Аннотация:In recent theorizing about language acquisition, children have often been credited with innate knowledge of rules that link thematic roles such as agent and patient to syntactic functions such as subject and direct object.These rules form the basis for the hypothesis that phrase-structure rules are established through 'semantic bootstrapping', and they are also invoked to explain the acquisition of verb subcategorization frames (for example, Pinker 1984).This study examines two versions of the hypothesis that linking rules are innate, pitting them against the alternative hypothesis that linking patterns are learned (as proposed, for example, by Foley and Van Valin 1984).The first version specifies linking rules through paired thematicsyntactic role hierarchies, and the second characterizes them as a function of verb semantic structure.When predictions of the two approaches are drawn out and tested against longitudinal spontaneous speech data from two children learning English, no support is found for the hypothesis that knowledge of linking is innate; ironically, in fact, the children had more trouble with verbs that should be easy to link than with those that should be more difficult.In contrast, the hypothesis that linking rules are learned is supported: at a relatively advanced age, the children began to produce errors that are best interpreted as overregularizations of a statistically predominant linking pattern to which they had become sensitive through linguistic experience.
