Auxiliary selection and the semantics of unaccusativity

Our focus in this paper is on those Romance and Germanic languages which exhibit alternation of 'have' and 'be' ('auxiliary selection') according to verb class. In these closely related languages 'have' occurs with transitives and unergatives, while 'be' occurs to varying degrees with unaccusatives. Crucially, the distribution of perfective 'have' and 'be' is structured within and across languages. Following Sorace [J. French Lang. Stud. 3 (1993) 71; Language 76 (2000) 859], we take the systematic variation in auxiliary selection to suggest that 'unaccusativity' is determined by a semantic notion whose components are organised along a typicality scale ranging from core to periphery. By pinpointing the semantic properties which are part of unaccusativity (the components of the scale), we seek to establish how they are combined and how the subsets of combined properties are ordered. On the basis of synchronic and diachronic evidence from a variety of languages with auxiliary alternation according to verb class we analyze perfective auxiliaries as morphosyntactic markers of tense and aspect. We conclude that these elements are realised by a morphological rule sensitive to the semantics of predicates. It is implicit in our account that analyses postulating a deterministic correspondence between perfective 'have' and an external argument fail to capture the crosslinguistic variation in auxiliary selection. © 2003 Elsevier B.V. All rights reserved.