Viñals, L. 1 , Mirkovic, J. 2, 3 , Gaskell, G. 2 & Davis, M. 1
1 Cognition and Brain Sciences Unit, Cambridge, UK
2 University of York, York, UK
3 York St John University, York, UK
The role of consolidation in allowing language learners to generalise input statistics to novel instances remains unclear. Mirkovic and Gaskell (in prep.) trained participants on an artificial language containing regular and irregular plural affixes. A subset of stems had an ambiguous phonological cue (e.g. arb: varb, farb) that was associated with a high token frequency affix in irregulars (varbesh) but a high type frequency affix in regulars (farbaff but also grollaff, shilnaff, etc.). When given novel phonologically ambiguous stems (harb, yarb), participants tested 24hr after training produced more token-based generalisations (irregularisations) than participants tested immediately after training. Here, we extend these findings using a within-subject/between-affix design better suited to planned brain imaging studies. Participants were trained on two sets of plural affixes, distinguished by grammatical gender, on two successive days. Token and type frequencies from Mirkovic and Gaskell were maintained. Generalisation to novel phonologically ambiguous stems showed greater influence of token frequency for affixes trained on the previous day than for affixes trained on the same day. This suggests a role for overnight consolidation in the extraction of statistical regularities underlying inflectional morphology. Enhanced token-based generalisation following overnight consolidation is explained within a complementary learning systems framework.