نقش حافظه اپیزودیک در قضاوت ارزشی در مورد نگرش کنترل: مطالعه بالقوه مربوط به رویداد

Event-related potentials (ERPs) are unique in their ability to provide information about the timing of activity in the neural networks that perform complex cognitive processes. Given the dearth of extant data from normal controls on the question of whether attitude representations are stored in episodic or semantic memory, the goal here was to study the nature of the memory representations used during conscious attitude evaluations. Thus, we recorded ERPs while participants performed three tasks: attitude evaluations (i.e., agree/disagree), autobiographical cued recall (i.e., You/Not You) and semantic evaluations (i.e., active/inactive). The key finding was that the parietal episodic memory (EM) effect, a well-established correlate of episodic recollection, was elicited by both attitude evaluations and autobiographical retrievals. By contrast, semantic evaluations of the same attitude items elicited less parietal activity, like that elicited by Not You cues, which only access semantic memory. In accord with hemodynamic results, attitude evaluations and autobiographical retrievals also produced overlapping patterns of slow potential (SP) activity from 500 to 900 ms preceding the response over left and right inferior frontal, anterior medial frontal and occipital brain areas. Significantly different patterns of SP activity were elicited in these locations for semantic evaluations and Not You cues. Taken together, the results indicate that attitude representations are stored in episodic memory. Retrieval timing varied as a function of task, with earlier retrievals in both evaluation conditions relative to those in the autobiographical condition. The differential roles and timing of memory retrieval in evaluative judgment and memory retrieval tasks are discussed.

Attitudes are distinct from other mental processes because they embody evaluative judgments based on internal scales that reflect an individual's values. By providing a basis for deciding whether to approach or avoid particular people, situations and objects, attitudes play an important role in everyday behavior. However, despite the fact that behavior is affected strongly by attitudes, relatively little is known about how they are represented and processed in the brain. Previously, we described the event-related brain potential (ERP) correlates of some executive control and affective processes used to make truthful and deceptive attitude evaluations (i.e., agree-disagree) (Johnson, Henkell, Simon, & Zhu, 2008). Here, we examine whether the representations retrieved during conscious attitude evaluations are stored in episodic or semantic memory. Although the processes used in attitude evaluations have been characterized using dual-process models for years (e.g., Breckler, 1984), researchers have only recently begun to specify how these processes are instantiated in the brain. In one model, reflexive responses are posited to be rapid, automatic and based on unconscious processes whereas reflective responses are thought to be slower, controlled and based on resource-demanding processes (Lieberman, 2003 and Lieberman, 2007). The purpose of the reflective system is to enable the automatic evaluations produced by the reflexive system to be overridden so that responses can be adaptively determined based on the current context and goals. In another model, both reflexive and reflective evaluations are posited to be made hierarchically using a common set of processes, with the primary difference between them being the duration of the evaluation process (Cunningham & Zelazo, 2007). In both these models, reflexive and reflective evaluations are linked to different memory systems. For example, reflexive evaluations are believed to depend on the amygdala due to its role in processing affective information and storing affective memories (e.g., Cunningham et al., 2004, Lieberman, 2003 and Stanley et al., 2008). By contrast, reflective evaluations are posited to rely on declarative memory and brain areas such as the hippocampus and dorsolateral frontal cortex (e.g., Cunningham et al., 2005, Cunningham and Zelazo, 2007, Lieberman, 2003 and Lieberman et al., 2004). One unresolved question concerns whether the declarative memory representations used for controlled attitude evaluations rely on stores located in episodic or semantic memory. Characterizing attitudes as a type of self knowledge, some researchers have argued that they are stored in semantic memory along with other types of knowledge (e.g., Ochsner & Lieberman, 2001). According to this view, although attitudes are at least partially based on autobiographical and/or episodic memories about specific life events, they ultimately represent schematic versions of these events in which the specific contextual information associated with the memories has been lost. The result is a relatively stable memory representation, or “tag,” containing information related to the valence and/or the type of response associated with the particular attitude item (e.g., approach or avoid). In this view, the acontextual nature of these attitude representations is seen as more closely resembling the characteristics of semantic memory than those of episodic memory (Tulving, 1983). Empirical support for this idea has come from studies of another type of self knowledge, that regarding one's personality traits. These studies have demonstrated that amnesic patients retain at least some ability to acquire new trait knowledge, despite having a diminished ability to make new episodic memories (e.g., Kihlstrom, Beer, & Klein, 2002; see Klein, Cosmides, Tooby, & Chance, 2002 for a review; Klein & Loftus, 1996). Thus, these researchers concluded that the ability to create new trait knowledge is relatively independent of episodic memory. Finally, a recent account by Wood and Grafman (2003) takes a different view that emphasizes the role of frontal brain areas as the location of storage and processing of social and self knowledge. An alternate approach is to use brain imaging techniques, such as those providing measures of hemodynamic (e.g., positron emission tomography, PET; functional magnetic resonance imaging, fMRI) or electrical activity (e.g., ERPs), to study how healthy individuals make attitude evaluations. In one notable fMRI study, Zysset, Huber, Ferstl, and von Cramon (2002) compared the brain activity when participants made evaluative judgments (i.e., agree/disagree) about a series of short sentences (e.g., I like Leipzig) to when they retrieved autobiographical (e.g., I have been to Leipzig) or semantic memories (e.g., Leipzig is a large city). These researchers reported that their evaluative and autobiographical conditions produced largely overlapping patterns of brain activity that were significantly different from that found during semantic retrieval. Moreover, the evaluative-episodic differences that were found are best characterized as quantitative rather than qualitative. For example, both conditions produced activation in anterior medial frontal and precuneus regions, with evaluative judgments showing greater activity in the former and autobiographical retrieval showing greater activity in the latter. Zysset et al. (2002) interpreted their results as reflecting the differing contributions of self-referential processing (anterior medial frontal cortex) and autobiographical retrieval (precuneus cortex) in these two conditions. Semantic retrievals, by contrast, failed to show activation in either of these two regions. Nevertheless, semantic retrievals and evaluative judgments did show similar patterns of activity in left inferior prefrontal cortex, a result assumed to reflect the need for selection processes in both tasks (Thompson-Schill, D’Esposito, Aguirre, & Farah, 1997). Finally, all three conditions showed similar increases in activity in left and right inferior frontal cortex and occipital cortex. In sum, these researchers concluded that the processes used during attitude evaluations were more similar to those used for autobiographical retrievals than to those used for semantic judgments. The study did not, however, determine the nature of the memory representations for attitude evaluations. Determining where attitude representations are stored is likely to be difficult based on hemodynamic data alone given the disparity between the speed of attitude evaluations and the sluggishness of the hemodynamic response. For example, temporal characterizations of hemodynamic responses in attitude studies typically show brain activations lasting many times the 1–2 s it actually took to complete the evaluations in these studies (e.g.,Cunningham et al., 2003 and Zysset et al., 2002). Although the timing of evaluation processes is an important component of models of attitude evaluation, the temporal resolution of the fMRI method is insufficient to specify either the duration or relative timing of the processes involved. For example, the main component of Cunningham and Zelazo's (2007) attitude model is an iterative re-processing loop in which memory representations can be retrieved and modified repeatedly prior to reaching a final evaluation (Cunningham & Zelazo, 2007). In view of the fact that their iterative loop is posited to cycle at a rate of several times per second, hemodynamic methods are ill suited for testing a least this aspect of their model. The high temporal resolution of the ERP can, however, reveal details about the timing of the processes used to make controlled attitude evaluations. Further, given the substantial body of research showing that episodic and semantic retrieval elicit different patterns of ERP activity, the question of where these representations are stored can also be addressed. Episodic retrieval, for example, elicits a specific pattern of ERP activity, which is revealed when ERPs elicited by items not in episodic memory (i.e., “new”) are subtracted from ERPs elicited by items in episodic memory (i.e., “old”). The resulting old-new difference ERP consists of a series of subcomponents, collectively known as the “old/new” or “episodic memory” (EM) effect, each of which correlates with a different aspect of the retrieval process and has its own unique spatio-temporal characteristics (for reviews see Friedman and Johnson, 2000, Johnson, 1995 and Rugg and Curran, 2007). The parietal EM effect, which is most relevant here because of its association with recollective processes, appears as additional amplitude for old items that is superimposed on the late positive component (LPC). The parietal EM effect is maximal over left parietal scalp between 500 and 800 ms after stimulus onset (e.g., Johnson et al., 1985, Smith et al., 1993, Van Petten, 1995 and Wilding et al., 1995). Although the latency of the parietal EM effect increases with retrieval difficulty (Johnson et al., 1985, Johnson et al., 1998 and Johnson et al., 2005a), its scalp topography, which reflects activity of the underlying neural generators, does not vary as a function of either retrieval timing or type of retrieval (i.e., recognition or recall) (Donaldson & Rugg, 1999). It should be noted that the parietal EM effect reflects recollection of item information because retrieval of associated source information elicits ERP activity that is maximal over right frontal scalp in a later time window (see Friedman & Johnson, 2000). Finally, when verbal stimuli are used, the ERPs elicited by “new” words, which by definition are only in semantic memory, reflect the brain activity associated with semantic retrieval. Recent studies aimed at identifying the functional significance of the parietal EM effect have found that its amplitude reflects the amount of information retrieved (Vilberg et al., 2006 and Vilberg and Rugg, 2009b). Further, by combining data from two brain imaging methods, Vilberg and Rugg (2009b) provided ERP evidence to support the idea that the parietal EM effect reflects the online maintenance of recollected episodic memories and fMRI evidence to show that this ERP activity is generated in left inferior parietal cortex. After ruling out other processes that it might represent (e.g., reorientation of attention to retrieved information), these investigators suggested that the parietal EM effect may functionally be part of Baddeley's (2000) proposed “episodic buffer” (Vilberg and Rugg, 2008 and Vilberg and Rugg, 2009a). Despite some similarities, there are important differences between autobiographical retrieval and attitude evaluations that affect how the stimuli in these tasks are processed. These differences arise from the fact that each task has a different goal and thus autobiographical retrieval and attitude evaluations can be characterized as being “memory” and “working-with-memory” tasks, respectively (cf., Moscovitch, 1992). For example, the retrieval cues used in autobiographical memory tests are “associative” because they are sufficient to elicit a retrieval, which fulfils the task demands. Thus, given that only response selection and execution follow retrieval in memory tests, retrievals necessarily occur late in the processing sequence and are followed by a brief retrieval-response interval. By contrast, the stimuli used to elicit controlled attitude evaluations can be characterized as “strategic” because the memories they cue only provide a starting point for the evaluation processes that ultimately produce the required good/bad judgment. Therefore, retrievals in working-with-memory tasks occur early in the processing sequence and are followed by longer and variable retrieval-response intervals during which the evaluation processes are conducted. This combination of an early retrieval followed by a variable retrieval-response interval embodies the tenets of Cunningham and Zelazo's (2007), (Cunningham, Zelazo, Packer, & Van Bavel, 2007) Iterative Reprocessing Model of attitude evaluations. According to that model, the length of the evaluation process, and thus the duration of the retrieval-response interval, lies on a continuum, with reflexive evaluations requiring only a few iterations and highly reflective evaluations requiring more iterations and considerable controlled processing. The assumption here is that, although evaluations can be made reflexively, controlled evaluations are necessary when it is desirable to modify those reflexive responses to reflect the current context and goals. Based on results from previous ERP studies, the presence of differential retrieval-response timing for attitude evaluations and autobiographical retrievals was expected to have large effects on LPC timing and waveshape. That is, since Kutas, McCarthy, and Donchin (1977) seminal study, it has been known that the LPC is elicited just prior to the response and thus that LPC peak latency and latency variability increase directly with mean reaction time (RT) and RT variability, respectively. Consequently, the latency and shape of the LPC in stimulus-synchronized averages mirrors the timing and shape of the RT distribution, with LPC latency variability being evident as decreased peak amplitudes and increased peak breadth. Hence, to create a more accurate view of LPC activity across conditions varying in mean RT and RT variability, researchers calculate response-synchronized averages (see Johnson et al., 2003 and Johnson et al., 2005a for comparisons of stimulus- and response-synchronized LPCs in an episodic recognition task). Therefore, the type of average in which the LPC is more peaked (i.e., stimulus- or response-synchronized) reveals whether the processing is more time-locked to the stimulus or the response, respectively. Given that the LPC results in all previous studies were obtained in tasks containing a brief LPC-response interval, the question of whether stimulus- or response-synchronized averages provided the best basis for representing the results was determined solely on the basis of whether mean RTs were short or long, respectively. However, the situation is more complex in the present experiment because comparing results from a “memory” task with those from a “working-with-memory” task raises the possibility of large differences in the retrieval-response intervals associated with each. Given that autobiographical retrieval falls into the memory task category, their brief retrieval-response intervals means that their LPCs should replicate past results and be represented most accurately in response-synchronized averages. However, because attitude evaluations fall into the working-with-memory task category, the need for early retrievals means that the LPC/parietal EM effect should be more synchronized to stimulus onset. Thus, the relative appearance and timing of the LPC in the stimulus- and response-synchronized ERPs provides an important indication of retrieval timing. The use of random-by-trial presentations of “old” and “new” stimuli in episodic recollection paradigms means that ERP signatures for episodic and semantic retrievals, respectively, are generated in a single condition. The resulting patterns of ERP activity can be used as “templates” against which ERPs elicited by retrieval of attitude representations can be compared. However, if ERPs elicited by attitude evaluations resemble those for episodic retrieval, then it is also necessary to address the question of whether this ERP activity represents recollection of the attitude representation or recollection of an autobiographical memory associated with the particular attitude item. For example, seeing “chocolate” might cue, either automatically or deliberately, an associated autobiographical memory (i.e., of a particular time when one ate chocolate). In such cases, if attitude representations reside in an episodic store, then two episodic retrievals should occur following a single stimulus; an early retrieval for the attitude representation and a late retrieval for the associated autobiographical memory. Conversely, if attitude representations reside in a semantic store then there would be a single, late episodic retrieval for the associated autobiographical memory. Distinguishing between these two possibilities can be accomplished using the ERP technique because, on the one hand, access to these two memory stores produces different spatial patterns of brain activity while, on the other hand, retrieval of attitude representations and/or associated autobiographical memories would produce brain activity with different temporal characteristics. Although Tulving (1983) divided declarative memory into episodic and semantic stores, some disagreement remains regarding the extent to which autobiographical and episodic memory are separable. Some investigators have suggested that there is little difference between these two types of memory (e.g., Burianova and Grady, 2007 and Nyberg et al., 1996) whereas others have posited that autobiographical memory is a subsystem of episodic memory (e.g., Piefke, Weiss, Markowitsch, & Fink, 2005). In support of the similarity position, results from hemodynamic studies consistently show that autobiographical and episodic retrieval activate many of the same brain areas (e.g., Burianova and Grady, 2007 and Cabeza et al., 2004; see Cabeza and St Jacques, 2007 and Gilboa, 2004 for reviews; Piefke et al., 2005). Nevertheless, others focus on the differences and point to findings showing that autobiographical memory is much more self-relevant and has stronger links to perceptual and emotional processes than episodic memory (e.g., Gilboa, 2004). This distinction has been upheld in studies demonstrating that, compared to episodic retrieval, autobiographical retrieval produces greater activation in visual areas, which is presumed to represent the reactivation of the perceptual aspects of the experience (e.g., Cabeza and St Jacques, 2007, Conway et al., 2001, Conway et al., 2003 and Gilboa et al., 2004). For the purposes of the present paper, given these overall similarities, we will consider autobiographical memory to be subsumed within the larger episodic division of declarative memory.

The present results are consistent with the idea that the representations used to make controlled attitude evaluations are stored in episodic memory because they elicited the same parietal EM effect, an index of episodic recollection, as autobiographical retrievals. These results were also supported by the findings that semantic evaluations of the same attitude items produced a significantly different pattern of LPC activity. However, the timing of episodic retrieval was markedly different across conditions, with attitude evaluations characterized by early retrievals followed by an extended retrieval-response interval and autobiographical retrievals characterized by late retrievals and a very brief retrieval-response interval. This pattern of results for attitude evaluations conforms to the tenets of the Iterative Reprocessing Model and illustrates the basic difference between memory and working-with-memory tasks. The differential timing of processes in working-with-memory tasks was confirmed and shown to be domain general as a result of the finding that semantic evaluations were also characterized by early retrievals and extended post-retrieval processing. Thus, the present study demonstrates the value of using the ERP as a tool for studying the nature and timing of processes involved in evaluative judgments.