How affective information impacts working memory and long-term memory in individuals with depressive tendencies?

Han Han; Huxi Li; Xinyu Yu; Nana Sun

doi:10.1186/s40359-025-03946-9

. 2026 Jan 16;14:214. doi: 10.1186/s40359-025-03946-9

How affective information impacts working memory and long-term memory in individuals with depressive tendencies?

Han Han ^1,^✉,^#, Huxi Li ^1,^#, Xinyu Yu ¹, Nana Sun ^2,^✉

PMCID: PMC12892467 PMID: 41545873

Abstract

Background

Depressive tendencies are considered a subclinical condition associated with increased risk for developing major depressive disorder. Previous research has shown that depressed individuals exhibit mood-congruent effects on long-term memory. Moreover, studies have indicated that the maintenance of working memory plays a crucial role in the formation of long-term memory. However, the relationship between affective working memory and long-term memory remains unclear. This study examines the connection between affective working memory and long-term memory in individuals with depressive tendencies.

Methods

Two experiments were conducted. Experiment 1 assessed immediate and delayed recall performance, with all participants completing both tasks. Building upon Experiment 1, Experiment 2 introduced systematic cognitive load manipulation (low vs. high), with participants randomly assigned to a low-load or high-load condition.

Results

Repeated-measures ANOVA revealed the following findings: (1) Affective information significantly impaired working memory performance, with the impairment from negative words further disrupting long-term memory. (2) In the low cognitive load condition, individuals with depressive tendencies exhibited significantly poorer working memory performance for negative information compared to healthy controls, although this discrepancy did not extend to long-term memory.

Conclusion

The results indicate that individuals with depressive tendencies exhibit deficits in working memory for negative information, whereas their performance in long-term memory does not differ from that of the healthy control group.

Keywords: Depressive tendencies, Affective working memory, Long-term memory, Mood-congruent bias, Emotional context insensitivity

Background

Depression is one of the major types of mood disorders, characterized primarily by marked and persistent low mood [1]. Depressive tendencies refer to a condition in which individuals experience depressive states that affect their daily functioning but do not meet the diagnostic criteria for depression, representing an intermediate state between depressive disorder and the absence of symptoms [2]. There is a stepwise, sequential relationship between depressive tendencies and major depressive disorder [3]. Consequently, individuals with depressive tendencies are considered a high-risk population for developing clinical depression. Investigating their cognitive characteristics and behavioral manifestations may improve the understanding of depressive risk factors.

Memory bias is a prominent cognitive characteristic of individuals with depression, often manifesting as a preference for recalling negative emotional information. The mood-congruent memory bias theory suggests that individuals are more likely to remember stimuli that match their current emotional state, compared to mood-incongruent information [4]. This theory has received substantial empirical support, particularly in studies on long-term memory in depressed populations. In autobiographical memory research, individuals with depression show greater difficulty recalling specific positive events than negative ones [5, 6]. Implicit memory, a form of long-term memory that operates outside conscious awareness [7, 8], shows similar patterns: individuals with depressive states consistently display mood-congruent biases [9–12], such that they tend to retrieve a greater amount of negative emotional information in implicit memory tasks [11]. LeMoult and Gotlib [8] propose that the preferential processing of negative information may sustain and worsen depressive symptoms. Specifically, when depressed individuals utilize existing cognitive schemas to interpret current experiences, their long-term memory biases toward negative information promote maladaptive processing of present stimuli. This cognitive pattern reinforces depressive symptoms, thereby creating a cycle that perpetuates negative memory biases [8, 13, 14]. Consequently, elucidating the mechanisms underlying long-term memory biases in individuals with depressive tendencies may provide critical insights for developing targeted interventions to disrupt this harmful cycle.

Long-term memory is responsible for the permanent storage and retention of information, transferring it from short-term memory. The longer an item remains in short-term memory, the stronger the memory trace becomes, increasing its likelihood of being transferred to long-term memory [15]. Building on the concept of short-term memory, Baddeley [16] proposed working memory, describing it as a limited-capacity system for temporarily maintaining and storing information, comprising a central executive system and several subsystems (e.g., the phonological loop and visuospatial sketchpad). The central executive system is the most critical component, responsible for the coordination of attentional resources [17]. This function not only involves the allocation of attentional resources but also the shifting of attention away from irrelevant tasks to focus on central task demands. In affective working memory, individuals experience conflicts in attentional resource allocation between processing the informational and emotional attributes of stimuli [18–20]. When excessive attentional resources are devoted to emotional processing, fewer resources remain available for the ongoing memory task, which may consequently impair working memory representations [19, 21]. Therefore, during the processing of emotional information, individuals must actively coordinate attentional resources—redirecting or focusing attention from emotional processing toward the memory task. However, Gotlib and colleagues, through a review of previous studies, found that individuals with depression exhibit difficulties coordinating task demands and emotional processing when handling negative emotional information [8, 22]. Specifically, when memory demands compete with emotional processing, individuals with depressive tendencies allocate more cognitive resources to negative emotional processing, leaving fewer resources available for the memory task than healthy controls. Consequently, their working memory performance for negative emotional information may be poorer.

Previous empirical studies have shown that deficits in the inhibition of negative information among individuals with depression exert a detrimental impact on working memory. Using a negative priming paradigm, Leung et al. [23] established two conditions—neutral words and negative words. In each condition, the same type of words first served as distractor stimuli in the priming phase and later as target stimuli in the subsequent test phase. In both the priming and target phases, the distractor and target stimuli were presented simultaneously. Participants were instructed to identify the target stimulus as quickly and accurately as possible by pressing the corresponding key, followed by a verbal report of the target. Results revealed that, compared to neutral words, individuals with depression exhibited longer reaction times in the negative-word condition [23], indicating that they had greater difficulty disengaging attention from negative emotional distractors during working memory processing. This finding has been replicated in ERP studies [24]. In a directed forgetting paradigm, participants first learned a list of positive and a list of negative words and were then informed that only one of the lists would be relevant for a subsequent recognition task, with the other list to be ignored. This design was used to examine the effect of emotional information on working memory updating. Results showed that when instructed to forget negative words, compared with the control group, individuals with depression exhibited longer reaction times [25] and lower accuracy [26] in response to negative probe stimuli, suggesting that negative words exerted a detrimental influence on working memory updating [25, 26]. Further neuroimaging and electrophysiological evidence supports these behavioral findings. Studies have shown that when individuals with depression attempt to identify or selectively forget negative words, brain regions associated with cognitive control show greater activation than in the healthy control group, such as the left inferior frontal gyrus [27, 28]. Moreover, event-related potential components related to sustained attention exhibit stronger amplitudes, whereas components related to cognitive control display weaker amplitudes [29], accompanied by poorer behavioral performance, such as lower accuracy and longer reaction times [27–29]. A meta-analysis of 65 studies conducted by Schweizer et al. [20] further revealed that individuals with mental health problems exhibited impaired working memory performance for emotional materials compared with healthy individuals. Collectively, these findings suggest that individuals with depression show cognitive control deficits in processing negative information, which in turn further impairs working memory representations. In addition, the Emotion Context Insensitivity (ECI) model posits that individuals with depressive symptoms display blunted emotional responses to both positive and negative stimuli compared with the healthy control group [30]. This model has been primarily supported by studies using physiological indicators [30, 31]. Accordingly, during the processing of emotional words, individuals with depressive tendencies may exhibit slower emotional processing and experience greater difficulty reallocating attentional resources from emotional to task-related processing, leading to impaired performance in immediate recall or working memory. In summary, the present study hypothesizes that individuals with depressive tendencies will demonstrate poorer working memory performance for negative information compared with the healthy control group.

LeMoult and Gotlib [8] suggest that cognitive control deficits in working memory among individuals with depression are closely associated with their preferential recall of negative emotional information in long-term memory. Specifically, during the working memory stage, depressed individuals allocate excessive attentional resources to negative information, engaging in self-referential processing that activates their pervasive negative cognitive schemas, thereby exhibiting mood-congruent biases and further influencing the manifestation of depressive symptoms. This process may involve elaborative processing [22], a mechanism that enhances memory by integrating multiple semantic features of stimuli into existing associative networks such as cognitive schemas [32]. Empirical evidence indicates that such elaborative processing primarily enhances long-term memory representations rather than working memory [33, 34]. Cotton and Ricker [35] further propose that working memory and long-term memory engage dissociable neural substrates, with early-stage long-term encoding partially overlapping with working memory networks, while working memory operations complete earlier in the processing timeline. Based on this, the present study proposes that the ECI model and mood-congruent bias describe processing at different stages. The ECI model characterizes the processing and storage of information in working memory among individuals with depressive tendencies, whereas mood-congruent bias pertains to the processing in long-term memory, in which individuals engage in self-referential processing of mood-congruent information, facilitating its successful storage in long-term memory. In short, the preferential recall of negative information in individuals with depressive tendencies may occur during long-term memory processing, whereas working memory processing is more likely to be negatively influenced by emotional content.

The primary focus of the present study concerns the processing of affective information in working memory among individuals with depressive tendencies. At the same time, emotional effects were also assessed in a long-term memory task as a point of comparison and supplementation, allowing for a more comprehensive understanding of emotional processing across different memory stages. Based on the ECI model and the characteristic deficits in cognitive inhibition observed in individuals with depression, the present study hypothesizes that, compared to neutral words, individuals with depressive tendencies will show reduced working memory for emotional words. Meanwhile, due to ruminative thinking and self-referential processing in these individuals, long-term memory representations are expected to exhibit mood-congruent biases. Accordingly, it is predicted that, during long-term memory processing, individuals with depressive tendencies will demonstrate superior processing of negative words relative to healthy controls. In summary, the present study proposes two hypotheses:

In line with the ECI model, individuals with depressive tendencies will exhibit reduced immediate recall accuracy for emotional words compared to neutral words, indicating impaired working memory maintenance.
Consistent with mood-congruent effects, individuals with depressive tendencies will show enhanced delayed recall accuracy for negative words relative to healthy controls.

The present study

To clarify whether the memory bias toward emotional information in individuals with depressive tendencies is an inherent trait rather than influenced by affective stimuli, this study included healthy individuals as the control group in both experiments.

In Experiment 1, three different types of words (neutral, positive and negative) were used as memory materials, and participants’ working memory and long-term memory were assessed through immediate and delayed recall tasks. If individuals with depressive tendencies recall more neutral than emotional (positive or negative) words in the immediate recall task, this would suggest weaker processing of affective information at the working memory stage, supporting the ECI theory. Furthermore, if individuals with depressive tendencies exhibit enhanced delayed recall of negative words compared to healthy controls, this would indicate enhanced consolidation of negative information in long-term memory, consistent with mood-congruent bias.

Building on prior evidence, Hu et al. [36] demonstrated using a directed forgetting paradigm that individuals with depressive tendencies had greater difficulty forgetting to-be-forgotten negative items than healthy controls, but increased cognitive load facilitated their directed forgetting of negative material. This suggests that high cognitive load may restrict the elaborative processing of negative information in depressive individuals [36]. In Experiment 2, this study manipulated cognitive load levels in a concurrent task (i.e., varying the presentation duration of digits following memory stimuli, as in Labaronne et al. [37]) to examine how memory biases in individuals with depressive tendencies change under cognitive load. It was predicted that increasing cognitive load would attenuate the negative memory bias in individuals with depressive tendencies, leading to performance comparable to healthy controls.

Experiment 1

Method

Participants

A priori power analysis conducted in G*Power 3.1 (effect size f = 0.25, Inline graphic = 0.05, power = 0.80) indicated that a minimum of 28 participants was required. To screen for depressive tendencies, participants were initially recruited online to complete the Self-rating Depression Scale (SDS), developed by W.K. Zung in 1965 [38]. The SDS consists of 20 items rated on a 4-point scale. Standard scores were derived by multiplying the raw total score by 1.25 and rounding to the nearest integer [38]. Based on Chinese norms [39], participants with SDS standard scores above 52 were classified into the depressive tendency group, whereas those scoring below 52 were placed in the control group.

The final sample consisted of 44 university students (29 female) aged between 18 and 24 years old (M = 20.09, SD = 1.84). All participants were native Chinese speakers with normal or corrected-to-normal vision. Each received ¥5 as compensation for completing the in-person experiment. In this experiment, the depressive-tendency group (n = 20) had a mean SDS standard score of M = 63.75, SD = 7.23, while the healthy control group (n = 24) had a mean SDS standard score of M = 41.67, SD = 6.65. The correlations between SDS scores and recall performance are presented in Appendix 1.

Materials and design

Drawing on previous research [13, 40], 105 words (35 positive, 35 negative, and 35 neutral) were selected from the Chinese Affective Bivalent Word Bank according to three criteria: valence, arousal, and word frequency. For valence, positive words had baseline scores >7, negative words <3, and neutral words ranged between 4-6. For arousal, both positive and negative words required scores >6, while neutral words <4. To balance word frequency, items with extremely high or low frequency were excluded. Subsequently, five psychology students rated these 105 words on a 5-point scale (1 = very negative, 5 = very positive). Significant differences in ratings were found across the three word categories (see Table 1). During the experiment, words were presented using a Latin square design to ensure equal positional frequency for each word type.

Table 1.

Descriptive statistics and ANOVA results of affective word characteristics

Affective information	N	Valence	Arousal	Frequency	Scores
Positive words	35	7.28 (0.18)^a	6.53 (0.31)^a	48 (76.87)^a	4.57 (0.40)^a
Neutral words	35	5.25 (0.46)^b	3.70 (0.32)^b	227.94 (497.94)^a	3.03 (0.35)^b
Negative words	35	2.66 (0.26)^c	6.62 (0.31)^a	30.57 (50.34)^a	1.34 (0.27)^c

Open in a new tab

Superscript letters (a, b, c) within each column represent results from Dunnett’s post hoc test. Different letters indicate significant pairwise differences (e.g., valence: Inline graphic ), whereas identical letters indicate non-significant differences (e.g., word frequency: ). For word frequency measures, , , , , , . The descriptive distribution of the words is shown in Fig. 5 in Appendix 1

The experiment employed a 2 (group: depressive-tendency group, healthy control group) Inline graphic 3 (affective information: positive emotional words, neutral words, negative emotional words) mixed design, with group as a between-subjects factor and affective information as a within-subjects variable.

Procedure

This experiment consisted of two phases: a practice phase and a formal experiment (see Fig. 1).

During the practice phase, participants fixated on a central cross (1000 ms), followed by the presentation of an affective word (1500 ms). After a 150 ms interval, a blank screen appeared for 3000 ms, following the timing parameters of Labaronne et al. [37]. The next word then appeared. After every five words, participants were required to recall the words in the order presented and confirm their recall by pressing the Enter key before proceeding. After completing three rounds of recall, participants performed 25 arithmetic judgment tasks (e.g., “Is 2 Inline graphic 5 = 15 correct?”). They were required to achieve at least 70% accuracy; otherwise, the arithmetic task segment was repeated. Upon completion, participants were asked to freely recall all words that had appeared during the experiment.

The formal experiment followed the same procedure as the practice phase, except participants completed 18 rounds of immediate recall tasks instead of 3.

Statistical analyses

According to the three-standard-deviation principle, values of immediate recall accuracy rates that fell beyond three standard deviations were treated as outliers and removed. In Experiment 1, all data points fell within three standard deviations, yielding a final sample of 44 participants for analysis.

Compared with traditional ANOVA, Bayesian ANOVA provides a more comprehensive description of the results. On one hand, Bayes factors allow for the direct quantification of the relative evidence between competing hypotheses, indicating not only the degree of support for a given hypothesis but also the comparatively weaker support for an alternative hypothesis. On the other hand, researchers are not restricted to dichotomous significance judgments; instead, the magnitude of the Bayes factor can indicate the strength of evidence supporting a particular hypothesis [41]. In the present study, Bayesian repeated-measures ANOVA was conducted using JASP [42]. Bayesian models were compared to the null model, and the “Across matched models” method recommended by Mathôt [43] was employed to calculate inclusion ( Inline graphic ) and exclusion () Bayes factors, which quantify the relative evidence for or against including each effect in the model. A Bayes factor ( or ) of 1 indicates equal support for the null () and alternative () hypotheses. A greater than 1 indicates increasing support for the alternative hypothesis ( Inline graphic over ), whereas a less than 1 favors the null. Conversely, (supporting over ) is the reciprocal of , such that values greater than 1 support the null hypothesis, and values less than 1 favor the alternative. The interpretation followed the guidelines of Jarosz and Wiley [44].

In this experiment, group and affective information were treated as predictors, and accuracy in the immediate and delayed recall tasks served as the dependent variables in a Bayesian repeated-measures ANOVA. To assess individual differences, such as working memory capacity, semantic memory ability, gender, and age, supplementary analyses were conducted (see Appendix 2). Following the approach of Labaronne et al. [37], individual differences in working memory capacity were examined by dividing participants into high- and low-capacity groups based on the median of total immediate recall accuracy. Participants with total immediate recall accuracy at or below the median were classified as the low-capacity group, while the remainder were classified as the high-capacity group. If immediate recall performance did not interact with other variables, this would indicate that individual differences in working memory capacity did not influence affective working memory performance. According to Baddeley [45], semantic processing is closely linked to long-term memory. Therefore, in the present study, the median of total delayed recall accuracy was used as an index of semantic memory ability for analysis. Finally, in both immediate and delayed recall analyses, gender and age were included as covariates to examine whether demographic variables influenced the results.

Results and discussion

Main analysis

At immediate recall (Fig. 2), the likeliest model included only the main effect of affective information ( Inline graphic ). The evidence for including affective information was decisive (). Recall performance for neutral words (M = 0.92, SD = 0.12) was higher than for positive emotional words (M = 0.89, SD = 0.10) and negative emotional words (M = 0.88, SD = 0.11). There was weak evidence against including the main effect of group ( Inline graphic ), and moderate evidence against the interaction between group and affective information (). These findings suggest no group differences in working memory performance.

At delayed recall, the likeliest model similarly included only the main effect of affective information ( Inline graphic ). The evidence for including affective information was extremely strong (). Recall performance for neutral words (M = 0.17, SD = 0.09) and positive emotional words (M = 0.16, SD = 0.09) was higher than for negative emotional words (M = 0.12, SD = 0.07).There was weak evidence against the main effect of group ( Inline graphic ) and moderate evidence against the interaction (). Thus, no group differences in long-term memory performance were observed.

Discussion of results

This experiment revealed that individuals indeed exhibit memory biases for affective information.

First, memory for affective information was not better but worse. On the one hand, this finding supports the ECI theory, suggesting that individuals with depressive tendencies show reduced responsiveness to affective information in working memory. On the other hand, the poorer performance could be attributed to affective words triggering more task-irrelevant cognitive processing, such as self-reflection or associative thoughts unrelated to the task. This may have led to reduced cognitive resources for task-relevant encoding [21].

Second, in contrast to prior evidence [13, 20], this emotional memory bias was observed in both healthy individuals and those with depressive tendencies, with no difference in memory for negative words between the two groups. According to Labaronne et al. [37], the processing time for a single stimulus is approximately 650 ms. This suggests that the word presentation and interval durations may have been too long, affording participants sufficient time for rehearsal and potentially leading to a ceiling effect. To address this, Experiment 2 introduced a concurrent task with varied time constraints to manipulate available cognitive resources.

Supplementary analyses revealed that working memory capacity, semantic memory ability, gender, and age did not significantly influence individuals’ affective memory performance (see Appendix 2). In addition, during the formal experiment, participants memorized a total of 90 words, and their recall accuracy in the delayed recall task was approximately 16% (see Table 2). This level of performance is consistent with previous findings [37, 46], indicating that the delayed recall task used in the present study provides a valid measure of long-term memory.

Table 2.

Recall accuracy across cognitive load (CL)

Task	Group	Experiment 1	Experiment 2
Task	Group		Low CL	High CL
Immediate recall	Depressive-tendency group	0.91(0.12)	0.78(0.11)	0.70(0.10)
Immediate recall	Healthy control group	0.89(0.09)	0.86(0.11)	0.72(0.15)
Delayed recall	Depressive-tendency group	0.17(0.08)	0.10(0.06)	0.12(0.06)
Delayed recall	Healthy control group	0.14(0.05)	0.12(0.06)	0.10(0.08)

Open in a new tab