Fixed wh-expressions in classroom second language acquisition: databases of computational properties or utterance schemas?

This study adopts concepts from two competing approaches to second language acquisition (SLA) (usage based vs. generative) to analyse the effect of formulaic expressions (FEs) on learners’ L2 syntactic development. Using spoken transcripts of the longitudinal Barcelona English Language Corpus (BELC), we identify four learned fixed wh-expressions (FEswh), which are all produced by learners in advance of respective L2 competence. We measure learners’ use of these expressions and the evidence of related computational properties (e.g., wh-movement, do-support) and utterance schemas (e.g., [WH + AUX DO + X]) outside this use across a 7-year data collection period.

Adopting a generative analysis, we find that an earlier and more frequent use of FEswh correlates with better L2 knowledge of the expressions’ associated computational properties. Then, adopting a usage-based ‘traceback’ methodology (e.g., Lieven et al., 2003; Eskilsden, 2020), we find that learners accurately produce some L2 interrogatives that share utterance schemas of previously used FEswh that appear in their production data ontogenetically. Utterance schema extraction and generalisation of model surface forms may therefore facilitate the development of the more general L2 feature specifications on the functional categories for which these surface forms exemplify. From this, we argue that such a unified account of learners’ L2 development can offer a better description of the trends observed in the corpus than either usage-based or generative models can do independently.