Abstract
In Ancient Japanese, the kakarimusubi construction of a constituent marked with the particle so agreeing with a predicate in the adnominal form (hereafter 'so adnominal form construction') was rarely used with sentence-final auxiliaries indicating conjecture (i.e., mu, ramu, kemu, mashi, beshi, and rashi). On the other hand, in Heian-period Japanese, the construction could take these auxiliaries except for rashi. The aim of this paper is to discuss why such a difference is observed between these two eras, and its relationship with their syntactic structures. This paper points out the following four observations: 1) The so adnominal form construction was used when the speaker regarded information for the grounds of the utterance as a fact; 2) The construction tended to end with stative predicates; 3) Until the late seventh century (the Hakuho period), adjectival predicates in main clauses were followed by conjecture auxiliaries under the same condition as the so adnominal form construction; 4) Until the late seventh century, adjectival sentences were chiefly used under the same condition as the so adnominal form construction: i.e., in utterances based on the recognition of certain information as fact. Based on these observations, the paper asserts the following two points: 1) The so adnominal form construction had common features with adjectival sentences with respect to their functional as well as their syntactic features. 2) There is a possibility that the change of the so adnominal form constructions was influenced by the change of the adjectival sentences.
