Usage-based approaches to the acquisition of L2 morphology

Usage-based approaches to language learning hold that we learn constructions (form-function mappings, conventionalized in a speech community) from language usage by means of general cognitive mechanisms (exemplar-based, rational, associative learning). The language system emerges from the conspiracy of these associations. Although frequency of usage drives learning, not all constructions are equally learnable by all learners. Even after years of exposure, adult second language learners focus more in their language processing upon open-class words than on grammatical cues. I present a usage-based analysis of this phenomenon in terms of fundamental principles of associative learning: Low salience, low contingency, and redundancy all lead to form-function mappings being less well learned. Compounding this, adult acquirers show effects of learned attention and blocking as a result of L1-tuned automatized processing of language. I review a series of experimental studies of learned attention and blocking in second language acquisition (L2A). I describe educational interventions targeted upon these phenomena. Form-focused instruction recruits learners' explicit, conscious processing capacities and allows them to notice novel L2 constructions. Once a construction has been represented as a form-function mapping, its use in subsequent implicit processing can update the statistical tallying of its frequency of usage and probabilities of form-function mapping, consolidating it into the system. I close with two investigations of how morphemes are more easily processed when they reliably occur with lemmas consistently conjugated in this form, and when these inflected forms are found in more formulaic contexts. This is evident (1) in a psycholinguistic investigation (with Wendy Guo) of L2 learners of English and their Elicited Imitation of sentences containing target morphemes, and (2) in a large-scale learner-corpus investigation (with Akira Murakami) into the accuracy of morpheme production in the writing of 140,000 ESL learners across 128 proficiency levels in the EF-Cambridge Open Language Database (EFCAMDAT).