Positive, But Not Negative, Emotion Retroactively Enhances Memory via Neural Reactivation

Emotion profoundly influences our memory for neutral events. For instance, positive emotions elicited by well-timed humor in a classroom can lead to stronger retention of the just-preceding lesson. This retroactive enhancement effect is well established: in animals, novel or emotional events (e.g., exploring a new environment or receiving a shock) can enhance memory for neutral stimuli encountered minutes to hours earlier (Dunsmoor et al., 2022). Similarly, in humans, emotional stimuli (e.g., frightening or pleasant pictures) can boost memory for neutral information presented seconds before those stimuli (Anderson et al., 2006). These findings indicate that emotional stimuli retroactively strengthen memory for immediately preceding neutral information.

The ability of both positive and negative emotional stimuli to retroactively enhance memory for a preceding neutral item is often attributed to emotional arousal, which strengthens recent memories by enhancing attention (Anderson et al., 2006; Mather and Sutherland, 2011). However, different emotional valences may influence memory via different attentional mechanisms. According to attentional theories, negative emotion narrows attention to salient details, potentially intensifying the encoding of the item's core features (Easterbrook, 1959). In contrast, positive emotion broadens attentional scope, potentially enriching the item's encoding with a wider array of contextual and associative details (Fredrickson and Branigan, 2005). This divergence raises a key question: Do these distinct attentional modes enhance memory through the same neural mechanism?

This question relates to two competing accounts of memory stabilization. The encoding variability hypothesis contends that superior memory arises from greater dissimilarity in neural representations across study episodes. This variability is proposed to result in the incorporation of diverse contextual information, thereby creating multiple retrieval routes and enhancing memory performance (Wirebring et al., 2015). In contrast, the neural pattern similarity hypothesis, supported by a growing body of empirical data, posits that memory is optimized when the same neural representation is reactivated across repetitions. This reactivation is thought to strengthen the initial memory trace, leading to more robust recall (Xue et al., 2010). Crucially, these competing hypotheses make divergent predictions about how attention enhanced by emotion modulates memory. If the broadening of attention by positive emotion fosters the integration of diverse contextual details, it might align with the encoding variability account. Conversely, if this broadened state instead facilitates the formation and consistent reactivation of a cohesive “item-context” unit, it would support the neural pattern similarity account. In contrast, for negative emotion, the narrowing of attention runs counter to the encoding variability account, as it inherently limits contextual diversity. Instead, this narrowed focus could logically favor either a highly stable item-focused representation or an intensified but less variable encoding—both of which would predict high neural pattern similarity. Thus, a key unresolved question is whether the retroactive memory boost from positive and negative emotions is mediated by a common neural mechanism (e.g., heightened arousal) or by valence-specific pathways that differentially engage neural pattern similarity versus variability.

A recent study by Pan et al. (2025) investigated how positive and negative emotions retroactively enhance memory for preceding neutral items through modulating neural pattern similarity. The procedure comprised two main phases: during encoding, participants viewed neutral squiggles, each immediately followed by a positive, negative, or neutral image, with each pair presented three times. After a 24 h delay, a recognition test required participants to judge each squiggle as “old” (i.e., previously viewed) or “new.” For squiggles judged “old,” participants rated the clarity of memory for the paired image (none, vaguely, or clearly). Throughout the experiment, continuous electroencephalogram (EEG) was recorded.

Pan et al. (2025) found that recognition accuracy (defined as the proportion of old items correctly identified minus the proportion of new items incorrectly judged as old) was higher for squiggles paired with positive images than for squiggles paired with neutral or negative images. To identify a neural correlate of this behavioral enhancement, the authors examined an electrophysiological signature of successful recollection—the parietal old/new event-related potential (ERP) effect. This effect is generally characterized by more positive potentials for old items versus new items, and it typically emerges between 400 and 800 ms after the stimulus. In this experiment, however, this recollection signal was significant only for squiggles previously paired with positive images, and it extended across a window of 400–1,200 ms, which was longer compared with the typical 400–800 ms window. No reliable old/new ERP difference was detected for squiggles paired with neutral or negative images.

The authors then tested whether the observed retroactive enhancement by positive emotion was achieved through increased representational similarity during the encoding of squiggles. Neural activity patterns for subsequently remembered squiggles in the positive emotion condition showed higher spatiotemporal pattern similarity (STPS) across repetitions than those for forgotten squiggles. This increase in STPS, which occurred within a 380–600 ms window after squiggle onset, was localized to the right-frontal region. The magnitude of this neural similarity effect positively correlated with later retrieval performance, suggesting that positive emotion enhances memory by promoting neural reactivation across repetitions. In contrast, although negative emotion enhanced overall neural similarity in the right-posterior region, this effect was dissociated from memory performance.

By providing neural evidence that positive emotion enhances memory through increased representational similarity, Pan et al. (2025) offer a mechanistic explanation for how positive emotions retroactively enhance memory. This extends prior work (Xue et al., 2010; Lu et al., 2015) by demonstrating that emotion valence moderates the link between reactivation strength (indexed by STPS) and memory performance. Specifically, they identified a right-frontal STPS effect (380–600 ms) for remembered items preceding positive emotional images. Previous studies suggest that the timing and frontal topography of this STPS effect likely indicate a controlled memory process, such as contextual reactivation (Lu et al., 2015). Importantly, this benefit was specific to positive emotion. Negative emotion did not improve memory or increase frontal STPS but enhanced pattern similarity in posterior sensory regions, which suggests an enhancement of vigilance rather than memory encoding. The frontal locus of the memory-related reactivation is consistent with this region's role in top–down cognitive control and the maintenance of coherent representations across time, processes distinct from the bottom–up sensory encoding functions of posterior visual areas.

Thus, although both positive and negative emotional valence increase neural similarity, these effects occur in distinct brain regions and reflect divergent processes: neural similarity within frontal regions, which support top–down control and the maintenance of coherent representations, signifies memory reactivation and predicts recall. Conversely, neural similarity within posterior sensory regions, responsible for bottom–up encoding, reflects enhanced perceptual processing or vigilance but does not forecast memory success. This observed valence-based dissociation—whereby emotion valence dictates the locus and functional meaning of neural similarity—provides a neural specification for the attentional scope framework (Easterbrook, 1959; Fredrickson and Branigan, 2005): specifically, broadened attention from positive emotion corresponds to frontal reactivation that strengthens memory, whereas narrowed attention from negative emotion corresponds to posterior vigilance-related similarity.

Although Pan et al. (2025) found that frontal neural pattern similarity is associated with successful remembering, what this pattern similarity reflects is still unknown. This issue is important to understanding the neural mechanisms by which positive emotion enhances memory. The authors attribute the effect to reactivation of the prior encoding episode, possibly including its emotional context. This interpretation is motivated by the evidence that participants reported recollecting the emotional context. However, a key neural finding appears inconsistent with this view: there was no increased similarity in neural patterns between different items paired with the same emotion. Thus, the combination of subjective recollection without corresponding neural convergence leaves open the precise nature of the reactivation: is it a purely item-specific trace or one that integrates the emotional context? The frontal and late-latency profile of the effect does argue against a purely perceptual account, focusing inquiry on whether the reactivation is primarily squiggle-specific or squiggle-emotion integrated. Future fMRI work using multivariate pattern analysis can directly test these alternatives by assessing whether memory success correlates with pattern similarity in emotion-related regions (e.g., the amygdala) or in associative binding regions (e.g., the hippocampus). Establishing a correlation between such deep-brain similarity and the frontal EEG signal reported here would bridge measures and clarify which account the scalp activity best reflects. This cross-modal link is essential for advancing from demonstrating reactivation to defining its representational content.

Notably, the authors found memory enhancement only for positive contexts, whereas Anderson et al. (2006) reported enhancement for both valences. This discrepancy may stem from foundational differences in the two paradigms. First, the stimulus nature differs: Anderson et al. used meaningful items (faces/houses), while Pan et al. used novel squiggles. Positive emotion's attentional broadening may be crucial for binding such amorphous items to a context. Second, the primary mnemonic demand differs. The repeated learning paradigm used by Pan et al. emphasizes the cross-temporal stabilization of a memory trace, a process indexed by frontal reactivation that is facilitated by positive valence. In contrast, Anderson et al.'s single-trial paradigm tests the postencoding consolidation of an event, where high arousal itself—regardless of valence—can act as a neuromodulatory trigger to strengthen the recent trace. Thus, positive emotion may be uniquely beneficial for tasks requiring the integrative linking of item and context over time, whereas arousal per se can boost the consolidation of a single episode. Future research systematically comparing these factors can delineate the boundary conditions of emotional effects on memory.

• Copyright © 2026 the authors