Contenu du sommaire : Constructions et interprétations de systèmes corrélatifs

In this paper, we examine the various uses of the terms correlation and correlative through the evolution of French Grammar. The adjective correlative is first used by Du Marsais in two different senses. When it is masculine, it denotes an 'identity relation' between two words linked by rules of grammatical agreement. In the feminine, the term means a bidirectional dependency between two clauses. These uses disappear during the XIXth century because ofthe hierarchical model of the complex sentence and semantic classification of the subordinate clauses. Under the influence of the Latin grammar, the study of the correlative systems develops in the XXth century as a particular form of subordination, as an anaphoric relation between morphologically similar terms. The correlation also contributes to transcend the opposition between parataxis and hypotaxis in the description of the complex sentence.

Basque correlatives and the matrix clauses that contain them are systematically compared withtheir counterpart in various languages. The dimensions of cross-linguistic variation describedare : (a) the paradigm the wh- items belong to (relative or interrogative ?), (b) their position inthe protasis, (c) the various mechanisms that possibly mark the protasis as dependent, (d) thepossibility of there being several wh- words or phrases ; but also (e) the option of having a linker between the correlative and the matrix, (f) the question of word order in the apodosis, and (g)what differences (if any) there are between such complex sentences according to the fact thatthe protasis contains a single wh- word or an NP modified by a wh- determiner.

This paper describes the correlative relatives of Kita Maninka, a West African language inwhich this type of relatives, which involves a relativizer originating from a former demonstrative, constitutes the only available relativization strategy. The author argues that nospecial notion of ‘correlative construction is necessary to account for these correlative relatives, which lend themselves to a straight forward description according to which there lativized clause behaves with respect to the matrix clause like an NP whose only particularity is that it cannot appear in positions in which it would receive a semantic role, andcan only appear in the extraposition construction.

Our study considers various correlative structures based on the relative and demonstrative stems *kwo- and *to-, which can express both coordinating and subordinating functions. In the domain of quantitative identity, correlative symmetric structures such as quam… tam, quantus… tantus correspond to the asymmetric (i.e. subordinating) correlative series quo magis… eo magis, which expresses a conditional implication. In the domain of temporal identity, the correlative series cum… tum is initially symmetric and coordinating, aimed to express a relation of coextensive simultaneity. The temporal correlation can also develop a value of conditional, subordinating implication in non-factual, iterative contexts of contingent simultaneity. Moreover, we analyze another structure (idem ac, atque ; idem qui, etc.), which presents a type of qualitative relation inside one clause to express identity or resemblance.

Our study concerns the structures which relate to the clauses introduced by quand (Quand-P,‘when-clauses'), which can be integrated in correlative structures with temporal value. We concentrate our work on the term correlate quand (‘when') and the temporal adverbscependant and alors (‘however/nevertheless' and ‘then'), which can take – in case they are not correlated to quand – concessive or opposite value as well. The aim is to analyze the value of these adverbs in a correlative structure quand p, cependant/alors q, and to compare them with the structure where one correlate term is missing (quand p, q/p, cependant q/p,alors q). We will try to underline the impact of the syntactic structure upon the selection of different values in our utterances.

The aim of this article is to examine the comparative correlative structure in Russian that is equivalent to the English construction The more you eat, the more you get fat. The correlationis established between two comparatives, which are introduced by two explicit lexical itemscem – tem that correspond respectively to the instrumental case of cto ‘what/that'and to‘that'. We demonstrate that existing analyses, which treat the element cem in terms of relative pronoun or conjunction, are not sufficient and that the relative nature of cem is questionable. We also argue that the relationship between the two correlated parts cannot be reduced to pure coordination or subordination. Some alternatives are proposed.

Subordinate comparative clauses mostly do not contain a verb and are then considered to be elliptic CPs. Nevertheless there are languages in which the comparative clause cannot always be analysed as an elliptic clause. This paper focuses on two cases in which an elliptic analysis is impossible. In the first case, occurring in Italian and Romanian, the comparative clause contains a pronoun in the accusative case instead of the expected nominative caseand the pronoun appears to receive its case from the complementizer, which thus acts as a preposition. In the second case, the comparative complement is marked by a preposition (cf.Italian or, with cardinals, in all Romance languages) or by a case (e.g. ablative in Latin,genitive in Russian and in Modern Greek). In the two cases the comparative complement cannot be analysed as a clausal constituent. The non clausal structure exhibits a kind ofshort cut : although a comparison always concerns two degrees of one or two different predicates, it is made directly between two alternative arguments of the predicate. This shortcut appears mostly when the compared arguments are the subject of the predicate.

Some kinds of hypothetical correlative structures are quite naturally associated withsituations of spoken dialogs as they share a number of specific features characteristic of thiskind of spoken discourse : 1) the fact that speaker and addressee interplay directly indiscursive exchanges, 2) the fact that the correlative relation between the hypothesisemerging from P and the consequence contained in Q is expressed through grammaticalconstructions and enunciative devices based on modal constituents, speech acts, imperativestructures, etc., 3) the fact that the correlative sequence must show a definite prosodicpattern (a suspensive pause on P preceding a falling intonation on Q). On the basis of thesedifferent factors, I will examine two main structure types, one with et ‘and' representing thehypothesis si p inferred from P, the other with ou ‘or' bringing about the hypothesis issuedfrom P but endowed with the opposite value.