Sleep–related consolidation of new form–meaning mappings: the acquisition of arbitrary and systematic mappings in adult language learning.

Six experiments investigated the role of sleep-related memory consolidation in learning new words (e.g. tib bisesh = queen). We tested the predictions of the Complementary Learning Systems model (CLS; McClelland et al., 1995) that sleep-related consolidation varies with arbitrariness in the form-meaning mapping. New determiners (tib, ked) and suffixes ( -esh, -ool) systematically mapped on to the referents' natural gender. Stem-meaning mappings (e.g. bis- = queen, jor- = cowboy) were arbitrary, as the meaning of the stem could not be predicted from its phonology. In Experiments 1 and 2 there was one determiner and two suffixes per gender (tib = female, ked = male; -esh, -eem = female, -ool, -aff = male). In Experiment 2 overnight polysomnography data was collected, to correlate slow wave sleep (SWS) with arbitrary mapping recall. In Experiments 3 and 4, there were two determiners and one suffix per gender (tib, paz = female, ked, jov = male; -eem = female, -ool = male). In Experiments 5 and 6, the systematic mapping included the suffixes only (without determiners), and the number of exemplars was increased in Experiment 6. The memory for the arbitrary mappings was tested in recall and recognition tasks. The knowledge of the systematic mappings was tested in generalisation tasks. As an exploratory investigation of Ullman's Declarative/ Procedural model (e.g. Ullman, 2001) we also correlated measures of arbitrary and systematic mappings with a declarative and a procedural task. As predicted by the CLS, there was evidence to suggest that sleep was beneficial for the memory of the arbitrary mappings, but not for the systematic mappings. Determiners required full systematicity to be extracted. Suffixes required increased exemplar variability and no determiners present to be extracted. The findings will be discussed in the context of models of memory consolidation in word learning.