Between headed and headless relative clauses

Most syntactic typologies of relative clauses recognize two distinct categories: ‘headed’ and ‘headless’ (or ‘free’) relatives, according to whether or not the relative clause is associated with a nominal element that refers to a category delimited by the relative (the ‘domain nominal’, e.g. Andrews 2007). To this, Citko (2004) adds a third, intermediate category of ‘light-headed’ relatives. However, this paper considers evidence from Hup, a Nadahup (Makú) language of the northwest Amazon, to argue that such a strictly categorical approach – even one that makes room for three categories – is descriptively and typologically inadequate. In particular, for languages like Hup in which relative clauses are nominalizations (in an appositional relationship to the domain nominal), elements occurring as domain ‘nominals’ may be shaped by processes of grammaticalization that give them a partly lexical, partly grammatical identity, and they may occupy different points along a lexical–grammatical continuum. In Hup, while relative clauses may be associated with a full noun phrase (i.e. headed) and with no domain nominal whatsoever (i.e. headless), they may also appear with a range of intermediate elements, including ‘bound’ nouns, which differ from full noun phrases in their syntactic status; classifiers or ‘class terms’; and the enclitic =d’ǝh, a semi-nominal element that also marks plural number. The Hup data suggest that the property of ‘headedness’ in relative clauses may be best represented as a gradient phenomenon, and that this approach is arguably descriptively richer and typologically more accurate than the alternative. Keywords: Headless (free) relatives; nominal classifiers; nominalization; Hup