On the interplay between syntax and statistical learning on errors in argument structure

Abstract

It has often been reported (cf. Bowerman, 1974) that L1 children or L2 learners tend to make errors in the use of verbs or predicates. They over-passivize, over-causativize, or over-intransitivize verbs or predicates when they have insufficient or incorrect knowledge of the language they try to acquire or learn. We examine these three types of errors, investigating what part of syntax is responsible for these errors, and how language development or learning proceeds on the basis of the language input available to L1 children or L2 learners to overcome these errors. We propose that all three cases of over-generalization are ascribed to the functional category, i.e. the little v, which is responsible for argument structure alternations. L1 children or L2 learners are presumably lacking or deficient in this functional category. However, they capitalize on statistical learning to learn the exact category of a verb or predicate on the basis of its distributional properties, identifying what verb or predicate can combine with a right kind of little v.