This paper claims that the English to-infinitives base-generated in the complement position of VP represent a type of syntactic (grammatical) aspect. The argument is partly motivated by a historical parallel: the fact that both to-infinitives and progressive participles have developed from prepositional phrases with spatial meanings. The semantics of the to-infinitive are analyzed along with those of the perfective and progressive constructions. It is argued that a verb that takes a to-infinitive as its complement describes an event or state with implicit relevance to the infinitive event or state, while the perfective participle has similar relevance to the matrix tense. The referential nature of verb-following to-infinitives and that of perfective participles are thus temporally symmetrical, suggesting that such to-infinitives form part of the syntactic aspect system of English. The ‘future- oriented’ use of toinfinitives is shown to be the syntactic representation of ‘prospective aspect’, discussed in Comrie (1976) and others. Following and modifying Felser (1999), Lasnik (2000), and others, it is assumed here that a typical English sentence has the structure [IP [NegP[ VP [AspP [VP]]]]], with infinitival to, progressive -ing, and perfective -en each checking their syntactic aspect feature in the head of AspP. The negation marker not, which normally precedes to, is assumed to occur either in the head of NegP or as an adjunct to AspP.
Koizumi and Tamaoka (2004a) (KT hereafter) reported an experiment addressing the controversy surrounding the word order of double objects constructions in Japanese (Hoji, 1985; Matsuoka, 2003; Miyagawa, 1997). Although their conclusions may turn out to be correct, we argue that their conceptual justifications are problematic and more experimental work is needed, especially in relation to Matsuoka's proposal. We are particularly concerned with KT's claim that scrambled word orders take longer to process because their syntactic representations are more complex compared to canonical orders. Such metrics based on competence models fail to address critical issues in performance raised in the 1970's and later (Fodor et al., 1974). One problem is that in sentence processing we must consider incomplete sentences, which are often ambiguous, whereas syntactic theories have the luxury of dealing only with complete structures. Therefore, the mapping from competence theories to performance is not as straightforward as KT imply. In this commentary, we discuss KT's claims within an incremental model of sentence processing and consider the kinds of hurdles one must overcome before empirical results can distinguish between competing syntactic proposals. Despite our misgivings about some aspects of Kt's work, we believe that their effort to bring syntax and language processing together is the kind of interdisciplinary work needed to attain a detailed understanding of human cognition, and we hope that our comments will clarify some issues regarding their contribution.