Dissociable age-dependent effects of emotion on scene and location memory

Abstract

Aging is associated with a decline in binding memories to their spatial context, including both spatial locations and background scenes. However, because emotions can enhance or impair memory depending on the type of spatial context, we investigated the effects of emotional modulation on spatial location and scene memory in adults between 21 and 67 years of age. 165 adults (83 female) were tested on their memory of scenes composed of emotional (or neutral) foreground images superimposed on neutral backgrounds, presented in one quadrant of the screen. In case of mental health-related individual variability, we also administered self-report surveys of depression and anxiety. First, we found that memory for the emotional foreground was more accurate but at the expense of the background scene. Across aging, we observed a decline in scene memory, particularly with positive foreground images. In contrast with memory for scenes, emotion enhanced memory for spatial location (the quadrant in which the scene was presented), but this benefit significantly decreased with age. Together, these findings provide evidence that aging not only impairs episodic memory accuracy overall but also alters the way in which emotions influence contextual binding. Finally, although we found that depression was correlated with poor memory and anxiety with better memory, neither significantly interacted with its emotional modulation.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-37242-2.

Introduction

The ability to recall past events is not only fundamental to normal day-to-day life¹ but also contributes to a sense of self². As we age, however, the accuracy of our autobiographical memories tends to decline³, sometimes selectively based on their emotional valence^4,5. While age-related changes in episodic memory are generally attributed to an impaired binding of memory components^6,7, such changes might be further exacerbated by alterations in the way emotions impact specific memory components, such as what happened (e.g., emotionally charged objects) and where it happened (e.g., background scene or location)^8–10.

Current theories of episodic memory emphasize the spatial framework upon which coherent memories are constructed^11–13. Previous research has shown that older adults exhibit deficits across various tasks involving spatial contextual memory^14–16. Furthermore, studies using scene-based memory tasks reported that older adults exhibit reduced sensitivity to spatial details within scenes and are more likely to forget the physical locations of targets, even on a computer screen^17–22.

Different types of spatial contexts, such as a scene’s visual background versus the physical location of an object, are thought to rely on dissociable neural pathways^23–26. Background scene processing is linked to a ventral route through the parahippocampal cortex (PHC), while spatial location processing involves a dorsal route through the retrosplenial cortex (RSC)^27–30. Given these dissociable pathways, emotional influences on memory binding may differ across these two systems. Indeed, emotional objects within a scene tend to draw attention away from peripheral background information^31–36, likely due to the amygdala’s preferential engagement with emotionally salient stimuli^37,38. This selective encoding may impair the integration of the background context into a coherent episodic memory^32,39,40. In contrast, the spatial location of emotionally significant objects is often remembered more accurately, likely due to enhanced amygdala–hippocampal interactions^41,42.

Despite past findings that have revealed how emotion can differentially affect what (objects) and where (space) components of memory, no single study has yet comprehensively investigated the combined influence of emotion on episodic memory binding across the two different spatial contexts—locations and scenes. To address this gap and investigate age-related changes in the emotional modulation of memory, we developed a computerized episodic memory task using scenes composed of emotional (or neutral) foreground images superimposed on neutral backgrounds, shown in distinct quadrants of the screen. Later, participants were asked to choose the scene they had previously observed during encoding from a selection of the target scene and three alternative combinations of target/lure foregrounds and backgrounds. After their selection, participants were asked to indicate the quadrant (location) in which the scene had originally appeared.

If location and background scene contexts are influenced by emotion in different ways, then participants should be able to remember the location of the foreground object, even if they do not remember the exact background of the scene. Moreover, given age-related decline in contextual memory and the benefits of emotion on location memory, older adults are expected to show greater deficits in background scene memory compared to location memory. In this study, we tested these hypotheses behaviorally on a sample of healthy online participants from 21 to 67 years of age. In addition, given past studies suggesting that individual affective health influences emotional memory, we administered self-report surveys of affective state and mental health (such as depression and anxiety) to explore how they may further shape emotional memory performance even in a nonclinical sample.

Method

Participants

A total of 165 eligible participants (83 women; age range = 21–67 years, M = 42.62, SD = 13.47) out of 191 participants (for detailed age distribution see supplementary materials Table S1), all of whom had completed education beyond a bachelor’s degree and reported no history of neuropsychiatric diagnosis, were recruited through the UK-based online platform Prolific (https://www.prolific.co/). Since data collected from online participants can vary widely in their experimental setting, making them noisier than that from a controlled laboratory experiment, we limited participation to users on desktop computers. Participants were required to have at least 95% approval rate (i.e., valid participation rate) on the Prolific website. All participants provided informed consent following a protocol approved by the Institutional Review Board at Seoul National University, and all procedures were performed in accordance with the Declaration of Helsinki. Participants who either scored below chance level (i.e., < 50% accuracy for either foreground or background recognition) or were designated as outliers in response time were excluded from the analysis.

An a priori power analysis was conducted using G*Power 3.1 to estimate the required sample size for a mixed design ANOVA (F-test). The analysis focused on detecting a group by emotional memory interaction effect (F-test within-between interaction) with the following parameters: a small effect size of 0.10, a significance criterion of 0.05, and a statistical power of 0.95. The design specification included 2 groups (between-subject factor) and 12 measurements (within-subject factor: 3 Emotions × 4 Memory Responses), assuming a correlation of 0.5 among repeated measures and a nonsphericity correction of 1. The calculation indicated that a total sample size of 106 participants was required to achieve the specified power.

Individual differences in mental health and mood

None of the participants in our study were diagnosed with affective, cognitive, or neurological disorders. The BDI-II (Beck Depression Index II) self-report survey measured severity of depressive symptoms. At testing, BDI-II scores ranged from 0 to 49, with 86 participants reporting minimal (BDI-II cut off range: 0–13), 23 mild (BDI-II cut off range: 14–19), 28 moderate (BDI-II cut off range: 20–28), and 28 severe (BDI-II cut off range: 29–63) levels of depression. A mean score of 15.4485 (SD = 12.886) was reported in this cohort. Along with the BDI-II, participants also completed the Patient Health Questionnaire-9 (PHQ-9), and the Center for Epidemiological Studies Depression Scale (CES-D). Mean scores were as follows: PHQ-9 = 6.897 (SD = 5.774), and CES-D = 17.091 (SD = 12.804).

The STAI-Y2 (State Trait Anxiety Index—trait anxiety) measured severity of the trait anxiety symptoms. At testing, STAI-y2 scores ranged from 20 to 76, with 58 participants reporting low anxiety (STAI-y2 cut off range: 20–37), 29 participants reporting moderate (STAI-y2 cut off range: 38–44), and 78 severe (STAI-y2 cut off range: 45–80). A mean score of 43.570 (SD = 14.173) was reported in this cohort.

For a general measure of mood, the Positive Affect and Negative Affect Schedule (PANAS) scale was implemented. Our sample showed a mean positive affect score of 24.273 (SD = 4.722) and a mean negative affect score of 21.048 (SD = 4.669).

Materials

The stimuli used in this study consisted of emotional (positive, negative) and neutral foreground images, paired with neutral background scenes. Each emotional valence category included 24 high-quality foreground images of people, along with a corresponding set of visually similar lure images. These were carefully selected from standardized affective picture databases, including the International Affective Picture Set (IAPS), the Geneva Affective Picture Database (GAPED), the Open Affective Standardized Image Set (OASIS), and Creative Commons sources from the internet^43–45. All images depicted distinct emotion-eliciting events and were manually cropped from their original backgrounds using Adobe Photoshop, while retaining their emotional content. For background stimuli, 72 neutral target scenes and an equal number of lures, matched in category (e.g., buildings, open fields, etc.) and low-level visual characteristics such as color and layout, were selected from the Correlated Photographs – Camera Edition panorama dataset, available on the Open Science Framework (https://osf.io/vy3sz/) and were used under the CC-BY 4.0 license⁴⁶. Before the main experiment, a separate group of 14 adult participants rated all stimuli for valence and arousal. Only images that showed clear agreement on emotional categorization were included in the final stimulus set (Valence: F(2, 71) = 227.81, p < 0.001).

Procedure

The experimental procedure consisted of two main components. First, participants completed psychological self-report surveys to evaluate general affective states (PANAS, Positive and Negative Affect Schedule), depressive symptoms (PHQ-9, CES-D, BDI-II), and trait anxiety (STAI-Y2). After the surveys were completed, participants proceeded to the emotional scene discrimination task (Fig. 1a). During the encoding phase of the memory task, participants viewed 72 complex scenes, each composed of an emotional foreground (positive, neutral, or negative) overlaid on a neutral background. Scenes were presented in a pseudo-random order, each appearing in one quadrant of the screen for four seconds. Participants rated the valence and arousal of each scene on a 5-point scale, following Russell’s circumplex model of affect⁴⁷. Valence ranged from 1 (“Unhappy”) to 5 (“Happy”) and arousal from 1 (“Calm”) to 5 (“Aroused”). After the encoding session, participants completed a 5-min interference task using the color Stroop test. A surprise recognition test followed, assessing memory for the scenes’ foreground, background, and spatial location (Fig. 1b, c). In each trial, participants were presented with four scene options combining different foreground and background elements: (1) correct (both foreground and background hits), (2) foreground-only (foreground hit and background miss), (3) background-only (foreground miss and background hit), and (4) miss (both incorrect). They had four seconds to select the scene they recognized from the encoding phase. The options were randomly placed in a cross-shaped array to avoid interference with spatial location memory. After making their selection, the chosen scene was shown at the center of the screen with four empty quadrants, and participants were asked to indicate the quadrant in which the scene had originally appeared. Following each recognition and location response, participants rated their confidence on a 3-point scale (1 = not confident, 3 = very confident). The test consisted of 72 trials, one for each encoded scene, presented in a randomized order.

Fig. 1.

Computerized emotional episodic memory task. (a) Each encoding trial included an emotional foreground superimposed upon a neutral background presented in one of four corners of the screen in a counterbalanced, pseudorandom order. At retrieval, participants were prompted to choose from four choices consisting of correct, foreground-only (with a novel background), background-only (with a novel foreground), and miss (with both novel foreground and background). After their response, they were asked to indicate which quadrant they remembered seeing the scene (location memory). (b) Response selection at recognition. (c) Location memory question example (used IAPS stimuli: 2005, 2800, 2095).

Task design and analysis

Emotion-dependent scene recognition was assessed by analyzing the proportion of correct responses, along with the distribution of incorrect response types: foreground-only, background-only, and miss. For location memory, we calculated the proportion of trials in which participants correctly identified the scene’s location. Trials in which both the foreground and background were wrong during the recognition phase were excluded, allowing us to evaluate location memory performance even in cases of partial scene recognition. This approach provided a more precise measure of participants’ ability to recall spatial context linked to encoded stimuli. Our rationale for employing a whole-scene recognition task, rather than the standard approach of separately probing foreground and background memory to assess emotional memory trade-off, was to dissociate a coherently bound emotional scene memory from a foreground-biased memory of the emotional object. Consequently, we hypothesized that demonstrating impaired recognition of the correct item-background pair for emotional scenes would be consistent with the trade-off effect, attributing the phenomenon more directly to a failure in remembering the contextually bound scene, even in its original form. Data were averaged across trials within each condition for each participant to increase the signal-to-noise ratio and to obtain robust estimates of memory performance, in line with standard practices in the field⁴⁸.

Results

Effects of emotion and age on scene recognition memory

First, we confirmed that participants’ ratings of emotional scenes significantly differed across emotional conditions, validating our emotional stimulus set (Fig. 2a, b, see supplementary materials Table S2 for detailed statistics). We compared the response rates across the four response types to assess overall memory performance. The vast majority of responses (over 80%) were concentrated in the correct and foreground-only categories, while the background-only and miss responses were infrequent and therefore insufficient for reliable statistical analysis (Fig. 2c). Moreover, analysis of the confidence rating yielded similar results (Fig. 2d, see supplementary materials Fig. S1 and Table S3 for detailed statistics). Therefore, subsequent analyses focused on comparing correct and foreground-only responses.

Fig. 2.

Valence and arousal ratings of the scenes, and the response rate of each option of the recognition memory test. (a) Valence ratings, (b) Arousal ratings, (c) Memory response rates. Error bars indicate the standard error of the mean. (d) Confidence ratings. Error bars indicate the standard error of the mean. Emotion conditions: Negative (“Neg”, red bars), Neutral (“Neu”, yellow bars), and Positive (“Pos”, blue bars). n.s, p > .0.10; †, p < .0.10; *, p < .0.05; **, p < .0.01; ***, p < .0.001.

We next examined recognition memory across emotional conditions using a two-way repeated-measures ANOVA with memory response type (correct, foreground-only) and emotion (negative, neutral, positive) as within-subject factors. A significant main effect of memory was observed, F(1, 162) = 43.55, p < 0.001, ηp² = 0.212, indicating that failure in scene recognition was primarily attributable to errors in background, rather than foreground, information. A significant main effect of emotion was also observed, F(2, 324) = 3.184, p = 0.043, ηp² = 0.019. Post hoc analyses (Holm–Bonferroni corrected) showed that scenes with negative emotional content were better remembered than both neutral (t(162) = 7.387, p < 0.001) and positive conditions (t(162) = 8.903, p < 0.001) (see supplementary materials Fig. S2).

Importantly, there was a significant interaction between memory response and emotion, F(2, 324) = 3.955, p = 0.020, ηp² = 0.024. To further explore this interaction, we conducted separate one-way repeated-measures ANOVAs for each memory response type. For correct responses, although paired tests indicated some emotion-related trends (Fig. 3a), the overall main effect did not reach statistical significance, F(2, 324) = 1.241, p = 0.290. In contrast, foreground-only responses showed a significant main effect of emotion, F(2, 324) = 6.342, p = 0.002, ηp² = 0.038. Post hoc tests (Holm–Bonferroni corrected) indicated significant differences among all three emotion conditions with the negative condition showing the highest response rates and neutral the lowest: negative vs. neutral, t(162) = 7.762, p < 0.001; negative vs. positive, t(162) = 3.879, p < 0.001; and positive vs. neutral, t(162) = 3.562, p < 0.001 (Fig. 3b). These findings suggest that when a scene is fully encoded, emotional valence has little impact on memory performance. However, when encoding is incomplete, emotionally salient foreground elements, particularly negative ones are more likely to be remembered than neutral ones, whereas background details tend to be forgotten.

Fig. 3.

Emotion condition comparisons of correct, foreground-only responses, and correlations with age. (a) Correct responses and their distribution for each emotion. (b) Foreground only responses and their distribution for each emotion. (c) Correlation between correct response rates and age. (d) Correlation between foreground only response rates and age. Error bars indicate the standard error of the mean. Emotion conditions: Negative (“Neg”, red bars), Neutral (“Neu”, yellow bars), and Positive (“Pos”, blue bars). n.s, p > 0.10; †, p < 0.10; *, p < 0.05; **, p < 0.01; ***, p < 0.001.

As expected, aging significantly decreased scene recognition performance, as indicated by a memory × age interaction, F(1, 162) = 9.277, p = 0.003, ηp² = 0.054. Correlation analyses (Pearson’s r, FDR-corrected) showed that age was negatively correlated with correct responses (r = −0.168, p = 0.031) and positively correlated with foreground-only responses (r = 0.205, p = 0.008), reflecting an age-related decline in detailed contextually bound memory (Fig. 3c, d).

Beyond this general decline, we also observed an emotion-specific modulation of the aging effect, as indicated by a significant three-way interaction between memory, emotion, and age, F(2, 324) = 3.084, p = 0.047, ηp² = 0.019. To determine which emotion condition drove this interaction, we conducted separate two-way ANOVAs (memory × age) for each emotion condition. The interaction between age and memory was significant for both positive scenes (F(1, 162) = 15.265, p < 0.001, ηp² = 0.086) and neutral scenes (F(1, 162) = 4.026, p = 0.046, ηp² = 0.024), but not significant for negative scenes. Follow-up correlation analyses (Pearson’s r, FDR-corrected) further confirmed this pattern: in the positive condition, age was negatively correlated with correct responses (r = −0.242, p = 0.002) and positively correlated with foreground-only responses (r = 0.235, p = 0.002) (Fig. 5a for the results divided by age groups). In contrast, no significant correlations were observed in the negative condition, and only a marginal correlation was found for foreground-only responses in the neutral condition (r = 0.152, p = 0.052).

Fig. 5.

Median split age group difference between correct, foreground-only, and location memory. (a) Correct and foreground-only, (b) Location memory. The red arrows highlight significant age correlation conditions. Error bars indicate the standard error of the mean. Emotion conditions: Negative (red bars), Neutral (yellow bars), and Positive (blue bars).

These results suggest that with age, the ability to retrieve integrated contextual details of the scene background deteriorates overall, while memory for positively-valenced foreground objects remains relatively preserved. This pattern aligns with prior findings on age-related shifts in emotional processing, in which older adults tended to show enhanced memory and processing of positive information^49,50. Nevertheless, the decrease in correct responses alongside the increase in foreground-only responses indicates an age-related loss in forming contextual associations in scene-based episodic memory even when positive events occur.

Effects of emotion and age on scene location memory

To examine spatial location memory across emotion conditions, we conducted a repeated-measures ANOVA to find a significant main effect of emotion, F(2, 324) = 5.388, p = 0.005, ηp² = 0.032. Post hoc tests (Holm–Bonferroni corrected) revealed that location memory was significantly better for both negative and positive conditions compared to the neutral condition (negative vs. neutral: t(162) = 2.793, p = 0.018; positive vs. neutral: t(162) = 2.638, p = 0.018) (Fig. 4a).

Fig. 4.

Emotion condition comparison of location memory, proportion of correct location memory between correct and foreground-only responses, and correlations with age. (a) Location accuracy and its distribution for each emotion, (b) Location memory for scenes in which participants made correct and foreground-only(fore-only) responses, and their distribution. (c) Correlation between location memory for positive scenes and age, (d) Correlation between location memory for negative scenes and age. Error bars indicate the standard error of the mean. Emotion conditions: Negative (“Neg”, red bars), Neutral (“Neu”, yellow bars), and Positive (“Pos”, blue bars). n.s, p > 0.10; †, p < 0.10; *, p < 0.05; **, p < 0.01; ***, p < 0.001.

To determine whether accurately remembering the background scene was associated with also remembering its location on the screen, we compared location accuracy between correct and foreground-only trials. Notably, no significant difference in location accuracy was found between the two response types, t(162) = 0.965, p = 0.336. This suggests that memory for emotional foreground content alone was sufficient to enhance memory of its location (Fig. 4b).

While there was no significant main effect of age on location memory, F(1, 162) = 1.990, p = 0.160, ηp² = 0.012, we observed a significant age × emotion interaction, F(2, 324) = 3.084, p = 0.047, ηp² = 0.019. Follow-up correlation analyses (Pearson’s r, FDR-corrected) indicated a significant negative correlation between age and location memory for negative scenes (r = –0.176, p = 0.024) but no significant effects for neutral (r = 0.034, p = 0.663) or positive scenes (r = −0.090, p = 0.252) (Fig. 4c, d. See Fig. 5b for the results divided by age groups). These results suggest that while aging does not broadly impair location memory on a computer screen, it diminishes the emotion-related enhancement of spatial memory, especially for negatively valenced content, which may reflect age-related reductions in negative emotional sensitivity^51,52.

Dissociable effects of depression and anxiety on emotional episodic memory

Given past reports that depression and anxiety are not only associated with aging and emotional processing^49,53 but also related to memory function even in non-clinical samples^54–59, we explored the potential influence of self-reported symptoms of depression, anxiety, and mood on emotional memory performance in our sample.

Taking into consideration that emotional effects on memory are also subject to age-related variation, we examined their association with age (Table 1). Results revealed significant negative correlations (Pearson’s r, FDR-corrected) between age and affect subscale of PANAS (Negative, r = −0.208, p = 0.007; Positive, r = −0.155, p = 0.047), as well as with depression scores from multiple surveys (PHQ-9: r = −0.153, p = 0.049; CES-D: r = −0.183, p = 0.018; BDI-II: r = −0.175, p = 0.025). Trait anxiety scores also decreased with age (STAI-II: r = −0.188, p = 0.015).

Table 1.

Correlation table showing associations between age and psychological survey scores.

n.s, p > 0.10; †, p < 0.10; *, p < 0.05; **, p < 0.01; ***, p < 0.001. All p values are FDR-corrected.

These findings are consistent with previous literature indicating that older adults tend to experience fewer negative emotions and report lower levels of depression and anxiety, potentially due to enhanced emotional regulation or reduced prioritization of negative emotion with age^60,61.

To further examine the influence of depression (BDI-II) and trait anxiety (STAI-Y2) on emotional episodic memory, we conducted a two-way repeated-measures ANOVA with memory response (correct, foreground-only) and emotion (negative, neutral, positive) as within-subject factors. We found significant main effects of both depression, F(1, 160) = 7.584, p = 0.007, ηp² = 0.045, and anxiety, F(1, 164) = 6.025, p = 0.015, ηp² = 0.036, which was maintained regardless of the specific depression measure used (see supplementary materials Table S4). Partial correlation analyses after controlling for anxiety (FDR-corrected) showed that depression was negatively correlated with correct responses (r = −0.191, p = 0.027). Conversely, anxiety was positively correlated with correct responses (r = 0.184, p = 0.030) after controlling for depression (Fig. 6a, b). Emotion-specific correlation analyses (FDR-corrected) revealed that anxiety was positively associated with correct memory in the neutral (r = 0.216, p = 0.027) and positive (r = 0.219, p = 0.027) conditions, but not negative condition (r = 0.045, p = 0.609). These results suggest that depression and anxiety have opposing relative effects on emotional episodic memory, with anxiety enhancing and depression impairing memory performance, particularly for positive and neutral content. In contrast, recent mood states, as measured by PANAS scores, did not significantly predict memory performance (See supplementary materials Table S5 for details). In addition, a significant three-way interaction was observed among memory, emotion, and anxiety (F(2, 320) = 4.717, p = 0.010, ηp² = 0.029), with the negative emotion showing an anxiety-related increase only for the foreground memory.

Fig. 6.

Dissociable effects of depression and anxiety on scene memory and location memory. Depression effect (controlling for anxiety) on (a) correct responses, (c) foreground-only responses, and (e) location accuracy. Anxiety effect (controlling for depression) on (b). correct responses, (d) foreground-only responses, and (f) location accuracy. Age and gender were common covariates. n.s, p > 0.10; †, p < 0.10; *, p < 0.05; **, p < 0.01; ***, p < 0.001. All p values are FDR-corrected.

For location memory, the results were consistent with those observed in scene recognition. A repeated-measures ANOVA revealed significant effects of depression (F(1, 160) = 4.340, p = 0.039, ηp² = 0.026) and anxiety (F(1, 160) = 7.172, p = 0.008, ηp² = 0.043) on spatial location memory. Correlation analyses (FDR-corrected) showed that while depression was negatively correlated with location memory accuracy (r = –0.179, p = 0.027), anxiety showed a positive correlation (r = 0.230, p = 0.015) (Fig. 6e, f). However, no significant interactions between emotion, age, and mental health scores were found.

When depression and anxiety were included as between-subject factors in the model, the previously significant three-way interaction among memory, emotion, and age was reduced to a marginal trend (F(2, 320) = 2.564, p = 0.079, ηp² = 0.016), possibly due to the decrease in affective scores in aging. Nevertheless, even when depression and anxiety were included as between-subjects measures, the interaction between age and memory responses remained significant (F(1, 160) = 9.170, p = 0.003, ηp² = 0.054), suggesting that the influence of mental health and age-related memory decline are independent. Overall, depression and anxiety exerted dissociable influences on emotional episodic memory, with depression impairing and anxiety enhancing memory performance across both scene and location tasks.

Discussion

The present study aimed to clarify how emotional episodic memory is influenced by different types of spatial contexts and how these effects interact with aging. First, we found that the scene’s background information was more prone forgetting compared to the foreground content. This is consistent with previous findings that showed that people often remember the objects in scenes better than the peripheral information^62,63, especially when there is not enough time to perceive all the details⁶⁴. Moreover, when participants failed to fully recognize the scene, memory for emotional foregrounds was better preserved than for neutral ones, while background (spatial) context was often generalized. This observation is also consistent with existing theoretical frameworks suggesting that emotional or arousing information is prioritized and recalled with greater accuracy than neutral content, at the expense of peripheral background information^40,65–70.

We found several interesting age-related changes in our study. First, we replicated the age-related decline in the ability to recognize background scenes reported in prior research^3,15,71. This may be attributed to reduced hippocampal function and structural atrophy that occur gradually across aging, which may have led to less detailed encoding and weaker integration of context-dependent memory^72–75, leading participants to use familiarity-based recognition rather than recollection. Interestingly, positive conditions showed even stronger memory decline in scene background memory, leaving them with a stronger bias for remembering foreground objects. This positivity-specific memory effect may reflect the combined effect of an age-related decline in episodic memories and heightened sensitivity to positive emotions compared to negative affective states. Together, these factors may ultimately contribute to the well-documented positivity bias observed in aging^76–79.

In contrast to the peripheral background information, location memory (spatial location of the scene) was enhanced by emotion, corroborating previous findings from emotional object-location memory studies^41,42,80,81. Interestingly, even when participants did not remember all the details of an event, they still remembered where it happened. This suggests that emotion influences background scene and spatial location memory through dissociable mechanisms, perhaps through different neural pathways involving the parahippocampal cortex (PHC) and retrosplenial cortex (RSC), respectively.

On the other hand, there was no overall age-related decline in location memory, possibly both because the present study was conducted in an on-screen space, rather than large-scale navigable spaces and because aging may affect the PHC later than other regions such as the RSC^82,83. Although we found an overall reduction in the emotional enhancement of spatial location memory, this decrease was more pronounced for scenes with negative emotional valence. This may be both because younger people showed a stronger enhancement effect with negative emotions and because older adults tend to experience stronger positive affect and attenuated responses to negative stimuli. Diminished negative affect in older adults has been linked to reduced amygdala activity compared to younger adults^84–86, potentially leading to a dampened emotional modulation of location memory⁶⁰.

Age-related decreases in amygdalar integrity has been well-documented. Notably, the degradation of amygdala function appears to have valence-specific effects, reducing responses to negative emotions while leaving responses to positive emotions relatively intact^51,52. These findings suggest that age-related decline in amygdala activity may particularly diminish reactivity to negative stimuli⁸⁷. That said, aging is not solely marked by decline in neural function; older adults often show improved emotional regulation, particularly in response to negative stimuli. This may be supported by increased prefrontal cortex activity during emotional processing, indicative of greater regulatory control^61,88.

It is noteworthy that our study sample featured a continuous and relatively uniform age distribution spanning young to middle-aged and older adulthood. Although our participants, on average, were younger than those typically sampled in aging research, we were still able to detect significant age-related effects. By including a robust representation of individuals from middle adulthood to those in their late 60s, our data allowed us to capture the onset and progression of these age-related changes, rather than just sampling from a traditionally “older” age range. Our findings, therefore, suggest that the decline in episodic memory is not an abrupt event that suddenly manifests in late life, but rather a more gradual process that begins during middle age.

Lastly, although exploratory, our findings suggest that depression and anxiety may differentially influence emotional memory performance. Depression was linked to a general decline in memory performane in our task, while anxiety was associated with improved memory for emotional foreground information. These findings are consistent with prior work suggesting that depression and anxiety have dissociable effects on cognition⁸⁹. Specifically, hippocampal dysfunction may underlie memory impairments in depression^55,90–92, whereas heightened vigilance and increased hippocampal volume associated with anxiety may enhance memory encoding^93–97. These effects were observed in a non-clinical sample, highlighting that emotional traits can influence memory even outside of clinical populations. Notably, the impact of individual differences in affective health on memory appears to be distinct from that of cognitive aging, underscoring the importance of considering these factors independently in emotional memory research.

The present findings should be interpreted in the context of several methodological considerations that also have implications for future research. Our decision to conduct this study online allowed for the recruitment of a large and diverse sample, enhancing the generalizability of our findings beyond the typical laboratory. Although this approach inherently involves a tradeoff with the consistency of a well-controlled experimental setting, our results were robust enough to emerge through the potential for increased noise, variability in screen size, and naturalistic environments. Nevertheless, a future replication in other settings, including a controlled laboratory, would serve to validate the generalizability of these effects. Another point to consider is that, while the current study relied on averaged metrics to compare across conditions within an individual, future studies should consider employing Linear Mixed Models (LMMs) to allow for a more granular analysis of trial-by-trial variability and provide a deeper understanding of interactions in emotion, memory, and aging.

Also, our exploratory analyses of depression and anxiety were conducted in a non-diagnosed, community-based sample. While this approach is valuable for understanding how individual affective characteristics influence memory in the general population, it limits the applicability of our findings to clinical disorders. For instance, emotional episodic memory may operate differently in individuals with diagnosed conditions such as Post-Traumatic Stress Disorder (PTSD), in which memory loss or distortion is often a core feature⁹⁸. Therefore, extending this paradigm to clinical populations is a crucial next step to determine whether these mechanisms represent a dimensional trait that spans the continuum from healthy to clinicaly diagnosed states.

Conclusions

In summary, this study demonstrated that while emotions modulate our episodic memory in context-specific ways, older adults are both affected by a general decline in episodic memory binding and a reduction in the emotional modulation of memory. Our cross-sectional findings reveal that emotional salience selectively influences memory for background scene contexts and spatial location, sometimes depending on the valence of the foreground objects, and that these effects are further modulated by individual depression and anxiety scores. By deepening our understanding of the neurocognitive underpinnings of how older adults may experience emotional episodic memory differently from younger adults, we hope this work will pave the way for future research that not only refines theoretical models but also informs practical strategies to support memory function across the lifespan.

Transparency and openness

Data were analyzed using JASP, and MATLAB^99,100. Adhering to the journal’s Transparency and Openness Promotion guidelines, we report all measurements and design elements and properly cite the methods used, according to the APA standards.

Author contributions

Minsok Koo and Sang Ah Lee both contributed to conceptualizing, investigating, and developing the methodology for this project. Minsok Koo was responsible for leading the data collection, analysis, validation, visualization, and writing of the original draft. Sang Ah Lee led the funding acquisition, resources, and supervision, and provided support for formal analysis and visualization. Both Minsok Koo and Sang Ah Lee contributed equally to the writing, review, and editing of the manuscript.

Funding

This work was funded by grants from the National Research Foundation of Korea (NRF) (No. 2021M3A9E4080780 and No. RS-2023–00265406), awarded to Sang Ah Lee by the Korean government (MSIT), the Creative Pioneering Researchers Program at Seoul National University (339–20230014), the Global-LAMP Program of the National Research Foundation of Korea (NRF) grant funded by the Ministry of Education (RS-2023–00301976), and Samsung Electronics Co., Ltd (A0426-20220108).

Data availability

Materials, including the data, scripts used for running the experiment, and data analysis scripts are available at https://osf.io/mdauz/?view_only=c599f977baf6441db4ffc6e07c78da37.

Competing interests

The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work the author(s) used ChatGPT in order to improve the language and clarity of the manuscript. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the published article.