Memory consolidation protects against interference in generalisation of language learning

Tamminen, J. 1 , Davis, M. 2 & Rastle, K. 1

1 Royal Holloway, University of London
2 MRC Cognition and Brain Sciences Unit

Associative learning (A-B) is vulnerable to interference from subsequent conflicting learning (A-C) unless the initial memory is allowed to consolidate first, preferably during sleep (Ellenbogen et al., 2006). We tested whether generalisation of newly learned language is similarly affected. In an artificial language learning approach participants learned new affixes (e.g., -tege) embedded in new words (e.g., ?sleeptege is a participant in a study about sleep?). Semantically consistent affixes referred to one semantic category (e.g., people) while semantically inconsistent affixes referred to two different categories (e.g., people and places). In Experiment 1 both affix types were trained in one training session. We combined the trained affixes with untrained stems (e.g., ?sailtege?) one week later in a semantic priming task to examine learning and generalisation of the affixes to new stem contexts. Priming was afforded only by the semantically consistent affixes. In Experiment 2 inconsistent affixes were trained with one semantic category on day 1 and with a different category on day 2. Priming was now seen in both consistent and inconsistent affixes. In the inconsistent condition offline memory consolidation appears to be necessary for the addition and generalisation of a second affix meaning without interfering with the first affix-meaning mapping.