Mirkovic, J. & Gaskell, G.
University of York
The Complementary Learning Systems model (McClelland 2013) suggests that the integration of the newly learned information into the existing neocortical memories depends on their consistency with the existing knowledge. We explored this hypothesis using the English past tense. We designed novel verbs which were phonologically similar to a neighbourhood of existing verbs with either predominantly regular past tenses (PLARE, cf. share, stare), or predominantly irregular past tenses (FLEEP, cf. sleep, keep). Participants learned past tenses for the novel verbs which were either consistent or inconsistent with the existing neighbourhoods (plared (consistent), plore (inconsistent); flept (consistent), fleeped (inconsistent)). The novel past tenses were learned in either the evening or the morning, and tested 12 hours later. Performance at training and at test was influenced by the consistency of the novel forms with the existing neighbourhoods, but this effect was modulated by whether the novel verbs were similar to the existing regular or irregular neighbourhoods. Crucially, only for the participants who were tested after a 12-hour delay which did not include sleep the consistency x regularity interaction changed over time. The findings will be discussed in the context of the CLS model of memory and psycholinguistic models of the English past tense.