Memory Control: A Fundamental Mechanism of Emotion Regulation

1Affective science has often sought to explain how external stimuli cause emotional reactions. Following from this, research on emotion regulation has emphasized (with notable exceptions,, ) the mechanisms that enable us to pursue our goals in the face of external emotional stressors. Understanding how people regulate exogenously elicited emotions (see Glossary) is, and should be, a fundamental goal of emotion regulation research. However, research on emotion regulation would benefit from a mechanistic understanding of an equally great – perhaps even greater – source of emotional states in our daily lives: memory. Outside the immediate experience of emotional events, any later impact – adaptive or maladaptive – that they have on our lives is mediated by memory. Like Proust’s madeleines, the triggers for this are often stimuli that most people would consider innocuous, but that by virtue of evoking an episodic memory in the individual have the power to elicit neural, physiological, and subjective reactions mimicking those of the original experience,. Thus, people can ruminate on past emotionally significant events, frequently leading to the reinstatement and perpetuation of (mostly negative) emotions from their past. Similarly, when people worry about their futures, they engage episodic prospection processes that can engender endogenous emotional reactions even though the eliciting events have not occurred. Such self-generated emotional states can be a major source of stress and are thought to play a central role in the etiology of psychopathology,. Indeed, this is likely the case even for disorders primarily characterized by aberrant reactions to stimuli, because even exogenously elicited emotions are robustly influenced by memory: emotional experiences are not mere reactions to the environment but are caused by both the properties of the stimulus and the context in which it occurs,,,,. This context consists not only of physical factors (e.g., whether one encounters a snake in a terrarium or on a forest path) but also one’s history with a stimulus and the expectations thus engendered; expectations that are rooted in internal predictive models,, which are themselves grounded in memories. Consequently, if it were possible to modulate the accessibility of memories, one could change either exogenously or endogenously generated emotional responses – in effect, regulating emotion,. Despite these illustrations of the important role of memories in shaping our emotional lives, how we regulate emotions by downregulating the memory traces that contribute to them has attracted comparatively little attention in research on emotion regulation.

2In this Opinion article we outline a view of the mechanisms supporting the regulation of mnemonic emotional material building on emerging findings from research on memory control (MC). Our starting point is the observation that, because our emotions are frequently driven by memories, controlling the accessibility of such memories to awareness should be an effective way of regulating one’s emotional reactions. Nevertheless, various factors have contributed to the view that attempts to voluntarily suppress thoughts and memories are unhelpful, and are perhaps harmful (Box 1). Contrary to this, we claim that controlling the occurrence and/or constitution of an upsetting memory – in short, MC – is the ultimate objective of successful emotion regulation of endogenous materials.A potential source of confusion about our argument is the tendency to equate DS with other, often counterproductive, suppressive phenomena. For example, in emotion regulation research ‘suppression’ usually refers to expressive suppression (ES), which involves inhibiting the behavioral expressions associated with emotional states (e.g., adopting a poker face to hide emotions). Owing to its small or even paradoxical effects on subjective emotion,, and because the trait tendency to use ES is associated with poor mental health outcomes, ES is often considered to be a maladaptive emotion-regulation strategy. However, DS clearly differs from ES in that it directly acts on the representation of the unwanted memory, rendering it less accessible. As such, DS is a fully cognitive strategy, involving different regulatory mechanisms.Another related line of research is thought suppression, as explored through the ‘white bear effect’. In this research participants are instructed to not think of a specific item (e.g., a white bear) over a period of 5 minutes, and report when it comes to mind. Research using this paradigm has shown that participants have difficulty avoiding thinking about the item, and this has been taken to show the inefficacy and counterproductive nature of thought suppression,. While DS conceptually resembles the ‘white bear’ variety of thought suppression, the experimental procedure used to investigate the latter has characteristics that make suppression inherently unlikely to succeed. Specifically, the white bear task makes explicit reference to a specific forbidden thought to be suppressed. However, achieving this stated goal is effectively impossible because simply remembering the purpose of the task requires that participants violate the task goal. This contrasts with the DS task, which does not integrate the specific memory/thought to be avoided as part of the task goal, rendering retrieval suppression possible. This ‘goal-integration theory’ may account for the apparent discrepancy between work on MC and the Wegner thought-suppression paradigm, and suggests that the white bear task might fail to capture the true utility of suppressive processes measured in work on MC. Consistent with this, a meta-analysis of 33 studies of psychopathology using the white bear task found no differences in suppression success or in rebound effects across patients and controls, raising concerns about whether this the task captures the mechanistic deficits that lead to intrusions in real life.Researchers have also investigated the relationship between thought suppression and psychopathology via the white bear suppression inventory (WBSI). Such work has contributed to the widely repeated conclusion that suppression, as a coping strategy, is associated with increased risk of psychopathology. Although the WBSI was intended to measure tendency to suppress thoughts, large-scale studies have now repeatedly found that the scale measures both the tendency to suppress thoughts and the degree to which they fail at doing so,,. Importantly, failure at, but not tendency to, suppress thoughts predicts psychopathology. Indeed, when successful thought-control ability is separately quantified it robustly predicts reduced symptoms of anxiety, depression, and other conditions characterized by intrusive symptomatology. These findings are consistent with the possibility that, in normative samples, thought suppression may play an important beneficial role in mental health.Alt-text: Box 1

3A potential source of confusion about our argument is the tendency to equate DS with other, often counterproductive, suppressive phenomena. For example, in emotion regulation research ‘suppression’ usually refers to expressive suppression (ES), which involves inhibiting the behavioral expressions associated with emotional states (e.g., adopting a poker face to hide emotions). Owing to its small or even paradoxical effects on subjective emotion,, and because the trait tendency to use ES is associated with poor mental health outcomes, ES is often considered to be a maladaptive emotion-regulation strategy. However, DS clearly differs from ES in that it directly acts on the representation of the unwanted memory, rendering it less accessible. As such, DS is a fully cognitive strategy, involving different regulatory mechanisms.

4Another related line of research is thought suppression, as explored through the ‘white bear effect’. In this research participants are instructed to not think of a specific item (e.g., a white bear) over a period of 5 minutes, and report when it comes to mind. Research using this paradigm has shown that participants have difficulty avoiding thinking about the item, and this has been taken to show the inefficacy and counterproductive nature of thought suppression,. While DS conceptually resembles the ‘white bear’ variety of thought suppression, the experimental procedure used to investigate the latter has characteristics that make suppression inherently unlikely to succeed. Specifically, the white bear task makes explicit reference to a specific forbidden thought to be suppressed. However, achieving this stated goal is effectively impossible because simply remembering the purpose of the task requires that participants violate the task goal. This contrasts with the DS task, which does not integrate the specific memory/thought to be avoided as part of the task goal, rendering retrieval suppression possible. This ‘goal-integration theory’ may account for the apparent discrepancy between work on MC and the Wegner thought-suppression paradigm, and suggests that the white bear task might fail to capture the true utility of suppressive processes measured in work on MC. Consistent with this, a meta-analysis of 33 studies of psychopathology using the white bear task found no differences in suppression success or in rebound effects across patients and controls, raising concerns about whether this the task captures the mechanistic deficits that lead to intrusions in real life.

5Researchers have also investigated the relationship between thought suppression and psychopathology via the white bear suppression inventory (WBSI). Such work has contributed to the widely repeated conclusion that suppression, as a coping strategy, is associated with increased risk of psychopathology. Although the WBSI was intended to measure tendency to suppress thoughts, large-scale studies have now repeatedly found that the scale measures both the tendency to suppress thoughts and the degree to which they fail at doing so,,. Importantly, failure at, but not tendency to, suppress thoughts predicts psychopathology. Indeed, when successful thought-control ability is separately quantified it robustly predicts reduced symptoms of anxiety, depression, and other conditions characterized by intrusive symptomatology. These findings are consistent with the possibility that, in normative samples, thought suppression may play an important beneficial role in mental health.

6Alt-text: Box 1

7More controversially perhaps, we further propose that MC mechanisms underlie the benefits of many of the volitional cognitive emotion regulation (CER) phenomena that have been identified, even when regulation is targeted at exogenously driven emotional reactions. Specifically, we claim that CER strategies can be classified by whether they employ two core mechanisms of MC that reduce access to memory traces that contribute to emotional responses. Importantly, we propose that the mechanisms of MC offer a novel account of how emotion regulation is implemented, that may help to understand its normal functioning and also psychopathological syndromes. This proposal is motivated by growing evidence that instructed MC abilities are related to beneficial outcomes (reviewed below) and the belief that similar mechanisms may support spontaneous MC, and therefore spontaneous CER, although further research will be necessary to test this assumption. As such, we propose that MC is crucial to regulating both mnemonically and exogenously elicited emotional states, and should be considered to be a fundamental process of volitional emotion regulation.

8MC can be defined as the capacity to volitionally influence the contents of thought in a goal-directed fashion by reducing the accessibility of memories. MC is often exerted reactively, in response to reminders that trigger the automatic retrieval of an unwanted memory or thought. Two mechanisms have been identified that enable MC when someone confronts an unwelcome reminder: (i) direct suppression (DS), involving the stopping or cancellation of the episodic retrieval process initiated by the cue, and the inhibition of the unwanted trace, and (ii) thought substitution (TS), involving the engagement of episodic retrieval, but instead being redirected towards alternative memories to occupy the limited focus of awareness, and to inhibit the to-be-avoided memory. Both of these mechanisms can be voluntarily deployed, and both induce forgetting of unwanted memories,,,, enabling motivated forgetting of both neutral and affectively charged memories,,,. Crucially, because these processes allow a person to avoid awareness of memories, they are sometimes mistakenly equated with other forms of cognitive avoidance that are often associated with poor psychological outcomes. The capacity to induce forgetting sets these mechanisms apart from conventional cognitive or behavioral avoidance: MC involves not simply avoiding reminders that trigger unwanted thoughts or memories but instead involves directly confronting those reminders and actively altering our cognitive response to them, and reducing the accessibility of the associated memory trace.

9DS of retrieval should not be confused with other forms of suppression that are often discussed in research on emotion regulation, and which have been argued to be ineffective and maladaptive (Box 1). Extensive evidence shows that suppressing retrieval in response to reminders is possible, and that doing so has persisting effects on the accessibility of suppressed traces,,,,,,,,,. DS begins when one encounters a reminder to an unwelcome memory or a thought. Such reminders are believed to trigger (owing to the affective qualities of the reminder, or of the unwelcome memory) inhibitory control to countermand the retrieval of the associated event, in a manner analogous to the way that control processes countermand motor processes to stop physical actions. Evidence indicates that repeated DS over multiple encounters with a reminder weakens the associated memory until it no longer intrudes,,,, ultimately impairing voluntary retrieval of the memory, consistent with an active memory-inhibition mechanism. Of course, memories vary in how amenable they are to suppression, and individuals vary in their ability to implement such suppression. Importantly, impaired DS is related to both subclinical traits associated with affective psychopathology (e.g., rumination and worry ), as well as pathology itself (e.g., post-traumatic stress disorder, PTSD, and depression ). Crucially, DS of aversive images can reduce one’s affective evaluation of those images at a later time (Figure 1). Interestingly, similar long-term effects arise for the CER strategy of reappraisal, raising the possibility that reappraisal engages regulatory mechanisms similarly to DS.

10One can also prevent recall of unwanted memories by diverting the retrieval process and recalling a substitute memory instead, preferably one that is innocuous or even positive,. Such TS may be especially useful when reminders are sufficiently powerful to make DS difficult, such as with traumatic memories or ruminative topics. Compared to DS, there has been less research on TS, but evidence suggests it is at least as effective as DS at causing voluntary forgetting,,. Interestingly, TS appears to cause forgetting even for individuals with impaired DS, such as those suffering from major depression, suggesting that they are supported by distinct mechanisms. In the context of CER, there is a conceptual resemblance between TS and the CER strategy of distraction,,. Although distraction can involve TS, distraction is a more general strategy that involves any reorienting of attention away from emotional stimuli to occupy awareness with non-emotional materials,. Although TS has a similar goal, it specifically involves the retrieval of alternative memories in response to the unwelcome reminders rather than merely a reorienting of attention. This generation process not only encodes an alternative path for retrieval, to follow in the future when that reminder is encountered, but also can inhibit the original memory, possibly by the mechanisms of retrieval-induced forgetting,. Thus, TS trains access to non-dominant associations for a given stimulus, in addition to occupying immediate awareness with distraction. For instance, if, upon seeing one’s ex-partner, one is reminded of upsetting memories about one’s lost love, one might instead recall the good times with the person, or even generate associations to new thoughts of a current, hopefully happier, love interest, thereby transforming one’s emotional reactions to the person,,. As we shall see later, this is also reminiscent of the cognitive operations supporting the CER strategy of reappraisal which, like TS, often entails the generation of an alternative representation of an unwanted percept or memory.

11DS and TS achieve forgetting of unwanted memories by distinct although partially overlapping neural mechanisms (Figure 1;, for detailed discussion). DS engages what may be a domain-general inhibitory control mechanism supported by the (primarily right) dorsolateral prefrontal cortex (dlPFC), that, in the context of retrieval suppression, dynamically interacts with the hippocampus to suppress its activity and disrupt retrieval,, (Figure 2A). This suppressive effect is especially pronounced when memories involuntarily intrude into awareness and need to be purged,. The pathway by which hippocampal activity is suppressed by prefrontal regions is presently unclear, although primate neuroanatomy suggests that it could involve anterior cingulate cortex (ACC) circuits central to top-down regulation. Whatever the pathway may be, evidence indicates that the concentration of γ-aminobutyric acid (GABA) in the hippocampus predicts individual differences in how effectively one forgets via DS, suggesting that GABAergic interneurons implement the proximate mechanism that suppresses activity in the hippocampus, and, by extension, supports DS. Interestingly, this suppressive mechanism appears to have a widespread effect on the retention of all recent events that depend on hippocampal activity for stabilization, inducing an amnesic shadow for events in the period before and after suppression. TS, by contrast, does not cause this shadow, again suggesting that its mechanisms are distinct. Because suppression reduces intrusions over repeated encounters with reminders, the need for prefrontal control declines adaptively, and more so for individuals who are better at suppression.

12Suppressing intrusive thoughts is not solely about modulating hippocampal activity and explicit memory, however. An important feature of DS is that it also suppresses regions outside the hippocampus that are involved in representing the specific content of a given memory, including, for example, fusiform cortex and parahippocampal cortex for objects and places, respectively. Correspondingly, DS disrupts implicit expressions of memory such as perceptual priming that are believed to rely on these neocortical representations,,. The need for inhibitory control to target both the hippocampus and neocortical regions is thought to arise because the unwelcome reminder elicits pattern completion of a memory in the hippocampus, in turn triggering re-entrant activation of cortical regions representing mnemonic content,,. Suppressing awareness, consequently, entails parallel targeting of inhibition in neocortical regions expressing this onset of activity. This reinstatement principle extends to emotional memories: during memory intrusions, DS of aversive scenes reduces activity in the amygdala and vmPFC – structures that are closely associated with the generation of subjective affect,,. Importantly, DS of memories of aversive scenes attenuates negative emotional judgments about those scenes when they are later encountered (Figure 1). This suggests that MC mechanisms, first identified outside the context of emotion regulation using purely neutral stimuli, can be deployed both to suppress mnemonic traces of stimuli as well as to directly suppress the affective connotations of the stimuli themselves. Indeed, this suggests that endogenously focused CER can be seen as a special case of the general reinstatement principle introduced above, such that suppression of affective connotations is a direct consequence of stopping the retrieval and neural reinstatement of affective mnemonic traces. If this account is correct, DS of affective material will, in effect, be an act of emotion regulation.

13In contrast to the strong right-lateralized activations observed for DS, TS is predominantly left-lateralized, centered on caudal prefrontal (cPFC) and ventrolateral prefrontal (vlPFC) cortex (Figure 2). The vlPFC supports selective retrieval, especially under conditions of high retrieval competition,, and coupling between cPFC and vlPFC predicts the amount of forgetting elicited by TS. Whereas the functional significance of this coupling remains unknown, one plausible interpretation is that cPFC entrains vlPFC to drive retrieval of alternative memories associated with the cue to supplant the dominant, to-be-avoided, memory. Supporting this, TS does not reduce hippocampal activity but can even increase it, especially under conditions of high competition, suggesting that the vlPFC interacts with the hippocampus to facilitate retrieval of the non-dominant memory. The pathway underlying this prefrontal–hippocampal interaction during TS has not been established, but related data suggest that it may take place via the uncinate fasciculus, which has been shown in humans and non-human primates to link these regions to the anterior hippocampus and to be related to controlled semantic retrieval,,.

14Taken together, these findings indicate that TS and DS constitute distinct routes to achieving MC, supported by distinct cortical networks, that exert opposing influences on hippocampal activity and different functional aftereffects. Importantly, the fact that they appear to be supported by separable mechanisms suggests that it may be possible to exert superior control over memories by using DS and TS in conjunction. In the following section we argue that this may be what occurs when people engage in CER using the ‘gold-standard’ emotion-regulation strategy of reappraisal.

15Reappraisal is by far the most-studied CER technique, and has been shown to be effective at both reducing negative and increasing positive emotion. Reappraisal involves an effort to re-evaluate a stimulus to alter its emotional impact,,,, most commonly by changing the narrative associated with a stimulus,. For instance, when confronted with blood and gore evoking strong visceral reactions of disgust, one can remind oneself that one is in a movie theatre watching a horror film, implying that the distressing images are almost certainly fake. Alternatively, if one happens to want the full horror-film experience, one might suppress thoughts of being in a cinema and instead focus on the narrative portrayal that led to the carnage, which would likely amplify one’s emotional reactions. Thus, by accessing different aspects of the information about an emotional event, one can alter its interpretation, reducing or increasing its emotional impact. Notably, while most frequently studied in the context of exogenous emotional reactions, reappraisal can be used to regulate both stimulus-based and memory-based emotional reactions,,,, meaning that it has a similar remit to MC.

16Mechanistically, reappraisal is best viewed as a complex process supported by multiple cognitive functions. Although there are many reasons why reappraisal might succeed in regulating emotional responses, we suggest that a key reason is that it relies upon the two MC mechanisms discussed above to reduce the current and future likelihood of retrieving representations of distressing interpretations. Specifically, we suggest that, as the reappraisal process unfolds, it involves (i) the initial DS of the representation of the primary affective narrative, and (ii) the generation of a substitute narrative in response to the stimulus, constituting a clear case of TS. Importantly, we suggest that this underlies not only simple reappraisals that may interfere with retrieval of primary emotional narrative but also more elaborate reappraisals that transform the meaning of the emotional event itself. Thus, for instance, in the preceding example in which emotional reactions to a horror movie were regulated, reappraisal plausibly involves (i) suppressing the primary emotional narrative of the external stimuli. and (ii) focusing on details of the scene or context that enable search of memory for an alternative interpretation (e.g., the blood is only ketchup) that can then substitute the primary narrative. Both of these are MC operations, and together they should reduce awareness of the immediate aversive interpretation and orient affective responses towards a more innocuous narrative as reappraisal takes place.

17A key finding distinguishing reappraisal from other forms of CER is that it has lasting effects on subjective responses to affective stimuli. This is explained by the current account because TS associates novel information generated during the reinterpretation process with the stimulus or memory. Thus, subsequent presentations of the stimulus should also evoke retrieval of the alternative interpretation, in the same way as substitute thoughts block the retrieval of unwanted memories in studies of MC. We argue, therefore, that the lasting effects of reappraisal on affective responses to stimuli or to memories can be properly regarded as motivated forgetting achieved via MC, possibly leveraging fundamental affective memory updating phenomena such as reconsolidation (Box 2).A phenomenon that has received increasing attention as a way to enact memory-based emotion regulation is reconsolidation,,. Research on reconsolidation suggests that, when both emotional and non-emotional memories are retrieved, they exhibit increased lability, leaving the memory trace vulnerable to interference. This retrieval-induced plasticity is thought to be important for enabling old memories to be updated by new information, but comes at the cost that retrieved memories require restabilization or reconsolidation to be available for retrieval at a later date. By interfering with this reconsolidation process using behavioral or psychopharmacological interventions, evidence suggests that it is possible to erase memory traces, including emotional threat memories in humans. Moreover, in the lability period it is also possible to update or alter memories. For example, a negative memory may be changed to be less negative by integrating new information, as is seen in extinction, a process that has been argued to be central to reappraisal, and arguably to CER efforts in general, to the degree that they have lasting changes on the affective reactions subsequently elicited by a memory.The conditions leading to forgetting in research on MC resemble those thought to be crucial for reconsolidation phenomena. For instance, like reconsolidation interference, MC also involves altering the retrieval process, with both phenomena eliciting forgetting. Inhibitory control may act like an endogenous ‘amnestic agent’ that disrupts reconsolidation. Interestingly, suppression-induced forgetting (SIF) arising via DS has been linked to actively inhibiting the memory from awareness when it intrudes,, indicating that the memory has already been (at least partially) retrieved. This suggests that SIF could involve interfering with the reconsolidation of memory traces. Similarly, SIF produced by TS (at least as we propose it to be employed during reappraisal) might involve reconsolidation mechanisms to ‘overwrite’ a primary emotional narrative. In line with this, it has been shown that reappraisal, but not distraction, has durable effects on neural reactions to emotional stimuli, albeit in the context of what is effectively consolidation of emotional memories following initial presentation and regulation of reactions to emotional stimuli. Because memory disruption effects are observed immediately, it is clear that mechanisms other than reconsolidation are also at play in memory inhibition phenomena. Nevertheless, for CER of endogenous emotional reactions, it is plausible that reconsolidation interference and updating also play a role, although specific research to explore this topic is needed.Alt-text: Box 2

18A phenomenon that has received increasing attention as a way to enact memory-based emotion regulation is reconsolidation,,. Research on reconsolidation suggests that, when both emotional and non-emotional memories are retrieved, they exhibit increased lability, leaving the memory trace vulnerable to interference. This retrieval-induced plasticity is thought to be important for enabling old memories to be updated by new information, but comes at the cost that retrieved memories require restabilization or reconsolidation to be available for retrieval at a later date. By interfering with this reconsolidation process using behavioral or psychopharmacological interventions, evidence suggests that it is possible to erase memory traces, including emotional threat memories in humans. Moreover, in the lability period it is also possible to update or alter memories. For example, a negative memory may be changed to be less negative by integrating new information, as is seen in extinction, a process that has been argued to be central to reappraisal, and arguably to CER efforts in general, to the degree that they have lasting changes on the affective reactions subsequently elicited by a memory.

19The conditions leading to forgetting in research on MC resemble those thought to be crucial for reconsolidation phenomena. For instance, like reconsolidation interference, MC also involves altering the retrieval process, with both phenomena eliciting forgetting. Inhibitory control may act like an endogenous ‘amnestic agent’ that disrupts reconsolidation. Interestingly, suppression-induced forgetting (SIF) arising via DS has been linked to actively inhibiting the memory from awareness when it intrudes,, indicating that the memory has already been (at least partially) retrieved. This suggests that SIF could involve interfering with the reconsolidation of memory traces. Similarly, SIF produced by TS (at least as we propose it to be employed during reappraisal) might involve reconsolidation mechanisms to ‘overwrite’ a primary emotional narrative. In line with this, it has been shown that reappraisal, but not distraction, has durable effects on neural reactions to emotional stimuli, albeit in the context of what is effectively consolidation of emotional memories following initial presentation and regulation of reactions to emotional stimuli. Because memory disruption effects are observed immediately, it is clear that mechanisms other than reconsolidation are also at play in memory inhibition phenomena. Nevertheless, for CER of endogenous emotional reactions, it is plausible that reconsolidation interference and updating also play a role, although specific research to explore this topic is needed.

20Alt-text: Box 2

21If our MC hypothesis of reappraisal is correct, reappraisal should recruit the neural systems supporting DS and TS. Figure 2 summarizes evidence on the neural bases of DS, TS, and reappraisal,. Crucially, the neural correlates of DS and TS were identified via simple paradigms involving MC of neutral stimuli, often simple word pairs. These methods were designed to isolate the fundamental mechanisms that enable a person to stop or redirect episodic retrieval. As such, overlap with reappraisal can be interpreted parsimoniously as the latter process engaging fundamental MC mechanisms. Figure 2C illustrates the overlap between the neural systems involved in reappraisal and DS (blue outline), TS (green outline), or both DS and TS (black outline), demonstrating that the brain regions engaged during these processes overlap considerably, with all major nodes of the reappraisal network being part of the MC networks. Notably, left lateral nodes of the reappraisal network are predominantly associated with TS, with the exception of inferior frontal/anterior insula, as well as the parietal nodes, which are also associated with DS. In the right hemisphere, there is extensive overlap between TS and DS, as well as with reappraisal nodes. This could indicate a hitherto unexplored involvement of DS processes in TS, plausibly related to an initial suppression of recall to better implement TS, similar to the role we propose for DS in reappraisal. Taken together, these observations are consistent with the proposal that the mechanisms supporting DS and TS constitute distinct component processes of reappraisal. It must be made clear, however, that this is a reverse inference; future work will need to experimentally manipulate the putative involvement of TS and DS in reappraisal and relate these findings to TS and DS abilities in more conventional MC tasks. If our account is correct, one would expect DS ability to be associated with the capacity to use reappraisal to inhibit emotional reactions, but not necessarily to the capacity to supplant them with alternative interpretations (as in e.g., positive reappraisal, ), which should instead be correlated with TS performance.

22Beyond its role in reappraisal, a link between MC and CER more broadly is suggested by the strong association between perseverative cognition and affective psychopathologies. Although psychopathologies cannot be reduced to single-factor explanations, they are frequently related to difficulties regulating emotional states elicited by memories or thoughts. The clearest examples of this come from anxiety disorders – that tend to be characterized either by worry about the future, or, as in the case of PTSD, by overwhelming emotional memories of past events. Indeed, both anxious worry and PTSD have been associated with MC deficits. Moreover, it has been shown that suppressing future worries engages MC networks to reduce activation in the hippocampus, amygdala, and vmPFC, presumably to interrupt episodic prospection. Importantly, the capacity to reduce apprehensiveness about feared events via suppression was modulated by trait levels of anxiety, supporting the notion that MC deficits play a part in anxious symptomatology. Similarly, depression is associated with elevated rumination, a problem of persisting recurring thoughts or memories that occur in a context-inappropriate fashion. Rumination is thought to be caused by an inability to disengage from endogenously generated negative autobiographical memories, raising the possibility that depression is associated with MC difficulties. As this observation suggests, depression is associated with both neural and behavioral abnormalities in MC, and the evidence also indicates that training MC might be an effective (adjunct) treatment for managing depression,. Conversely, good trait MC may impart resilience to emotional stressors, as seen in a recent study showing that behavioral and electrophysiological metrics of DS predict the number of intrusions experienced by participants in the week following viewing of a strongly negative film. Similarly, the ability to engage in motivated forgetting using DS is related to the ability to forgive others for their transgressions. Experience of contending with intrusive memories may also improve MC and future resilience: in healthy samples, retrieval suppression ability increases with the degree of adversity someone has experienced. Thus, at least for DS, the evidence suggests a broader link between MC abilities and the capacity to limit the impact of negative events on one's affective constitution, which could be conducive both to CER and to resilience.

23Adopting a MC perspective may improve our understanding of the neurocognitive mechanisms supporting CER in general. Although a strong case can be made for thinking of reappraisal as a MC phenomenon, generalizing this proposal to CER strategies more broadly is difficult owing to lack of data on other regulation strategies. For example, a recent meta-analysis found that fivefold more neuroimaging experiments had studied reappraisal versus all the other strategies combined. This problem is compounded by the wide variety of CER strategies proposed, with little agreement on protocols for studying them, much less for comparing their mechanisms of action.

24One approach to exploiting what is known about the architecture of MC would be to characterize CER strategies according to the involvement of MC mechanisms. In Table 1 we offer an initial conjecture on how MC might play a role in a range of common emotion-regulation strategies (adapted from ), suggesting potential contributions of DS and TS to each. This list is not intended to be exhaustive, but instead aims to provide some initial testable hypotheses: if, as we suggest, MC processes contribute to emotion regulation, individual differences in DS and TS ability should predict the capacity and/or preference for deploying different CER techniques. At present, it is unknown how strongly correlated DS and TS abilities are, but, given the partially dissociable neural systems underlying each, and given the fact that TS may be easier for people deficient in inhibitory control, the ability to wield these mechanisms may be at least partially distinct. Individual differences in DS and TS should predict differing engagement of their respective brain networks during reappraisal. Thus, by investigating reappraisal implementations and other forms of CER (Table 1) that could engage MC processes, the proposed role of the MC architecture could be tested. Moreover, as discussed above, efficacy at implementing DS or TS should predict the specific implementations of reappraisal that people spontaneously adopt: people with strong DS abilities may find it easier to dismiss their appraisals, whereas those lacking DS ability may fall back on TS, preferring to seek new appraisal frames. Assessment of MC abilities could therefore serve to guide targeted interventions aimed at either (i) improving deficient CER abilities, perhaps via training, or instead (ii) tailoring interventions to capitalize on the existing strengths of each individual.

25Aside from understanding the broad component processes of CER strategies, the MC literature offers novel hypotheses about the neural machinery of emotion regulation. For example, research on MC suggests that CER research should investigate prefrontal influences on the medial temporal lobe (MTL) more broadly, and not simply on the amygdala. If the initial suppression of an appraisal engages DS, for example, hippocampal activity may be suppressed, contributing to successfully reorienting to and encoding a novel appraisal. Similarly, the evidence discussed above that DS efficacy is related to differences in local inhibitory circuits within the hippocampus, suggests that there could be a connection between hippocampal GABA and the ability to implement CER techniques that rely on suppression of emotional memories. Parallel GABAergic mechanisms might be important in forms of CER that emphasize amygdala inhibition and the suppression of negative affective states, but might be less important for CER focused on enhancing positive emotion.

26Emotion regulation is the ability interrupt or alter the generation of emotional states. Because many emotions we experience in daily life stem from recall of emotionally charged memories, the capacity to control whether and when such recall occurs must be a central emotion-regulation mechanism. In this article we have explored how this mechanism might be employed and support different forms of CER. Although we hope to have offered a convincing argument in support of this perspective, there remains a great deal of work to be done (see Outstanding Questions) to map out how MC supports emotion regulation. For instance, while we believe that MC capacities play a central role in CER, there are several other factors (e.g., working memory capacity, emotional constitution, and executive control abilities aside from inhibition, to name but a few) that influence the capacity to engage in CER. An important topic for future research will be to map out how such factors and MC abilities for both emotional and non-emotional memories interact in determining individual strengths and weaknesses in CER. Importantly, however, we predict that MC will emerge as a key factor because of the central role of memory in emotional processing. Ultimately, therefore, we believe that an understanding of the human ability to surmount the compulsions of our past must be grounded in a mechanistic account of how our memories of that past are controlled.To what degree are the psychological and neural mechanisms supporting affective control coextensive with those supporting general MC?Can the emotion associated with an event be inhibited without affecting its mnemonic content? Does inhibiting an emotion fundamentally require inhibition of the thoughts and memories that precipitated it?Do differing approaches to MC vary in whether the forgetting they produce leaves emotional responses intact and free to influence behavior?Do emotional or high arousal stimuli/memories require special regulatory mechanisms to control, or are they on a continuum with other non-emotional thought contents?Do MC abilities predict preferences for, and the efficacy of, different emotion-regulation strategies?Does training someone on MC improve their emotion-regulation ability?Does reappraisal induce forgetting of stimulus or event details that are not consistent with the new narrative generated?Are MC abilities similarly important for explicit and implicit forms of emotion regulation?

27To what degree are the psychological and neural mechanisms supporting affective control coextensive with those supporting general MC?

28Can the emotion associated with an event be inhibited without affecting its mnemonic content? Does inhibiting an emotion fundamentally require inhibition of the thoughts and memories that precipitated it?

29Do differing approaches to MC vary in whether the forgetting they produce leaves emotional responses intact and free to influence behavior?

30Do emotional or high arousal stimuli/memories require special regulatory mechanisms to control, or are they on a continuum with other non-emotional thought contents?

31Do MC abilities predict preferences for, and the efficacy of, different emotion-regulation strategies?

32Does training someone on MC improve their emotion-regulation ability?

33Does reappraisal induce forgetting of stimulus or event details that are not consistent with the new narrative generated?

34Are MC abilities similarly important for explicit and implicit forms of emotion regulation?