Cliticization to NP and Lexical Phonology

The term clitic is usually applied to morphemes whose distribution is syntactic in nature but which, rather than being treated as phonological words themselves, attach to a neighboring word. They are thus syntactic words that are, from a phonological point of view, parts of other words. The sense in which clitics become part of other words is generally left rather vague, and in many cases it appears that the only property that clitics have that distinguishes them from ordinary words is their lack of stress. More generally, the relevant notion of "phonological word" is usually one equivalent to "minimal prosodic phrase". The existence of clitics of this type is not particularly problematic. Within the theory of Lexical Phonology (Kiparsky 1981, Mohanan 1981), and less explicitly, within other lexicalist theories, there is another sense to the notion in which a morpheme may belong to another word, namely that it might give evidence of belonging not merely to a phonological word but to a lexical word, that is, that it might give evidence of being attached in the lexicon by word-formation rules. Were morphemes of this type to be found it would be necessary either to abandon the division between lexical and post-lexical processes or to modify the relationship between morphology and syntax in such a way as to provide a means for the lexically attached morpheme to have its effect in the syntax. In the present paper I propose to show that a clitic with exactly these properties occurs in Tongan. Like the English genitive /z/, this Tongan morpheme appears at the extreme right of a Noun Phrase. But perhaps unlike the English genitive, whose status is unclear, there is strong evidence that this Tongan morpheme is lexically attached. In the following sections I will demonstrate the phrasal distribution of this morpheme, argue that it is lexically attached, and finally consider how this apparent paradox may be resolved.