Simulating Children's Null Subjects: An Early Language Generation Model

This paper reports work in progress on a sentence generation model which attempts to emulate certain language output patterns of children between the ages of one and one-half and three years. In particular, the model addresses the issue of why missing or phonetically "null" subjects appear as often as they do in the speech of young Englishspeaking children. It will also be used to examine why other patterns of output appear in the speech of children learning languages such as Italian and Chinese. Initial findings are that an output generator successfully approximates the null-subject output patterns found in English-speaking children by using a 'processing overload' metric alone; however, reference to several parameters related to discourse orientation and agreement morphology is necessary in order to account for the differing patterns of null arguments appearing crosslinguistically. Based on these findings, it is argued that the 'null-subject phenomenon" is due to the combined effects of limited processing capacity and early, accurate parameter setting.