6.1. Expletive NSs

The analysis of the sub-corpus showed that 1789 sentences out of 4996 (35.81%) feature an expletive subject, and as mentioned above, the absolute majority of the NSs are expletives. Indeed, null expletives constitute 20.51% of the total expletive subjects, while only 0.9% of the referential subjects are left unexpressed.

Data show that existential (il) y a constructions constitute the great majority, accounting for 79% of the occurrences. Impersonal expressions featuring (il) faut follow with 16.22%, while sentences with (il) fait represent a smaller proportion at 2.7%. Light verb constructions with (il) arrive and fixed expressions such as s’(il) vous plaît and (il) vaut mieux exhibit minimal occurrences in the dataset. These varying rates of occurrence underscore the differential distribution and usage patterns of expletive structures in French colloquial speech, reflecting the diverse linguistic contexts in which they appear.

To conclude, according to this semantic and IS-approach that we assume, specificational constructions are realized as copular sentences in which new information serves as the predicate of a SC, and the presupposed part of the sentence is topicalized and resumed by a subject pseudo-expletive pronoun in the subject position.

Given this resumptive function, the subject of a specificational sentence (i.e., ce in French) is not an expletive but a referential pronoun, and as such, it can be hardly silent in a pro-drop language like French. This proposal provides a feasible explanation for the fact that the pronoun ce is always present in the corpora examined. As a matter of fact, even when a co-indexed Topic is apparently not realized, it is in fact present, albeit silently, as it can be deduced from the context (it can thus refer to something being talked about, as in the case of sentence (37b) above).

On the contrary, the expletive il in predicational sentence is indeed an expletive pronoun, which is merged in the subject position because no argument can move there to meet EPP requirements. In particular, this happens when the theme argument is propositional (hence, too ‘heavy’ to move in subject position, as with a raising verb like seem in English), or in presentational sentences, in which new information is a theme selected by the verb. This is exactly the case of the expletive il in il y a constructions in French.

The basic semantic and informational distinction characterizing expletives in specificational and predicational sentences can be thus feasibly assumed to be the explanation for their different behaviour and allow for a principled distinction of expletive drop in a non-pro-drop language like French.

As a final support to this proposal, it can be noticed that expressions such as s’(il) vous plaît and (il) vaut mieux can be considered examples of phrasemes, namely fixed expressions whose meaning is not directly derived from the meanings of their individual components. In these expressions, the expletive il does not contribute an independent semantic value to the overall meaning of the expression but is instead part of a formulaic structure that expresses a particular meaning as a whole. It can thus be argued that this lack of individual semantic contribution makes the presence of the subject il optional or redundant, thus contributing to its omission. This semantic distinction across syntactic structures underscores the complex interplay between formal realizations, semantics, and discourse in the possibility of subject omission, even in a non-pro-drop language like French, shaping the distribution of the Null and overt Subjects observed in the analysed data.

6.2. Referential NSs

6.3. Null Subjects in Non-Pro-Drop Languages: A Comparison Between English and French

In particular, Cote used some 10 per cent of the Switchboard Corpus (telephone conversations performed by pairs of native speakers of English adults aged 20–40 on a variety of everyday topics), thus collecting a total of 190 NSs (out of 243 conversations), which have been examined taking into consideration several factors: (a) the form of the subject, (b) the person/number of the subject, (c) the source of the subject (i.e., whether it is referential, deictic, discourse deictic, or expletive), (d) the so-called “centering” transition of the utterance, (e) the turn position of the utterance (i.e., either initial or final), (f) the discourse segment position of the utterance, and (g) the clause type and the sentence type.

Turning to results, it is interesting to notice that the most frequent type of NSs in Cote’s corpus of English conversations is not that of expletives, but of referential pronouns (63%). Indeed, null expletives are only 37% of total NSs. However, even if null expletives are not as frequent as in French, they are still much more frequent than their explicit realization even in English, which, in fact, are only 8.9% of total expletive subjects.

Unfortunately, since the author was mainly interested in the discourse-related aspects of subject omission, no specific distinction is provided between expletives. Therefore, no one-to-one comparison can be carried out with respect to our French data. On the other hand, an interesting comparison can be provided between expletive and referential NSs. Indeed, the omission of 1sg and 3sg pronouns appears to be rather frequent in English phone calls, which is contrary to French face-to-face conversations. In particular, null 1sg deictics reach 26% of the total NSs and null 3sg referential pronouns reach 16.6%. In this latter case, it might be interesting to notice that null third person referential subjects were noticeably lacking in animate referents: only 3 out of 30 examples referred to animate entities.

In light of the present comparison, we can thus confirm that subject omission is possible in non-pro-drop languages and it is neither sporadic nor occasional. In particular, expletives appear to be the most frequently omitted type of pronouns. Nevertheless, a crucial distinction emerges between French and English, according to which the omission of referential subjects seems to be significantly more frequent in the latter.

Nevertheless, we surmise that this distinction can be attributed to the different context of conversations examined in the relevant corpora: face-to-face (French) vs. phone call conversations (English). Indeed, it is plausible to suppose that a phone conversation between two persons is dedicated and thus concentrates on some specific entity which is taken as the Topic of the relevant discourse and maintained as continuous, somehow “stimulating” its repetition across sentences. On the other hand, Topics can vary during a conversation among friends or be missing while building a piece of furniture.

Of course, these are just feasible assumptions which need to be resumed and confirmed in future comparative studies.

6.4. Coalescence: Corpus Support for a Restricted but Highly Frequent Phenomenon

A side note is reserved in this section to highlight the high frequency of morpho-phonological reduction phenomena occurring in conjunction with specific person–verb inflection associations. We refer to the incorporation of the 1sg subject je into the following verb, with the consequent creation of a single lexical form, with related morpho-phonological changes.

Indeed, corpus analysis shows that this phenomenon does not occur for all the occurrences of the 1sg pronoun and that when it occurs assimilation proceeds in both directions. Specifically, the postalveolar fricative [ʒ] of the pronoun je determines a change in the place of articulation of the following dental fricative [s], which, in turn, determines the regressive assimilation of the voiceless quality, thus obtaining a postalveolar fricative [∫]. On the other hand, the mode of articulation (fricative) remains unchanged.

Resuming the cases of French illustrated above, corpus analysis shows that these realizations occur almost exclusively with the auxiliary être ‘to be’ and with the verb savoir ‘know’ (occasional occurrences have been found with je serre [∫ɛr] ‘I squeeze’ (1 out of 2) and je dis [ʒi] ‘I say’ (1 out of 1)). Nevertheless, though restricted to these verbs, their frequency is remarkably high: 95% for je suis (74 occurrences out of 83), and 73% for je sais (45 out of 70). Additional evidence that this phenomenon cannot be (solely, at least) attributed to the fall of the final schwa is provided by the presence of a few occurrences of the 2sg pronoun tu ‘you’ and the verb savoir ‘know’ (5 out of 9), obtaining [tɛ] from tu sais.