This paper proposes a semantic analysis of comparison constructions in Japanese which is crucially different from the standard semantics of comparatives as developed for English and related languages. The interpretation of the Japanese comparison construction is determined to a larger extent by pragmatic strategies, as opposed to compositional semantics. The syntactically provided item of comparison (the constituent accompanying yori) does not, in contrast to an English than-clause, have a degree semantics; it ultimately contributes an individual. From this item the real comparison has to be inferred. We argue that Japanese does not have English-style degree operators and probably lacks abstraction over degree variables in the syntax altogether. The proposed analysis accounts for a number of empirical differences between Japanese and English. A more general outcome is that the semantics of comparison is subject to crosslinguistic variation. A parameter of language variation is suggested as the source of the differences we observe.

No abstract

The acquisition of word order has been one of the central issues in the study of child language. One striking finding from the detailed investigation of various child languages is that from the earliest observable stages, children are highly sensitive to the basic word order of their target language. However, the evidence so far comes mainly from the acquisition of rigid word-order languages. In light of this background, this study presents new evidence that such early sensitivity to basic word order can be observed even in the acquisition of Japanese, a free word-order language.

Stowell (1981Stowell ( , 1982 proposes that there is a parameter that determines the availability of both preposition stranding (P-stranding) and the Vparticle-NP construction. While the V-particle-NP construction requires only the positive setting of the relevant parameter, P-stranding requires several additional properties to be identified. Stowell's analysis predicts the following: English-learning children should never acquire P-stranding significantly earlier than the V-particle-NP construction. The investigation of ten longitudinal corpora for English from CHILDES has borne out this prediction. Given the recent proposal by Snyder (1995Snyder ( , 2001) that the verb-particle construction follows from the Compounding Parameter, which determines the availability of various complex-predicate constructions, the findings suggest that P-stranding is also dependent on this parameter, and thus support its global nature.*