any immediately obvious unmarked (= both H and L) alternative, speakers using these structures have no option other than to adopt an overtly H or L style. This contrasts with the situation illustrated in (7) and (8), where in (7) the unmarked variant in (7a) alternates with a highly marked H variant in (7b), and in (8) the unmarked variant in (8a) alternates with a highly marked L variant in (8b).Footnote 6
(7)
(8)
Massot (Reference Massot2008) has a particular interest in the data in (1)–(8), namely the issue of how the attested variation is related to the grammar encoded in speakers’ minds. According to one model, which might be labelled variationist,Footnote 7 speakers have access to a single grammar which internally accounts for the attested variation: a number of loci of variation are embedded within the grammar, and for each one speakers make a choice on the hoof as to which (stylistically marked or unmarked) available variant to use in a particular utterance. Such an approach predicts that all logically possible combinations of variants are in principle available, and, in the case of the variation illustrated in (1)–(6), that speakers can freely combine marked H and marked L variants. Significantly, though, Massot (Reference Massot2008) shows that free co-occurrence of variants is not found. For example, with respect to the variation illustrated in (1) and (2) above, the variationist approach predicts that all four combinations of negation with ne (H) versus negation without ne (L), and of no dislocation (H) versus dislocation (L), should be attested. Yet, they are not, as shown by the judgements in (9):
(9)
Footnote 8 While the grammatical (9a) illustrates the choice of two H options (no dislocation; ne present) and (9d) the choice of two L options (dislocation; ne absent), the ungrammatical (9b, c) represent the selection of two mismatching options, one H and one L. This pattern of (un)grammaticality (or at least unacceptability) is not expected within the variationist model of variant selection within a single grammar, and Massot (Reference Massot2008) therefore rejects that model.
3 FRANCE AS DIGLOSSIC
As an alternative to the variationist approach based on choices available within a single grammar Massot (Reference Massot2008) suggests that the attested variation is a reflection of the fact that individual speakers have access to two distinct ‘French’ grammars, one of which is sociolinguistically/stylistically marked L, while the other is marked H. The data in (9) are therefore captured by two grammatical differences between the two varieties.
Various labels have been used in the literature to characterise H and L grammars/varieties, for example, le français avancé, le néo-français, colloquial French, contemporary French or spontaneous French (L) contrasting with written French, modern French, or normed French (H). Massot (Reference Massot2008) settles on the terms français démotique (FD) for the L grammar, and français classique tardif (FCT) for the H variety. FD and FCT are characterised in (10a) and (10b) respectively:
- (10)
a. FD: an innovative but socially stigmatised vernacular acquired early in the naturalistic setting of the home and in which the speaker has a stable, active competence which might be described as that of a native speaker.
b. FCT: a conservative and more prestigious variety learnt later, under the influence of schooling and literacy, conscious instruction from caregivers and the normative tradition, and not necessarily to the same degree of stability (perhaps even only with a passive competence), and hence characterised by uncertainty and hypercorrection.
For Massot FD and FCT co-exist in each speaker's mind as two distinct and autonomous grammars. As we have seen, there is overlap (extensive but not total) between FD and FCT, and so some surface forms will be generated by both grammars and have the same grammatical properties in each: such areas of overlap, such as those in (7a) and (8a), will be sociolinguistically / stylistically unmarked (both H and L); forms generated by FCT but not FD, such as those illustrated in (1a)–(6a) and (7b), will be prestigious, valued, normative (H); forms generated by FD but not FCT, such as those illustrated in (1b)–(6b) and (8b), will be stigmatised, non-normative (L).
Massot's two-grammar account of the judgements in (9a–d) is based on the notion that only one of the two available sets of features (= grammars) can be accessed at one time, at least within the context of a single clause,Footnote 9 and that the choice which the speaker makes will be made on the basis of a sociolinguistic assessment of the context. For our purposes, the model explains why, within the relevant unit, a speaker is unable to combine uniquely FD (L) features with uniquely FCT (H) features. In section five I suggest that the incompatibility of marked L and H features within a single clause can be explained without recourse to two grammars. But first, in section four, I propose a refinement to Massot's approach by considering surface forms which are generated by the two grammars posited, but which have different underlying structures depending on the source grammar.
4 ONE SURFACE FORM, TWO GRAMMATICAL STATUSES
Massot assumes, along with Zribi-Hertz (e.g. Reference Zribi-Hertz2011: 240–1) for example, that a surface form generated by both FCT and FD (and therefore stylistically/sociolinguistically unmarked) has the same grammatical properties in each. This is true of the structures illustrated in (7a) and (8a). However, it is not necessarily the case: one and the same surface form can have quite different grammatical properties in FCT and FD. For example, in her discussion of matrix wh interrogatives in Francilian French Hamlaoui (Reference Hamlaoui2011: 134–5) contrasts in situ object wh questions as illustrated in (11) in the dialect described in Boeckx (Reference Boeckx1999) with her own dialect:
(11) T(u) as acheté quoi?
In Boeckx's variety an answer to the question in (11) along the lines of Rien ‘Nothing’ is infelicitous, while in Hamlaoui's it is perfectly fine. In Boeckx's grammar (11) asks for the value of x, such that x identifies an entity (and therefore presupposes that such an entity exists); in Hamlaoui's grammar it asks for the set of individuals (which can be empty). However this difference is captured in terms of abstract representation (an issue which is beyond the scope of this article), it is clear that one and the same surface form has distinct linguistic properties depending on which grammar generated it.
A second example of one form having different structures depending on its source grammar is noted by Massot himself (Reference Massot2008: 208–22), in respect of negation. Massot gives the examples in (12) (his (19), p. 213):
(12)
I have tried to capture the subtle semantic contrast in the (somewhat awkward!) translations. The important feature is that ne is optionally present in (12a), but necessarily absent in (12b). Massot's perfectly plausible approach is to say that, in the ne-less version of (12a), pas raises to NegP, thereby encoding wide-scope sentential negation, which means that Neg° is generated but has a null realisation (see (13a)), while in the necessarily ne-less (12b), pas has narrow scope over évident, occupies a relatively low position in surface syntax (hence to the right of (= lower than) être) and is not associated with any NegP projection, hence the unavailability of ne (see (13b)).
(13)
These two approaches to the syntax of pas mean that a superficially simple negative sentence like (14) is compatible with two analyses, one in which pas is associated with NegP and Neg° is null, the other in which pas is a low, VP adverb, and in which NegP is not generated.
(14) Il vient pas.
In other words, a single surface form like (14) has different structures depending on its source grammar, and is therefore not in any real sense part of any overlap.
A final surface form which has two underlying structures depending on the grammar which generates it is est-ce que/qui ‘is it that’,Footnote 10 found in interrogative sentences. Diachronically, est-ce que/qui is transparently the output of an overt (and presumably genuine — see footnote 5) syntactic inversion process which has applied to c'est que/qui ‘it is that’, a cleft structure marking pragmatic focus. Synchronically, things are not so straightforward. Est-ce que/qui is found in both FCT and FD, and in that sense might a priori be thought of along the same lines as (7a) and (8a): these can be generated by both grammars, and there is every reason to believe that they have the same grammatical status in each. This is Massot's approach to est-ce que/qui (see Massot Reference Massot and Barra-Jover2010, Fig. 1). Yet for est-ce que/qui the approach is problematic. The problem does not lie so much within FCT: since this grammar is known to retain ‘inversion’ (see (3a) and footnote 5), we can assume that est-ce que/qui in FCT is an instance of an ‘inverted’ cleft. Rather, the problem lies within FD: FD has not retained ‘inversion’, and in fact est-ce que/qui is presented in (3b) as the FD alternative to the ‘inverted’ FCT form. Est-ce que/qui is therefore deemed to have a different grammatical status in FD to the one it has in FCT. Rowlett (Reference Rowlett2007) adopts an analysis of est-ce que/qui in FD as a grammaticalised complementiser drawn from the lexicon ready-made (Martinet Reference Martinet1960), and merged directly in a head position within the left clause periphery, without marking pragmatic focus. Thus, unlike Massot's (Reference Massot and Barra-Jover2010) approach, in which est-ce que/qui has the same syntactic and, presumably therefore, also pragmatic properties in both FCT and FD, two distinct analyses are posited for est-ce que/qui which differ both syntactically (result of ‘inversion’ (FCT) or atomic complementiser (FD)?) and pragmatically (focus-cleft (FCT) or not focus-cleft (FD)?):
- (15)
a. Est-ce que tu pars? In FCT: biclausal; the consequence of the Q feature being merged on I*° in the matrix clause; an ‘inverted’ cleft expressing pragmatic focus.
b. Est-ce que tu pars? In FD: monoclausal; the consequence of the Q feature being merged on C*°; therefore not inverted; therefore not expressing pragmatic focus.
The interaction between these two dimensions of variation is illustrated in (16) in the context of the question ‘Who can you see?’ (Rowlett Reference Rowlett2007: 210).Footnote 11
(16)
Footnote 12Qui est-ce que tu vois? appears twice in (16) (italicised): bottom left it is generated by the FCT grammar within an inverted cleft with wh fronting; top right it is generated by an FD grammar within an uninverted non-cleft with wh fronting and the atomic complementiser est-ce que drawn from the lexicon. The question has a different pragmatic status depending on which grammar generated it.Footnote 13
The two analyses of est-ce que/qui, one for FCT on the basis of syntactic ‘inversion’, another for FD without such a device, address several issues and these are set out in the following sections. The issues relate to double clefting, tense marking, ‘inversion’ in subordinate interrogatives and the parallel between est-ce que and si ‘if’. In each case, it is shown how the data could not be captured if it were assumed that the same surface form has the same syntactic properties in both FCT and FD.
4.1 Double clefting
Consider the examples in (17) and (18), generated by an FD grammar, and taken from the bottom right-hand corner cell in the table in (16):
(17) Qui est-ce que c'est que tu vois?
(18) Qui c'est que c'est que tu vois?
If est-ce que in (17) were the result of ‘inversion’ in FD, as it is in FCT, then this example would be derived from the uninverted (and non-wh-fronted) structure in (19):
(19) C'est qui que c'est que tu vois?
This is problematic since the prospective underlying structure in (19) is characterised by two instances of clefting, a feature which is hard to motivate on pragmatic grounds. The same is also clearly true of (18). If, in contrast, and as suggested here, c'est que/qui and est-ce que/qui are used within an FD grammar as an atomic complementiser drawn directly from the lexicon, without the pragmatic force of a cleft, then there is no need to derive (17)/(18) from (19). Examples (17)/(18) are single, pragmatically motivated cleft sentences containing a formally complex, but syntactically atomic, complementiser.
4.2 Tense marking
The two analyses of est-ce que/qui, one within FCT and another within FD, make contrasting predictions in respect of tense marking. In FCT est-ce que/qui is an inverted cleft, and est a finite verb. As such, tense variation is expected (together with sequence-of-tense implications), as in the examples in (20), which run parallel to Qui est-ce que tu vois?, from the table in (16):
(20)
The phenomenon is quite rare, probably due to the anchoring of clefts to utterance time,Footnote 14 but the following are attested examples found online:
(21)
As expected given their FCT origin, the examples in (21) are stylistically marked as H (in addition to expressing pragmatic focus).
In FD est-ce que/qui is an atomic complementiser drawn ready made from lexicon; it does not express pragmatic focus; it does not contain a finite verb, and so tense-related variant forms are not expected. In other words, not only are the examples in (21) expected to be stylistically marked as H, the FD equivalents are expected to contain one of the forms in the right-hand column of the table in (16), irrespective of the tense of the lexical verb, and this prediction is borne out, as illustrated in the examples in (22):
(22)
4.3 ‘Inversion’ in subordinate interrogatives
The dual analysis of est-ce que/qui explains the mystery of the contrast in (23) (from Jones Reference Jones1999: 190):
(23)
Example (23a) shows that ‘inversion’ is impossible in subordinate interrogatives, in FCT and FD alike. This has been accounted for theoretically by claiming that, in both grammars of French (as indeed in standard English), in selected interrogative contexts like indirect questions the crucial Q feature is merged high in the left periphery, that is, higher than in matrix interrogatives, and ‘inversion’ is therefore unmotivated. On that basis, the grammaticality in FD (although not in FCT) of (23b), containing est-ce que/qui, is possibly unexpected. Crucially, though, this is only the case if est-ce que/qui reflects ‘inversion’ in FD as it does in FCT. But, if, as suggested here, FD is not characterised by ‘inversion’, and if est-ce que/qui in FD is instead an atomic complementiser rather than the output of ‘inversion’, then nothing more needs to be said: ‘inversion’ in (23a) is ungrammatical for both grammars because Q is merged high in subordinate interrogatives; est-ce que in (23b) is grammatical (in FD) because it is not generated by ‘inversion’ but is instead one of several available lexical realisations of the interrogative complementiser (alongside Ø, que and c'est que):
(24)
4.4 ‘Est-ce que’ vs. ‘si’
Finally, the claim that est-ce que/qui functions in FD as an atomic complementiser, rather than as the output of ‘inversion’, explains the two ways in which it parallels the uncontroversial complementiser si ‘if’. The first is that est-ce que/qui is a direct FD equivalent of si ‘if’ as a marker of an indirect yes–no question, as in (25):
(25)
The second is illustrated in (26), which Goosse (Reference Goosse, Antoine and Cerquiglini2000: 114) characterises as oral and regionally marked (hence FD in our terms).
(26)
Assuming a parallel structure across the two conjuncts of the conjunction, est-ce que and si both appear to function as a complementiser introducing a matrix interrogative. Crucially, such a parallel analysis would not be possible if est-ce que/qui were analysed here as the output of ‘inversion’.
In short, therefore, there are at least four reasons not to conclude that the surface form est-ce que/qui has the same grammatical status in its two source grammars, and reasons to believe that superficially identical negative and matrix wh-interrogative forms have different underlying properties depending on which grammar generates them. Massot's diglossic approach to syntactic variation in French consequently needs to be revised to allow this: the term ‘overlap’, sometimes represented as intersecting sets, underestimates what is actually going on. In section five I show that the idea of one and the same form having different structures depending on which grammar has generated it is actually key to understanding diglossia.
5 TWO GRAMMARS OR NOT TWO GRAMMARS?
So far we have identified syntactic variation within modern French, rejected the variationist account of it on the grounds that it fails to explain why some combinations of variants are not attested, seen the diglossic approach as a potential alternative account which does not have this weakness, and added that the model needs to be flexible enough to allow, where appropriate, a surface form to have distinct properties in the two posited grammars (as well as allowing for structural ambiguity within one and the same grammar; see footnote 13). In this section I consider issues which suggest that the two-grammar model is too powerful for French. Those issues are: (a) the overlap between the two posited grammars; (b) the differences between the two grammars; and (c) the cognitive status of the two grammars. I then argue that these issues point to a rethink, not in terms of two grammars, but instead in terms of a single grammar together with what I shall term a grammatical ‘bolt on’. The issues, and the alternative approach, are discussed in the following sections.
5.1 Overlap
The first issue which I suggest is problematic for the two-grammar approach to variation in French is the extent of the overlap between the two proposed grammars. The overlap between FD and FCT is massive, covering all levels of grammatical description (as well as the lexicon). In fact it is so extensive that not only are FD and FCT both commonly simply known as ‘French’ among non-linguists, but the existence of the difference in terms of distinct and discrete varieties is in fact not even generally acknowledged, at least not in the way that standard and dialectal Arabic are recognised as distinct varieties. And indeed the absence of this acknowledgement plays to the variationist approach to the empirical phenomena we are concerned with, in terms of grammar-internal flexibility.
The massive FCT–FD overlap in the specific context of French raises the more general question of whether there is a limit to the extent to which two or more varieties can overlap in terms of their grammatical (not to mention lexical) properties, yet still reliably be acquired by children, and subsequently maintained independently in the speaker's mind/brain. From the perspective of acquisition, the child, having first acquired FD, would need to notice, in the linguistic data to which s/he is exposed (e.g. through formal schooling), a surface distinctiveness above a threshold quality and quantity in order to conclude that the data was the output of a system distinct from FD.Footnote 16 Where a child is growing up in a bilingual environment as more typically conceived of, exposed to and acquiring two quite different languages (as opposed to different varieties) in a naturalistic setting either from birth or at least from an early age, this is not an issue since the PLD generated by the two systems are presumably sufficiently qualitatively and quantitatively distinct. But it is not so clear in the case of two varieties of the same language which show significant overlap. The variationist approach to the empirical data we have been considering, based on a single grammar, does not have this problem: the different ‘varieties’ of French are as strikingly similar to one another as they are because they are generated by the same (albeit flexible) grammar. In the diglossic account of variation in French, where two distinct grammars are posited, the question is, Why are FD and FCT so similar?
5.2 Non-overlap
The second issue for the two-grammar approach relates not to their overlap but to how they differ, that is, their non-overlap. Massot (Reference Massot2008) makes clear that the FCT–FD distinction is not the same as the distinction between spoken and written language: uniquely FD and uniquely FCT features appear in written and spoken language alike; there is nothing inherent in the (distinctive) properties of FCT or FD which predisposes them to one particular channel; neither is there anything particular about speech or writing which explains why certain grammatical features are characteristic of FD or FCT. There would appear, therefore, to be no reason to assume anything other than random differences between FD and FCT, both in terms of the precise points of grammar where the differences are located, and in terms of the precise nature of the difference at each point. As one reviewer put it, in the two-grammar model FD could be as different from FCT as French and Chinese.
Yet FD and FCT are not as different as French and Chinese, and it is conceivable that the grammatical differences between FCT and FD are not random, either in terms of position or nature; it is possible that they are instead ‘micro-’ reflexes of an overarching ‘macro-’ contrast which can therefore be characterised at a higher level of abstraction. A suggestion that this might be the case comes, for example, from comments like those made by committed variationist Françoise Gadet: while not claiming that [FD] is a simplified version of [FCT], Gadet (Reference Gadet1997) characterises [FD] in terms of a séquence progressive, fixed word order, analyticity, invariability. An indication of the nature of the non-random relationship between (?some of the properties of) FD and FCT is hinted at in the use of the labels ‘conservative’ for FCT and ‘innovative’ for FD. By definition relative, the labels suggest that properties of FD are a diachronic development of those of FCT. This is the case of the two analyses of est-ce que/qui: the status of the form in FD (with Q in C*° underlyingly) indicates an innovative reanalysis, the result of the grammaticalisation of the FCT form (with Q in I*° underlyingly but linked to C*° via a representational chain). This also applies to the distinction between bipartite and postverbal monopartite negation within the context of Jespersen's negative cycleFootnote 17: a form like Il vient pas ‘He's not coming’ has a different structure depending on whether the grammar generating it is at one stage within the cycle (NegP projected with a null head: Il ne/ø vient pas) or at the next one (NegP not projected: Il vient pas; see example (13)). And it applies to the reanalysis of subject proforms as agreement markers within the context of van Gelderen's (Reference van Gelderen2011: 38ff.) agreement cycle: subject pronominals are reanalysed as verbal agreement markers which can then co-occur with, for example, DP or emphatic pronoun subjects without pragmatic topicalisation or syntactic dislocation; co-ordinated VPs have obligatorily repeated ‘subjects’, as in (27a), and il can be anaphoric on indefinite antecedents, as in (27b):
(27)
While it would doubtless be inappropriate to suggest that FD is a straightforward descendent of FCT, if they are deemed to be the output of two independent grammars, then the question to be answered is, Why does so much of the FD–FCT contrast appear to be amenable to a single overarching generalisation of FD being innovative with respect to FCT?
5.3 Cognitive status of FCT
The third issue to be addressed by the two-grammar model has to do with the cognitive status of the two posited grammars, in particular, the status of FCT. FD is characterised in (10a), and its status as a cognitively real grammar, the product of the language acquisition device, encoded in the mind/brain of speakers therefore seems unproblematic. FCT, in contrast, is characterised in (10b). It is not acquired early in a naturalistic environment, and is dependent on literacy and formal schooling; speakers commonly only have a passive competence of FCT, characterised by uncertainty (e.g., le fait que ‘the fact that . . .’ followed by ind rather than standard sub) and hypercorrection (e.g., après que ‘after . . .’ followed by sub rather than standard ind, by analogy with avant que ‘before . . .’ which has sub in the standardFootnote 18). The status of FCT in the mind/brain of speakers is therefore less clear, and doubts have been cast as to its cognitive reality and coherence for some time (Bauche Reference Bauche1928; Côté Reference Côté1999). FCT has also been the subject of artificial interference from the Académie française, for example, with respect to present- and past-participle agreement as well as the mood found in finite subordinate clauses selected by concessive conjunctions or bridge verbs (including in respect of the judgements given above; Rowlett, Reference Rowlett2007: 7). From such perspectives, as well as from the perspective of Ferguson's original sociolinguistic notion of diglossia amounting to relatively stable co-existence of two varieties, the question is, Can FCT really be thought of as (the output of) a separate, independent grammar?
5.4 An alternative approach
In this section I endeavour to capture the essential insight of the diglossic approach to variation in French – namely the judgements in (9) – while addressing the issues raised above. The idea is as follows. As per (10a) the child is exposed to PLD in the naturalistic context of the home and family, and acquires the grammar of a vernacular variety labelled here FD. Later, as per (10b), the child notices that they are being exposed to a subtly different kind of linguistic input, labelled here FCT, which is incompatible with the (FD) grammar previously acquired (containing non-dislocated topical nominal subjects, negative ne,Footnote 19 for example).
Over time the child develops a sensitivity to the relative socio-stylistic status of FD and FCT, and employs them differentially on the basis of assessments of situational appropriateness. The child (or, more specifically, their LAD) notices, however, that FCT is not randomly different from FD. In fact, in line with the notion that FD represents a diachronic innovation with respect to FCT, guided (in terms of direction of change if not in terms of rate of change) by UG-internal principles (van Gelderen, Reference van Gelderen2011: 377), the child ('s LAD) notices that the properties of relevant functional items in FCT correspond to a predecessor variety of FD.Footnote 20 Since the child can conceptualise the relationship between FD and FCT in this way, they do not need to posit/acquire a full, stand-alone grammar for FCT alongside the one for FD; rather, they can (and, for reasons of economy, therefore should) define/acquire FCT in the relative terms of how it compares to FD. I label the specification of FCT a grammatical ‘bolt on’ in order to capture the two ideas (a) that it is not defined autonomously in absolute terms, but only in terms relative to FD, and (b) that it is more concise/economical than a full grammar. The ‘bolt on’ contains a grammatical specification of where and how FCT differs from FD, but has nothing to say about where they are the same, and therefore is not a fully specified grammar. For the child who has already acquired FD, acquisition of FCT is therefore acquisition of this ‘bolt on’ rather than of a second grammar (or lexis), and the child's LAD does not the need to go through the steps in (28):
- (28)
a. creating a second grammar initially by copying the first;
b. modifying the copy as required, in order to take account of the problematic input; and
c. maintaining the two grammars independently of each other and indefinitely, not only in terms of where they differ, but also in terms of where they overlap.
And code-switching between FD and FCT is not switching between two independent grammars, but rather engaging (or disengaging or failing to engage) the ‘bolt on’, which speakers do on the basis of an assessment of the situation.
The answers I propose to the questions posed at the end of each of the sections above are therefore as follows:
- (29)
a. FCT is as similar as it is to FD because, rather than being an independent grammar, FCT has FD at its core, and only differs from FD to the limited extent that the content of the grammatical ‘bolt on’ specifies that it differs.
b. The differences between FD and FCT are amenable to analysis in terms of overarching principles because they are the expression at the micro level of a single set of differing properties (or maybe even just one differing property) specified at the macro level within the grammatical ‘bolt on’ rather than within an independent grammar.
c. FCT is not a separate, independent grammar; rather, it is essentially FD which has been modified on the basis of the content of the grammatical ‘bolt on’.
The ‘bolt on’ approach to French diglossia crucially differs from the ‘inherently variable’ approach since the former, unlike the latter, accounts for the non-randomness of the location/nature of the variation, as well as the judgements in (9). The approach raises a number of issues:
1. For example, by closely examining the nature of what is referred to above as the non-overlap, which for FCT is encoded within a grammatical ‘bolt on’, and by endeavouring to capture with minimal mechanical apparatus, we can explore the architecture of grammar across various levels of description.
2. Beyond accounting for French diglossia, the notion of a grammatical ‘bolt on’ can offer an insight into situations of grammars in competition and/or periods of grammatical change in progress. The ongoing co-existence of FD (as a regular grammar) and FCT (as a ‘bolt on’) can be regarded as an (extended) period of change in progress. This should come as no surprise since the grammaticalisation theory and the model of grammars in competition rely on the possibility of a single form being compatible with more than one analysis, and of the diachronic drift from one analysis to another being driven by UG-internal principles. In the specific context we have been considering, the lifespan of the change in progress has been (artificially?) extended as a consequence of the social–political–cultural status of FCT within the specific context of the recent history of France (see Milroy and Milroy's Reference Milroy and Milroy1985 notion of the ideology of the standard).
3. By conceptualising the various dimensions of the FCT–FD contrast in terms of an immediately adjacent position around a cyclic pattern of development, or along a cline, a framework is provided for linking the dimensions with each other: we might think of the separate cycles as being exponents of a single overarching cycle. Rowlett (Reference Rowlett1996, Reference Rowlett1998: chapter 3) tries to link the negative cycle and agreement cycle along these lines. The theoretical challenge is determining just how minimal the specification contained within the ‘bolt on’ can be.
4. Since the decision whether or not to activate the FCT ‘bolt on’ is culture-specific and can also be the expression of individual personality, it allows for differential patterns of linguistic behaviour across French-speaking communities, as illustrated in individual/group-based differential rates of occurrence of marked H and marked L variants.
5. Also, within the overall Francophone community it is possible that different smaller-scale speech communities have slightly different versions for FCT and FD. For example, De Cat (Reference De Cat2007) notes that ne retention is higher amongst French speakers in Belgium than in France, while Goosse (Reference Goosse, Antoine and Cerquiglini2000: 118) notes that negative imperatives like the one in (30) are not found in Belgian French:
(30) Parle-moi pas!
These two observations are plausibly linked. FD and FCT for speakers of Belgian French could both be more conservative than they are for speakers in France, as illustrated by the table in (31):
(31)
All other things being equal, this would lead us to expect Belgian French to have more ne retention (Belgian speakers using FCT would have overt ne while their French counterparts might have a null negative head) but no negative imperatives like (30) (the Belgian FD grammar projects NegP, which is necessarily absent in structures like (30); see Rowlett forthcoming). But in both the French and the Belgian scenarios, FCT is ‘one stage more conservative’ than FD, and so the same overall approach to the relationship between FD and FCT can be adopted.
6 SUMMARY AND CONCLUSION
The starting point for the present article has been (a) the attested dimensions of (primarily syntactic) variation in French, (b) the (uncontroversial) notion that variation along some dimensions is socio-stylistically marked H or L, and (c) the (more controversial) contention that much of this variation is not random. Under (c) in particular, it has been suggested (Massot Reference Massot2008) that marked H features do not co-occur with marked L features. This state of affairs is accounted for by Zribi-Hertz (e.g. Reference Zribi-Hertz2011; this volume) and her collaborators in terms of diglossia, that is, the idea that speakers have internalised two grammars, one of which (FCT) is associated with the marked H features, while another (FD) is associated with marked L features, and that, at least within the context of a single linguistic unit, which Massot claims is the clause, speakers use one grammar or the other, but not both, and therefore are prevented from combining marked H and marked L features. The existence of grammatical features which are H/L-neutral, which can combine with marked H or marked L features, is accounted for by assuming overlap between FD and FCT.
On this basis, I endeavoured to demonstrate that the notion of overlap between FD and FCT and the idea that a variety of surface forms can be generated by both FD and FCT fail to acknowledge the fact that some superficially identical surface forms have a different set of grammatical properties depending on whether they are generated by FD or FCT. The example of this that we saw in detail was the interrogative est-ce que/qui, which is an atomic complementiser in FD, but the output of a syntactic ‘inversion’ process applied to a biclausal cleft structure in FCT. Other examples included negation and in situ wh questions.
I then questioned the notion that FD and FCT are the products of two separate grammars in the mind/brain of speakers. This line of enquiry was based on (a) the extent of the overlap between FD and FCT (massive), (b) the way FD and FCT are acquired and their cognitive status for speakers (see (10a, b)), as well as (c) the suggestion that FD and FCT do not differ randomly. In respect of (c) it was suggested that FD was grammatically innovative. This means that where FD and FCT differ, FCT is conservative with respect to FD, and that where a single superficial form has a different structure in FD and FCT, the form in FCT is again conservative with respect to FD. Further, innovation–conservatism along each dimension of variation was conceptualised as adjacent positions around a cyclic pattern of diachronic development: the relevant feature specifications of FD are the natural successors of those of FCT as determined by internal economy principles and learner bias (van Gelderen, Reference van Gelderen2011).
Instead, therefore, of positing two grammars in the speaker's mind/brain, one for FD and another for FCT, I suggest that a more economical approach to the FD–FCT distinction is to assume that, subsequent to the acquisition of an FD, the child acquires an FCT ‘bolt on’ which, rather than capturing a full FCT grammar, is dependent on the FD grammar and merely captures how FCT differs from FD. The child posits a ‘bolt on’ of this kind because the linguistic input which is incompatible with their FD grammar is nevertheless recognised as being structurally related to the grammar of FD.
Having acquired the FCT ‘bolt on’ and evaluated its socio-stylistic status, the speaker is then in a position to decide when and where to activate the FCT bolt on, on the basis of his/her assessment of the socio-stylistic context. If the FCT ‘bolt on’ is not activated, the ‘pure’ FD grammar is used, and unmarked as well as marked L variants are produced; if the FCT ‘bolt on’ is activated, the marked FD (= L) settings/values are overridden and the marked FCT (= H) variants are found. Those (unmarked) features of genuinely ‘superordinate’ French, common to FD and FCT in superficial form and underlying structure, are unaffected by whether or not the FCT ‘bolt on’ is activated.
Looking ahead, the next stage of reflection within this approach to diglossia would be to enter into the detail of, for example, van Gelderen (Reference van Gelderen2011) with her analysis of grammatical cycles in terms of macroparameters of agreement and structural case and, where appropriate, minor cycles related to interpretable or uninterpretable features in specifier and head position within NegP, DP and other relevant functional projections. The challenge for the ‘bolt on’ alternative to a second grammar is to push the economy approach to the FD–FCT relationship as far as possible with the difference encoded at as high a level of abstraction as is able to capture the empirical facts. I hope to be able to address this in future work.
