Neural correlates of retrieval processing in the prefrontal cortex during recognition and exclusion tasks

Abstract

Event-related fMRI was employed to contrast the neural activity elicited in prefrontal cortex during recognition memory and exclusion tests. The study phases preceding each memory test were identical, involving the presentation of study items (visually presented words) in one of two study contexts. For the recognition test subjects were required to respond positively to all old items regardless of study context, and to respond negatively to new items. For the exclusion task, positive responses were required to old items presented in one of the study contexts only; negative responses were required both to unstudied items and studied items from the alternative context (non-targets). No prefrontal region demonstrated greater activity for new items in the exclusion task. Thus, there was no evidence that retrieval cues were processed differently according to the specificity of the sought-for information. In several regions, most notably bilateral anterior prefrontal cortex, activity was greater for old than for new items regardless of task. Activity in right dorsolateral prefrontal cortex was also greater for old than for new items; these effects however were larger in the exclusion task. The findings are consistent with previous reports that activity in anterior prefrontal cortex elicited by recognition retrieval cues is sensitive to retrieval success, and extend these findings to the exclusion task. The findings for the right dorsolateral cortex add further weight to the proposal that this region supports post-retrieval monitoring of retrieved information.

Introduction

Since the earliest functional neuroimaging studies of memory it has been apparent that tasks requiring explicit memory retrieval (‘direct’ memory tasks) engage the prefrontal cortex (for reviews of this early work see [6], [12]). Initial accounts of prefrontal activation during retrieval focused on what, at the time, was the most consistently observed effect—the activation of right anterior (brodmann area (BA) 10) and dorsolateral (BA46) cortex. This activity was variously proposed to be a correlate of retrieval ‘mode’ [32], retrieval ‘effort’ [42] or retrieval ‘success’ [40]. It is now clear that retrieval-related activity in different prefrontal regions is differentially sensitive to variables such as type of retrieval task [41] and the nature of the retrieved information [27], and accounts of these findings are beginning to be framed with reference to the different cognitive operations thought to support episodic retrieval [11]. For example, it has been proposed that activity in right ventrolateral cortex supports processing of retrieval cues prior to a retrieval attempt, whereas activity in dorsolateral cortex reflects monitoring operations conducted on the products of a retrieval attempt [13].

The starting point for the present experiment are studies in which retrieval-related activity was contrasted according to whether test items required a simple yes/no recognition decision, or a judgement based on retrieval of specific information about an item’s study context (for example, whether the item was presented to the left or right of fixation, a test of ‘source memory’). These two tasks potentially differ in a variety of ways in terms of the demands made upon retrieval processing, of which two are particularly relevant here. First, the level of specificity of the information that must be retrieved is different. In the case of recognition, any information about the recent occurrence of a test item can support a positive judgement; indeed, it has been argued that recognition judgements can be supported by information (termed ‘familiarity’) which is entirely bereft of contextual information, and derived from retrieval processes independent of those supporting ‘true’ episodic memory [1], [29]. By contrast, judgements about study context (as in source memory tasks) necessarily require retrieval of episodic information sufficiently specific to support the relevant discrimination. Because of these different informational demands, subjects may process test items differently in a test of recognition memory than in a test of source memory.

Yes/no recognition and source memory tests also differ with respect to the amount of ‘post-retrieval’ processing [36] that is needed to permit an accurate judgement. In the case of recognition, the correct response can be selected merely on the basis of whether a test item was experienced at study. Thus, evaluation of the products of a retrieval attempt is limited to an assessment of evidence signalling prior occurrence. When that evidence is strong, little evaluation will be required prior to response selection. By contrast, in tests of source memory evidence that a test item belonged to the study phase, however strong, is insufficient; the content of the information providing that evidence must be evaluated prior to response selection. Other things being equal, therefore, tests of source memory will require more extensive evaluation of retrieved information than tests of recognition memory.

A third way in which recognition and source memory tasks differ is in terms of their general difficulty. Other things being equal, source judgements on previously studied items are more demanding, and take longer to perform, than recognition judgements on the same items. While this difference is not inherent to the two kinds of task (unlike those noted above), it will always be present unless specific measures are taken to control for or equate difficulty.

Only a few studies have contrasted retrieval-related prefrontal activity according to whether the retrieval test required a simple recognition judgement or a judgement based on specific details about the study context. In an early event-related fMRI study, Nolde et al. [30] reported that test items subjected to source memory judgements elicited more activity in left dorsolateral and anterior prefrontal cortex than did items in a recognition test. Nolde et al. [30] interpreted these findings as supporting the proposal [31] that left prefrontal cortex is selectively engaged by the requirement to retrieve specific rather than general information about a prior episode. Unfortunately, activity was not assessed separately for studied and unstudied test items, making it impossible to determine whether the additional left prefrontal activity was elicited by all items, or only those eliciting episodic retrieval (i.e. studied items). This is a crucial issue [36]: the first of these alternatives would suggest that the left prefrontal activity is sensitive to differences in the way retrieval cues are processed according to the nature (in this case, the specificity) of the sought-for information. By contrast, if the activity was elicited by studied items alone, this would suggest that the activity is associated with post-retrieval processes such as evaluation of the retrieved information.

Two studies employing blocked experimental designs contrasted recognition and source tasks. In Rugg et al. [39] the study phase involved the presentation of words to the left or right of fixation with the instruction to make a different judgement (animacy or pleasantness) according to the side of presentation. Three kinds of test block were contrasted: low-density recognition memory, when a recognition judgement was required on a list of predominantly new (unstudied) items; high-density recognition memory, requiring recognition judgements on mainly old (studied) items; and high-density source memory, when the requirement was to judge whether each item (most of which were old) had been presented at study on the left or right. Consistent with previous results [40], [41], the contrast between high and low-density recognition revealed greater activity in several prefrontal regions, the most notable of which were bilateral anterior and left inferior prefrontal cortex. The contrast between the source and high-density recognition conditions revealed greater activity for the source task in, among other areas, bilateral anterior insula, left anterior prefrontal cortex (medial to that observed for the contrast between recognition blocks), and right dorsolateral prefrontal cortex. In Henson et al. [17] study words were presented in two temporally segregated lists, and displayed either above or below fixation. Recognition memory for test items was contrasted with an ‘exclusion’ condition. In exclusion tasks [22] studied items belonging to a designated source (e.g. above fixation, or in list 1) must be accepted as old (‘target items’), whereas those from an alternative source (‘non-targets’) must be rejected along with new items. This task thus shares with source memory the requirement to recollect the context in which a studied item was presented in order to allow the correct response to be selected. Henson et al. [17] reported that relative to recognition, the exclusion task was associated with elevated activity bilaterally in dorsolateral prefrontal cortex.

The findings of Rugg et al. [39] and Henson et al. [17] suffer from the same limitations of interpretation as those noted earlier for Nolde et al. [30], namely, that it is not possible to determine whether the reported task effects reflect differences in activity elicited by both studied and unstudied items, or by studied items only. None the less, it is noteworthy that in several event-related studies of recognition memory (see [37] for review) some of the task-sensitive regions identified by Rugg et al. and Henson et al. have consistently been reported to show differential activity according to the status of the test items eliciting the activity. Most notably, several studies have reported effects in the vicinity of the right dorsolateral regions identified by Rugg et al. [39] and Henson et al. [17]. On the basis of the finding that activity in this region was greater for items associated with recognition judgements accorded low rather than high confidence ratings, Henson et al. [18] argued that the right dorsolateral region supports operations necessary for the monitoring and evaluation of the products of a retrieval attempt, as suggested previously [19]. On the assumption that source memory makes greater demands on monitoring and evaluation than recognition (see above), this interpretation is consistent with findings that source memory tasks engage this region to a greater extent than do tests of recognition memory.

In the final study to be mentioned here [34], event-related fMRI was employed to investigate the activity elicited by test items according to whether the task required retrieval of ‘general’ information (was this picture seen at study?) or ‘specific’ information (is this picture larger or smaller than that shown at study?). Ranganath et al. [34] identified a left anterior prefrontal region where activity was greater during the specific task. This effect was found for both studied and unstudied items, consistent with a role for the left anterior region in modulating the processing of retrieval cues according to the demands of the retrieval task (in the terminology of Rugg and Wilding [36], in modulating ‘retrieval orientation’). Intriguingly, the region identified by Ranganath et al. [34] is very close to left anterior prefrontal areas reported to exhibit greater activity for studied than unstudied items (so-called ‘old/new’ effects) in some studies of recognition memory (e.g. [7], [25], [28]; see [37] for review). These findings suggest a specific role for this region in processing retrieved information, additional to any more general role it may play in the processing of test items.

In the study reported here we followed up the findings reviewed above by using event-related fMRI to investigate activity elicited by studied and unstudied test items in both a recognition memory and an exclusion task. The design of the study allowed several of the issues raised by previous studies to be addressed, including: (i) whether activity elicited by unstudied test items in left anterior prefrontal cortex varies according to task (cf. [34]), and (ii) whether, as might be expected given the findings of Rugg et al. [39] and Henson et al. [17], right dorsolateral old/new effects are greater in the exclusion than the recognition task. In addition, the employment of the exclusion task allowed an assessment of the extent to which old/new effects depend upon the allocation of old and new items to different response categories. Since excluded non-targets (old items) and unstudied (new) items receive the same response, any differences in the activity they elicit must be due to something other than differential response selection.

Finally, we attempted to control for the confound—present in all previous studies—between task and difficulty (other things being equal, source memory judgements are more difficult than yes/no recognition). We tried to achieve this by administering each task at two difficulty levels, so that task and difficulty would vary orthogonally. To anticipate the results, this aim of the study was unsuccessful because the difficulty manipulation had little effect on performance and, correspondingly, only minimal effects on brain activity. The results are therefore reported without regard to this manipulation. A brief report of findings from the exclusion task has been published previously [38].