Mirkovic, J. 1, 2 & Gaskell, G. 2
1 York St John University, UK
2 University of York, UK
Complementary Learning Systems (CLS) theory (McClelland et al. 1994; McClelland, 2013) suggests that statistical properties of new mappings crucially determine the role of the hippocampal and neocortical systems in learning. Specifically, due to the distributed nature of its representations, the neocortex is assumed to favor regular, systematic mappings, whereas the hippocampal system, using sparse representations, is thought to be more involved in learning and sleep-related consolidation of random, arbitrary mappings.
Two experiments manipulated statistical properties of novel mappings using artificial languages. In Experiment 1, participants learned novel word-forms that were systematically or arbitrarily related to their meanings. In Experiment 2, native English speakers learned novel verbs that mapped either consistently or inconsistently to the properties of English verbs. In both, participants learned the novel language and then spent time awake or asleep. The effect of sleep depended on the properties of the newly learned mappings: More systematic mappings, and the mappings consistent with the existing knowledge, showed no benefit of sleep, whereas more arbitrary or inconsistent mappings revealed better performance after sleep than wake. These findings demonstrate that statistical properties of new mappings influence the involvement of the two memory systems in learning new linguistic knowledge, consistent with the CLS.