Computational complexity and the acquisition of functional categories by French-speaking children with SLI

Abstract

Previous research conducted with French-speaking normally developing children and children with specific language impairment has shown a striking asymmetry between present and past tense production - the passe compose - in favor of the present. Based on specific assumptions about the nature of finiteness and the projection of the passe compose in French, we have explained this asymmetry in terms of the computational complexity hypothesis (Jakubowicz and Nash 2001). According to this hypothesis, kernel functional categories (i.e. INFL in French) are easier to compute than supplementary functional categories that are added to the functional skeleton of the clause (i.e. PAST in French). Alternatively, given that in the French past-tense construction the finiteness of the clause is expressed by an auxiliary in its present form, while the verb denoting the event surfaces as a participle, children's difficulty in using the past tense could be due to the absence of overt morphological marking that signals past meaning in the auxiliary. The research reported in this article was designed to determine the adequacy of this alternative theoretical interpretation, which we call the morphological salience hypothesis, for explaining the difficulties of French-speaking children with respect to the passe compose. To this effect, a study based on elicited production of the past tense and the pluperfect was conducted with two groups of normally developing children aged three and four respectively and a group of children with specific language impairment who ranged in age from 5;5 years to nine years of age. Results showed (i) that normal children are significantly less accurate on the pluperfect than on the past tense, and (ii) that the pluperfect is completely unavailable to the SLI group, while past tense is (almost) systematically or optionally avoided depending on the children. We argue that these findings provide further support for the computational complexity hypothesis