Event-related potential investigations of source and item memory

Abstract

Memory for the context of a learning episode (source memory) was investigated in four experiments using event-related brain potential (ERP), accuracy and reaction time measures. Four questions were posed about the relationship between item and source memory: (a) whether source information is automatically retrieved when an item is remembered; (b) whether item and source retrieval involve different brain activity; (c) whether ERPs recorded during source retrieval are unique or generalizable across sources; and (d) whether perceptual source attributes such as the voice of a spoken word are qualitatively different from self-generated attributes such as one's own actions. The results showed that studied items elicited more positive ERPs than unstudied items at all scalp sites, beginning relatively early after stimulus onset. Source information was retrieved only when required by the assigned task, and only after item information had been recovered from memory. When source information was requested, studied items elicited an additional, late prefrontal positivity which was independent of the accuracy of the source memory judgement. This result suggests that the prefrontal effect reflects the mere attempt to retrieve source information. The success or failure of source retrieval was evident at more posterior scalp sites; this effect also began several hundred milliseconds after the initial differentiation between studied and unstudied items. ERPs recorded posterior to prefrontal cortex were also sensitive to source content. In two experiments, participants studied objects. In the source memory test, they were asked to recall what they did with the objects during the encoding phase: performed an action, watched the experimenter perform an action, imagined an action, or estimated the cost of the object. Objects studied under different encoding tasks elicited different patterns of brain activity during the source retrieval task, but not when the objects were simply judged as studied or unstudied. Even during the source memory test, the ERPs elicited by perform-encoded objects were similar to those elicited by watch-encoded objects, although both were different from imagine-encoded and cost-encoded items. Thus, memory for self-generated actions and observed actions did not show a fundamentally different pattern of brain activity in the present work.