3.1 ICs with non-expressed IL as arbitrary lexical gaps (Auger, Reference Auger, Ashby, Mithun, Perissinotto and Raposo1993)

Auger (Reference Auger, Ashby, Mithun, Perissinotto and Raposo1993) applies the set of criteria established in Zwicky and Pullum (Reference Zwicky and Pullum1983) to differentiate between phonological clitics and inflectional affixes to clitic subject pronouns in CF. In the context of this application, she claims that the non-expression of IL in CF represents a ‘morphological argument’ for an approach in terms of inflectional prefixes. Specifically, Auger (Reference Auger, Ashby, Mithun, Perissinotto and Raposo1993: 179) considers the observation that constructions with non-expressed clitic subject pronouns are encountered only with ‘a few impersonal verbs in Colloquial French’ an instantiation of Zwicky and Pullum's (Reference Zwicky and Pullum1983) criterion B and argues that such constructions constitute arbitrary lexical gaps. These are claimed by Zwicky and Pullum (Reference Zwicky and Pullum1983: 504) to be ‘more characteristic of affixed words than of clitic groups’.

Yet, there are good reasons to believe that ICs with non-expressed IL in CF do not constitute arbitrary lexical gaps. In their initial abstract definition, Zwicky and Pullum (Reference Zwicky and Pullum1983: 505) consider arbitrary lexical gaps ‘cases where [. . .] some single expected host-clitic combination fails to occur’. Later on, in the context of their discussion of the morpho-syntactic status of the English negative formative n't, Zwicky and Pullum (Reference Zwicky and Pullum1983: 507) expound on what exactly they consider to be arbitrary lexical gaps: combined forms which ‘do not exist at all’ such as *mayn't and *amn't. According to Zwicky and Pullum (Reference Zwicky and Pullum1983), then, arbitrary lexical gaps are instances in which one element categorically fails to combine with another one, and this on no principled grounds. Against this background, the non-expression of IL in ICs in CF may hardly be considered an instance of Zwicky and Pullum's (Reference Zwicky and Pullum1983) criterion B. For as follows both from section 2 and from the relevant indications given by Auger (Reference Auger, Ashby, Mithun, Perissinotto and Raposo1993) herself, in CF, the non-expression of clitic subject pronouns in finite constructions without any other subject is neither categorical, nor is it arbitrary.

It is widely acknowledged that in CF, clitic subject pronouns are never categorically non-expressed in finite constructions, but only optionally so (Koch and Oesterreicher, Reference Koch and Oesterreicher2011: 117). We shall provide independent empirical evidence, since the optionality of the non-expression of clitic subject pronouns is generally deduced from introspection only. Given that in the context of the approach to clitic subject pronouns in CF in terms of inflectional prefixes, the notion of (more or less constant) code-switching between CF and SF is generally adduced (Culbertson, Reference Culbertson2010; Zribi-Hertz, Reference Zribi-Hertz2011), the methodology of introspection is quite problematic in this context. In order to be able to give such empirical evidence, an independent morpho-syntactic device is needed which allows for the unequivocal classification of a given IC as either CF or SF. In fact, such a device is readily at hand, namely the (non-)expression of the preverbal negative particle ne in the context of sentential negation: while in SF, this particle is categorically expressed, usually together with the postverbal adverbial negative marker pas, it is generally non-expressed in CF, pas then being the sole marker of sentential negation (Söll, Reference Söll and Hausmann1985: 119; Rowlett, Reference Rowlett1998).

With this background in mind, we have searched three modern synchronic data corpora of spoken FrenchFootnote ⁶ for instances of negated finite constructions with the impersonal verb falloir ‘to have to’ and pas as the sole marker of sentential negation. The search results, given in Table 1 and illustrated in (4), show that all of the three data corpora feature instances of both the expression and the non-expression of IL in the constructions under investigation.

Table 1.

Sequences of ‘(IL +) finite form of falloir ‘to have to’ + pas ‘not’

1. (4)

   

We take this finding as an unequivocal indication that in CF, clitic (expletive) subject pronouns are never categorically non-expressed in finite constructions.

Note, furthermore, that with the exception of constructions such as in (1a) which are generally taken to constitute a highly restricted conventionalised set, the non-expression of clitic subject pronouns in finite constructions without any other subject is solely encountered with (certain) impersonal verbs. We take this as a clear indication that in CF, the non-expression of these pronouns is not arbitrary, but follows from principled grounds. In light of these considerations, then, the non-expression of IL in ICs in CF may be shown not to represent arbitrary lexical gaps.

3.2 ICs with non-expressed IL as precursors to a new change (Fonseca-Greber, Reference Fonseca-Greber2004)

Fonseca-Greber (Reference Fonseca-Greber2004: 86) considers the non-expression of IL in ICs in CF both a clear indication of the (near) completion of the morphologisation of clitic subject pronouns and the ‘leading edge of a new change’ (cf. also Haas, Reference Haas1909: 53; Seefranz-Montag, Reference Seefranz-Montag1983; Reference Seefranz-Montag and Fisiak1984; Culbertson, Reference Culbertson2010). Specifically, she conceives of this non-expression of IL as ‘the internal restructuring of the new prefixal verb paradigm through the emergence of a zero morph, restricted for the present to [a] communicatively clear subset of the 3Sg.: impersonal verbs’ (Fonseca-Greber, Reference Fonseca-Greber2004: 86). Both conceptually and empirically, however, this claim does not stand up to closer examination.

From a conceptual point of view, it is entirely unclear why, as noted by Fonseca-Greber (Reference Fonseca-Greber2004: 81) herself, ‘[a]t a time when clitic use is on the rise in most linguistic contexts [. . .] certain impersonal verbs appear sans clitic, apparently in flagrant violation of the general trend toward morphologised verbal prefixes’. Rather than being a consequence of the alleged morphologisation of clitic subject pronouns in CF, the non-expression of IL in ICs represents a clear contradiction, and this all the more so, since, as pointed out by Fonseca-Greber (Reference Fonseca-Greber2004: 84), ‘morphologization is approaching completion in the 3^rd person’, while in the 1^st and 2^nd person, it is arguably complete. In fact, under the assumption that the non-expression of clitic subject pronouns is a consequence of the near completion of their morphologisation, one would naturally expect those pronouns to be the precursors to a subsequent change which have already been completely morphologised, namely those of the 1^st and / or 2^nd person; yet, this is not the case. Fonseca-Greber (Reference Fonseca-Greber2004: 84) tries to circumvent this conceptual objection by noting that ‘[m]orphologization is complete with 3Sg. verbs that disallow a referential subject, as in weather verbs [. . .] and impersonal verbs’. Even if one were to grant this circumvention, the proposed analysis is still conceptually problematic. For why should of all verbs (certain) impersonals be the precursors to a subsequent change? Since, Fonseca-Greber (Reference Fonseca-Greber2004: 83) argues, ‘impersonal verbs [. . .] (a) exist only in the 3^rd person and (b) cannot take a referential subject’.Footnote ⁷ Rather than corroborating her analysis, however, this functional explanation severely undermines it.

First of all, if expletive subject pronouns were functionally that useless, it is unclear why they were integrated into the language earlier on. Second, under the view that clitic subject pronouns in CF are agreement marking prefixes, one would naturally expect these elements to be categorically expressed, since they evidently fulfill a specific functional role. Third, the non-expression of IL is not restricted to ICs with genuine impersonal verbs such as falloir ‘to have to’, but is also encountered in ICs with verbs such as suffire ‘to be sufficient’ whose conjugation is not restricted to the 3^rd person singular and which also combine with subject elements other than IL. Fourth, among the constructions with genuine impersonal verbs, IL may be non-expressed with a small subset of these verbs only. Fifth, if IL was functionally useless in constructions with certain impersonal verbs, it is entirely unclear why in these very constructions, its non-expression is not categorical or, put differently, why IL is (still) expressed, indeed frequently, as follows e.g. from the data given by Fonseca-Greber (Reference Fonseca-Greber2004) herself.Footnote ⁸

In addition to these conceptual drawbacks, the claim that ICs with non-expressed IL constitute the precursor of a new change in CF also faces a severe empirical drawback, since such ICs are not recent phenomena. In fact, the non-expression of IL in ICs in CF is generally taken to represent the continuation of a grammatical trait of older stages of French in which all ICs may have ILnon-expressed (Brunot, Reference Brunot1891; Haas, Reference Haas1909; Brugmann, Reference Brugmann1917), as illustrated in (5a) for weather verbs, in (5b) for verbs whose conjugation is restricted to the 3rd person singular and which solely combine with IL, and in (5c) for verbs whose conjugation shows no such restrictions (Zimmermann Reference Zimmermann2012).

1. (5)

   

Note, however, that virtually no empirical evidence has been provided so far to corroborate this general claim, at least as far as the period from the mid-17th to the mid-19th century is concerned. In fact, Steinmeyer (Reference Steinmeyer1979: 219f.), referring back to relevant observations by grammarians, is the only researcher who provides a few scattered examples of ICs with non-expressed ILfor the second half of the 17th century as well as for the middle and end of the 18th century.

To compensate for this general lack of empirical evidence for the common claim of the continuation of the non-expression of IL in ICs in CF, we have, in light of the general absence of diachronic corpora of CF, sifted through numerous pieces of literary work from the mid-17th to the end of the 19th century in search of instances of constructions with the 3rd person singular present indicative impersonal verb form faut ‘must’ and non-expressed IL. The pieces of literary work we looked at principally constitute comedies which are generally considered most likely to contain instances of CF (Steinmeyer, Reference Steinmeyer1979: 219; Gadet, Reference Gadet1997: 11). For this literary genre stands out due to both its more or less consistent conversational / dialogical nature and its frequent recourse to characters representing members of the lower classes which are commonly associated by contemporary authors with a general lack of education and, more importantly in this regard, the occasional ignorance of SF. This ignorance is evidently a direct reflection of the lack of education and, to all appearances, a prominent means of illustrating this deficiency (Siede, Reference Siede1885: 1). Independent evidence for the soundness of such a conjecture comes from relevant rebukes by contemporary grammarians such as Chiflet (Reference Chiflet1659). Because of its relevance, his rebuke proves to be all the more pertinent in the present discussion, as it unequivocally indicates that the non-expression of ILis indeed a trait of the spoken language and continues right into spoken Modern French: ‘Never say Faut faire cela [‘One must do this’], Faut prendre garde [‘One must watch out’] without adjoining Il, as do the people from the Provence’.Footnote ⁹ (Chiflet, Reference Chiflet1659: 85).

Contrary to what one may expect, however, our examination of various pieces of literary work from the period under investigation exposes that the non-expression of ILwith faut ‘must’ is comparatively seldom encountered. In fact, relevant instances are few and far between, a state of affairs which most probably results from the general expectation that authors adhere to (the speech ideal of) SF (Diehl, Reference Diehl1895: 1; Rohlfs, Reference Rohlfs1928: 4). In any case, what is crucial in this regard is that instances of ICs with non-expressed ILare encountered from the mid-17^th to the end of the 19^th century, and this consistently so. Lack of space prevents us from listing all 113 pieces of literary work established in which at least one instance of constructions with faut ‘must’ and non-expressed ILis encountered as well as giving all of the 306 relevant instances. By way of illustration, we shall therefore only provide one example of the non-expression of IL in constructions with faut ‘must’ for every quarter of a century of the period under investigation.

1. (6)

   

In light of this unequivocal empirical evidence showing that ICs with non-expressed ILrepresent the continuation of a grammatical trait of older stages of French as well as in light of the conceptual drawbacks elaborated on above, the non-expression of IL in ICs in CF may be compellingly shown not to constitute the precursor of a new change.

3.3 Intermediate summary

The preceding discussion of the different conceptions of the non-expression of IL in ICs has shown that the linguistic reality of this intriguing characteristic of CF is even more intricate than previously assumed. Specifically, this non-expression of ILis neither categorical, nor arbitrary. What is more, it represents the continuation of a grammatical trait of older stages of French in which the non-expression of subject pronouns was generally possible. From these findings we shall therefore conclude that the non-expression of IL may not be readily considered (further) evidence for approaching the status of clitic subject pronouns in CF in terms of inflectional affixes.