Event related potential correlates of recollection and familiarity

Five experiments employed event-related potentials (ERPs) to investigate dual-process models of recognition memory. Each experiment consisted of a three phase design in which two lists of words were presented in two temporally segregated study tasks. During the third, 'test', phase, subjects were required to respond on one key to old items from a specified study phase, and to respond on another key both to old items from the alternate study phase and to new items. As recognised items require differential responding depending on their source, it is argued that recollection-based recognition of a studied item allows subjects to respond accurately whereas familiarity-based recognition does not. Four principal patterns of neural activity were observed throughout the course of the five experiments. Items recognised on the basis of familiarity elicited greater positivity than new items at frontal sites between 300-600 msec. Recollected items were associated with a second phase of positivity between 500-800 msec, maximal at parietal sites. The finding that these two patterns of neural activity were qualitatively distinct supports dual-process models of recognition memory which state that recollection and familiarity are independent. Recognised items also elicited greater positivity than new items over right frontal and frontopolar sites from 800 msec until the end of the recording epoch (approximately 1400 msec). It is argued that this ERP effect reflects processes that evaluate and monitor the products of retrieval. Finally, a fourth pattern of neural activity is reported in which ERPs associated with studied items are more negative going than those associated with new items, maximal at mid and right parietal sites between 800-1400 msec. It is suggested that this ERP effect may reflect the response conflict experienced when recognition does not determine the response. The implications of these findings for models of recognition memory are discussed