This study examines the early acquisition of verb-argument structure within the theoretical frameworks of both nativist and empiricist approaches to language acquisition. The aim is to evaluate the relative strengths and weaknesses of each approach and establish to what extent each approach provides an accurate description of the developmental data. Data collect ion took the form of naturalistic audio-tape recordings of individual mother-child dyads engaged in normal everyday interaction. Twelve predominantly middle-class families took parr in the study. The children are all first-borns and monolingual. At the beginning of the study. the children were aged between 21-24 months with MLUs of between 1.00-2.50. Each child was taped for two separate hours in each three week period for the duration of one year. In total, 395 hours of data were collected. The data was transcribed in CHAT format using the CHILDES system of transcription (MacWhinney. 1995). Three separate analyses were carried out to assess the role of performance limitations in the acquisition of verb-argument structure (Valian, 1991), the role of semantically simple or 'light' verbs in early speech (Pinker. 1989). and the development of argument structure from a constructivist perspective (Tomasello, 1992). In all cases, the method of analysis used was fine-grained and operated at the lexical level rather than at the level of abstract grammatical classes. The findings of these studies suggest that analysis at the general level, as adopted by nativist ihcorixts, typically results in children being credited with an abstract knowledge of grammatical categories and rules which is not supported by fine-grained analysis of [he data. When the data is examined at the lexical level. there is evidence to suggest that children acquire verb-argument structure in a lexically-specific manner based around individual verbs and other lexical items. Thus. in no sense do children show evidence of operating with innate grammatical knowledge. Furthermore, the particular verbs and structures the children acquire early in development are closely related to the verbs and structures used by their mothers. with frequency of use playing a particularly important role. These findings suggest that children are unlikely to learn language guided by innate grammatical knowledge of the type assumed by nativist theorists. Instead. it is proposed that the process of language acquisition may depend on a distributional processor which is sensitive to the distributional regularities of the input. This process would predict that children will first learn lexically-specific patterns of high frequency in the input. and only over time will children come to acquire the more complex grammatical classes assumed to exist in adult language.