Attention and successful episodic encoding: an event-related potential study

Abstract

Event-related potentials (ERPs) were used to delineate the cerebral processes occurring when information is encoded into episodic memory and to determine how these processes are affected by divided attention. ERPs were recorded during encoding under focused or divided attention, and were selectively averaged on the basis of their retrieval during later free recall and recognition tests (with remember–know judgments). Items retrieved with conscious recollection of the encoding episode (remembered, recalled) were distinguished at encoding from later missed items by an enhanced left fronto-temporal negative wave (N340), a negative posterior sustained potential and a positive frontal sustained potential. These effects occurred independently of the level of attention. Items later retrieved on the basis of familiarity (known) elicited a larger N340 than missed items, but did not demonstrate the increased sustained potentials. We suggest that item-specific conceptual processing (N340) is sufficient to produce familiarity-based recognition, but additional elaborative processing (sustained interaction of frontal and posterior regions) is necessary for conscious recollection. The effect of divided attention on these processes was related to the difficulty of the secondary task, with the more difficult task causing greater and earlier interference.

Introduction

Attention is essential to the formation of a declarative memory trace. Reducing the attentional resources available for encoding information by distraction from a secondary task impairs performance on explicit tests of recall and recognition (e.g. Ref. [9]) but not on implicit tests such as word-fragment completion (e.g. Ref. [65]). In addition, when explicitly recognized items are categorized according to whether the retrieval experience is one of consciously ‘remembering’ the details of an encoding episode, or simply ‘knowing’ that it previously occurred [89], a secondary task at encoding specifically decreases the proportion of ‘remembered’ items (e.g. Ref. [28]). This impairment may occur because the secondary task diverts time and attentional resources away from deep semantic processing. Support for this view comes from studies demonstrating that processing information at a physical or structural level rather than at a deeper, semantic level influences episodic memory performance in a manner similar to divided attention: free recall is impaired [11], and the proportion of items remembered during recognition is reduced [25], [27], [73]. Yet, the stage of processing at which the interference occurs may depend on the degree to which these resources are reduced. Moderate resource reductions may only interfere with relational and elaborative processing [8], whereas more severe reductions may affect basic semantic or even phonemic processing [54].

Moscovitch and Umilta [53] proposed a model of memory and consciousness that describes how reduced resources might interfere with the cerebral processes responsible for episodic memory formation. They propose that the medial temporal lobe/hippocampal complex (MTL/H) only has access to information that has been consciously apprehended. The prefrontal cortex controls the extent to which incoming information is processed into consciousness through the mechanism of selective attention. When attention is divided across multiple tasks, fewer stimuli will reach the level of conscious awareness necessary for apprehension by the MTL/H. Furthermore, those stimuli that do reach this threshold may not be sufficiently encoded because the multitasking required by divided attention tasks occupies the supervisory capacities of the prefrontal cortex and interferes with the ability of this region to engage in the level of elaborative, semantic processing necessary for successful episodic encoding.

Many aspects of this model have been supported by blood flow studies in normal adults. Recent event-related fMRI studies of memory have shown maximal activation of frontal and parahippocampal cortices when items were subsequently recognized with either conscious recollection of the encoding episode (i.e. ‘remembered’) or with high confidence [3], [93]. Items later recognized without either conscious recollection (i.e. ‘known’) or with low confidence did not differ from missed items. Furthermore, PET studies examining the effects of divided attention on episodic encoding have found that reduced resources affect activity in similar regions. Using a visuomotor secondary task and an auditory memory task, Fletcher and co-workers ( [20]; see also Ref. [83]) found a specific decrease in left prefrontal activity (BA 46) at encoding as the demands of the secondary task increased. More recently, Iidaka et al. [37] found that divided attention at encoding reduced blood flow to the left prefrontal and inferior temporal cortices, although it minimally affected the encoding-related activity of the left hippocampus. Taken together, these studies provide support for the view that divided attention interferes with the ability of the prefrontal cortex to carry out the semantic, elaborative processing necessary for successful episodic encoding.

Event-related potentials (ERPs), which measure neural activity that is time-locked to the processing of an individual stimulus, also have demonstrated differential neural activity associated with successful episodic encoding. Specifically, words later retrieved on a subsequent test of explicit memory typically elicit more positive-going potentials at encoding than words later forgotten (see Refs. [40], [79] for review). Differences between the ERPs for words later recalled and those not recalled have been termed ‘differences based on later memory’ or Dm effects [5], [62], [63]. Although Dm effects involve multiple separate components, the most prominent of these are a transient posterior positive wave in the latency range of 300–800 ms and a sustained frontal positivity beginning at about the same time and lasting longer. These components are sensitive to different types of encoding instructions and therefore, may index different cognitive processes. For example, studies have shown that the transient posterior positive wave is more predictive of successful recall for orthographically distinctive words that are studied under rote memorization instructions [17], [18], whereas the sustained frontal positive wave is generally more predictive when words are studied under instructions that emphasize elaborative strategies such as semantic association, semantic verification or imagery [19], [62], [91], [94]. These results suggest that the transient posterior positivity may represent item-specific evaluation and processing of distinctive information, whereas the sustained frontal positivity may represent inter-item associative encoding.

Although ERPs have not yet been used to examine episodic encoding in relation to attention, evidence suggests that both Dm effects index effortful processes that would be influenced by the demands of a secondary task. Fabiani and co-workers related the transient posterior positive wave to the P300 or P3b wave of the ERP, a centro-parietal positivity that follows the detection of task-relevant events [13], [66]. The P3b is sensitive to attentional manipulation and indeed, may primarily index the amount of attention allocated to processing a stimulus [21], [31]. For example, in a target detection task, the amplitude of the P3b is reduced in proportion to the degree to which attentional resources are diverted by a secondary task [96]. Some have argued, however, that the transient memory-related wave is not equivalent to the P3b because this Dm effect has a more widespread distribution [22], [31] and is sensitive only to subsequent memory rather than to perceptual processing in general [62]. One way to reconcile these different interpretations is to consider the possibility that under certain circumstances, the amount of attention deployed in the processing of an item may predict subsequent memory for that item. Stimuli that are physically distinctive may elicit an orienting response that increases item-specific attention, and when elaborative strategies are not employed, this increase may be sufficient to influence and predict subsequent memory. However, elaborative strategies also require the sustained allocation of processing resources. Thus, it is also likely that attention is a necessary condition for the semantic, organizational processing associated with the sustained frontal positivity.

In the present study, we varied attention at encoding by having participants attend only to a series of visually-presented words (focused attention) or divide their attention between the words and an auditory-motor task (divided attention). As in neuroimaging studies of divided attention [37], we varied attentional demands by using easy and difficult versions of the secondary task. ERPs at encoding were averaged both as a function of subsequent memory performance and as a function of attention at encoding. These two averages provide somewhat different information about the encoding process. Averaging as a function of subsequent memory performance provides specific information about those encoding processes that are critical for successful storage in episodic memory. Averaging as a function of attention provides information about the general encoding mode, or neurocognitive set, during conditions of reduced attentional resources, as well specific information about the stages of processing where reductions in attention influence encoding.

We categorized our ERP findings on subsequent memory performance derived from both free recall and remember–know recognition tests. We chose to use both types of test as the basis for selective averaging at encoding for three reasons. Firstly, in previous studies, Dm effects were generally largest and most reliable when based on free recall, smaller and less reliable when based on recognition, and essentially undetectable when based on implicit memory tests [45], [55], [61], [64]. These findings suggest that Dm effects differentially index the conscious, controlled processes that predict successful episodic encoding. By this view, the weak relationship between Dm effects and recognition in these studies may have occurred because retrieval success on recognition tasks can arise from both conscious, controlled processes, which are associated with measurable Dm effects, and unconscious, automatic processes, which are not. This view also would predict, however, that robust Dm effects could be found when based on recognition performance if items retrieved on the basis of conscious recollection (R) and familiarity (K) were averaged separately at encoding.

It is therefore surprising that Smith [84] found no significant ERP differences at encoding between subsequently remembered items and subsequently known items, even though R and K responses could be dissociated at retrieval [14]. Investigators have disputed whether R and K responses index categorically different cognitive processes [29], [39], [76] or whether they simply differ in terms of trace strength (e.g. Refs. [12], [36]), however all appear to agree that R responses involve the conscious recollection of the encoding context. Thus, if memory-related ERPs index processes related to episodic encoding, one would predict that they would differentiate items on the basis of subsequent R and K responses, either qualitatively — if they represent different cognitive processes, or quantitatively — if they exist as points along a unitary dimension. A more recent ERP study, however, found that younger adults showed greater sustained frontal positivity for subsequently remembered items than item subsequently known or missed [23]. Older adults did not show this differentiation, perhaps because of a failure to use elaborative encoding strategies, a deficit in attentional resources, or more lenient retrieval criteria. Thus, a second goal of the present study is to evaluate further whether R and K involve differentiable neural activity and whether ERP components associated with these responses are influenced by changes in attentional resources.

Finally, although successful recall and ‘remembering’ are both thought to arise from successful episodic retrieval [95], free recall and recognition tests have inherently different retrieval demands and are sensitive to different encoding processes. Whereas free recall benefits from relational processing that strengthens inter-item associations and facilitates strategic retrieval, recognition benefits more from item-specific processes that enhance the distinctiveness and discriminability of the individual item [16], [49]. With regard to R responses, some studies have emphasized their dependence on the type of elaborative, associative encoding that also benefits free recall [26]. Others have focused on the relationship of R responses to processing of salient or distinctive attributes of the individual item, which would largely benefit recognition [74]. The most parsimonious view is that both relational and distinctive item-specific processing contribute to the elaborative encoding that results in an R response. Therefore, in the present study, rather than give participants any specific encoding instructions that might result in preferential relational or item-specific processing, we simply encouraged participants to memorize the items. Yet, we averaged the encoding ERPs on the basis of whether the item was later recalled and remembered, remembered but not recalled, known, or missed. If posterior and frontal memory-related components index item-specific and relational processing, respectively, these components may be differentially sensitive to episodic encoding predictive of R items that can only be recognized, versus R items that can be both recalled and recognized.

In summary, we hypothesized that we would find a series of components in the ERPs recorded at encoding that would be sensitive to both the level of attention at study and the level of conscious recollection at retrieval. With regard to the effects of memory, we predicted that a posterior transient wave would be related to item-specific processing contingent on initial conscious apprehension of the incoming information, and that a sustained frontal wave would be related to subsequent elaborative processing. This prediction would be borne out by greater posterior positivity for both subsequently recalled and/or remembered items than missed items, but greater frontal positivity only for items that were both recalled and remembered. With regard to the effects of attention, we predicted attenuation of the sustained frontal positivity under dual-task relative to single-task conditions, regardless of the difficulty of the secondary task. We expected, however, that the more difficult secondary task would affect initial conscious apprehension of the item, and therefore would also attenuate the earlier transient posterior wave. Finally, because our recordings were more widespread over the scalp than in previous ERP studies of this type, an additional goal of this study was to explore other aspects of the encoding waveform that might be related to subsequent memory or attention.