MIT CogNet, The Brain Sciences ConnectionFrom the MIT Press, Link to Online Catalog
SPARC Communities
Subscriber : Stanford University Libraries » LOG IN

space

Powered By Google 
Advanced Search

 

Motivations for Scrambled Sentences In Japanese

 Hiroko Yamashita
  
 

Abstract:

Unlike the topic construction, just what triggers triggers speakers/writers to produce sentences with noncanonical word-order ('scrambled' sentences) in Japanese has not yet been thoroughly investigated. Based on an analysis of written texts, the current study argues that scrambled sentences in Japanese are motivated by the efficiency of a speaker/writer's working memory load and discourse factors.

A total of 2,635 sentences from eighteen different articles, ranging from extremely colloquial texts (transcriptions of forums) to formal texts (political essays), were analyzed. Less than 1% of the sentences in the texts contain a scrambled constituent, which is congruent with previous corpus analyses in Japanese (e.g., Hawkins, 1994; Hinds, 1983). The most remarkable trend in the type of scrambled constituents observed involves 'heaviness,' i.e., modification by a long modifier or a by a subordinate clause. Such a tendency is observed both in external scrambling and internal scrambling. I argue that the heavy constituent is scrambled by the speaker/writer to avoid the production of center-embedded sentences (cf. Wasow, 1997). Because Japanese is a strictly verb-final language in which all arguments precede the verb, center-embedding caused by a heavy object forces the matrix arguments appearing before the object to be stored in working memory while the heavy argument is completed. For example, both John-ga 'John' and Bill-ni 'to Bill' in (1) must be stored without theta assignments while the heavy accusative-marked object is completed.

(1) John-ga Bill-ni [totemo mezurasikute kookana Edo-zidai-no katana] -o ageta. -nom -dat very rare expensive Edo period-gen sword-acc gave 'John gave Bill a very rare and expensive sword from the Edo period.'

By scrambling the heavy constituent to the sentence-initial position, or as close as possible to it, the number of arguments in the matrix clause to be stored decreases.

Another characteristic of scrambled sentences is that the scrambled constituents tend to refer to the immediately preceding context. Such a tendency can result from production and discourse factors. A phrase referring to what has just been evoked is more 'accessible' in production (e.g., Bock and Warren, 1985) but it also augments the smooth flow of the discourse.

The fact that almost all (95%) scrambled constituents share the characteristics of heaviness and/or a reference to the immediately preceding context indicates that occurrences of scrambled sentences are strictly regulated by syntactic characteristics, efficiency in production, and the priority of information in the discourse.

Bock and Warren (1985). Conceptual accessibility and syntactic structure in sentence formulation. Cognition, 21, 47--67.

Hawkins, J. (1994). A Performance Theory of Order and Constituency. Cambridge University Press, Cambridge, England.

Hinds, J. (1983). Topic continuity in Japanese.Article title. In Givon, T. (ed.), Topic Continuity in Discourse: A Quantitative Cross-Language Study. John Benjamins, Amsterdam/Philadelphia, 47--93.

 
 


© 2010 The MIT Press
MIT Logo