Abstract
Since the advent of psycholinguistic research, there has been a major debate about a speaker's ability to produce novel morphologically inflected forms. Two opposing accounts of morphological productivity have been proposed: one attributing the productivity to the application of rules (Prasada&Pinker, 1993) and one attributing it to analogy based on stored exemplars (Bybee&Slobin, 1982).
With its relatively clear distinction between regular and irregular patterns of inflection, English past-tense morphology provides a particularly suitable framework to decide between these two approaches. Consequently, two models have been proposed: the single-route model (e.g.
Bybee&Moder, 1983), which posits that both regular and irregular past-tense forms are generated by analogy across stored exemplars (e.g. heal/healed - steal/*stealed, see (ii)), and the dual-route model (e.g. Prasada&Pinker, 1993), which posits that regular forms are generated via the application of a default rule (-ed) and irregulars are generated by analogy (see (i)).
This study builds on previous work by Albright and Hayes (2003) and Blything et al. (2018) and uses an elicited production paradigm to investigate which of the two models best accounts for L2 learners' morphological productivity. 88 adult English L2 learners with L1 Czech at A1-C1 proficiency levels and a control group of 9 native speakers heard sentences with someone performing a novel action described with a nonword (e.g. The baby likes to bize.
Look, there he is bizing. Everyday he bizes.).
Past-tense forms were then elicited by prompting the participant to describe what the agent "did yesterday." Produced forms were recorded and analysed with a binomial linear mixed-effects model in the R environment. The results showed that different language levels perform differently.
For native speakers, the likelihood of a verb being produced in regular past-tense form was positively associated with its phonological similarity to existing regular verbs (in line with the single-route model and the findings of Albright and Hayes (2003) and Blything et al. (2018)). However, L2 learners showed lesser dependence on the verbal similarity to regulars.
The A1-, A2-, and B1-level participants did not rely on the nonword's similarity to existing regulars or irregulars to produce the inflected form. In contrast, a main effect of similarity-to-regulars and similarity-to-irregulars was observed with the B2-level and C1-level participants, respectively.
The results indicate that the L2 acquisition of the English past-tense is characterized by a progressive development from the mastery of mechanistic rules (~ the dual-route mechanism at the A1, A2, and B1 levels) to the refinement of their application by spotting analogical patterns of existing verbs (~ the single-route analogical mechanism at the B2 and C1 levels, ~ native speakers). The second-language speakers thus show dynamic changes in the development of inflectional morphology that come closer to native speakers with the higher proficiency of B2 and C1 levels.