Abstract
Moving to the retrospective perspective, we must first identify some salient properties of adult knowledge of language, and then determine how they manifest themselves in children. A central property highlighted in studies of formal linguistics (but observed long before) is the unlimited character of linguistic expressions : in our everyday linguistic behavior, we are constantly confronted with new utterances, which we can produce and understand at any time, and the set of possible utterances is unlimited. This capacity can be reduced to the mastery of a recursive computational mechanism that generates an unlimited number of linguistic structures, a mechanism that is formalized by the very simple assembly operation (" merge ") in the minimalist program (Chomsky, 1995). Such a mechanism generates hierarchical representations, expressed by syntactic trees : indeed, all syntactic, morphosyntactic and meaning-interfacing processes are sensitive to such hierarchical structuring, rather than to simple linear organization.
Are children, in the course of language acquisition, also sensitive to hierarchical structuring ? Experimental studies show that they are. The formation of yes-no questions in English is a classic example. The rule in question moves the auxiliary to the initial position(John is leaving → Is John leaving?). Yet this rule, which could be conceived as a simple linear operation, is in fact applied with respect to the hierarchical structure in adults. What do children do ? Experimental work by Crain and Nakayama (1987) has shown that children, as soon as they are tested on these structures, automatically adopt a hierarchical version of the rule, without taking into account the simpler, purely linear formulation. This conclusion is also valid for interface processes that calculate referential dependencies between nouns, pronouns and anaphors (reflexive, etc.). Here too, comprehension studies show that children, as soon as they come to be tested, adopt procedures sensitive to hierarchical representations, expressed in terms of the c-command relation (a formal relation of structural prominence : Reinhart, 1976). In all these cases, obviously, an internal force pushes the child (and the adult) to adopt hierarchical representations and operations for language, discarding a priori linear formulations, however simpler. In fact, the adoption of hierarchical representations stems from the assumption that the fundamental mechanism for building structures is the recursive operation of assembly, an operation that creates tree-like syntactic representations, necessarily organized hierarchically.
