Testing the long-before-short performance preference in Japanese with a Murakami novel

Abstract

The long-before-short preference of a head-final language is supported by online experiments (Yamashita & Chang 2001). The length-correlated scrambled order of subject, object, and indirect object supports Hawkins’ prediction, which suggests that for a head-final language like Japanese, a longer constituent will be produced earlier than a shorter one, to minimize the constituent recognition domain between the first argument and the verb (Hawkins 1994, 2004; Hawkins 2014). This study differs from Yamashita & Chang’s by investigating the relative length of the wo-suffixed NP and a non-argument constituent, the adverbial phrase. We address Kondo & Yamashita (2011) observation that the verb in Japanese serves more to summarise than to give a road map within the unfolding VP as in English, along their conclusion of a long-before-short performance preference in a spoken corpus of Japanese. The sentences below show the permutable word order of a wo-suffixed NP and an adverbial phrase. (7) a. その 部分-を ひとしきり 読む this part-wo for.a.while read “read this part for a while” b. ひとしきり 世間話-を している for.a.while gossip-wo doing ”gossiping for a while” The freedom of ordering alternatives under study is confirmed in 401 sentences extracted from Haruki Murakami’s novel Norwegian Wood, showing a frequency of fifty three percent for a wo-phrase-last arrangement, when no other constituents intertwine before the verb is presented. The adverb “for a while”allows either position as in (7). It is observable in our data that certain adverbial phrases are topical and resist scrambling, and they are excluded from our study. The ordering alternations demonstrate consistent volume in growing length differential on a logarithmic scale of base 2, and the two orders remain proportional (Fig 1). Only twenty two percent of the sentences present a short-before-long arrangement, and in these cases the length differential is limited to nine moras at most. A further seventeen percent have equal lengths. In the remaining sentences where the length differential is positive, i.e. the long constituent is presented before the short, the length differential soars as high as 34 moras. The results support Hawkins’ prediction that the long-before-short reference is proportional to the differential in length between the two constituents. Reference: Hawkins, John A. 1994. A performance theory of order and constituency. Cambridge: Cambridge Univ. Press. Hawkins, John A. 2004. Efficiency and complexity in grammars. Oxford: Oxford Univ. Press. Yamashita, Hiroko & Franklin Chang. 2001.“Long before short”preference in the production of a head-final language. Cognition 81(2). B45–B55. https://doi.org/10.1016/ S0010-0277(01)00121-4.