The subject-object asymmetry revisited: Experimental and computational approaches to the role of information structure in children’s argument omissions

In two studies, we investigated the relation between information structure and argument omission in German child language in order to quantify to what extent the subject-object hypothesis (i.e., subjects are omitted more often than objects) is influenced by discourse pragmatics. Twenty four children took part in an elicited production study in which they produced transitive subject-verb-object and object-verb-subject sentences. Both constructions are instances of a topic-comment information structure. The results showed that 3;6 year-old children omitted subjects and objects alike when the arguments assumed topics status and were placed in utterance-initial position. In a second study, we then assessed whether a model of language learning implemented with a recency bias (resulting in learning from the end of utterances) would produce similar omission rates of initial arguments. The model was found to be sensitive to the frequency with which both word orders occurred in the input: initial objects were omitted more often than initial subjects, the pattern found in German caregiver speech. The results suggest that argument omission is heavily influenced by information structure and that a subject-object asymmetry per se does not exist.