Available evidence suggests that the AMY is involved not only in successful encoding of basic emotional stimuli, but also in that of stimuli with social and personal relevance (Harvey et al., 2007; Kleinhans et al., 2007; Botzung et al., 2010a; Tsukiura, 2012; see also Adolphs, 2010; Cunningham and Brosch, 2012). For instance, among a sample of dedicated basketball fans, AMY activity was preferentially involved in successful encoding of highly emotional memories with personal significance, especially those regarding positively valenced plays (e.g., a player from the fan’s team making a shot; Botzung et al., 2010a). Interestingly, increased activity in the AMY was also identified during imagination of positive future events compared to negative ones (Sharot et al., 2007b; Sharot, 2011). This evidence suggests that basketball fans may be more likely to consider positive plays as more personally significant than negative ones, and thus this enhanced self-relevance could lead to increased AMY activity linked to encoding of positive personal events (but see Northoff et al., 2009). Moreover, there is also evidence suggesting that the AMY is involved in encoding social information (e.g., faces, impressions), regardless of its valence (Said et al., 2009; Schiller et al., 2009; Vrtička et al., 2012). Because of its intrinsic motivational value and importance for survival, it is possible that social relevance in the environment is detected by the AMY, irrespective of basic emotional properties, such as valence and arousal (Harvey et al., 2007; Vrtička et al., 2012). These findings are consistent with evidence identifying a more general involvement of the AMY in social cognition and behavior, in tracking the subjective significance or relevance in the environment based on the current goals (Adolphs, 2010; Cunningham and Brosch, 2012).

Besides the AMY, the involvement of medial/orbital PFC regions has also been implicated in encoding social relevance (Harvey et al., 2007; Vrtička et al., 2012; Gutchess et al., 2015; see also Meyer and Lieberman, 2012; Meyer et al., 2015). For example, increased activity in the medial PFC was related to enhanced memory for the information encoded with reference to oneself – i.e., the self-reference effect (Macrae et al., 2004; Yaoi et al., 2015) – and also to encoding of impressions of other people based on face-behavior associations (Mitchell et al., 2004; Gilron and Gutchess, 2012; Cassidy et al., 2013). In addition, activity in the medial orbitofrontal cortex (OFC) was related to encoding of faces signaling positive social cues (Tsukiura and Cabeza, 2008, 2011a,b). These findings are consistent with the role of medial PFC regions in representations of one’s own and others’ minds (Northoff et al., 2006; Wagner et al., 2012), as well as in encoding and integration of the subjective value of stimuli (Delgado et al., 2016). Moreover, increased activity in the medial OFC, along with its increased functional connectivity with the HC, have been observed during successful encoding of socially rewarding stimuli (i.e., smiling and attractive faces) (Tsukiura and Cabeza, 2008, 2011a). Conversely, encoding of socially negative stimuli (e.g., untrustworthy or unattractive faces) was mediated by increased activity in and connectivity between the Insula and the HC (see also Botzung et al., 2010a; Tsukiura et al., 2013). Finally, activity in the anterior temporal lobe (ATL) structures along with the ATL connectivity with the AMY, MTL, and medial PFC regions have been implicated in social emotional memory encoding, with the ATL playing a role in representing and storing person identity information (Tsukiura et al., 2010; Olson et al., 2013; Collins and Olson, 2014; Spreng and Andrews-Hanna, 2015).

Similar to the evidence regarding encoding, retrieval of social emotional memories has been linked to the involvement of the AMY, HC, and PFC. Increased AMY activity was associated with retrieval of faces that had previously been encoded with emotional descriptions of behaviors (Somerville et al., 2006), and with social fairness learned in the context of an economic game (Singer et al., 2004). Furthermore, increased activity in, and interactions between, the AMY and the HC have been associated with false recollection of episodes resulting from conformity to others’ behavior (Edelson et al., 2011), thus suggesting that social interaction could lead to long-lasting memory alterations through the engagement of AMY-HC mechanisms. Regarding the PFC involvement, the right medial PFC seems to mediate retrieval of information related to social contexts, while the left medial PFC seems to underlie retrieval of self-generated information (Mano et al., 2011), thus suggesting lateralization of social vs. self-referential processing during episodic memory retrieval within the right and left mPFC, respectively. Finally, also similar to encoding, increased activity in the ATL and its increased functional connectivity with the HC have been linked to retrieval of person identity information (Tsukiura et al., 2008, 2011; Collins and Olson, 2014).

The available evidence reviewed so far in this section suggests that encoding and retrieval of social information are associated with the involvement of brain regions typically implicated in basic emotion processing, social cognition, and memory. In integrating the differential contributions of these regions to the mechanisms underlying social emotional memory, Tsukiura (2012) has recently posited that the effect of emotional information on memory for faces may be mediated by interactions between affective (AMY, medial OFC, and Insula) and memory [HC, fusiform face area (FFA)] systems (Figure 1). In particular, the AMY detects general emotional properties of the faces, and interacts with the medial OFC and the Insula to process positive and negative signals from them, respectively. These regions of the affective system then modulate activity in the HC and the FFA, which have previously been implicated in encoding and retrieval of faces in general (Prince et al., 2009), to enhance memory for faces with affective information (Tsukiura, 2012). This framework is consistent with the notion that emotional memory processes are subserved both by bottom-up (e.g., AMY-HC) and top-down (e.g., MTL-PFC) mechanisms (LaBar and Cabeza, 2006; Dolcos et al., 2012), and also with an emerging account that the AMY serves as a functional hub across large-scale functional networks involved in various facets of social cognition and behavior (Bickart et al., 2014).

Moreover, evidence from recent studies examining large-scale functional networks of the brain suggests that complex processes involved in social emotional memory may be subserved by contributions from a network of regions whose activity represents a “default state” of the brain. These studies identified substantial spatial overlap between regions typically activated in tasks involving emotion processing, social cognition, and episodic memory, and those consistently activated when individuals are at rest – i.e., the default-mode network (DMN) (Schilbach et al., 2012; Spreng and Mar, 2012; Li et al., 2014). For instance, a recent meta-analysis identified the common involvement of the cortical midline structures [i.e., the dorsal subregion of the medial PFC (dmPFC) and precuneus] across studies examining emotion processing, social cognition, and the DMN, thus suggesting that introspective processes may be a common denominator across the three domains (Schilbach et al., 2012). Furthermore, extending this evidence to the memory domain, a recent study revealed that successful encoding of others’ faces in a self-referential manner (i.e., by imagining a potential friendship) was associated not only with increased activity in the (d)mPFC, but also with increased encoding-related functional connectivity between the cortical midline structures (i.e., the mPFC and precuneus) and HC (Yamawaki et al., 2017). Importantly, the dmPFC/precuneus and HC are considered part of the dorsomedial/core and medial temporal subsystems of the DMN, respectively (Spreng and Andrews-Hanna, 2015). Therefore, these findings suggest that complex memory encoding that involves self-referential processes in social context is associated with enhanced interactions among different functional subsystems within the DMN.

Finally, further expanding the traditional view of the DMN, an emerging view in the literature posits that the AMY, through its high intrinsic and task-evoked connectivity with some DMN regions, is part of a functional network involved in both maintenance of the brain’s default state and various socio-affective processes – i.e., the extended social-affective default network (Amft et al., 2015; see also Schilbach et al., 2008; Bickart et al., 2014). Taken together, emerging evidence suggests that social emotional memory is subserved by the complex interaction of functional systems/networks involved in a multitude of mental processes. A highly interactive nature of the neural mechanisms underlying social emotional memory reviewed here highlights the critical importance of examining measures of both neural activity and connectivity in future investigations on related topics.

Overall, recent functional neuroimaging evidence concerning the role of social information in emotional memory emphasizes the contributions of the AMY, HC, PFC, along with other regions involved in emotion processing and social cognition (e.g., Insula, ATL), to both encoding and retrieval of social emotional memories. The involvement of the AMY/HC and the medial PFC was linked to memory for positive stimuli with increased personal relevance, whereas the HC and Insula were linked to memory for negative social stimuli. Additionally, the AMY, medial PFC, and ATL were also associated with encoding and retrieval of social information, regardless of their valence, and these findings are consistent with evidence regarding the role of these regions in social cognition and behavior, in general. Emerging evidence from the investigations of large-scale functional networks has begun to reveal the relationships between the DMN and various social-affective and cognitive processes related to social emotional memory, with the midline cortical structures (dmPFC/precuneus) possibly serving as functional “hub” regions that integrate these processes. In sum, encoding and retrieval of social emotional information seem to involve complex interactions among both distinct and overlapping neural networks subserving basic emotion processing, social cognition, and memory processes, which may collectively allow us to integrate information from internal and external environments to adaptively navigate through the complex social landscape.

Available evidence suggests that the engagement of cognitive reappraisal during memory encoding enhances subsequent memory for the reappraised stimuli (Richards and Gross, 2000; Dillon et al., 2007; Liu et al., 2015), even following extended intervals of 1–2 weeks (Kim and Hamann, 2012; Ahn et al., 2015), whereas the engagement of suppression tends to impair memory for the suppressed items (Richards and Gross, 1999, 2000; Dillon et al., 2007; Dunn et al., 2009). One potential explanation for the enhancing vs. impairing effects of reappraisal and suppression on subsequent memory, respectively, concerns the involvement of semantic elaboration involved in these processes. More specifically, it has been suggested that reappraisal requires significant stimulus elaboration, which leads to a deeper level of encoding of and better memory for the stimuli. However, suppression inhibits the emotional experience and/or expression, and therefore leads to impaired subsequent emotional memory (Dillon et al., 2007).

Consistent with these behavioral effects, evidence from neuroimaging studies points to dissociable neural engagement linked to successful encoding of reappraised vs. suppressed items (Hayes et al., 2010; Binder et al., 2012). In particular, the memory enhancement by reappraisal was associated with increased co-activation of the HC and left lateral PFC (inferior frontal gyrus, IFG) (Hayes et al., 2010), whereas the memory impairment by suppression was associated with decreased activity in the HC and its decreased connectivity with the lateral PFC (Binder et al., 2012). More recently, the idea of enhanced subsequent memory by semantic elaboration (and by accompanying self-generated emotions) has been extended to encoding of neutral stimuli. Specifically, subsequent memory for neutral stimuli was significantly enhanced when participants encoded the stimuli by imagining emotional background stories compared to passively viewing them (Kaneda et al., 2017). Furthermore, encoding of neutral stimuli via semantic elaboration, compared to passive viewing, was linked to increased activity in the left IFG and dmPFC, and functional connectivity between these regions as well as between the left IFG and right HC predicted subsequent memory for the neutral stimuli encoded via emotional semantic elaboration (Figure 2). These findings suggest that top-down emotion generation by semantic elaboration can enhance memory for perceptually neutral stimuli, and that this effect seems to be mediated by similar HC-PFC mechanisms involved in the impact of reappraisal on encoding of emotional stimuli. Overall, the available evidence regarding the impact of ER on emotional memory encoding demonstrates that the dissociable effects of reappraisal and suppression are linked to opposing patterns of interaction (increased vs. decreased, respectively) among brain regions involved in basic memory processes (HC) and in higher-order cognitive control and emotion processing (PFC).

It has been known for a long time from the attention literature that emotional arousal tends to narrow attention to central cues, at the expense of peripheral, irrelevant cues (Easterbrook, 1959). Such prioritization of resources toward central aspects of information can result in enhanced memory for the central aspects, compared to the peripheral or contextual aspects that are attended less (Burke et al., 1992; Christianson, 1992; Buchanan and Adolphs, 2002). Therefore, emotional arousal or salience seems to weaken the integration of central aspects with peripheral contextual information into a unified memory representation (Mather, 2007). This phenomenon has also been referred to as the emotion-induced memory trade-off, in which memory is enhanced for the central emotional content of a stimulus but impaired for the associated neutral contextual information (Kensinger, 2009).

Among the various accounts for these opposing effects of emotion on memory (Mather, 2007; Kensinger, 2009; Chiu et al., 2013), one suggests that emotional arousal has dissociable effects on memory binding depending on certain aspects related to an event. In particular, Mather (2007) proposed that, whereas emotion can enhance memory for within-object features, it impairs memory for between-object features, the latter of which do not take advantage of emotional arousal (see also Kensinger, 2009 for a similar distinction between the intrinsic vs. extrinsic information). These assumptions were derived from a number of studies using either within-item or between-item binding tasks (reviewed in Chiu et al., 2013). For instance, memory was enhanced for specific intrinsic features, such as color (Doerksen and Shimamura, 2001; D’Argembeau and Van der Linden, 2004) or location (Mather and Nesmith, 2008; Nashiro and Mather, 2011) of emotional stimuli, compared to the same features being associated with neutral stimuli. By contrast, memory was impaired for extrinsic features and contextual details of emotional stimuli, when emotional objects were accompanied by neutral scenes (Kensinger et al., 2007a), when neutral objects were superimposed on emotional scenes (Touryan et al., 2007b), or when emotional scenes were associated with a colored frame (Rimmele et al., 2011).

Mather’s (2007) view regarding these opposing effects of emotion on memory has been recently postulated in the so-called Arousal-Biased Competition (ABC) Theory (Mather and Sutherland, 2011), which underscores the role of another important factor contributing to memory binding, namely, attentional priority. According to this account, emotional arousal enhances processing of stimuli with the highest priority (either due to bottom-up salience or top-down relevance), and impairs processing of those with lower priority (Mather and Sutherland, 2011). Thus, emotional arousal may enhance associative memory for features of high-priority items (e.g., color or location of an item) and impair memory for neutral items, when presented at the same time (or closely in time, see Sakaki et al., 2014) with emotional items. For instance, presentation of emotional stimuli enhanced or impaired memory for the preceding neutral stimuli, depending on whether these neutral stimuli were highly prioritized or not, respectively (Figure 4; Sakaki et al., 2014).

Importantly, the ABC theory also seems to account for previous behavioral evidence identifying enhanced memory for neutral words presented in the context of emotional words (Guillet and Arndt, 2009), or when neutral objects were placed on emotional scenes (Smith et al., 2004b, 2005). In general, memory should be impaired for neutral stimuli following an attention narrowing stimulus (Easterbrook, 1959) or a between-object feature (Mather, 2007). In these studies, participants were instructed to learn word-word associations (Guillet and Arndt, 2009) or to mentally integrate or connect objects and scenes during encoding (Smith et al., 2004b), which might have given such associations high “attentional priority,” resulting in facilitation of binding between emotional and neutral information. Taken together, these findings illustrate both enhancing and impairing effects of emotion on memory binding, and that these opposing effects can be explained by differences in attentional deployment toward, and prioritization of, emotional items and the associated features (Easterbrook, 1959; Mather, 2007; Kensinger, 2009; Mather and Sutherland, 2011).

An alternative (but not mutually exclusive) explanation for the opposing effects of emotion on memory binding has been recently proposed by Chiu et al. (2013), who posit that the enhancing or impairing effects may be linked to whether or not items are represented in an unitized manner in memory. This view was derived from a growing body of literature pointing to distinct MTL regions supporting item (perirhinal cortex) vs. relational (parahippocampal cortex/HC) memory representations (Cohen et al., 1999; Davachi, 2006; Ranganath, 2010). Previous fMRI studies (Haskins et al., 2008; Staresina and Davachi, 2010) have shown that the perirhinal cortex may also contribute to simpler forms of associative learning based on unitization, which involves representation of separate components (e.g., an object and its color or location) as a single unit (Graf and Schacter, 1989). Therefore, memory for items and “unitized” items can be mediated by similar mechanisms (perirhinal cortex), unlike memory representations that involve more complex associations, for instance, in temporal, spatial, and/or situational domains, which rely on the HC (Konkel and Cohen, 2009). This idea is also supported by recent ERP evidence showing differential ERP Old/New effects related to familiarity vs. recollection for high-unitized and low-unitized associations (Diana et al., 2011).

Based on the extant evidence, Chiu et al. (2013) suggest that emotion leads to enhancement of “item-only” or unitized memory representations, but it impairs more complex HC-dependent relational representations. Therefore, emotion leads to memory enhancement in tasks where the nature of the item-source association is more intrinsic (e.g., color or location) and therefore allows a single representation; however, it impairs memory when complex HC-dependent relational memory representations are required, for instance, in tasks using object-background associations (Kensinger et al., 2007a) or item-pairs (Mather and Knight, 2008). Importantly, when separate objects or features receive equivalent attention, for instance, by instructions to mentally integrate or connect certain items (Smith et al., 2004a,b; Ventura-Bort et al., 2016b), this can facilitate unitized processing and subsequent remembering. There is evidence showing that such integrative mental imagery relies more on amygdalar and parietal processing, and less on frontal and hippocampal processing, thus suggesting that unitized emotional associations may be less mediated by HC-dependent mechanisms (Murray and Kensinger, 2014; see also Ventura-Bort et al., 2016b for discussion of the involvement of familiarity-based recognition and valence). More research is required, however, in order to substantiate this view, given that this idea has not been empirically tested using systematic manipulations of unitization.

Overall, the enhancing and impairing effects of emotion on memory binding that have been identified in the literature so far seem to be explained by differences in attentional deployment toward emotional items and associated features during encoding, whether information is processed in a unitized or complex manner, or differences related to consolidation processes (retention retrieval). The available evidence suggests that emotion may facilitate memory for contextual details or other surrounding stimuli when viewed as intrinsic or unitized to the emotional event itself, via the involvement of the AMY and its interaction with perceptual regions (see also Mather et al., 2015 for the possible role of the locus coeruleus) that promote memory binding in the MTL regions. Successful retrieval of emotion-associated contextual information seems to be associated with increased activity in the AMY and memory-related MTL regions as well as enhanced electrocortical activity related to recollection.

Available evidence suggests that general personality traits, such as extraversion, neuroticism, and anxiety, as well as more specific traits indexing habitual engagement of ER, modulate the neural mechanisms underlying the impact of emotion across different processes, including perception, attention, emotional response, and memory (Canli et al., 2002b; Hamann and Canli, 2004; Touryan et al., 2007a; Hooker et al., 2008; Drabant et al., 2009; Eden et al., 2015). For example, extraversion has been linked to enhanced recall of positive memories (Mayo, 1983; Rusting, 1999) and more positive affective states following recall of positive AMs (Denkova et al., 2012). Extraversion has also been associated with a positive affective bias in memory, which in turn is associated with enhanced AMY engagement during memory encoding (Haas and Canli, 2008). In contrast, neuroticism is linked to enhanced recall of negative information (Bradley and Mogg, 1994), including negative personal memories, as shown by an increased frequency of rehearsing negative AMs in women (Figure 6) and greater proportion of recollecting negative AMs in men (Denkova et al., 2012). Furthermore, extant evidence suggests that neuroticism modulates the neural mechanisms that are altered by the impact of emotion on attention and memory (Ormel et al., 2013). The negative affective bias in emotional memory related to neuroticism might be due to increased activation of brain networks engaged during attentional processes, driven largely by greater recruitment of the AMY to negative information during memory encoding (Haas and Canli, 2008). Consistent with this idea, individuals with higher levels of neuroticism showed increased AMY and HC engagement when learning associations between fearful and neutral stimuli, which in turn was associated with enhanced memory for the neutral stimuli learned in association with the fearful stimuli (Hooker et al., 2008). Additionally, neuroticism was linked to reduced connectivity within the top-down executive network (Carballedo et al., 2015).

Trait anxiety is also linked to enhanced encoding and retrieval of negative stimuli (Russo et al., 2006) and modulates AM. Highly anxious individuals seem to have difficulties suppressing the effects of negative AMs, which can be observed in a decrease of the fading of emotional experience linked to recollection of negative personal memories over time (Walker et al., 2014). Furthermore, trait anxiety appears to modulate the neural mechanisms subserving emotional learning, including AMY’s engagement. For example, highly anxious individuals showed enhanced AMY response when exposed to items previously learned in association with negative emotional stimuli (Eden et al., 2015). Overall, this emerging evidence suggests that neuroticism and anxiety are associated with greater engagement of the bottom-up systems subserving emotional memory, leading to enhanced encoding and retrieval of negative associations. Additionally, trait anxiety is linked to reduced resting-state connectivity between the AMY and the dorsal attention network as well as to increased connectivity between the AMY and the ventral attention network (He et al., 2016). This evidence appears consistent with the attentional control theory (Eysenck et al., 2007), which proposes that anxiety modulates the dynamics between the top-down and bottom-up systems of attention in the brain.

Furthermore, available evidence suggests that habitual engagement of emotion control can also influence the effect of emotion on cognition. On the one hand, cognitive reappraisal, which is positively related to extraversion and negatively to neuroticism, is linked to more positive affective states and overall greater psychological well-being (Gross and John, 2003), as well as with a bias in remembering positive personal memories (Denkova et al., 2012). On the other hand, expressive suppression, which is negatively related to extraversion, is associated with increased negative affect and greater tendency toward anxiety symptoms (Llewellyn et al., 2013). Habitual engagement of suppression has also been associated with reduced confidence in memory accuracy and decreased experience of emotional and sensory details during autobiographical retrieval (Rubin and Seigler, 2004), as well as with increased recollection of negative AMs and enhanced negative affective states following retrieval of unpleasant memories in women (Denkova et al., 2012).

As alluded to in the previous section, extant evidence points to sex differences in emotion processing (Stevens and Hamann, 2012; Andreano et al., 2014) and in emotional memory (Cahill, 2003; Hamann, 2005; Andreano and Cahill, 2009). For example, compared to men, women remember more emotional personal/AMs (Seidlitz and Diener, 1998; Davis, 1999), and show enhanced overall brain response during encoding of emotional memories (Canli et al., 2002a). Previous research on sex differences in the underlying mechanisms of emotional memory also highlights a hemispheric asymmetry in the recruitment of AMY, with the left AMY linked to successful emotional memory encoding in women, and the right AMY linked to successful encoding in men (Cahill et al., 2001; Canli et al., 2002a; but see Fischer et al., 2007). There is also evidence that consolidation of emotional memory is altered by sex differences (Mackiewicz et al., 2006), and that the sex-related lateralization in the AMY associated with emotional memory encoding appears to also be influenced by sex differences in the stimuli being encoded. More specifically, there is evidence pointing to preferential engagement of the left AMY during successful encoding of female emotional faces in women, and to preferential engagement of the right AMY during successful encoding of male emotional faces in men (Armony and Sergerie, 2007).

The idea that women preferentially process same-sex stimuli is consistent with findings that women tend to have increased preference for socially relevant stimuli, in general (e.g., faces and people as opposed to scenes) (Proverbio et al., 2008a,b). This has led to the idea that feminine and masculine roles as developed by society, rather than the sex per se, contribute to the observed differences in emotional memory (Cahill et al., 2004) between women and men. This interpretation is consistent with findings showing sex differences in AMY activity associated with opposing effects in memory for peripheral vs. central details. Namely, the engagement of the left and right AMY was associated with enhanced consolidation of peripheral vs. central aspects of emotional events in women and men, respectively (Cahill and van Stegeren, 2003). Furthermore, there is also evidence that those participants with greater feminine traits recalled more the peripheral features of emotional information, whereas those with greater masculine traits recalled more the central aspects (Cahill et al., 2004). Overall, these findings demonstrate that femininity is linked to relatively enhanced vs. impaired recall of peripheral vs. central aspects of emotional information, respectively, with evidence that the former involves recruitment of the left AMY.

Emerging evidence also shows that women recall fewer positive and more negative AMs than men do (Denkova et al., 2012), although women and men show similar assessment of phenomenological properties (e.g., vividness and arousal) of their personal emotional memories (Young et al., 2013). Furthermore, available evidence suggests that recall of emotional AMs and post-retrieval mood is differently influenced by habitual ER, in women and men (Denkova et al., 2012). That is, habitually engaging suppression as a strategy in women may be inefficient and could come at the cost of overall enhanced recollection of negative AMs and greater post-retrieval negative mood (Denkova et al., 2012). At the neural level, retrieval of emotional AMs has been shown to be related to common and dissociable brain activity in women and men. More specifically, although women and men showed similar activation of the PFC, AMY, and memory-related MTL regions during retrieval of emotional AMs, remembering negative AMs was associated with greater engagement of brain regions such as the AMY, whereas remembering positive AMs was associated with greater engagement of lateral PFC, in women compared to men (Young et al., 2013). Overall, the evidence regarding sex differences in emotional memory is consistent with the idea that women perceive, express, and experience emotions to a greater extent than men do (Briton and Hall, 1995; Meyers-Levy and Loken, 2015), and with evidence highlighting both advantages and disadvantages associated with sex differences in emotion processing, in general. While there is evidence in support of enhanced emotional competence in women (Hall and Matsumoto, 2004; Montagne et al., 2005; Collignon et al., 2010), there is also evidence pointing to women’s increased vulnerability to affective disturbances (Nolen-Hoeksema, 2001; Kessler, 2003; Bekker and van Mens-Verhulst, 2007).

Healthy aging is characterized by both an overall preserved ability to process and remember emotional information (Phillips et al., 2002; Keightley et al., 2006; Mather, 2006; St. Jacques et al., 2009, 2010; Dolcos et al., 2014) and by an enhanced ability to spontaneously control emotion (Gross et al., 1997; Mather and Knight, 2005; Dolcos et al., 2014). These are reflected in the so-called age-related positivity effect – i.e., the tendency to enhance positive emotions and reduce negative emotions (Mather, 2016). Thus, examination of age-related differences in emotion processing and emotional memory also has the potential to help clarify the underlying mechanisms of emotional memory in mood and anxiety disorders, and to identify possible therapeutic interventions in these and other related disturbances.

In the memory domain, many studies have provided evidence supporting the positivity effect observed in healthy older adults, as reflected in their tendency to remember more positive and fewer negative memories for personal or laboratory events (e.g., Charles et al., 2003; Kennedy et al., 2004; Comblain et al., 2005; Schlagman et al., 2006; Fernandes et al., 2008; Spaniol et al., 2008; Ebner and Johnson, 2009; St. Jacques et al., 2009). At the neural level, the AMY-MTL and PFC have been highlighted as key regions subserving the memory-enhancing effect of emotion in aging, with the PFC mechanisms playing a crucial role (reviewed in Dolcos et al., 2012). For example, Murty et al. (2009) and St. Jacques et al. (2009) both identified reduced connectivity between the AMY and HC, paired with enhanced connectivity between the AMY and lateral PFC, associated with successful encoding of negative information in older compared to younger adults (Figure 7). Additionally, recent investigations of effective connectivity during emotional encoding identified both (1) greater positive modulation of MTL engagement by PFC, and (2) greater intra-regional connectivity within PFC (e.g., orbital, medial, and lateral), during the encoding of positive information in older than in younger adults (Addis et al., 2010; Waring et al., 2013). Finally, older adults who showed greater connectivity between AMY and mPFC at rest also showed (1) increased mPFC engagement during the encoding of emotional faces, and (2) a greater positivity effect in a subsequent memory recognition test (i.e., more likely to remember positive as opposed to negative faces), whereas these patterns were not shown in younger adults (Sakaki et al., 2013). Taken together, these findings emphasize an age-related strengthening in AMY-PFC connectivity that is overall linked to enhanced emotional memory encoding, and also associated with the positivity effect observed in older adults. Importantly, age-related decrease in the engagement of the AMY-MTL mechanisms coupled with the increased involvement of the AMY-PFC mechanisms in emotional memory is also consistent with the PASA (Posterior–Anterior Shift in Aging) model in the cognitive aging literature, characterized by reduced contribution of the posterior brain regions compensated by greater recruitment of the PFC regions (Davis et al., 2008).

In addition, extant behavioral findings suggest that older adults have greater susceptibility compared to young adults to the aforementioned memory trade-off effect between emotional/central vs. non-emotional/peripheral aspects of an event (Kensinger et al., 2005, 2007b; Nashiro and Mather, 2011). These age-related differences have been associated with older adults’ relatively decreased tendency to engage in particular encoding strategies (e.g., broad allocation of attention to contextual features of stimuli) that help younger adults reduce the negative influence of the trade-off effect. This suggests that older adults may have particular difficulty in disengaging attention from emotionally salient features of stimuli (Kensinger et al., 2005, 2007b). Furthermore, there is evidence that emotional arousal enhances memory for information about intrinsically linked contextual aspects (e.g., stimuli and their location) through memory binding in younger adults but not in older adults, thus suggesting that possibly limited cognitive resources in older adults may lead them to remember only the gist but not the associated details (Nashiro and Mather, 2011). This evidence is consistent with previous research documenting age-related impairments in inhibition and attentional control operations (reviewed in Fabiani, 2012), and various forms of associative memory (Old and Naveh-Benjamin, 2008). Finally, although evidence regarding the underlying mechanisms of age-related deficits in associative memory is currently limited, a recent investigation of the effect of age on memory processes found a significant link between associative memory performance and HC volume in older adults, consistent with the idea that this region is involved in representations of item-context relations (Henson et al., 2016).