Quantifier float and information processing: A case study from Japanese

Previously proposed syntactic and semantic analyses of quantifier float (Q-Float) in Japanese have failed to explain why there is a mismatch in the syntactic range of Q-Float between speakers' judgment (nominative and accusative arguments) and discourse (absolutive arguments). This paper describes quantifier float in Japanese in terms of a mechanism of cognitive attention and claims that two conditions hold: (i) a host noun to be matched with the floated quantifier must be activated in attention, and (ii) the host should require a greater activation cost than the other potential quantifier host in order to achieve optimal processing. The former sets up a preliminary condition which predicts an acceptable range in the speaker's judgment, and the latter limits Q-Float to the range in which the processing of the sentence is optimized. The second condition above predicts not only the ergative distribution of Q-Float in discourse but also the long-observed subject-object asymmetry in sentences with scrambled Q-Float.