The 1990s witnessed a major expansion in research on children's morphosyntactic development, due largely to the availability of computer-searchable corpora of spontaneous speech in the CHILDES database. This led to a rapid emergence of parallel findings in different languages, with much attention devoted to the widely attested difficulties in inflectional morphology in the speech of two-year-olds. First written in 1995, and framed within the terms of contemporary syntactic theories, this article argues that cross-linguistic differences in the distribution of children's morphosyntactic errors provide important clues to the source of the errors, in particular whether they are morphological or syntactic in origin. The article takes as its starting point some striking previous findings that children's verb inflection errors are systematically correlated, on a sentence-by-sentence basis, with errors in the use of overt subjects, and with the use of syntactically complex constructions such as wh-questions. The article shows that these correlations are found in some languages but not in others, and argues that these differences are predictable, based on the verb movement and case licensing properties of individual languages. The article argues that children's errors reflect a combination of grammatical and speech production deficits.