Separating retrieval strategies from retrieval success: an event-related potential study of source memory

Abstract

Event-related potentials (ERPs) were recorded while subjects performed two different source memory retrieval tasks. Each task was preceded by a study phase in which subjects heard an equal number of words spoken in a male or a female voice. A cue preceding each word indicated whether the subjects should make an active/passive (action) or a pleasant/unpleasant (liking) judgment to the word. In one retrieval task (the voice condition), subjects made a three-way distinction between new (unstudied) words, and words that had been spoken by the male or the female voice at study. In the second retrieval task (the task condition), subjects distinguished between new words, and words to which they had made an action or a liking judgment. All test words were presented visually. In keeping with previous findings, the differences between the ERPs to correct memory judgments for old and new items were characterised by two temporally and topographically dissociable modulations, with right-frontal and left-parietal maxima respectively. These old/new effects displayed different sensitivities to successful retrieval of either voice or task information, providing further evidence that they index functionally dissociable processes. The direct comparison of the ERPs to correct rejections in the voice and task retrieval conditions revealed reliable differences over frontal scalp, suggesting that, irrespective of whether retrieval is successful, neural processing differs according to the source retrieval demands of the task.

Introduction

Source memory refers to memory for information that specifies the conditions under which memories were acquired [15]. This includes details of spatial, temporal and social context, and information about the modalities through which events were initially perceived. The results of studies with brain damaged patients 11, 32, 34, as well as behavioural [9]and event-related potential studies in normal subjects [41], are consistent with the view that memory for source involves at least two distinct sets of processes. One influential proposal is that these processes are associated with information retrieval, and with post-retrieval mechanisms necessary for recovery of accurate source information 23, 35.

This two-stage description of memory for source is evident in the source monitoring framework (SMF) of Johnson and colleagues 15, 17, which emphasises the strategic use of memory. In the SMF it is assumed that source judgments can be made by assessing the distribution of various qualitative characteristics of a memory trace, which should differ for memories of different origins. For example, the memory trace for a heard word will contain more perceptual information than the memory trace for a word that a subject imagined hearing [13]. Consequently, one strategy that subjects could adopt in order to distinguish real (heard) from imagined events is to make an assessment based on the amount of perceptual detail in the memory trace. Evidence consistent with the view that strategies of this form are used in source memory tasks comes from two studies by Johnson, Foley, Suengas and Raye [14]. The first required recall of memories for previous life events, and for imagined ones. Subjects reported that memories for real events were associated with clearer contextual information, and greater perceptual detail, than memories for imagined events. In the second study, a different group of subjects identified these same forms of source information as those they would employ in order to distinguish real from imagined events 10, 15.

While these findings are consistent with the view that subjects engage in some forms of strategic mnemonic processing, the locus at which these strategic processes act is unclear. According to a post-retrieval explanation, the engagement of any such processes is contingent upon successful recovery of information from memory. For example, following retrieval, attention may be directed to task-relevant source information. An alternative account is that selective mnemonic processing of this form operates independently of retrieval success. According to this view, the selective processing of mnemonic content could engage processes that direct retrieval, or processes that monitor for retrieval of certain mnemonic contents. If this second view is correct, then in tasks where subjects discriminate between new items and more than one class of old item, these strategic mnemonic processes will act over new as well as old test items. Consequently, evidence consistent with this account would stem from findings that the processing afforded correctly identified new items (correct rejections) in source memory tasks varies according to what source information is to be retrieved.

Since event-related potentials (ERPs) can be separated on a trial-by-trial basis according to factors including item type and behavioural response, they provide a means for assessing the extent to which the processing of correct rejections varies with retrieval demands. However, to this point there have been no ERP studies that have compared ERPs to correct rejections that were evoked under different source retrieval conditions. Indirect evidence supporting the view that the processing of new items in source memory tasks varies according to what information subjects use for source discriminations comes from one recent ERP study of source memory [16]. In that study, two groups of subjects performed different encoding tasks, prior to completing the same source retrieval task. The authors reported that distinct modulations over frontal and posterior scalp sites differentiated the ERPs to correct rejections from the two groups. In addition, the same modulations differentiated the ERPs to correct source judgments. On the basis of these findings, Johnson and colleagues [16]concluded that, since these modulations were common to the ERPs associated with correct judgments to old and new stimuli, they reflected global shifts in strategy that varied according to the different encoding tasks that had been performed by the two groups.

The experiment reported here differs from that of Johnson and colleagues in that retrieval orientation at test is manipulated directly via task instructions. All subjects performed the same encoding tasks, followed by two source retrieval tasks that required retrieval of different aspects of the source information that had been encoded at study. ERPs were recorded to visually presented words in separate test blocks that required source judgments based either on voice or task information. The voice and task information had been encoded in a task where subjects heard words, an equal number of which were spoken by a male/female voice. To each word subjects made one of two judgments—an active/passive judgment (action task), or a pleasant/unpleasant judgment (liking task). The type of judgment varied across trials, and was determined by a cue that preceded each spoken word. In one test condition, subjects made a three-way distinction between new words, and old words spoken at study by the male or female voice. In the second condition, subjects distinguished new words from old words to which either an active/passive or a pleasant/unpleasant judgment had been made at study. Subjects were informed of the retrieval orientation for each test block only after completion of the preceding study phase.

Voice and task information were chosen as source variables because electrophysiological correlates of successful retrieval of task and voice have been reported in previous ERP studies of source memory 39, 41. In addition, they were selected since they should impose quite different retrieval demands, thereby increasing the likelihood that different strategic mnemonic processes are engaged in the two cases: while the voice condition entails retrieval of perceptual information, the task condition requires retrieval of information related to forced-choice semantic judgments that were made at study.

This design therefore allows comparisons between ERPs to correct source judgments and correct rejections that are evoked in tasks requiring retrieval of either voice or task information. In conjunction, the paired comparisons of the ERPs to correct source judgments and correct rejections should shed light on the form and locus of processes that are involved in the selective processing of mnemonic content. In particular, any reliable differences between the ERPs to correct rejections evoked in the voice and task conditions would indicate that source retrieval demands influence the neural response to classes of stimuli that are presumably not associated with retrieval of mnemonic content. Findings of this form would therefore be consistent with the view that the memory-related processes involved in selectively processing different types of source information are not restricted to a post-retrieval locus.