Null objects in second language acquisition: grammatical vs. performance models

Abstract

Null direct objects provide a favourable testing ground for grammatical and performance models of argument omission. This article examines both types of models in order to determine which gives a more plausible account of the second language data. The data were collected from second language (L2) learners of Spanish by means of four oral production tasks and a grammaticality judgement task. The results reveal that null objects in oral production are rare events limited to pragmatically appropriate contexts, that is, when the referent is easily recoverable from preceding discourse. The results of the grammaticality judgement task indicate that beginning level learners frequently accept sentences containing null objects with specific antecedents, while more proficient learners categorically reject such argument omissions. It is suggested that lower proficiency learners may rely primarily on semantic strategies in parsing and evaluating sentences, while advanced learners are more sensitive to syntactic violations. A performance account is ultimately adopted to explain the data given the lack of a clear null object stage in development, the presence of self-corrections, and the discourse-constrained nature of object omissions.