Abstract
To investigate the conditions under which two consecutive words (or 'bunsetsu') are dephrased, two factors, grammatical relationship and mora-size, were analyzed using the Corpus of Spontaneous Japanese (CSJ). The results showed the highest rate of dephrasing when an adjective and a noun produced a noun phrase, or a verb and a noun produced a noun phrase. It is considered that a phonological word which is syntactically preposed into the modifier position of a noun has a focus, in that the modifier expresses exhaustive identification among the members of a set of alternative entities, and this focus has an effect on dephrasing. The analysis based on mora-size, from six to thirteen morae, showed that the rate of dephrasing was in inverse proportion to the mora-size of two words. An increase in the mora-size lowered the rate of dephrasing. It is considered that there is a tendency to avoid constructing extremely large minor phrases. Multiple logistic regression analysis was performed to determine the variables that predict dephrasing. The model with the lowest AIC included grammatical relationship and mora-size as significant predictors, but not the interaction of them. The correct prediction rate of dephrasing by means of the logistic regression model was 65.58%.
