In recent years a growing number of theoretical and empirical studies of first language acquisition have cast doubt on the hypothesis that acquiring language is a deterministic process in which the role of experience is restricted to triggering innate principles of grammatical content. The aim of this thesis is to explore areas of language where input -based learning demonstrably plays a role and to find learning mechanisms that account for the construction of observed overgeneral grammars and the process of their restriction. The thesis is a comparative study of the acquisition of argument structure in English and in Hungarian. The detailed analysis of spontaneous speech samples of two -- year -old children reveals that the omission of subjects, objects and prepositions at the so- called telegraphic stage of English child language cannot be explained either by limitations in processing capacity or by postulating an incomplete Universal Grammar. It is suggested that children's implicit arguments and oblique noun phrases lacking case or prepositional marking need not be analysed as syntactically ill- formed, since they conform to permissible abstract structural configurations. The errors may instead be attributed to overgeneral or indeterminate rules of pragmatics, which are fuzzy and variable in the mature grammar. It is shown that the nature of the children's intake of the primary linguistic data is a good predictor of the nature and extent of overgeneralisation or indeterminacy and of the speed with which the rules are fine -tuned to match the target.