There has been a long debating between ``Dual-Mechanism'' (DM) models and ``Single-Mechanism'' (SM) models in morphological processing. DM models claim that regular productive morphological processes (e.g. walk-walked) are dealt with by rules while irregular ones (e.g. go-went) are by associative memory (Pinker and Prince, 1992 among others). In contrast, SM models, mostly connectionist (e.g., Elman et al., 1996), argue that associative memory would deal with both types of processes. However, most studies were using inflectional morphology in English and similar languages, then it is natural to question whether or not DM models fit in with other types of morphology. One of them tested in this study is derivational morphology.
In past research, Sugioka et al. (2001) showed the double dissociation between -sase causative and lexical-causative in Japanese causative formation; while -sase causatives are dealt with by the computational rule (V-(s)ase) (which `is syntactic in its effects but morphological in its form' (Sugioka et al., 2001)), lexical causatives are stored in the associative memory. This is because only a suffix -sase attaches to the verb root in -sase causatives (Sugioka et al., 2001), whereas it is not predictable which lexical causative affix attaches to which verb (Jacobsen, 1992). However, due to the small number of participants (3 Broca's aphasic patients) and the lack of frequency control for experimental items, their argument is less convincing.
Thus in my experiment, I modified Sugioka et al.'s study - (1) to increase the number of participants, (2) to use high and low frequent verbs, and (3) to test normal Japanese native speakers. I conducted the filling-gap task - subjects fill in the appropriate causative forms as below;
The results clearly indicated a double dissociation - high frequency lexical causatives were performed significantly better than low frequency ones, while no such differences appeared in high and low frequency -sase causatives. In addition, the overregularization was observed in both low frequency lexical and -sase causatives. I argue that these findings are the strong evidence for DM models.
References
Elman, J., Bates, E., Johnson, M., Karmiloff-Smith, A., Parisi, D., & Plunkett, K. (1996). Rethinking innateness: A connectionist perspective on development. Cambridge, MA: MIT Press.
Jacobsen, W.M. (1992) The transitive structure of events in Japanese. Tokyo : Kurosio.
Pinker, S. and A. Prince (1992) `Regular and irregular morphology and the psychological status of rules of grammar', Proceedings of the 17th Annual Meeting of the Berkeley Linguistics Society. Berkeley, CA.
Sugioka, Y., Ito, T., & Hagiwara, H. (2001) Computation vs. Memory in Japanese Causative formation: Evidence from Agrammatic Aphasics. Cognitive studies, 8 (1), 37-62.