Learnability issues in the acquisition of the dative alternation in English

This dissertation provides a model of the learnability of datives, as in Mary gave a book to John / Mary gave John a book. Two general approaches are examined: The first approach holds that speakers have a dative rule which relate the two sentences above. The other approach holds that speakers learn the dative item- by-item. Rule-based studies of dativization have proposed several characterizations of the dative rule. These characterizations, however, have to treat many cases as exceptions. It is argued that, from a learnability perspective, lexical exceptions pose potential problems for the rule-based approach. One problem is that it is not clear how children would know these exceptions; another, is how they would achieve the adult grammar. The lexical-listing approach, on the other hand, is inadequate on empirical grounds, since children overgeneralize, suggesting some kind of a production process. I propose to modify the lexical-listing approach (Baker, 1979) to handle the problem of overgeralization. Essentially, the present proposal explains children's overgeneralization as the result of “performance economy,” rather than the product of an internalized rule of grammar. According to this view, both child and adult speakers are led by a general tendency to choose a reduced form over an expanded form (cf. Chomsky's, 1981 Avoid Pronoun Principle). (In the case of dative, the double-object form is taken to be a reduced form of the to dative form.) To explains how children “unlearn” overgeneralization, the present proposal invokes a form of indirect negative evidence (Chomsky, 1981). This proposal is supported by three experiments. The experiments elicit grammaticality judgments from adult speakers and from children between 4 and 6 years of age. The findings indicate that children accept verbs in the double-object form which are not accepted by adults. The findings also reveal that adults vary on their judgments on what is characterized in the literature as report-type verbs. These findings are explained under the learning model proposed.