Overgeneralization of modal verbs in child language as a result of non-conventional form-meaning mapping

A typical characteristic of child language is overgeneralization; children use specific forms with more meanings and/or functions than the ones that are possible in adult language (see e.g., Clark, 1979). Modal verbs seem particularly interesting in this respect, as they have many different meanings in child speech, which are not all possible in adult language. For example, the Dutch modal verb moeten (‘must’) has a broad variety of modal meanings in adult language, such as necessity, obligation, etc. but in child language, it expresses also non-adult meanings, for example possibility, ongoingness, future etc. (see Jonkers, 2014). This overgeneralization of particular modal verb forms over different modal meanings in child language can be seen as a form-meaning mismatch. However, these form-meaning asymmetries can also be explained by the fact that not all modal nuances are already acquired by children. The too general interpretation of modal verbs in child language might hence indicate that only some semantic features of modal verbs are acquired, and others are not fully analyzed yet (see e.g., Clark 1991). In the case of the modal verb moeten (‘must’), children might interpret this verb as expressing a general notion of ‘non-factuality’, without further specifications (see Jonkers, 2014). Modal verbs in child language can hence still be interpreted as form-meaning pairs, but on a different level than in adult language. These non-conventional form-meaning pairs can be seen as a result of analogy-finding processes of children to deal with input variety (see e.g., Clark & Kelly, 2006; Goldberg 2006; Tomasello, 2003). The subtle semantic contrasts of modal meanings in adult language makes acquiring them challenging, which leads to the construction of a basic, general meaning to account for all occurrences, neglecting the polysemy in adult language. Consequently, overgeneralization of modal verbs in child language could be analyzed as form-meaning mapping in which not all semantic features are already elaborated