Kuniyoshi Ishikawa Laboratory
Contents
Ph.D. dissertation: Predicates, Events, and Discourse:
Representing the So-Called Head-Internal Relatives in Japanese
2003
The structure of the HIR Construction has been associated with that of a relative clause. The Japanese HIR is also treated as equivalent to a canonical relative clause at some level of representation in Ito (1986), Ishii (1988), and Watanabe (1991). With its unique properties being investigated, however, I will argue that the alleged HIR in Japanese is not a relative clause, as has been claimed. I will instead claim that it is a prenominal complement clause with a contentless nominal -no as a head. Before discussing the construction in Japanese in detail, I will quickly give an outline of what is generally termed as an HIR in some other language. First of all, in Japanese and other languages such as Quechua languages and Lakhota, both HIRs and relatives lack an overt Wh-relative marker, in contrast to relatives in a language such as English. Second and more importantly, HIRs do not indicate any explicit relativized noun outside the clause, which contrasts with canonical relatives in the same language. The following examples illustrate these points.
(1)a. Canonical Relative Clause in Ancash Quechua
[NP [S' nuna ti ranti-shqa-n] bestya i ] alli bestya-m ka-rqo-n
man buy -Prft-3 horse(NOM) good horse-Evd be-Pst-3
'The horse the man bought was a good horse.'
b. Head-Internal Relative(HIR) clause in Ancash Quechua
[NP nuna bestya-ta ranti-shqa-n] alli bestya-m ka-rqo-n
man horse(ACC) buy-Prft-3 good horse-Evd be-Pst-3
'The horse the man bought was a good horse.'
[ from Cole 1987: 279, 5 and 1 respectively]
The examples in (1) taken from Ancash Quechua (Cole1987) show that there is no obvious syntactic clue to identify a missing argument, i.e. a subject or an object, of the matrix verb in the HIR sentence (1b), since there is no relativized noun appearing to the right of the embedded NP nuna bestya-ta ranti-shqa-n. In contrast, in the canonical relative clause (1a), the apparent relativized noun bestya 'horse' occurs outside of the modifying relative clause.
Past analyses of the HIRs involve LF movement of an internal noun out of the subordinate clause domain. Thus, Williamson (1987), Cole (1987), and Basilico (1996) basically regard the structure of the HIRs as identical to that of canonical relatives in that they both involve movement of an internal noun. In (1b), Cole (1987) argues that bestya 'horse' in the HIR clause moves covertly at LF to the right, i.e. outside of the subordinate domain nuna bestya-ta ranti-shqa-n. Consequently, according to his account, the structures of the HIRs and relatives would be equivalent at LF. They would both have the structure in (1a) at LF.
However, in this dissertation I argue that Japanese HIR counterparts are not relative clauses but untensed prenominal complements. The association of the Japanese counterpart with the HIR will be claimed to be immaterial to the discussion of the construction in question. I instead claim that access to discourse is required for their interpretation due to the lack of tense indication with the alleged HIR in Japanese. I present a model of discourse representation by means of events, and show that Japanese HIRs can adequately be interpreted through the event representations.