From language-specific constraints to implicational universals

This article seeks to shed more light on the well-studied, yet still challenging, dative alternation. It starts from the cognitive-typological suggestion of Croft (2001, 2003) that language-internal variation is subject to the same constraints as cross-linguistic variation (the semantic map model), and that careful languagespecific research may therefore reveal facts about language in general. I argue that there is a parallel between dativisability and passivisability. Then, using a sample of active tokens from the British National Corpus of ditransitive give in both the indirect-object and double-object constructions and comparing these to a matched sample of passive examples, I evaluate the effect on passivisability — and hence dativisability — of the semantic parameters proposed in previous scholarship. The results are stated as a set of implicational universals. They should hold for all languages that feature the alternation, and make diachronic predictions as well. In addition to the semantics — which has been discussed in many previous studies — I argue that token frequency also plays a role in promoting dativisability — which has never been suggested before. The conclusion identifies some general implications for theoretical linguistics and for the practice of research on language structure and meaning.