ABSTRACT
In this chapter, I review studies conducted in SLA that are relevant to the connectionists–symbolist debate. First, studies relevant to the single- vs. dual-mechanism model (i.e. the regular–irregular dissociation issue) are reviewed, including the issues on (1) the hypothesis that frequency effects are observed for irregular morphology but not for regular morphology, and how such dissociations are represented in our mind/brain, (2) whether neighborhood effects are observed only for irregulars but not regulars, (3) whether level-ordering (predicting differential effects on regular vs. irregular in morphological operations, which disallows regular morphology in compounds (*mouse eater)) constrains L2 acquisition, and (4) how the difference between denominal and deverbal verbs is treated by L2 learners. Second, research on the aspect hypothesis that is relevant to the connectionist–symbolist debate is reviewed, focusing on (1) regular–irregular dissociation in the acquisition of past and progressive marking and (2) connectionist modeling of L2 Italian acquisition. Third, a study comparing universal grammar and connectionist approaches in the acquisition of Japanese syntax was reviewed, and, finally, connectionist research on the acquisition of gender (and case) marking, which is considered to be soft rules, involving semi-regularities, which are difficult to characterize by a rule-based approach.
The studies reviewed in this chapter are generally more consistent with the connectionist approach than the symbolic approach, either contradicting some crucial assumptions of the symbolic model (e.g. frequency effect for regulars; proficiency effect for the supposedly innate level-ordering principle), or mixed results (sometimes regular–irregular dissociation is found, but not always). The latter finding is precisely the strength of the connectionist approach. Unlike the symbolic approach, which makes discrete predictions, the connectionist approach is very flexible because it assumes graded (rather than all-or-nothing) representations and interactions of multiple factors to contribute to acquisition outcome.
