Putting Emotional Memories in Context: The Constructionist Model of Emotional Memory

Despite the generalizations made in the above section, which were necessarily brief, it is important to acknowledge that investigations of the mnemonic effects of valence and arousal have given rise to a rich and nuanced line of scientific inquiry, as exemplified by the many theoretical perspectives that exist in this space (Bowen et al., 2018; Mather & Sutherland, 2011; Talmi, 2013; Yonelinas & Ritchey, 2015). Accounts such as Bowen et al.’s NEVER model (negative emotional valence enhances recapitulation; Bowen et al., 2018), for example, argue that rather than having identical effects on memory, positive and negative stimuli of comparable arousal may impact certain stages of memory differently, highlighting the possibility that important valence asymmetries may exist depending on which aspects of memory are tested (for additional discussions of possible valence asymmetries, see Kensinger, 2009a; Kensinger & Ford, 2020; Kensinger & Schacter, 2016; Larson & Steuer, 2009; for additional discussion of how emotion affects memory consolidation and retrieval, see Kensinger & Ford, 2020; Ritchey et al., 2008, 2013; Yonelinas & Ritchey, 2015). Additionally, and as alluded to earlier in our discussion of memory trade-off effects, perspectives such as Mather et al.’s arousal-biased competition theory emphasize that emotional arousal does not always result in a straightforward enhancement of memory but instead enhances the perception and memory of some event details at the expense of others (Mather & Sutherland, 2011), and that under certain testing conditions (e.g., when viewing degraded line drawings of previously seen images), may even result in worse memory for emotional relative to neutral objects (Faul et al., 2025). As yet another example, recent research has expanded on past findings of emotionally enhanced vividness by demonstrating that emotion has dissociable effects on the reconstruction of visual properties such as color, contrast, and hue (Faul et al., 2025). Furthermore, there are almost certainly exceptions to the generalizations made in the preceding section of our review; the effects of valence and arousal outlined above may not always co-occur (e.g., in some contexts, affect-related memory effects may be uncorrelated with affect-related attention effects; Monkman et al., 2025). Thus, the goal of our model is to generate novel research questions by drawing on ideas from affective science that are complimentary and orthogonal to existing ideas within this literature; we do not aim to encompass or replace the many nuanced perspectives within the field of emotional memory that have been invaluable to scientific progress within this area. Although it is clear that understanding the mnemonic effects of valence and arousal is essential to understanding emotional memory, we argue that this approach can be supplemented by drawing upon theories from affective science that provide mechanistic accounts of why valence and arousal, as well as categorical emotion states, might influence memory. In this regard, the CMEM is particularly novel in its predictions regarding the mnemonic effects of discrete emotions such as anger, disgust, and fear.

In addition to the studies referenced here on the effects of valence and arousal on episodic memory, a smaller number of studies have investigated the mnemonic effects of specific discrete emotions (e.g., fear). The term discrete emotions is often used to refer to affective states that are experienced or perceived as bounded in time and place; linked to specific eliciting events; associated with specific types of meaning-making, and perhaps behaviors; and sometimes included in English-language categories such as “fear” or “afraid” (see Hoemann et al., 2025; Lindquist, 2013; Russell, 2003). For example, Chapman et al. (2013) found that participants had better memory for normatively disgusting images compared with normatively frightening images on tests of both recall and recognition. This finding has since been documented by others and termed the disgust advantage (Chapman, 2018; Chapman et al., 2013; Ferré et al., 2018; Marchewka, Wypych, Michałowski, et al., 2016; Moeck et al., 2021; Riegel et al., 2022; West & Mulligan, 2021, Experiment 3). Others have found that experiencing sadness is associated with increased memory for situational outcomes (Levine & Burgess, 1997; Levine & Edelstein, 2009; Levine & Pizarro, 2004), whereas experiencing anger is associated with increased memory for agents’ goals when memory is tested for narratives (Levine & Burgess, 1997). Thus, in addition to emotion affecting memory through valence and arousal, there is some evidence that discrete emotions affect memory as well.

The disgust advantage has largely been interpreted from the perspective of basic-emotion theories, according to which specific discrete emotions (e.g., fear, anger, disgust, etc.) represent distinct patterns of coordinated behavioral and physiological response in reaction to events within the environment (Ekman, 1999; Ekman & Cordaro, 2011; Tracy, 2014). It is important to note that these emotions are claimed to be basic in that they are thought to have evolved (via strictly genetic transmission during brain evolution; see Lindquist et al., 2022) to allow our ancestors to respond to recurring environmental challenges in adaptive ways. Furthermore, basic emotions are said to be universal across human cultures and are thought to be associated with unique physiological and behavioral response patterns. Although there is disagreement about which specific emotions are considered basic (Ortony & Turner, 1990), the ones most commonly agreed upon are fear, disgust, anger, sadness, happiness, and (sometimes) surprise (Ekman & Cordaro, 2011; Lindquist et al., 2012; Siegel et al., 2018).

Explanations of the disgust advantage draw on basic-emotion theories in that such explanations propose that disgusting and frightening stimuli differ in their memorability because of differences in the evolutionary functions served by these emotions. For example, Chapman (2018) theorized that disgusting information may be particularly memorable because, relative to physical threats that evoke fear, disease-related (i.e., disgusting) threats may be more subtle in nature and may require additional attention to evaluate. This additional attention at encoding, it is reasoned, may explain why disgusting stimuli are ultimately better remembered than frightening stimuli (but see Chapman et al., 2013; Moeck et al., 2021). Similarly, neuroimaging researchers have drawn upon basic-emotion theories when interpreting the neural correlates of disgust-related memory (Marchewka, Wypych, Michałowski, et al., 2016; but see Marchewka, Wypych, Moslehi, et al., 2016).

Causal-appraisal theories of emotion have also been used to interpret the effects of discrete emotions on memory (Kaplan et al., 2012, 2016; Levine & Burgess, 1997; Levine & Edelstein, 2009; Levine & Pizarro, 2004). Briefly, causal-appraisal theories posit that emotions result from one’s interpretations (i.e., appraisals) of the current situation, with specific patterns of appraisals resulting in specific discrete emotions, which in turn have specific and consistent effects on behavior and physiology (Roseman & Smith, 2001).³ For example, according to causal-appraisal theories, sadness results from the appraisal that something important has been lost and encourages passivity, whereas fear is thought to result from the appraisal that the current situation is dangerous and encourages avoidance (Moors, 2014; Roseman & Smith, 2001). Memory researchers adopting a causal-appraisal perspective have argued that discrete emotions and their corresponding appraisals determine which aspects of an event are goal congruent and ought to be prioritized by memory (for reviews, see Levine & Edelstein, 2009; Levine & Pizarro, 2004).

Notably, memory researchers adopting basic-emotion or causal-appraisal perspectives interpret discrete emotion effects within a given experiment as evidence that such emotions affect memory in specific and consistent ways across instances. That is, whether finding differences between discrete emotions in terms of memorability (Chapman, 2018; Chapman et al., 2013; Ferré et al., 2018; Marchewka, Wypych, Michałowski, et al., 2016; Moeck et al., 2021; Riegel et al., 2022; West & Mulligan, 2021, Experiment 3), the neural correlates of memory (Marchewka, Wypych, Michałowski, et al., 2016), or their effects on the aspects of a narrative that are remembered (for reviews, see Levine & Edelstein, 2009; Levine & Pizarro, 2004), such differences are interpreted as the effects of these specific emotions, resulting from preprogrammed patterns of context-free cognition and relationships between a given emotion and memory that are one to one. Variability in these models—whether across individuals or contexts—is seen as error rather than an interesting effect to be explained. Such interpretations stand in stark contrast to those that follow from constructionist theories of emotion.

As reviewed above, prior emotional-memory research has largely taken one of two approaches to investigating the effects of emotion on memory, when the effects of affective features such as valence and arousal or the effects of discrete emotions have been considered. Often, perspectives applied to the emotional-memory literature focus on the role of valence and arousal or discrete emotions, but not to consider both in tandem. With this in mind, the consideration of well-developed theories from affective science that provide mechanistic accounts of both affect and discrete emotion may be of great benefit to emotional-memory researchers by facilitating the generation of novel, theoretically informed research programs. In the current article, we consider the utility of one class of emotion theories that has received considerable support in recent years: psychological constructionism (Barrett, 2014, 2017a, 2017b; Lindquist, 2013), which sees emotions as emergent phenomena produced by more basic psychological and neural mechanisms that are highly idiographic and situated in context and culture.

Given that the theory of constructed emotion makes a strong prediction that access to conceptual knowledge is necessary for the experience of discrete emotions (Barrett, 2014, 2017a, 2017b; Lindquist, 2013; Wilson-Mendenhall et al., 2011), the CMEM predicts that any effect of discrete emotions on memory (see Fig. 1) should be reduced under conditions in which conceptual knowledge is reduced or absent. Three specific hypotheses follow from this prediction, which we articulate first before describing the methods by which these predictions might be tested. First, it should be the case that in the absence of the availability or accessibility of conceptual knowledge, participants do not exhibit the sort of discrete emotion concept–congruent memory biases described in Prediction 1, as these effects should depend on the categorization of stimuli using discrete emotion concepts. In other words, although the CMEM predicts that attributing a specific emotional concept (e.g., fear) to an event will result in emotion concept–congruent memory biases, these biases should appear only under conditions or in populations in which conceptual knowledge about these emotions is accessible. We acknowledge that participants may still show memory effects of valence and arousal, but if Prediction 1 is correct, emotion-concept knowledge should bias memory beyond valence and arousal.

Second, during memory tests that require participants to make fine-grained mnemonic discriminations between stimuli belonging to different discrete emotion categories (e.g., recognition, source memory, the Mnemonic Similarity Test; Stark et al., 2013), performance should suffer when emotion knowledge is reduced. This prediction follows from the idea that perceiving stimuli as belonging to different conceptual categories results in the perception of greater differences between these stimuli (Brosch et al., 2010; Harnad, 2003). To the extent that the perception of such differences results in stimuli associated with different emotional categories being represented with increased between-stimulus distinctiveness, this ought to benefit memory discrimination. We refer to this phenomenon as concept-dependent distinctiveness (see Fig. 1). Consistent with the prediction that the perception of stimuli as being associated with different emotion concepts will increase between-stimulus mnemonic discrimination, a recent study found that providing participants with different emotion labels for faces (“angry” vs “fearful”) does indeed result in faces paired with one label being remembered as more dissimilar than faces with a different label (McMullen et al., 2025). Similarly, in a short-term memory experiment by Fugate et al. (2010), participants who had been taught to identify different primate facial-muscle movements were better able to discriminate between different chimpanzee behaviors, suggesting that acquiring conceptual knowledge about nonhuman facial behaviors results in enhanced short-term memory discrimination for such faces. What remains untested is the CMEM’s novel prediction that the spontaneous process of emotion categorization as mediated by conceptual knowledge will result in the same sort of mnemonic discrimination that has been observed in studies in which emotion concepts are provided by the experimenter.

Last, to the extent that discrete emotions have different effects on memory in certain contexts, reducing emotion knowledge should reduce such differences. For example, discrete emotion effects such as the disgust advantage (Chapman, 2018; Chapman et al., 2013; Ferré et al., 2018; Marchewka, Wypych, Michałowski, et al., 2016; Moeck et al., 2021; Riegel et al., 2022; West & Mulligan, 2021, Experiment 3) should be reduced when access to conceptual emotion knowledge is reduced, because such effects are predicted to be mediated by differences in the features associated with different emotion concepts, which in turn result from concept-dependent emotion construction.

Empirically, predictions regarding the necessity of conceptual knowledge for emotional memory effects may be tested using methods from affective science in which (a) populations deficient in conceptual knowledge about emotions are compared with populations with intact conceptual knowledge or (b) experimental manipulations that modify the availability of conceptual knowledge are applied. With regard to the first approach, in which the emotional memory of populations with deficient conceptual knowledge is examined, one such population is those high in alexithymia, a personality trait characteristic of individuals who possess a paucity of conceptual knowledge about discrete emotions (for reviews, see Lindquist & Barrett, 2008b; Taylor & Bagby, 2000). Alexithymia is associated with factors that can impair cognitive development during childhood more generally, such as low socioeconomic status (Karukivi & Saarijärvi, 2014) and deficits in the development of language (Hobson et al., 2018, 2019; K. S. Lee et al., 2025). Thus, the hypotheses discussed regarding alexithymia specifically likely apply to individuals with a relative paucity of conceptual knowledge about emotions due to any cause. Because of their impaired conceptual emotion knowledge, individuals with alexithymia experience difficulties in identifying and describing their emotions (Sifneos, 1973; Taylor & Bagby, 2000). In addition, although such individuals do experience affective sensations (see Lindquist & Barrett, 2008b), they lack the conceptual emotion knowledge necessary for categorization, and as a result experience disruptions in discrete emotion perception (Donges et al., 2014; Lane et al., 2000; Nook et al., 2015; Parker et al., 1993).

A population deficient in conceptual knowledge, broadly speaking, is patients with semantic dementia (also known as the temporal lobe variant of frontotemporal dementia), a neurodegenerative disorder characterized by deficits in conceptual knowledge caused by neurodegeneration within the anterior temporal lobes (Hodges & Patterson, 2007). Because conceptual knowledge (including knowledge about emotions) is impaired in patients with semantic dementia, TCE predicts that these individuals’ emotion construction will be impaired as well. Indeed, as with individuals with alexithymia, patients with semantic dementia show marked disruptions in discrete emotion perception (Bertoux et al., 2020; Calabria et al., 2009; Kamminga et al., 2015; Kumfor & Piguet, 2012; Lindquist et al., 2014; Macoir et al., 2019; Rosen et al., 2004).⁸

In addition to testing the above predictions in populations deficient in conceptual knowledge, another way to test whether conceptual knowledge is necessary for the effects of emotion on memory is to manipulate the availability of such knowledge experimentally. For instance, by using a procedure known as semantic satiation, it is possible to temporarily inhibit the availability of semantic concepts (for a review, see Black, 2003). Semantic satiation entails having participants repeat a given word many times aloud in a row (e.g., 30 repetitions; Gendron et al., 2012; Lindquist et al., 2006). Eventually, this repetition results in the decoupling of a word’s phonological form from its semantic meaning, rendering the word’s associated concept temporarily inaccessible (Black, 2003). Consistent with the claim that conceptual knowledge is necessary for emotion construction (Barrett, 2014, 2017a, 2017b; Lindquist, 2013; Wilson-Mendenhall et al., 2011), the semantic satiation of emotion words such as “anger” has been shown to disrupt the perception of discrete emotions in facial-muscle movements (Gendron et al., 2012; Lindquist et al., 2006).⁹

Another method of experimentally manipulating the availability of conceptual knowledge is through neurostimulation. Research has shown that by applying repetitive transcranial magnetic stimulation (rTMS) to healthy individuals’ temporal lobes, researchers are able to experimentally induce impairments in conceptual knowledge similar to those seen in patients with semantic dementia (Campanella et al., 2013; Lambon Ralph et al., 2009; Pobric et al., 2007). Similarly, applying transcranial direct current stimulation (tDCS) to the temporal lobes has been shown to affect the availability of conceptual knowledge (for a review, see Joyal & Fecteau, 2016; Pobric et al., 2007). Notably, both temporal rTMS and tDCS have already been shown to affect episodic-memory effects dependent on the availability of conceptual knowledge as discrete emotion effects are predicted to be, with prior research demonstrating that both rTMS and anodal tDCS reduce false memories for semantically related information in the Deese-Roediger-McDermott (DRM) paradigm (Boggio et al., 2009; Díez et al., 2017; Gallate et al., 2009). To our knowledge, past research has not addressed this question with regard to emotional memory, but this would be an interesting avenue of future research.

Having reviewed the predicted effects of conceptual emotion knowledge on the mnemonic effects of discrete emotions, we now demonstrate how these predictions and methods may be combined to generate specific, falsifiable research hypotheses. For instance, by combining the prediction that conceptual emotion knowledge is necessary for concept-congruent memory biases (Prediction 1) with the fact that individuals with alexithymia are thought to be deficient in conceptual emotion knowledge, the CMEM generates the hypothesis that individuals high in alexithymia will show reduced concept-congruent memory biases relative to those low in alexithymia. Similarly, because discrete emotion effects (such as the disgust advantage) are predicted to depend on the availability of discrete emotion concepts such as disgust, and because the availability of such semantic concepts may be inhibited through temporal neurostimulation, the CMEM generates the hypothesis that the disgust advantage should be reduced under conditions of inhibitory temporal neurostimulation, compared with sham stimulation. Although by no means an exhaustive combination of the predictions and methods listed above, these hypotheses illustrate the value of adopting a constructionist perspective when considering the conditions under which discrete emotions may or may not affect later memory.

Whereas constructionist theories make strong predictions about the role of conceptual emotion knowledge in discrete emotion construction, the role of conceptual knowledge in the perception of affect (i.e., valence and arousal) is an ongoing question. First, it should be emphasized that conceptual knowledge is not thought to be necessary for the perception of affect in an absolute sense, as evidenced by the fact that one may come to view stimuli as simply being positive or negative because of relatively automatic processes that do not require the experience of categorical emotion states, such as classical conditioning (Lindquist, 2013). The question, however, is whether the construction of discrete emotion states might influence the specificity or intensity with which one experiences sensations of valence and arousal during an emotional episode.

Articulation of constructionist theory provides some suggestion that categorization may fundamentally transform the meaning of affective sensations. Barrett (2005), for example, wrote that by parsing affective sensations into discrete emotional experiences through the process of emotion construction, affect is imbued with “intention or emotional aboutness” (p. 270). Such intention and aboutness may mean that the affective features of valence and arousal are represented in a more specific, stimulus-directed manner when the perceiver has access to conceptual knowledge. Consistent with this assertation, some evidence has suggested that in addition to experiencing disruptions in discrete emotion perception, both individuals with alexithymia (De Gucht & Heiser, 2003; Grabe et al., 2004) and semantic dementia (Mendez, 2021) tend to attribute affective sensations to physiological causes (i.e., somatization) rather than to external stimuli. This raises the possibility that sensations of valence and arousal may be experienced as less stimulus-specific and more generalized by populations with deficits in conceptual knowledge. Additional empirical support for the notion that conceptual knowledge may interact with the representation of affect is seen in studies demonstrating divergent effects of affect on behavior (Lindquist & Barrett, 2008a; Lee et al., 2018) and experience (Lee et al., 2018) depending on the presence of accessible concept knowledge of “fear.” Additionally, to the extent that qualities such as negative valence and high arousal are associated with one’s representation of specific emotion concepts such as fear, it may be predicted that the perception of a stimulus as being frightening will result in the stimulus being perceived as more negative and arousing than it would have been in the absence of access to conceptual knowledge of “fear.” Suggestive of this hypothesis is evidence that people primed with the concept “fear” (vs. a neutral control) show a greater startle reaction to aversive images (Oosterwijk et al., 2010). Furthermore, people who are higher in emotional granularity—who differentiate more among their emotions in daily life and who are thought to experience their emotions more discretely through greater access to concept knowledge (Hoemann, Nielson, et al., 2021; Lindquist & Barrett, 2008b)—show even greater sympathetic arousal during a stressor than individuals who are lower in differentiation (Bonar et al., 2023). This might tend to support the possibility that discrete emotional construction may amplify the experience of affect.

If access to conceptual knowledge renders affective sensations that are more specific and intense, reductions in conceptual knowledge should reduce what we refer to here as the “typical effects of emotion on memory,” which have historically been attributed to valence and arousal (i.e., emotionally enhanced memory, emotionally enhanced vividness, memory trade-off effects). Given the state of the current literature on this topic and the logic outlined above, we believe the assumption underlying this specific prediction to be sensible at present. At the same time, it should be emphasized that this assumption is modular with respect to the validity of the CMEM, broadly speaking. Should future research in affective science indicate that the availability of conceptual knowledge does not affect the specificity or intensity with which one perceives valence and arousal, the rest of CMEM’s predictions will be unaffected.

Thus, the CMEM predicts that the typical effects of emotion on memory should be weaker for individuals with reduced conceptual emotion knowledge. Indeed, ample evidence already exists to support this prediction. With regard to semantic dementia, evidence suggests that—in contrast to healthy older adults and patients with Alzheimer’s disease—patients with semantic dementia do not show emotionally enhanced memory (Kumfor et al., 2013) and are less likely than controls to recall emotional details when retrieving recent autobiographical memories (Irish et al., 2011). With regard to alexithymia, ample evidence suggests that individuals with alexithymia show reduced memory for emotional but not neutral information (for a review, see Apgáua & Jaeger, 2019). For example, Meltzer and Nielson (2010) found that participants high in alexithymia recalled fewer negative words than participants low in alexithymia. Similarly, Ridout et al. (2020) found evidence that alexithymia is associated with reduced memory of emotional facial expressions and emotionally charged social interactions (see also Donges & Suslow, 2015). Such results suggest that, as predicted by the CMEM, the beneficial effects of emotion on memory are selectively reduced for individuals with alexithymia.

To date there does not appear to be a single commonly accepted explanation for reduced emotional memory in individuals with alexithymia. Some have argued that these deficits result from the fact that these individuals perceive emotional information as less salient because of their less well-integrated emotion schemas (Luminet et al., 2006; Ridout et al., 2020; Suslow et al., 2003); in contrast, others have explained these effects as being due to deficits in cognitive control (Dressaire et al., 2015), or, in the case of memory for emotional facial expressions, as due to the challenges that individuals with alexithymia experience in communicating their own feelings (Donges & Suslow, 2015). The CMEM offers a unifying explanation that is largely consistent with the impoverished-emotion schema-based explanation of these findings (but see footnote 6). That is, the CMEM predicts that alexithymia selectively impairs memory for emotional information because deficits in conceptual emotion knowledge disrupt the attribution of meaning to affective stimuli that would normally occur during emotion construction (Barrett, 2006, 2017a; Barrett et al., 2015). Thus, although individuals with alexithymia may have a preserved experience of affect (Lindquist & Barrett, 2008b), the experience of affective sensations alone may be insufficient to allow one to perceive external stimuli as emotional in a meaningful sense. Such stimuli may therefore not be processed in ways which would normally result in emotionally enhanced memory (see Fig. 1). Empirically, the validity of this explanation could be assessed by testing whether the reduced emotional memory of individuals with alexithymia is mediated by reductions in their personal ratings for normatively emotional stimuli of valence, arousal, and specificity or distinctiveness.

An alternative—or perhaps complementary—explanation of such effects is that linking stimuli to a specific discrete emotion concept, as mediated by conceptual knowledge, results in memory enhancements beyond what one would observe if the same stimulus had been perceived as simply being valenced and arousing. As discussed above, the attribution of emotional meaning to stimuli might be expected to enhance their degree of self-relevance (Barrett, 2017a; Feldman et al., 2024; Shaffer et al., 2022), which, if true, ought to result in better memory (Czienskowski, 1997; Symons & Johnson, 1997). Additionally, the perception of a stimulus as being associated with a specific discrete emotion may result in memory enhancements by increasing the distinctiveness with which said stimulus is represented (see also Fig. 1). Broadly speaking, such possibilities illustrate that, at present, it is unclear how much of the typical effects of emotion on memory are due to valence and arousal per se as opposed to some combination of affect and discrete emotion. It is our hope that the current article provides a framework for future investigations to tease apart these issues.

With regard to experimental methods of modifying the availability of conceptual knowledge, the CMEM predicts that such methods should constrain the effects of emotion on memory. Because no study to our knowledge has investigated whether methods such as semantic satiation of emotion concepts or neurostimulation of brain regions implicated in conceptual knowledge impacts the typical effects of emotion on memory, this novel prediction remains to be tested. However, Gendron et al. (2012) used a repetition-priming paradigm in which they showed that semantic satiation of the word “anger” prior to viewing a posed, scowling face prevented that face from priming itself moments later. That is, the inhibition of conceptual knowledge presumably influenced perceptual encoding of the first instance of the face, which impaired short-term visual memory. Future research should extend these effects to paradigms in which long-term encoding and retrieval could be examined.

Although some prior research suggests that populations with reduced conceptual knowledge show reductions in emotionally enhanced memory, it remains to be seen whether the participant-level characteristics and experimental manipulations discussed here also reduce the vividness (Bowen et al., 2018; Buchanan, 2007; Kensinger & Ford, 2020; Kensinger & Schacter, 2016; Levine & Pizarro, 2004; Ochsner, 2000; Phelps & Sharot, 2008; Talmi, 2013) and memory trade-off effects (Dolcos et al., 2017; Kensinger, 2009a, 2009b; Kensinger et al., 2007; Levine & Edelstein, 2009; Levine & Pizarro, 2004; Mather & Sutherland, 2011) associated with emotion. Although preliminary evidence indicates that individuals with alexithymia remember emotional information with a reduced degree of recollection (Luminet et al., 2006; Vermeulen & Luminet, 2009), it is currently unknown whether semantic dementia/aphasia, semantic satiation, or neurostimulation of the temporal lobes have similar effects on the phenomenology with which emotional events are remembered. Furthermore, to our knowledge no study has examined whether the memory trade-off effect is dependent on conceptual knowledge as predicted by the CMEM using any of these methods. Broadly speaking, the utility of the CMEM with respect to the mnemonic effects of valence and arousal can be seen as specifying for whom and under which circumstances the typical effects of emotion on memory may or may not manifest.

The theory of constructed emotion predicts that a stimulus’s affective and emotional value will depend on situational factors, so that a stimulus might be perceived as more or less arousing, positive, negative, or evocative of a certain emotion depending on the situation within which it is perceived (Barrett, 2006). In line with this prediction, there exists an extensive body of evidence demonstrating that context has a strong influence on the emotions perceived in facial-muscle movements (Aviezer et al., 2008, 2009, 2011, 2017; Ensenberg-Diamant et al., 2025; Goel et al., 2024; Hassin et al., 2013; Israelashvili, Hassin, & Aviezer, 2019). Also, consistent with the prediction that situational variability influences emotion construction, research has shown that participants’ affective and physiological responses to emotional stimuli depend on the situational context within which they are framed (Fernandes et al., 2019; Foti & Hajcak, 2008; Kirk et al., 2020; Santos et al., 2023). For example, Fernandes et al. (2019) found that participants rated images of hands covered in ambiguous substances as more negatively valenced, arousing, and disgusting if these images were framed as being photos of someone suffering from a gastrointestinal disease. Given that contextual framing manipulations have been shown to impact emotion perception and affect perception, it may be predicted that such factors will also affect memory. For instance, because contextual-framing manipulations imbue stimuli with affective and emotional meaning, it may be predicted that nonemotional stimuli will exhibit emotional-memory effects if framed in a way that makes them emotional within a given situational context. Consistent with this prediction, nonemotional stimuli that are framed as having been touched by contaminated hands are better remembered than those framed as being touched by noncontaminated hands (Bonin et al., 2019; Fernandes et al., 2017, 2021). Conversely, it may also be possible to reduce or eliminate the emotional-memory effects associated with normatively emotional stimuli by framing them in contexts that reduces their emotionality (e.g., by presenting an image of a car accident as an artificial recreation for a movie rather than as an actual car crash).

In addition to demonstrating that situational context can imbue stimuli with affect and emotion, research drawing on constructionist theories has also demonstrated that situational features explain heterogeneity that exists between different instances of the same emotion. For example, Wilson-Mendenhall et al. (2011) reported that functional magnetic resonance imaging (fMRI) data demonstrate that the brain regions associated with fear and anger differ within these emotions depending on whether fear and anger were experienced in situations associated with physical danger or with social evaluation. Likewise, Kveraga et al. (2015) found that the threat appraisals and fMRI activity associated with fear differed depending on whether participants were responding to direct threats, indirect threats, or past threats (see also Wang et al., 2024). Last, McVeigh et al. (2024) found that the relationship between patterns of physiological activity and levels of fear evoked by videos of spiders, heights, and social evaluation were best accounted for by a model that took into account differences between the three situation types as well as interindividual variability.

Taken together, these results suggest that important differences exist between instances of the same emotion depending on the situational context within which they are experienced. Given these results, it may be predicted that, rather than specific emotions such as fear or disgust exhibiting one-to-one effects on memory, the mnemonic effects of a given emotion will depend critically on the situation within which it is experienced. For example, given research demonstrating that instances of fear differ depending on whether the perceiver is experiencing fear of physical danger or fear of social evaluation (Wilson-Mendenhall et al., 2011), one might predict that the specific contextual features encoded into memory will differ in these situations, with sources of physical danger (e.g., weapons, predators) being better remembered for the former and the reactions and behaviors of others being better remembered for the latter. Broadly speaking, the general prediction that the effects of emotion on memory depend on situational context dovetails with prior research demonstrating that situational context is selectively encoded into memory during emotion perception (Barrett & Kensinger, 2010), highlighting the possibility that contextual features and emotion may interact in important ways with respect to episodic memory.

Constructionist theories predict that individuals differ considerably in how they construct their emotions. As briefly discussed above, one construct that has received considerable attention as a source of interindividual variability in emotion is emotional granularity,¹⁰ or the extent to which an individual represents emotions in a situation-specific and differentiated manner (Barrett & Bliss-Moreau, 2009; Hoemann, Nielson, et al., 2021; Lindquist & Barrett, 2008b; Tugade et al., 2004). A common approach to measuring granularity is to ask participants to rate episodes from their daily lives in terms of various emotional terms (e.g., anger, fear, sadness) across multiple events. Higher intercorrelation between ratings of different same-valenced emotions within episodes indicates lower granularity insofar as this indicates that individuals do not differentiate among same-valence emotion categories when rating events during their daily lives. For example, whereas a highly granular individual might feel angry after failing to receive a promotion, afraid after getting bad news about health, and sad after receiving negative feedback, a less granular individual may not make such fine-grained distinctions and may instead report similar levels of each emotion within these events. The concept of emotion granularity highlights the fact that people may not experience the same emotions in reaction to the same stimuli. Rather, constructionist theories predict that individual differences in constructs such as emotional granularity result in meaningful variation in emotion construction between individuals (Barrett, 2017a; Hoemann, Nielson, et al., 2021; Lindquist & Barrett, 2008b). Differences in emotional granularity are thought to arise from variation in the structure of individuals’ conceptual representations of emotions. More specifically, it is theorized that highly granular individuals either possess more differentiated emotion concepts or are more likely to use those concepts in the moment, allowing these individuals to represent their emotions in a nuanced, situated manner (Hoemann, Nielson, et al., 2021; Lindquist & Barrett, 2008b).

Research suggests that the effects of emotional granularity go beyond verbal reports of emotional experiences and extend to the level of differentiation with which emotional stimuli and events are processed and perceived. For example, Israelashvili, Oosterwijk, et al. (2019) found that individual differences in negative emotional granularity (i.e., the extent to which one’s experiences of different negative emotions are differentiated) were associated with higher performance on a task of facial emotion recognition. Additionally, an electroencephalography (EEG) study in which participants viewed emotional images found that granularity was related to differences in alpha and gamma synchrony (associated with access to conceptual knowledge and affective processing, respectively; Benedek et al., 2011; Güntekin & Başar, 2014; Jaušovec & Jaušovec, 2005), and that highly granular participants exhibited increased N2 amplitudes (associated with executive control; Folstein & Van Petten, 2008) and late positive potential amplitudes (associated with motivated attention; Hajcak et al., 2009) compared with less granular participants (J. Y. Lee et al., 2017). These findings demonstrate that individuals who differ in granularity process the same emotional stimuli differently, and they were interpreted by J. Y. Lee et al. (2017) as evidence that highly granular individuals process emotional information in a more complex and differentiated manner. Similar associations between granularity and differentiation in emotion representation have been demonstrated in daily life, with highly granular individuals demonstrating higher experiential diversity (Hoemann, Lee, et al., 2023) and higher context specificity in emotion-related cardiorespiratory physiology (Hoemann, Khan, et al., 2021) than individuals low in granularity. Taken together, such research suggests that highly granular individuals tend to represent emotional stimuli and events with greater differentiation and context specificity than those low in granularity.

If highly granular individuals represent emotional stimuli in a more differentiated (i.e., distinctive) manner (J. Y. Lee et al., 2017; Lindquist & Barrett, 2008b; Wang et al., 2020), such individuals should show larger benefits of emotion on memory. This prediction is motivated by classic research demonstrating that stimuli that are processed in a more distinctive manner are ultimately better remembered, as such processing results in the formation of memory traces that are more discriminable (Hunt & McDaniel, 1993; Jacoby & Craik, 1979). Additionally, because emotionally enhanced memory is thought to be partially mediated by distinctiveness (Talmi, 2013; Talmi & McGarry, 2012; Talmi et al., 2007), any variable that increases the distinctiveness of emotional stimuli ought to increase their mnemonic advantage.

The prediction that highly granular individuals will show increased emotional memory was investigated in the first author’s doctoral dissertation (West, 2023). In this study, participants completed multiple end-of-day reports of their emotional experiences during episodes from their daily lives as well as laboratory tests of episodic memory for emotional images and DRM word lists.¹¹ Results showed that participants high in granularity for negative emotions in daily life exhibited stronger valence-based memory effects for emotional images than those low in negative granularity. Such results suggest that the effects of emotion on memory depend on one’s level of granularity, at least for certain visual stimuli. In light of these preliminary results, researchers are encouraged to conduct additional investigations on the mnemonic effects of granularity on emotional information. In particular, because granular individuals have been conceived of as experts in representing complex and nuanced emotions (Hoemann, Nielson, et al., 2021), future research should investigate the possibility that granularity exhibits stronger effects on memory for information that requires more nuanced emotion construction, such as emotionally ambiguous materials (Brainerd, 2018; Brainerd et al., 2021) or emotionally complex autobiographical memories.

Although we have focused mostly on how emotional granularity might affect memory at the behavioral level, it is also possible that differences in granularity relate to the neural processes mediating emotional memory as well. As already discussed, prior EEG research has shown that individual differences in granularity are related to electrocortical measures such as event-related potentials and event-related synchrony when processing emotional images (J. Y. Lee et al., 2017; Wang et al., 2020). Because such research suggests that differences in granularity are associated with variability in the neural representation of emotional information, it is possible that granularity is likewise associated with variability in the neural correlates of emotional memory. For instance, given that granularity is thought to allow for more distinctive representation of emotional information, it might be predicted that item-specific patterns of neural activation within brain regions involved in affect and emotion representation will be more differentiated from one another (i.e., less interchangeable) for highly granular individuals, and that this increased neural distinctiveness will predict enhancements in emotional memory. This prediction could be evaluated using multivariate analyses of fMRI data such as representational similarity analysis (Dimsdale-Zucker & Ranganath, 2018; Kriegeskorte et al., 2008; Sommer & Sander, 2022), and if true, would suggest that highly granular individuals’ increased distinctiveness in their neural representation of emotion is associated with enhanced memory for emotional information.

Individual-level characteristics other than emotional granularity likely also impact emotional memory through their effects on emotion construction. An essential prediction of the theory of constructed emotion is that, because a stimulus’s physiological consequences are predicted on the basis of one’s learning history, the computation of valence, arousal, and emotion will depend on the perceiver’s prior experiences (Barrett et al., 2015; Lindquist, 2013). By highlighting the importance of idiosyncratic variability in emotion and affect perception, constructionist theories underscore the need to take seriously the possibility that individuals may differ considerably in their perception of identical emotional stimuli. Indeed, recently Westlin et al. (2026) demonstrated that emotion labels based on normative ratings of images and film clips do a poor job of describing the emotional experiences of individual participants, providing strong evidence against the basic-emotion view that affective stimuli reliably and exclusively evoke specific discrete emotions across individuals.

To the extent that such variability affects cognitive processes, such as memory, it may be warranted to adopt an idiosyncratic approach to operationalizing affect and emotion—one in which experimenters rely not on normative ratings to define the properties of their stimuli but instead on the idiosyncratic, person-specific perceptions of the individuals within a given study. Having observed substantial idiosyncratic variability in discrete emotion perception (see also Ensenberg-Diamant et al., 2025), Westlin et al. (2026) likewise advocated for “idiographic analytical approaches that explicitly model subject-level variability” (p. 13). Unpublished work in our lab suggests that this approach may indeed be fruitful, as an image-recall experiment revealed that participants’ own idiosyncratic ratings of valence were substantially better at predicting their subsequent recall performance than normative ratings of the same images (West, 2023).¹² Ongoing work in our lab is investigating this issue.

Central to the CMEM is the prediction that the factors that impact emotion construction will impact emotional memory. Thus, although cognitive factors such as distinctiveness and relatedness are typically conceived of as inherent properties of emotional stimuli (Talmi, 2013; Talmi & Moscovitch, 2004; Talmi et al., 2007), the CMEM raises the possibility that such properties might be better thought of as emergent properties arising during emotion construction and that factors such as emotional granularity may place important perceiver-dependent constraints on such properties. The hypotheses regarding individual differences in emotion construction that follow from the theory of constructed emotion and from the CMEM may prove particularly valuable in advancing memory research, given that the majority of emotional-memory research to date has investigated how emotional stimuli are remembered on average, without considering the ways that individuals differ in their emotion perception (but for exceptions, see Dolcos et al., 2017, 2020; Hamann & Canli, 2004; Waring et al., 2009).

There are no doubt additional sources of variability in emotion construction that should be investigated in future research. After all, variability in emotion is not restricted to the level of the individual or the current situation. In the following section, we consider the mnemonic effects of emotional variability at another level of analysis: that of culture.

Because constructionist theories propose that emotion construction depends on learned emotion concepts, these theories predict that differences in the nature of the emotion concepts transmitted to individuals by their culture should affect emotion construction. In other words, like situational factors and individual differences in conceptual knowledge, culture provides an ongoing context within which emotion construction is situated (Gendron et al., 2018). Thus, unlike basic-emotion theories that predict that emotional experiences and expressions are universal (Ekman, 1999; Ekman & Cordaro, 2011; Tracy, 2014), constructionist theories predict that the emotions one experiences and perceives in others depend on social concepts that are culturally relative (Barrett, 2014, 2017a). Consistent with this claim, research has shown that when using unconstrained methods, the emotions individuals perceive in facial-muscle movements show substantial cross-cultural variability (for reviews, see Gendron, 2017; Gendron et al., 2018). In this section, we consider the mnemonic implications of such variability.

A review of the extensive literature on cross-cultural differences in emotion is beyond the scope of the current article. Instead, we focus on one way that cultures differ with respect to emotions to demonstrate the utility of considering cross-cultural differences in emotional-memory research. In particular, we consider cross-cultural differences in the extent to which emotional experiences and perceptions are influenced by ongoing social context. Because most of the cross-cultural emotion research on this topic has focused on the perception of facial expressions, such research is the focus of this section. As with any research described here related to facial emotion perception (e.g., research demonstrating that emotion concepts bias memory for facial expressions, Doyle & Lindquist, 2018, Experiments 1 and 2; Fugate et al., 2018; J. Halberstadt, 2005; J. B. Halberstadt & Niedenthal, 2001; J. Halberstadt et al., 2009), we predict that the same mechanisms at play during emotion construction that lead to cross-cultural differences in facial emotion perception will also impact emotional reactions and memories for nonfacial stimuli (see footnote 7).

Cross-cultural research has suggested that Westerners tend to view emotions as phenomena that occur within individuals, whereas Easterners tend to view emotions as phenomena that occur between people within the context of social interactions (for a review, see Mesquita, De Leersnyder, & Boiger, 2016). In order to investigate the influence of social context on cross-cultural differences in emotion perception, Uchida et al. (2009) conducted a series of studies in which American and Japanese participants made judgments about Olympic athletes. In support of the claim that Easterners are more likely to think of emotion as something that occurs between people, the authors found that Japanese participants inferred more emotion in athletes who were shown with others, whereas Americans inferred more emotion in athletes who were pictured alone. This suggests that the degree to which an event is perceived as emotional depends on an interaction between one’s culture and the social context of the event. Also consistent with the idea that cultures differ in the extent to which social context influences emotion construction, research has demonstrated differences between cultures in the extent to which emotion perception is affected by the emotions of task-irrelevant others. More specifically, research has shown that whereas Easterners’ emotion ratings of a target face are influenced by the expressions of peripheral, task-irrelevant people, Westerners’ ratings are not (Masuda et al., 2008, 2012).

Taken together, prior research has demonstrated that cultures differ in the extent to which emotion construction is influenced by the ongoing social context of an event. If one takes the constructionist approach of conceiving of emotions as emergent phenomena constructed using culturally relative emotion concepts (Barrett et al., 2015; Wilson-Mendenhall et al., 2011), whether one conceptualizes emotions as social or asocial phenomena might be thought of as an aspect of conceptual emotion knowledge that differs between cultures. Because the CMEM predicts that variability in the conceptual representation of emotions should affect emotional memory, this model predicts that such cross-cultural differences will impact the way that members of different cultures remember emotional information and events.

Two predictions regarding emotional memory can be made from the premise that cultures differ in the extent to which emotion construction is influenced by social context. First, if cultures vary in the extent to which the presence of others influences the degree to which an event is perceived as emotional (Levenson et al., 1992; Uchida et al., 2009), then the presence of others will affect participants’ emotional memory differently, depending on their culture. For example, compared with American participants, Japanese participants may exhibit stronger emotional memory effects for social compared with asocial stimuli. Second, if Easterners’ emotion construction is more influenced by social context than Westerners’ (Masuda et al., 2008, 2012), then details of the social context of an emotional event will have a stronger influence on Easterners’ memories than Westerners’. In particular, this view predicts that stimuli surrounded by people reacting in emotional ways will show stronger emotional memory effects for Easterners compared with Westerners. Similarly, it is predicted that Easterners will exhibit stronger concept-congruent memory biases for stimuli surrounded by people displaying stereotypical expressions of a specific emotion (e.g., fear) than will Westerners (see Prediction 1).

Although far from exhaustive, the predictions discussed in this section regarding the potential influence of culture on emotional memory illustrate the utility of considering the mnemonic consequences of cross-cultural differences in emotion construction. In addition to testing such predictions, future work should consider as well the role of other well-established cross-cultural differences in emotion. For example, research has shown that culturally reinforced emotions tend to be experienced with greater prevalence and intensity (Boiger & Mesquita, 2012; Mesquita, Boiger, & De Leersnyder, 2016; Mesquita, De Leersnyder, & Boiger, 2016). Thus, it may be that the effects of a given emotion on memory will be stronger in cultures where that emotion is more salient compared with cultures where it is less salient. For instance, because the experience of anger is thought to be up-regulated in the United States, whereas shame is thought to be up-regulated in Japan (Boiger & Mesquita, 2012; see also Kirmayer, 1991; Kleinknecht et al., 1997; Tanaka-Matsumi, 1979), it can be predicted that anger will have a stronger effect on the memories of American individuals, whereas shame will have a stronger effect on the memories of Japanese individuals.

Although it is generally the case that emotional information is seen as more conceptually interrelated than neutral information (Talmi & Moscovitch, 2004; Talmi et al., 2007), the extent to which this is true may depend on whether or not emotional stimuli are thought of as exemplars of a superordinate and distinct emotion category. Because research suggests that certain cultures may not possess an explicit concept of emotion in their lexicon (e.g., Samoans, Tahitians, Gidjingali aborigines; see Russell, 1991), it may be that such cultures do not perceive emotional stimuli as being conceptually interrelated in the same way that Western cultures do.¹³ If so, this would have important implications for emotional-memory research, given that increased relatedness is thought to partially explain the emotionally enhanced memory effect (Talmi, 2013). In particular, the lack of a distinct concept for emotions among Samoans suggests that they may not represent emotional stimuli as being conceptually related (Russell, 1991) and may not show the corresponding mnemonic effects of relatedness for emotional stimuli (Russell, 1991; Talmi, 2013; Talmi & Moscovitch, 2004; Talmi et al., 2007).

As illustrated by the fact that cultures differ in whether they have a distinct word for “emotions” (Russell, 1991), cross-cultural differences in emotion lexicons provide important information regarding differences in the nature of emotion concepts between cultures. In this way, emotion lexicons may be thought of as indicators of the collective emotion knowledge propagated through a cultural context. Consequently, an examination of the emotion words present in a culture’s lexicon will likely prove fruitful when generating hypotheses regarding cross-cultural differences in emotional memory.

As an example, because the CMEM predicts that the mnemonic effects of discrete emotions depend on the nature of the discrete emotion concepts accessed during emotion construction, differences between cultures in these emotion concepts as reflected by their emotion lexicons may result in differences in memory for stimuli associated with those emotions. As a specific example, the absence of a translationally equivalent word for the English “disgust” in Polish (Ba˛k, 2023; Russell, 1991; Wierzbicka, 1986) suggests that this specific emotional state (as conceptualized by English speakers) may be less salient in Polish culture. If so, this raises the intriguing possibility that although stimuli conceptualized as disgusting by Americans (Chapman, 2018; West & Mulligan, 2021, Experiment 3), Canadians (Chapman et al., 2013), Australians (Moeck et al., 2021), and Spaniards (Ferré et al., 2018), are more memorable than those conceptualized as frightening (i.e., the disgust advantage), this difference in memorability may be reduced for Polish participants. Although one research group has demonstrated some evidence that disgusting stimuli are better remembered than frightening stimuli by Polish participants (Marchewka, Wypych, Michałowski, et al., 2016; Riegel et al., 2022; but see Marchewka, Wypych, Moslehi, et al., 2016), it is an open question for future cross-cultural research whether such effects are equal to or smaller than those observed in other cultures when using the same procedures and materials. This prediction raises the broader possibility that the effects of specific discrete emotions on memory are not universal, but rather situated within cultural contexts that differ in how emotions are learned, reinforced, and enacted.

The prediction that cross-cultural differences in emotion construction will result in differences in emotional memory has important implications for the generalizability of emotional-memory research. Central to constructionist theory is the idea that cross-cultural variability is a meaningful and fundamental aspect of emotion (Barrett, 2017a; Boiger & Mesquita, 2012; Gendron, 2017; Gendron et al., 2018; Hoemann, Gendron, et al., 2023; Mesquita, 2022; Mesquita & Boiger, 2014; Mesquita, Boiger, & De Leersnyder, 2016; Mesquita, De Leersnyder, & Boiger, 2016). Likewise, the CMEM predicts that the emotion concepts that enable emotion construction are culturally situated and that cross-cultural differences in emotion concepts will affect how emotional stimuli are remembered. Care must be taken when attempting to generalize emotional-memory findings beyond the Western samples that dominate the extant research. Indeed, recent research has begun to investigate the ways in which culture places boundary conditions on emotional memory effects. For example, although American participants show impaired memory for neutral backgrounds previously paired with emotional images, this memory trade-off effect appears to be absent for Turkish participants (Gutchess et al., 2018). These findings provide support for the idea that certain memory effects that are often thought of as general facts about emotional memory may in fact apply only to some cultures but not to others. Such findings caution against the assumption that findings regarding emotional memory are universal, and they reinforce the need to determine whether the effects and theories that describe Westerners’ emotional memories apply to non-Western cultures as well.