Separating functions and positions: Evidence from structural priming in Japanese

Hiroko Yamashita,¹ Franklin Chang² & Yuki Hirose^3
1 Rochester Institute of Technology, ² University of Illinois at Urbana-Champaign, ³ The University of Electro Communications, Tokyo

hyamash@mail.rit.edu

The theory of grammatical encoding divides sentence production into two levels of processing: functional and positional (Bock & Levelt, 1994).  At the functional level, syntactic functions are assigned to event roles; at the positional level, the linear order among phrases is determined.  Although explaining free word-order languages has been the primary motivation for the functional/positional distinction, there has been relatively little work done in those languages.  We report a study of structural priming (Bock & Loebell, 1990), speakers' tendency to reuse previously-produced sentence structures, in Japanese.  The results we obtained support this distinction.

Japanese marks syntactic functions with particles (wa-topic, ni-indirect object, o-direct object) and allows the order of phrases to vary.  Thus, unlike previous production studies which use fixed word-order languages, word-order and syntactic functions can be independently probed.  Furthermore, manner and temporal adjuncts are composed of NP and adjunct particle ni, which makes the surface phrase order identical to that of dative sentences.

Using the paradigm in Potter & Lombardi (1998), Japanese speakers produced a list of prime and target sentence pairs presented quickly phrase by phrase.  The argument order of targets was AGENT-wa PATIENT-o RECIPIENT-ni (wa-o-ni), followed by a verb.  We examined how often the speakers traded the PATIENT and RECIPIENT of the targets when preceded by three types of prime sentences: 1) the same order as the target (wa-o-ni), 2) RECIPIENT before PATIENT (wa-ni-o), and 3) a manner/temporal adjunct in place of RECIPIENT (wa-adjunct_ni-o).

If speakers switch the targets to the wa-ni-o order after producing the wa-ni-o primes, compared to the wa-o-ni primes, it cannot be attributed to syntactic function assignments.  The target and prime share identical syntactic functions and differ only in linear order.  The wa-adjunct_ni-o condition, on the contrary, has the same linear order as the wa-ni-o primes, but it has one less syntactic function.  If this condition primes the wa-ni-o order differently from the wa-ni-o primes, it cannot be due to linear order.  Rather, it possibly reflects differences in syntactic functions.

Table 1 demonstrates that, indeed, the wa-ni-o condition primes the wa-ni-o order more often than the wa-o-ni condition.  The wa-adjunct_ni-o condition primes significantly less than the wa-ni-o condition.  We therefore found independent priming at both positional and functional levels using a single structure.

Table 1:
Percentage of switch to wa-ni-o structures

References

Bock, K., & Levelt, W. (1994).  Language production: Grammatical encoding.  In Gernsbacher, Morton Ann (Ed). (1994), Handbook of Psycholinguistics, pp. 945-984.  San Diego, CA, US: Academic Press, Inc.

Bock, J. K.,& Loebell, H. (1990).  Framing sentences.  Cognition, 35: 1-39.

Potter, M. C. & Lombardi, L. (1998).  Syntactic priming in immediate recall of sentences.  Journal of Memory and Language, 38: 265-282.