Commenting on Goldberg's (1995) ‘construction grammar’, Tomasello (1998) proposes a model of language acquisition in which children move from highly specific utterance–event pairings to abstract, verb-general structures. Despite their many strengths, models of this kind predict considerably more overgeneralization of the argument structures of verbs than seems to occur. In recognition of this, the paper explains (and supports with data from a previously unpublished study of 44 children aged 2;0 to 4;4) how processes which are side effects of the emergence of the verb form class could counter the overgeneralizing tendencies. It is argued that these processes are consistent not just with the model proposed by Tomasello but also (in large part) with the grammatical theory developed by Goldberg.