Les études plus ou moins récentes consacrées expressément ou en partie au système des pronoms déterminatifs en français parlé européen ou canadien ont mis à jour l’étonnante diversité des formes qui peuvent coexister au sein de ce système. Si ces travaux contiennent parfois en plus d’un simple inventaire des formes relevées, des explications sur les origines possibles de ces formes, ils laissent cependant de côté plusieurs questions qui intéressent de près la sociolinguistique contemporaine. C’est ainsi par exemple que très peu de renseignements sont fournis relativement à l’effet des caractéristiques sociales des locuteurs (âge, sexe, appartenance socio-professionnelle, etc.) sur leur fréquence d’emploi des différentes formes determinatives. Or de tels renseignements pourraient permettre de dégager les tendances évolutives du système pronominal déterminatif dans les différents dialectes du français. Par le biais de cette étude nous voudrions combler en partie cette lacune en examinant non seulement la morphologie mais aussi l’évolution des pronoms déterminatifs dans le français parlé de Welland (Ontario). Le français parlé à Welland se prête à une telle étude dans la mesure où il possède lui aussi un système pronominal déterminatif dans lequel les formes standard côtoient de nombreuses formes non standard.

Over the past several years, a considerable amount of research on language comprehension has been carried out under the assumption that comprehension crucially involves language-independent cognitive strategies interacting with grammatical properties specific to a given language. Accordingly, the two factors of grammatical structure and cognitive strategies interact to render certain structures relatively more difficult to process than others. For example, it has been suggested that centre-embedded structures are more difficult to process than right-branching structures because centre-embedding interrupts the main clause and imposes a burden on short term memory, thereby making it relatively more difficult for the hearer to obtain closure on the entire sentence (Kuno 1974). Much attention within the cognitive strategies paradigm has been addressed to the comprehension of English complex sentences, and in particular to those sentences containing relative clauses. Since English permits relative clauses to be attached to NPs playing almost any grammatical role, and since, within a relative clause, the relativized NP can itself play almost any role, sentences containing relative clauses have provided a useful arena for testing various proposed cognitive strategies insofar as they relate to complex structures. In particular, much research has been addressed to SVO sentences in which a relative clause is formed on either the subject or object NP, and in which the relative pronoun is either subject or object.

Si la linguistique, en tant que discipline scientifique, évacue la vieille tentation de l’anomalisme, elle doit parfois, comme dans le cas de l’étude de la morphologie des temps, reconstruire une régularité parodoxale, une régularité remettant en cause 1’ ‘évidence.’ C’est bien le geste qu’effectue E. Benveniste dans un article justement célèbre, “Les relations de temps dans le verbe français” (1966). La régularité ‘évidente’ était la symétrie des formes simples et des formes composées, qui dans tout manuel scolaire se présentent comme suit:

In the course of his discussion of the relative merits of transformational generative phonology (TGP) and natural generative phonology (NGP), Piggott (1980) introduces as evidence in favour of TGP two phonological changes that took place between Proto-Algonquian and Ojibwa in the following chronological order.: (1) final lax vowel deletion, and (2) the deletion of resulting word-final glides *-w and *-y. By change (1), Proto-Algonquian *elenyiw-a ‘man’ mutatis mutandis became *i/iniw, and by change (2), *iliniw became *i/ini, modern Algonquin dialect of Ojibwa anini. Much of Piggott’s argumentation in favour of TGP rests on certain types of exceptions to these historical processes. In the first type, a Proto-Algonquian form *ne-wi-nteko-wil ‘I am a Windigo [cannibal monster],’ and other verb forms of similar morphemic constituency, underwent final lax vowel deletion and became *ni-wīndigōw, but failed to undergo final glide deletion. This was because the final *-i, unlike the final *-a in *elenyiw-a, an inflectional suffix which does not occur anywhere in the paradigm of the modern word anini, was a derivational morpheme and part of the verb stem, and was retained in non-final position. There it stands in morphophonemic alternation with its own absence, as shown by forms such as ni-wīndigōwi-min < *ne-wi·nieko·wi-Hmena ‘we are Windigos.’ The underlying form of the singular verb is hence ni-wīndigōwi, and the fact that the surface word-final glide -w was not elided is adduced as evidence that the phonological change is better described in reference to the abstract, underlying form rather than the surface form. Since TGP can distinguish between underlying and surface word-final glides and NGP presumably cannot, Piggott claims that TGP can better characterize this change than can NGP.

Number discord between subject and verb raises a theoretical problem for which linguists have, as yet, provided no satisfactory answer. Since facts of both morphology and syntax are involved, it is a problem which brings into focus the relation between what Guillaumians call tongue and discourse. First I shall outline the problem by giving a number of examples, and then try to suggest one avenue which might be explored in the hopes of finding a solution.

Historically, languages seem to have two kinds of affixes: those that stay put as unproductive (English -ity, Hindi -at, and German -sal) and those that go on to increase their lexical domain (English -hood and non-, Hindi be-, German -ung). When a native affix attaches to non-native lexical items it could not previously attach to, we say that the lexical items in question have moved into the native classification (cf. Allen 1978; Aronoff 1976). English -hood, originally restricted to native bases, now attaches to latinate bases such as priest and state, implying that state and priest have been totally nativized. When a non-native affix increases its lexical domain, we say that the original restriction on it has been removed. Hindi be, originally restricted to [+Persian] words can now attach to almost any adjective (cf. Singh, forthcoming). Non-native affixes either stay put as unproductive (cf. English -ity) or may generalize to cover the entire lexicon (cf. English non- and Hindi be-). When they do generalize so pervasively, they can be attached productively to newly borrowed words. Attachment to newly borrowed words, however, generally implies free attachability with native bases (cf. the Russian nominal suffix -izm discussed in Shanskii, 1968, particularly pp. 100 ff).