Résumé Cette étude explore les effets du degré d'adhésion d'un locuteur A à une proposition sur le choix de la forme syntaxique utilisée par un locuteur B pour modifier cette proposition. L'hypothèse est que plus A se montre certain de ce qu'il avance, plus B préférera la forme clivée à la forme simple pour modifier cette croyance. Cette hypothèse est vérifiée dans une première expérience d'appariement : la forme clivée est associée au contexte où la croyance contestée a été exprimée par A avec un haut degré de certitude et la forme simple au contexte où A s'est montré plus hésitant. Dans une seconde expérience (tâche de production), les sujets reproduisent cette tendance quand l'élément modifié est autre que l'agent. Le degré de conviction de A influence également le caractère complet ou elliptique de la phrase clivée et son accompagnement par une phrase négative. Les résultats sont discutés en termes de facilitation du traitement de l'information par l'interlocuteur. Mots clés : phrase clivée, fonction contrastive, production.

The present study focuses on the effects of contextual demands on the selection of a particular syntactic device, in asking whether French-speaking children from 4 to 10 years old will spontaneously produce it-cleft sentences if there is a functional necessity arising from the context. Taking into account recent studies that have specified the discourse function(s) served by this marked sentence form, it was hypothesized that the cleft formulation would be more likely than its uncleft counterpart whenever the child's intention was to contrast their own belief or knowledge with that of their addressee. The study showed this to be the case when the matter of the disagreement concerned the agent of an action: that situation elicited an overwhelming majority (from 80% to 97%) of cleft constructions, even from the youngest children. On the other hand, when the matter of the disagreement concerned the patient, there were only a few cleft constructions, even with the oldest children; contrastive stress on the object constituent was the predominant device employed for marking information in that situation. The high proportion of clefts in the Agent condition, and the high proportion of stressed object constituents in the Patient condition, both differ from previously reported data. These differences are discussed with reference to differences in task requirements and to differences between the prosodic constraints of French and English.

Contextual factors influencing the acceptability and understanding of cleft and corresponding uncleft sentences were investigated in four experiments. Experiments 1 to 3 used a sentence-choice paradigm within short dialogs between two interlocutors. Experiment 1 showed that the cleft construction was highly preferred over the corresponding uncleft in contexts where the speaker's utterance was to provide new information which was incompatible with the addressee's belief; on the contrary, in contexts where there was no incompatibility between the interlocutors' positions, the standard uncleft formulation was preferred. Experiment 2 showed that, in the belief-change contexts, unclefts preceded by an interjected No were preferred to unclefts not preceded by an interjected No. Experiment 3, however, contradicted the idea of a functional equivalence between clefts and unclefts preceded by an interjected No. Experiments 1 to 3 actually indicate that clefts do something that the unclefts do not, be they preceded or not by an interjected No, when contrastive information has to be conveyed. Using a simple comprehension time paradigm, Experiment 4 found that contextually appropriate formulations were understood faster than contextually inappropriate ones; cleft sentences, in particular, were understood faster than the corresponding unclefts when used for conveying contrastive information. These effects were analyzed in reference to H. H. Clark and S. E. Haviland's (in R. O. Freedle (Ed.), Discourse Production and Comprehension (Vol. 1), Norwood, NJ: Ablex, 1977) sketch of the Given-New Comprehension strategy applied to denials.