Neural correlates of stereotype processing in major depressive disorder: An event-related potential study

Abstract

BACKGROUND

Stereotype is a specific social cognitive schema that often leads to cognitive biases and prejudice, negatively impacting social interaction. Previous studies have focused almost exclusively on healthy populations, so it remains unclear how these processes unfold in major depressive disorder (MDD).

AIM

To examine the neural mechanisms of stereotype activation in MDD and differences between patients and healthy individuals during stereotype processing.

METHODS

Twenty-eight patients with MDD and 28 healthy controls enrolled in this study. All participants completed the gender stereotype task, with behavioral and electroencephalogram data recorded. We focused on analyzing behavioral data and event-related potential (ERP) components, including N170, N400, and P600. Furthermore, we analyzed the correlation between clinical features and ERP data in patients with MDD, with the false discovery rate correction applied.

RESULTS

Behaviorally, both patients with MDD and healthy control participants had significantly longer reaction times when processing stereotype-incongruent information than congruent information (all P < 0.05). The ERP results showed that both groups elicited higher N170 amplitudes when processing stereotype-incongruent information (all P < 0.05). Moreover, male participants exhibited larger N170 amplitudes than females when processing such incongruent information (all P < 0.05). The MDD group showed larger N400 amplitudes than healthy controls when processing stereotype-incongruent information (all P < 0.05). Additionally, correlation analysis between clinical characteristics and ERP data in the MDD group revealed that disease duration was positively correlated with N400 amplitude (female stereotype-male face; false discovery rate-adjusted P = 0.036).

CONCLUSION

Processing counter-stereotypical information may require greater cognitive resources, as reflected by an enhanced N170 effect, particularly in males. Patients with MDD may also need additional cognitive resources to integrate counter-stereotypical information, as reflected by an enhanced N400 effect, which appears to be more pronounced with longer illness duration.

Key Words: Major depressive disorder; Event-related potential; Social cognition; Stereotype; Neural mechanism

Core Tip: This study combined the gender stereotype task paradigm with event-related potential technology to explore stereotype processing in patients with major depressive disorder (MDD) and whether there are different neural mechanisms between patients with MDD and healthy individuals. We found that processing counterstereotypical information may require greater cognitive resources, as reflected by an enhanced N170 effect, particularly in males. Patients with MDD may also need additional cognitive resources to integrate counter-stereotypical information, as reflected by an enhanced N400 effect, which appears to be more pronounced with longer illness duration.

INTRODUCTION

According to the 2023 epidemiological data from the World Health Organization report, the global prevalence of major depressive disorder (MDD) is estimated at 3.8%, affecting approximately 280 million individuals. Critically, approximately 30%-35% of individuals with MDD attempt suicide annually[1]. The incidence of MDD has been steadily increasing. Projections indicate that by 2030, as many as 300 million individuals could be affected, potentially making MDD the leading cause of global disability-adjusted life years[2].

Depression impairs individual functioning across various aspects, affecting both external social functioning and internal cognitive systems. Specifically, it may exacerbate interpersonal conflicts, degrade occupational performance, hinder social activities, and even lead to self-harm behaviors. Meanwhile, cognitive functions such as processing speed, decision-making capacity, and sustained attention often exhibit significant declines in individuals with MDD. It is worth noting that these impairments often remain even when emotional symptoms are no longer prominent, which greatly hinders the return of social cognitive function and worsens long-term outcomes in individuals with MDD[3].

Stereotype is a specific social cognitive schema, referring to the generalized and fixed beliefs about the characteristics of members of a social group and the perceived causes of these attributes[4]. Specifically, when an individual perceives a member of a social group, the categorical representation of that group is automatically activated. Neural excitation then spreads from the individual’s group concept node to the associated stereotype feature nodes, thereby enabling rapid processing of stereotype-consistent information, namely, categorical cues that activate the stereotype. This process largely guides and shapes the individual’s social cognition and behavioral responses. However, stereotypes often lead to cognitive biases and prejudice. From the perspective of social cognition, stereotypes function as cognitive shortcuts by simplifying complex social information. However, the rigid nature of stereotypes distorts individuals’ judgments of others and becomes internalized as stable, biased schemas. This cognitive bias often makes it difficult for people to accurately understand others’ feelings and communicate effectively during social interactions, thus having a negative impact on individuals’ social adjustment[5]. In fact, many stereotypes do not originate from direct personal experiences but are acquired through interpersonal communication[6]. This suggests that interpersonal communication plays a crucial role in the consensual process by which stereotypes are formed, revised, and maintained in social groups. Through verbal, behavioral, and non-verbal cues, individuals convey and receive information related to stereotypes about members of different social groups. This exchange of information affects how individuals perceive those around them and gradually shapes their stereotypical thinking. Interpersonal communication is involved not only in the initial formation of stereotypes, but also in their subsequent modification and persistence. Under certain conditions, sustained and meaningful interpersonal contact may facilitate stereotype revision and facilitate a more accurate understanding and evaluation of others. Such social interactions promote more detailed and impartial cognitive development, thereby reducing the prejudice and discrimination that arise from stereotypes. This framework is particularly relevant to depression, a disorder characterized not only by persistent negative affect but also by impaired social cognition and interpersonal functioning[7]. Under the pathological condition of depression, the dynamic updating mechanism of stereotypes may be systematically destroyed. Individuals with depression frequently exhibit social withdrawal, reduced emotional expression, and negative cognitive biases. These symptoms may limit their access to a variety of social information and also bias their processing toward negative cues that are biased towards existing stereotypes, further exacerbating the rigidity of social cognition[8]. This kind of cognitive rigidity not only prevents patients from obtaining effective social support and deepens their social isolation but may also gradually impair the cognitive flexibility they need to update and revise their social schemas through interpersonal interaction. As a result, patients may become trapped in persistent negative and oversimplified cognitive patterns. These patterns constitute a critical factor contributing to the chronicity of depression and the impairment of social functioning.

A large body of research have demonstrated that the activation of stereotypes occurs automatically, mainly through rapid and unconscious processes. The content and valence of these activated stereotypes inevitably influence subsequent information processing[9], highlighting the existence of a stereotype activation effect[10-13]. Due to the limitations of behavioral experimental techniques previously available in this field, researchers have a limited understanding of when stereotype activation begins and the cognitive processing course of its automatic activation. With the development of neuroelectrophysiological technology, event-related potential (ERP) has provided an effective approach to addressing this issue. Owing to their high temporal resolution, ERP allow researchers to capture neural activity features underlying cognitive processing in real time. This capability enables the identification and separation of individual cognitive processes from the initial stages of processing[14]. So, researchers have employed ERP to investigate the continuous cognitive processes that occur when stereotypes are automatically activated. They aim to examine how these processes shape subsequent cognitive processing and behaviors, as well as the mechanisms that support these effects. For example, Brusa et al[15] employed ERP to examine the neural correlates of implicit ethnic stereotype activation. Results revealed that stereotype-incongruent information elicited an enhanced N400 component, reflecting the difficulty of semantic integration. In addition, this neural response is significantly correlated with individual prejudice scores. In another study, the researchers showed participants sentences containing occupational nouns associated with gender stereotypes, along with inverted pronouns that matched or did not match these nouns. They found that the ERP effects elicited by stereotype incongruence are very similar to the P600 component associated with syntactic violations[16]. Similarly, Canal et al[17] reported that conflicts arising from gender stereotypes can elicit a pronounced P600 effect, suggesting that this component reflects the reintegration of semantic information with stereotypical knowledge. In a recent study, researchers adopted an image priming paradigm. In this paradigm, participants were first presented with various types of priming images (e.g., gender-related images), followed by gender stereotype information that was congruent or incongruent with the priming images. Research has found that gender stereotype conflict elicits an N400 effect with distinct topographical distributions: Upper and secondary categories activated left frontal regions, whereas typical examples and counter-examples engaged right frontal regions[18]. These research findings indicate that electrophysiological measures are effective in capturing the processing of stereotypes while avoiding interference from social expectation deviations. ERPs provide a powerful tool for examining implicit stereotypes and reveal neural processes that occur under unconscious or unintentional control. This offers researchers a unique neuroelectrophysiological perspective on the automatic nature of social cognition.

However, we observed that most existing ERP studies on stereotypes have excluded individuals with a history of mental illness or current psychological issues, which limits the applicability of their research findings to clinical populations. To fill this gap, the present study first incorporates patients with MDD into the ERP research framework concerning stereotype processing. We aim to compare the ERP components and time-frequency features of patients with MDD and healthy controls when they process stereotype-congruent and incongruent social information. It should be noted that we selected gender stereotypes as the stimulus material for the present study, since they are among the most widely studied and culturally stable forms of social stereotypes. Across cultural contexts, the social category of “gender” possesses a relatively stable and similar conceptual core across most cultural systems. For instance, regardless of variations in national, ethnic, or cultural backgrounds, traits such as “gentle” and “virtuous” are generally stereotyped as feminine attributes, whereas “masculine” and “strong” are commonly stereotyped as masculine attributes. This crosscultural stability accounts for why stereotypes of males and females show such strong similarities in their defining features and content. This relative cross-cultural consistency makes gender stereotypes a useful entry point for examining the core neurocognitive mechanisms underlying social category processing while reducing the confounding influence of cultural variation. Based on this theoretical framework, our study employed the gender stereotype task (GST) to investigate the neural correlates of stereotypes in patients with MDD and healthy controls, as well as potential differences between the two groups. In this study, we focused on analyzing three ERP components, including N170, N400, and P600. The N170 is an early indicator of face structure encoding[19]. A body of evidence suggests that this component is also sensitive to socio-semantic information conveyed by faces. For instance, Hügelschäfer et al[20] found that the N170 shows a modulation to gender-incongruent faces during a gender categorization task. This suggests that the early structural encoding stage indexed by the N170 is already influenced by the processing of social category information. More directly, Rodríguez-Gómez et al[21] demonstrated that the amplitude of the N170 to faces was larger when they were preceded by gender-stereotyped statements compared to neutral ones, indicating that stereotyping exerts an early effect on face processing. These findings support the view that the N170 is sensitive not only to the physical properties of faces but also to the stereotypical information associated with them. The N400 is widely recognized as an index of semantic integration and expectancy violation. It was first described by Kutas and Hillyard[22] in a seminal study using sentence reading paradigms. Subsequent research has established that N400 amplitude is highly sensitive to the congruency between incoming information and pre-existing semantic or conceptual expectations, reflecting the ease with which a stimulus can be integrated into its preceding context[23]. Within the domain of social cognition, ERP studies have consistently shown that violations of gender stereotypes evoke robust N400 effects[24]. Moreover, a recent study using a sentence-face priming paradigm found that female faces preceded by stereotypically gender-incongruent occupational primes (e.g., “surgeon”-female face) elicited larger N400 amplitudes than those preceded by congruent primes[25], demonstrating that N400 sensitivity to stereotype violations extends beyond purely linguistic stimuli to cross-modal social information. The P600 was first reported by Osterhout and Holcomb[26] as a positive shift elicited by syntactic violations, and was originally interpreted as indexing syntactic reanalysis or revision processes in response to grammatical anomalies. In a seminal ERP study, Osterhout et al[16] presented participants with sentences containing occupational nouns that carry strong gender stereotypes and found that a mismatching reflexive pronoun (e.g., “the nurse prepared himself for work”) evoked a P600 effect that closely resembled the one typically seen in syntactic anomaly paradigms, suggesting that stereotype-incongruent information may trigger similar controlled reanalysis and repair mechanisms as grammatical errors. More recently, Canal et al[17] further reported that conflicts arising from gender stereotypes also produced a pronounced P600 effect, which they interpreted as reflecting the reintegration of semantic information with stereotypical knowledge. In addition, Proverbio et al[27] observed enhanced occipito-parietal P600 responses to items that violated gender stereotypes, notably in male participants. In summary, this study selected the N170, N400, and P600 components to examine the temporal dynamics of stereotype processing across early perceptual processing, mid-stage hybrid process of semantic retrieval and integration, and late-stage semantic reprocessing and conflict monitoring. We formulated two hypotheses: (1) If stereotype activation occurs at the post-perceptual processing stage, then early ERP components such as the N170 component may show no group differences, whereas the N400 and P600 components may exhibit significant group differences; and (2) In patients with MDD, variations in the N170, N400, and P600 might be associated with their behavioral performance as well as to the severity of their depressive symptoms.

MATERIALS AND METHODS

Participants

Before collecting data, we used G*Power (version 3.1) to conduct a power analysis to ensure sufficient statistical power. The analysis was based on a medium effect size (f = 0.25), an alpha level of α = 0.05, and a power of 1-β = 0.80. Considering a 15% attrition rate, we determined that at least 54 participants would be required to ensure sufficient statistical power for detecting a medium effect size.

This study recruited participants from inpatients diagnosed with MDD from July 2022 to July 2023 in the Affiliated Wuxi Mental Health Center of Jiangnan University. The following criteria were used for inclusion: (1) Met the diagnostic criteria for depression according to Diagnostic and Statistical Manual of Mental Disorders, fourth edition; (2) Aged between 18 years and 55 years; (3) The Hamilton Depression Scale 24-item version (HDRS-24) scores > 20; and (4) Had no record of electroconvulsive therapy treatment during the six months before enrollment. The following criteria were used for exclusion: (1) Met the Diagnostic and Statistical Manual of Mental Disorders, fourth edition criteria for other axis I conditions, including bipolar disorder, mixed episodes, or psychotic symptoms; (2) Serious physical disease and neurological disease; (3) History of substance or non-substance abuse; and (4) Women in a state of pregnancy or lactation. We also recruited healthy controls via online and social media advertisements, matching them with the patient group in terms of gender, age, and education level. All participants were right-handedness, native Mandarin speakers, and had normal vision and hearing. The study approved by the Ethics Committee of the Affiliated Wuxi Mental Health Center of Jiangnan University (Approval No. WXMHCIRB2022 LLky012). Written informed consent was obtained from each participant.

Clinical assessments

Each participant filled out a custom questionnaire including basic demographic and clinical data, such as name, sex, age, education, medical history, therapeutic modalities, and medication regimens.

We used the Mini International Neuropsychiatric Interview, Hamilton Anxiety Rating Scale (HAMA), and HDRS-24 to confirm diagnoses and assess symptom severity of the patients, while healthy controls were screened using the Mini International Neuropsychiatric Interview to exclude various psychiatric disorders. All diagnostic assessments were conducted by associate chief psychiatrists who had received standardized training.

Measurements

GST: GST is a dedicated conflict control paradigm that elicits controlled processing of implicit gender stereotypes[21]. This paradigm can be viewed as a variant of the Implicit Association Test, a widely recognized measure of implicit social cognition. The task consisted of 180 sentences broadly describing an individual's personality, interests, and social expectations. Notably, none of the sentences specified the gender of the subject (with the subject denoted as “Ta”). The sentences were divided into three sets (60 each): (1) Female-biased sentences; (2) Male-biased sentences; and (3) Neutral sentences without gender bias. Sample sentences for each condition are presented in Table 1. In the task, we selected 80 neutral emotion images (40 female and 40 male faces) as stimuli from the Chinese Facial Affective Picture System. Before the formal experiment, participants need to complete five practice trials to familiarize the procedure. At the beginning of every trial, a “+” fixation cross appeared in the center of the computer screen to ensure participants sustained attention. Meanwhile, the researchers kept a close watch on the participants’ behavior for the duration of the experiment. Afterwards, a sentence representing either female gender bias, male gender bias, or no gender bias was randomly presented. After 4000 milliseconds, a neutral emotion image appeared at the computer center. As a control condition, participants were presented with either a male or female neutral emotion image that was either congruent or incongruent with the gender bias. Upon emotion image presentation for 1000 milliseconds, a prompt word appeared on the computer screen. Based on the contextual relationship, participants were required to respond “yes” or “no” to the presented prompt word by pressing the corresponding keys on the computer keyboard (A for “yes”, L for “no”). Each emotion image was presented three times under three different conditions: Gender-biased congruent sentence-face pairs, gender-biased incongruent sentence-face pairs, and neutral (non-gender-biased) sentence-face pairs. The whole task has 240 trials and takes about 35 minutes to complete. All emotion images included in the paradigm had comparable valence value, with a mean valence value of 5.63 for female faces and 5.71 for male faces. Detailed task instructions were provided to ensure that participants fully comprehended the procedures, thereby enhancing their engagement and helping them sustain focused attention throughout the experiment (Figure 1).

Figure 1 Two examples of a trial sequence of the gender stereotype task. A: A trial sequence with female-biased sentence-female face; B: A trial sequence with female-biased sentence-male face. ITI: Inter-trial interval.

Table 1 Example sentences used in the experiment.

	Sentence
Female bias	He/she change clothes into different styles every day
	His/her favorite toy is a Barbie doll
	He/she accidentally hit a tree while driving
	He/she cried with grievance after being criticized by the boss
	He/she spent half a day shopping at the mall
Male bias	He/she smoke alone at the end of the hospital corridor
	He/she played basketball with friends for a whole afternoon on the weekend
	He/she kicked over the trash can by the roadside when angry
	He/she drunk outside until late at night before going home
	He/she played video games with friends at home in the evening
Gender-neutral	He/she speak two languages
	He/she live near the school
	He/she commute by bus
	He/she passed the Mandarin proficiency test
	His/her home is 30 kilometers away from the city center

Electroencephalography data recording

Electroencephalography (EEG) data recording was performed using the international 10-20 system with 64 Ag/AgCl active electrodes fixed to an elastic cap (EasyCap, Herrsching, Germany). The EEG signals were digitized at 500 Hz. The data was filtered with a low cutoff set at 0.1 Hz and a high cutoff set at 30 Hz. Additionally, all electrode impedances were reduced to below 5 kΩ before data collection.

ERP analysis

EEG data were preprocessed with Brain Vision Analyzer 2.1 (Brain Products GmbH, Munich, Germany). Baseline correction was first performed to eliminate low-frequency drifts and limit their effect on signal quality. We re-referenced the EEG data to the mean of A1 and A2 (the two mastoids), which provided a stable reference point for the subsequent analysis. We filtered the data using a band-pass between 0.1 Hz and 30 Hz (24 dB/octave) and also applied a 48-52 Hz notch filter to get rid of power-line noise. After data preprocessing, continuous data were visually inspected, and segments with severely degraded signal quality during recording were manually rejected. The rejection criteria included any baseline drift exceeding ± 100 μV or visually identifiable low-frequency slow drift that could not be corrected by filtering (lasting longer than 500 milliseconds with amplitude fluctuations exceeding ± 75 μV).

To maintain overall data integrity, we applied interpolation to any channels that had poor signal quality. Then, independent component analysis was run to find and take out components from eye blinks, eye movements, heartbeats, and muscle activity. We then segment the continuous data into epochs that extend from -200 milliseconds to 800 milliseconds around the presentation of the stimulus. Baseline correction was performed using the pre-stimulus interval from -200 milliseconds to 0 millisecond. Finally, epochs corresponding to identical experimental conditions were averaged to generate individual ERP waveforms for each participant.

Statistical analysis

All statistical analyses were conducted using SPSS 26.0 (IBM Corp., Armonk, NY, United States). Continuous variables were presented as mean ± SD when they met the assumptions of normality and homogeneity of variance; otherwise, they were expressed as median with interquartile range (25^th percentile, 75^th percentile). Categorical variables were described as n (%).

A three-way mixed-design analysis of variance (ANOVA) was conducted on reaction times for GST behavioral data, comparing individuals with depression and healthy control groups. In this analysis, participant group (MDD, healthy controls) was included as a between-subjects factor, while stereotype compatibility (consistent, inconsistent, neutral) and gender bias type (male, female) were included as within-subjects factors.

Guided by earlier literature and the overall ERP waveforms we recorded under each condition, we concentrated our analysis on the N170, N400, and P600 components. These have been shown to reflect cognitive conflict processing in the GST task. The N170 time window (130-230 milliseconds) was defined on the basis of classical electrophysiological evidence on face processing reported by Bentin et al[19]. This temporal range has also been consistently adopted in subsequent ERP studies, including Molinaro and Carreiras[28], where it was used to capture early visual processing and structural encoding processes. The N400 time window was defined according to its original characterization as an index of semantic integration[22], as well as its application in stereotype-related ERP research[15]. The P600 time window (500-800 milliseconds) was selected based on prior work linking this component to syntactic reanalysis and late integrative processes[26]. For N170, we measured their peak amplitudes and latencies at four occipito-temporal electrode sites (P7, P8, O1, and O2). For N400, measurements were taken at eleven fronto-central electrode sites (F3, FC3, C3, Fz, FCz, Cz, F4, FC4, C4, CPz, and Pz). For P600, measurements were taken at nine centroparietal electrode sites (F1, Fz, F2, FC1, FCz, FC2, C1, Cz, and C2). Repeated-measures analyses of variance (ANOVAs) with 2 (group: MDD group vs healthy controls) × 3 (stereotype compatibility: Congruent, incongruent, neutral) × 2 (gender bias type: Male, female) were performed to analyze the differences in the amplitude and latency of N170, N400, and P600, respectively. To control for the potential confounding effects of antidepressant medication on ERP components, fluoxetine-equivalent dose (mg/day) was included as a covariate in the statistical models. Fluoxetine-equivalent doses were standardized based on previously published equivalence conversion tables for antidepressant medications, allowing for comparability of dosage across different drug types[29].

It should be noted that participant gender was not part of the original research hypotheses nor included in the a priori sample size estimation. Consequently, all analyses involving participant gender (including the ANOVA with participant gender as a between-subjects factor) should be considered exploratory. Similarly, correlation analyses examining the relationships between clinical characteristics (such as disease duration, HDRS-24, and HAMA scores) or fluoxetine-equivalent dosage and ERP components were not included in the original sample size estimation and should also be treated as exploratory. Spearman correlation analyses were performed, and to mitigate the risk of false positives arising from multiple comparisons, the false discovery rate (FDR) procedure was applied, with FDR-adjusted P < 0.05 considered statistically significant.

For all ANOVA models, post-hoc pairwise comparisons were adjusted using the Bonferroni correction to control for type I error. In cases where the assumption of sphericity was violated, the Greenhouse-Geisser correction was applied to adjust the degrees of freedom. All statistical tests were two-tailed, with the significance level set at α = 0.05 (P < 0.05). For significant effects, we reported partial (ηp²) as the effect size.

RESULTS

Demographics and clinical sample characteristics

A total of 60 participants were initially enrolled in this study. However, two participants failed to complete the entire task, and two more were excluded due to having fewer than 20 valid trials. As a result, the final analysis included 28 patients with MDD and 28 healthy controls, the male-to-female ratio was balanced at 1:1. In the MDD group, the mean disease duration was 45.43 months (SD = 28.833), the mean HDRS-24 score was 29.79 (SD = 7.156), and the mean HAMA score was 10.61 (SD = 5.301). Approximately 50% of the MDD patients in this study exhibited mild anxiety symptoms (HAMA ≥ 7); however, individuals meeting diagnostic criteria for independent anxiety disorders were excluded. All the patients with MDD were treated with antidepressant medications. The specific medications and dosages were as follows: 8 patients were treated with sertraline (mean dose: 168.75 ± 34.72 mg/day), 7 with escitalopram oxalate (15.71 ± 4.50 mg/day), 5 with mirtazapine (26.0 ± 11.94 mg/day), 3 with venlafaxine (200.0 ± 43.30 mg/day), 3 with fluoxetine (23.33 ± 5.77 mg/day), and 2 with paroxetine (25.0 ± 7.07 mg/day). Based on previously published conversion methods, the mean fluoxetine-equivalent daily dose was calculated to be 40.36 ± 14.53 mg/day[29].

Shapiro-Wilk normality tests of the age data indicated that the data conformed to a normal distribution. An independent samples t-test was employed, and the results suggested no statistically significant difference in age data between the two groups (P > 0.05). Years of education data deviated from normality, and the Mann-Whitney U test was employed. The results showed no statistically significant difference in education duration between the two groups (P > 0.05). The gender distribution fits the χ² test’s conditions. The test found no significant difference between the two groups (P > 0.05; Table 2).

Table 2 Demographic characteristics and clinical information of participants, mean ± SD/median (interquartile rage).

Variable	MDD (n = 28)	HCs (n = 28)	Test statistic
Age (years)	33.86 ± 7.895	33.06 ± 6.232	t = -0.377, P = 0.708
Sex (male/female)	14/14	14/14	χ2 = 1.000, P = 1.000
Education (years)	14 (9-15.5)	15 (12-15)	U = 355.0, P = 0.088
Course of disease (month)	45.43 ± 28.833
HDRS-24	29.79 ± 7.156
HAMA	10.61 ± 5.301
Medicine (S/E/M/P/V/F)	8/7/5/2/3/3

HDRS-24: Hamilton Depression Scale 24-item version; HAMA: Hamilton Anxiety Rating Scale; S: Sertraline; E: Escitalopram oxalate; M: Mirtazapine; P: Paroxetine; V: Venlafaxine; F: Fluoxetine; MDD: Major depressive disorder; HCs: Healthy controls.

Behavioral task performance

Reaction times: The analysis revealed a significant main effect of stereotype compatibility (F_2,108 = 65.363, P < 0.001, ηp² = 0.548), and a significant effect of gender bias type (F_1,54 = 6.916, P = 0.011, ηp² = 0.114), while the main effect of participant group was not significant (F_1,54 = 0.702, P = 0.406, ηp² = 0.013). The interaction between stereotype compatibility and participant group was not significant (F_2,108 = 0.753, P = 0.473, ηp² = 0.014), nor was the interaction between gender bias type and participant group (F_1,54 = 2.303, P = 0.105, ηp² = 0.041). A significant interaction was observed between stereotype compatibility and gender bias type (F_2,108 = 4.105, P = 0.048, ηp² = 0.071). The three-way interaction among participant group, gender bias type, and stereotype compatibility was not significant (F_2,108 = 1.477, P = 0.134, ηp² = 0.016). Further analysis revealed significant differences in reaction times across different control conditions for both patients with MDD and healthy controls (P < 0.05). Specifically, reaction times were significantly longer in stereotype-incongruent conditions (male stereotype-female face, female stereotype-male face) compared to stereotypecongruent conditions (male stereotype-male face, female stereotype-female face; Supplementary Table 1 and Figure 2).

Figure 2 Reaction times between major depressive disorder group and healthy controls under different control conditions. MDD: Major depressive disorder; HCs: Healthy controls.

ERP results

N170: A 2 (group: MDD group vs healthy controls) × 3 (stereotype compatibility: Congruent, incongruent, and neutral) × 2 (gender bias type: Male, female) mixed-design ANOVA was run on N170. There was a significant main effect of stereotype compatibility (F_2,108 = 8.302, P = 0.001, ηp² = 0.129) and a significant main effect of gender bias type (F_1,54 = 7.401, P = 0.009, ηp² = 0.121), while the main effect of participant group was not significant (F_1,54 = 0.121, P = 0.730, ηp² = 0.002). The interaction between stereotype compatibility and gender bias type was significant (F_2,108 = 7.113, P = 0.001, ηp² = 0.116). However, the interaction between stereotype compatibility and participant group was not significant (F_2,108 = 1.078, P = 0.344, ηp² = 0.020), the interaction between gender bias type and participant group was not significant (F_1,54 = 2.109, P = 0.152, ηp² = 0.038), and the three-way interaction among participant group, gender bias type, and stereotype compatibility was not significant (F_2,108 = 0.674, P = 0.512, ηp² = 0.012). Further analysis revealed statistically significant differences in N170 amplitudes across conditions for both patients with MDD and healthy controls (P < 0.05). Specifically, N170 amplitudes were significantly higher for the male stereotype-female face condition compared to all other conditions (female stereotype-male face, male stereotype-male face, female stereotype-female face, and neutral; Supplementary Table 2 and Figure 3).

Figure 3 N170 amplitudes at P7 electrode. A: N170 amplitudes at P7 electrode of the major depressive disorder group under different control conditions; B: N170 amplitudes at P7 electrode of the healthy controls under different control conditions.

Additionally, we conducted a 2 (participant gender: Male, female) × 3 (stereotype compatibility: Congruent, incongruent, and neutral) × 2 (gender bias type: Male, female) repeated-measures ANOVA on the N170 amplitude with participant gender as a between-subjects factor. The results indicated a significant main effect of participant gender (F_1,54 = 11.306, P = 0.001, ηp² = 0.173), a significant main effect of stereotype compatibility (F_2,108 = 7.991, P = 0.001, ηp² = 0.129), and a significant main effect of gender bias type (F_1,54 = 7.336, P = 0.009, ηp² = 0.120). A significant interaction was found between stereotype compatibility and participant gender (F_2,108 = 7.071, P = 0.001, ηp² = 0.116), while the interaction between gender bias type and participant gender was not significant (F_1,54 = 0.797, P = 0.453, ηp² = 0.015). The three-way interaction among participant gender, gender bias type, and stereotype compatibility was significant (F_2,108 = 5.675, P = 0.005, ηp² = 0.095). Further analysis results showed that male participants exhibited higher N170 amplitudes than female participants in response to both male stereotype-female face (F_1,54 = 10.034, P = 0.003, ηp² = 0.157) and female-stereotypical male face (F_1,54 = 8.802, P = 0.004, ηp² = 0.140; Figure 4).

Figure 4 N170 amplitudes between male and female participants. A: Male and female participants for male bias-female face; B: Male and female participants for female bias-male face; C: The corresponding topographic map of Figure (A); D: The corresponding topographic map of Figure (B). The blue area indicates the time window that was used to inspect the N170 amplitude (130-230 milliseconds).

For N170 latency, there was a significant main effect of gender bias type (F_1,54 = 5.436, P = 0.023, ηp² = 0.091), while the main effect of participant group was not significant (F_1,54 = 1.641, P = 0.206, ηp² = 0.029), and main effect of stereotype compatibility was not significant (F_2,108 = 0.068, P = 0.934, ηp² = 0.001). The interaction between stereotype compatibility and participant group was not significant (F_2,108 = 0.827, P = 0.440, ηp² = 0.015), the interaction between stereotype compatibility and gender bias type was not significant (F_2,108 = 2.819, P = 0.640, ηp² = 0.050), the interaction between gender bias type and participant group was not significant (F_1,54 = 2.375, P = 0.129, ηp² = 0.042). The three-way interaction among participant group, gender bias type, and stereotype compatibility was not significant (F_2,108 = 0.214, P = 0.808, ηp² = 0.004).

N400: A 2 (group: MDD group vs healthy controls) × 3 (stereotype compatibility: Congruent, incongruent, and neutral) × 2 (gender bias type: Male, female) mixed-design ANOVA was run on N400. There was a significant main effect of participant group (F_1,54 = 8.937, P = 0.004, ηp² = 0.142), a significant main effect of stereotype compatibility (F_2,108 = 10.624, P = 0.002, ηp² = 0.164), and a significant main effect of gender bias type (F_1,54 = 4.413, P = 0.037, ηp² = 0.068). The interaction between stereotype compatibility and gender bias type was significant (F_2,108 = 7.899, P = 0.024, ηp² = 0.051), also significant between participant group and gender bias type (F_1,54 = 6.330, P = 0.015, ηp² = 0.105). The three-way interaction among participant group, gender bias type, and stereotype compatibility was significant (F_2,108 = 8.323, P = 0.006, ηp² = 0.134). Further analysis results showed that patients with MDD showed larger N400 amplitudes than healthy controls in response to both female stereotype-male face (F_1,54 = 4.738, P = 0.042, ηp² = 0.065) and male stereotype-female face (F_1,54 = 5.721, P = 0.006, ηp² = 0.179; Supplementary Table 3 and Figure 5).

Figure 5 N400 amplitudes at healthy controls electrode. A: The major depressive disorder group and the healthy controls for female bias-male face; B: The major depressive disorder group and the healthy controls for male bias-female face; C: The corresponding topographic map of Figure (A); D: The corresponding topographic map of Figure (B). The blue area indicates the time window that was used to inspect the N400 amplitude (350-500 milliseconds). MDD: Major depressive disorder; HCs: Healthy controls.

For the latency of the N400, the main effect of stereotype compatibility was significant (F_2,108 = 6.913, P = 0.027, ηp² = 0.084), while the main effect of participant group was not significant (F_1,54 = 0.207, P = 0.651, ηp² = 0.004) and the main effect of gender bias type was not significant (F_1,54 = 0.372, P = 0.544, ηp² = 0.007). The interaction between stereotype compatibility and participant group was not significant (F_2,108 = 0.928, P = 0.398, ηp² = 0.017), the interaction between stereotype compatibility and gender bias type was not significant (F_2,108 = 0.606, P = 0.547, ηp² = 0.011), the interaction between gender bias type and participant group was not significant (F_1,54 = 1.613, P = 0.210, ηp² = 0.029). The three-way interaction among participant group, gender bias type, and stereotype compatibility was not significant (F_2,108 = 0.270, P = 0.765, ηp² = 0.005).

P600: A 2 (group: MDD group vs healthy controls) × 3 (stereotype compatibility: Congruent, incongruent, and neutral) × 2 (gender bias type: Male, female) mixed-design ANOVA was run on P600. There was a significant main effect of participant group (F_1,54 = 5.365, P = 0. 024, ηp² = 0.090), while the main effect of stereotype compatibility was not significant (F_2,108 = 0.143, P = 0.866, ηp² = 0.003) and the main effect of gender bias type was not significant (F_1,54 = 0.408, P = 0.526, ηp² = 0.007). The interaction between stereotype compatibility and participant group was not significant (F_2,108 = 0.033, P = 0.968, ηp² = 0.001), nor between stereotype compatibility and gender bias type (F_2,108 = 0.237, P = 0.790, ηp² = 0.004), nor between gender bias type and participant group (F_1,54 = 0.094, P = 0.761, ηp² = 0.002). The three-way interaction among participant group, gender bias type, and stereotype compatibility was not significant (F_2,108 = 0.745, P = 0.463, ηp² = 0.014).

For P600 latency, the results showed no significant main effect of stereotype compatibility (F_2,108 = 0.416, P = 0.661, ηp² = 0.084), no significant main effect of participant group (F_1,54 = 0.099, P = 0.754, ηp² = 0.008), and no significant main effect of gender bias type (F_1,54 = 0.126, P = 0.724, ηp² = 0.002). The interaction between stereotype compatibility and participant group was not significant (F_2,108 = 0.361, P = 0.550, ηp² = 0.005), the interaction between stereotype compatibility and gender bias type was not significant (F_2,108 = 0.189, P = 0.828, ηp² = 0.003), the interaction between gender bias type and participant group was not significant (F_1,54 = 0.181, P = 0.672, ηp² = 0.003). The three-way interaction among participant group, gender bias type, and stereotype compatibility was not significant (F_2,108 = 1.554, P = 0.216, ηp² = 0.028).

Medication-adjusted results: After controlling for fluoxetine-equivalent dose, previously observed group differences for N400 remained in the same direction and significance. The dose covariate showed no main effect and no interactions with group, stereotype consistency, or gender bias type on amplitudes or latencies (all P > 0.05) (Supplementary Table 4).

Correlation analysis between clinical characteristics and ERP data

In the MDD group, we performed Spearman correlation analyses between disease duration, HDRS-24 scores, HAMA scores, and N170/N400 amplitudes, with FDR correction applied. Before correction, disease duration was positively correlated with N400 amplitude in the female stereotype-male face condition (r = 0.560, P = 0.003); HAMA score was negatively correlated with N170 amplitude in the neutral-female face condition (r = -0.440, P = 0.029). After FDR correction, the correlation between disease duration and N400 (female stereotype-male face) remained significant (FDR-adjusted P = 0.036), whereas the correlation between HAMA and N170 was no longer significant (FDR-adjusted P = 0.177). The same FDR procedure was applied to the Spearman correlations between fluoxetine-equivalent dose and N170/N400 amplitudes. Before correction, fluoxetine-equivalent dose was positively correlated with N400 amplitude in the male stereotype-male face condition (r = 0.368, P = 0.021), and with N170 amplitude in the neutral-female face condition (r = 0.403, P = 0.032). After FDR correction, none of these correlations remained significant (all FDR-adjusted P > 0.05; Supplementary Table 5 and Figure 6).

Figure 6 Correlation analysis between clinical characteristics and N400 amplitude in the major depressive disorder group after the false discovery rate correction. Spearman correlation analysis between disease duration (months), Hamilton Depression Scale 24-item version scores, Hamilton Anxiety Rating Scale scores, and N400 amplitude. ^aP < 0.05. HDRS-24: Hamilton Depression Scale 24-item version; HAMA: Hamilton Anxiety Rating Scale.

DISCUSSION

As one of the most prevalent mental disorders, MDD is characterized by impairments in social cognitive functioning as a core symptomatic feature. Of these impairments, social and interpersonal relationship issues are particularly notable[30]. Such difficulties may be exacerbated by the negative cognitive biases and distorted social appraisals characteristic of MDD. Patients often hold or internalize rigid and negative stereotypes about themselves (e.g., “I am incompetent and unlikable”) and about others (e.g., “People are judgmental and rejecting”). The development of maladaptive stereotyped interpretations creates problems that damage social connections between people and result in lower access to social support networks, which function as primary protective elements during depression recovery. However, the neglect of this population in existing research on the neural mechanisms of stereotypes prevents a comprehensive understanding of the cognitive neural basis of stereotype processing and its specific manifestations in mental disorders. To address this gap, the present study included individuals with MDD as participants and used gender stereotypes, the most common form of stereotypes in human society, as an example. Specifically, by employing stereotype trait words as the core content of stereotype representation, this study aimed to investigate the cognitive mechanisms underlying stereotype activation effects in MDD.

We used a GST paradigm to compare 28 MDD patients with 28 healthy individuals. Behavioral data showed that both groups showed similar differences when processing gender stereotype-consistent vs inconsistent information. Specifically, reaction times were significantly longer for stereotype-inconsistent information than for stereotype-consistent information. According to stereotype processing theory, when individuals are presented with information that is consistent with existing social semantic networks, stereotypes are automatically activated and rapidly mapped onto established cognitive schemas, resulting in relatively efficient processing. In contrast, when incoming information conflicts with automatically activated stereotypes (i.e., in stereotype-incongruent conditions), additional cognitive control resources are required to suppress prepotent responses and to re-integrate and reinterpret the conflicting information. This process substantially increases cognitive load and consequently prolongs reaction times[31]. Therefore, the prolonged reaction times observed in the incongruent condition in our study may reflect the sustained engagement of conflict monitoring and inhibitory control systems during semantic processing. This result is consistent with several previous studies[10,12,32-34], suggesting that the effect of semantic conflict on processing speed appears to be relatively stable across different samples and task designs.

Previous ERP studies on gender stereotypes have examined various topics, including automatic processing of gender information[35], effects of gender stereotypes on attention and memory[36], and the impact of stereotype threat on cognitive performance[37]. These studies have shown that multiple ERP components are involved in the process of stereotype activation[38-40], including N2, N170, N400, and P600[41-44]. Based on the grand average ERP waveforms and topographic maps obtained under different experimental conditions in the present study, we focused our analysis on three ERP components: N170, N400, and P600.

The N170 is a negative waveform that peaks roughly 200 milliseconds following stimulus onset at occipito-temporal regions. Studies have shown that faces typically elicit higher N170 amplitudes than any other visual stimulus category, which is why it is often referred to as the face-selective N170[45,46]. In the course of our daily life, stereotype activation is frequently based on facial information[47]. This process inevitably begins with social categorization (e.g., gender, age, race) and subsequently leads to the activation of associated stereotypes. In other words, when we socially categorize a person based on their facial information, this process directly triggers the activation of the associated stereotypes. In most cases, facial information is treated as an independent object of study, meaning that when investigating facial characteristics, researchers isolate facial information from other contextual factors. However, in recent years, some studies have begun to explore whether face processing is influenced by the different situational and contextual backgrounds in which faces are presented[48-52]. Diéguez-Risco et al[53] used emotionally biased sentences as situational context and presented participants with face images whose emotional expressions were either congruent or incongruent with the preceding sentences. They measured participants’ behavioral responses to facial expressions and recorded EEG to examine neural activity during processing. Their findings indicated that when faces appeared within an emotionally biased sentence context, the N170 amplitude was enhanced. This suggests that situational context influences the processing of facial expressions and that this influence occurs at an early stage of cognitive processing. It is well known that in the field of cognitive psychology, stereotypes and prejudice are closely interrelated. The human cognitive system tends to quickly capture information and patterns, which may lead us to fall into the trap of stereotypes, focusing only on certain features while overlooking others. In such a process, the formation of prejudice becomes inevitable. A study using a digital categorization task and eye-tracking techniques to explore whether prejudice drives attentional bias toward outgroups found that the N170 component showed larger amplitudes to outgroups toward which participants held prejudice, regardless of whether these outgroups had been habituated. In contrast, N170 amplitudes responses to non-prejudiced outgroups and ingroups were smaller, and these effects correlated with implicit prejudice measures. These findings suggest that the N170 component may serve as an implicit electrophysiological marker of prejudice[44]. In our study, the same N170 effect was observed across both participant groups, as reflected by enhanced N170 amplitudes elicited by genderstereotype-incongruent information. Furthermore, when we divided all participants into male and female subgroups, we interestingly found that male participants exhibited significantly larger N170 amplitudes than female participants in response to stereotype-incongruent information. This finding suggests that participant gender may be a potential moderating factor in stereotype processing, with male participants potentially exhibiting greater neural sensitivity when processing stereotype-incongruent information, reflected in enhanced engagement during early stages of facial structural encoding or face-social semantic matching, or in the allocation of greater perceptual resources at early processing stages. This observation is consistent with the results of Rodríguez-Gómez et al[21], who used a face-matching task to directly compare the N170 component between male and female participants after gender stereotype priming and found that the modulation of early face processing by stereotypes was significantly stronger in males, providing direct neurophysiological evidence that males are more sensitive to stereotype conflicts at early perceptual stages. However, it should be emphasized that analyses involving participant gender were not part of the original hypotheses or a priori power calculations in our study. Therefore, this finding should be interpreted as exploratory. In addition, the relatively balanced but limited sample size (28 males and 28 females) may still be underpowered to detect reliable gender interaction effects. Future studies should predefine participant gender as a moderating variable in larger samples and further examine the robustness and clinical relevance of this N170 gender effect by integrating measures of clinical symptoms, social functioning, cognitive control, as well as explicit and implicit indices of stereotype processing.

The N400 is a negative waveform that peaks roughly 400 milliseconds following stimulus onset[54] and is widely recognized as a biomarker index of semantic understanding and language processing[55,56]. Specifically, the latency and amplitude of the N400 component can reflect the process of semantic cognitive processing: The latency reflects the speed of semantic cognitive processing, while the amplitude reflects the difficulty of semantic cognitive processing. Studies have shown that when facing a situation of semantic conflict, the brain needs to invest more cognitive resources to integrate the conflicting information, resulting in an increased N400 amplitude. Moreover, the difficulty of semantic cognitive processing is positively correlated with the N400 amplitude[23,57]. Van Berkum et al[58] presented participants with speakers of different genders along with sentences they uttered (e.g., a female speaker saying a line typically spoken by a male character, or a male speaker saying a line typically spoken by a female character) to observe the brain’s response to the mismatch between speaker and information. The results showed that when the speaker’s gender was inconsistent with the information, the N400 amplitude increased significantly. Wang et al[59] explored the cognitive processing of gender stereotypes by observing electrophysiological changes in the brain. The results showed that when participants heard gender-stereotype-inconsistent word pairs (e.g., male - nurturing), the N400 amplitude increased significantly. It is not difficult to find that previous studies have largely focused on the N400 effect of gender stereotypes only in healthy populations. In the present study, we included patients with MDD as participants for the first time. We found that patients with MDD exhibited a significantly larger N400 amplitude in response to gender-stereotype-incongruent information compared with healthy controls. This finding suggests that when processing stereotype-incongruent information, patients with MDD allocate more cognitive resources for semantic integration, indicating a stronger stereotype activation effect. Moreover, it supports the classic theory that the N400 component can serve as a marker of the activation of gender-stereotype-incongruent information. In addition, all the patients with MDD in this study were receiving antidepressant treatment, and there was substantial heterogeneity in both medication types and dosages. Therefore, the observed differences in N400 between the MDD and healthy controls cannot fully exclude potential confounding effects of medication. To address this issue, we re-conducted a repeated-measures analysis of covariance (ANCOVA) on N400, with fluoxetine-equivalent dose included as a covariate. The results showed that the group differences in N400 amplitude between the MDD and healthy control groups remained significant after controlling for medication dosage. However, given the relatively small sample sizes in certain medication subgroups, these between-group effects should still be interpreted with caution. To better disentangle disease-specific effects from medication effects, future studies should prioritize recruiting medication-naïve first-episode MDD patients or adopt longitudinal designs (e.g., comparing pre-treatment baseline and post-treatment measurements) to examine changes in ERP components following pharmacological intervention.

The P600 is a positive waveform that peaks approximately 600 milliseconds after stimulus onset and is predominantly distributed over the centroparietal regions of the scalp[60]. Similar to the N400 component, the P600 component is also associated with semantic integration, as well as with syntactic reanalysis and conflict monitoring[61-63]. Evidence suggests that the P600 component is also a sensitive electrophysiological indicator of stereotypes. Osterhout designed a series of experiments that presented situations violating gender stereotypes (e.g., a man doing housework or a woman holding a leadership position) while observing participants' electroencephalographic activity. The study found that when the stereotypical gender of a preceding role noun did not match the gender of the following reflexive pronoun, the P600 amplitude increased significantly, and this increase was more pronounced in female participants[16]. In another study, the same P600 effect was observed. The researchers recorded participants’ electroencephalographic activity to investigate the time course between gender stereotype activation and lexical semantic activation during language processing. They found that when nouns conveyed gender information through their associated stereotypes, the P600 amplitude was significantly increased for gender stereotype mismatches. This similarity has been interpreted as suggesting that gender stereotype information is encoded in the grammar, thereby producing a “syntactic” P600 effect[17]. In the present study, stereotype congruency did not show any main effect or interaction on P600 amplitude or latency. Instead, we observed only a significant group effect, with reduced P600 amplitudes in the MDD group relative to healthy controls. A possible explanation for this finding lies in differences in task paradigms. In previous studies reporting P600 effects, most evidence comes from within-language sentence-processing paradigms. In these designs, gender stereotype expectations are established incrementally within sentence context, and violation information is also presented linguistically in subsequent words or sentence continuations. Such settings are more likely to elicit the P600 effect, which has been associated with semantic integration, conflict resolution, and late-stage reanalysis processes. In contrast, our study employed a sentence-picture matching paradigm in which participants were required to integrate gender stereotype information from sentences with subsequently presented facial images. In this case, violation information is not embedded within linguistic structure but is instead conveyed through visual facial stimuli. As a result, processing is more likely to rely on early face perception and cross-modal semantic matching, rather than the late-stage reanalysis processes typically observed in language-based paradigms. Consistent with this view, a recent systematic review suggests that P600 effects are more reliably observed when stereotype information contributes to discourse-level inferential coherence[24]. In the present paradigm, stereotype information does not directly predict subsequent syntactic or linguistic elements, which may explain the absence of a stereotype-related P600 effect. Regarding the main effect of group, with reduced P600 amplitudes in the MDD group, this finding may reflect a general reduction in cognitive resources and impaired controlled processing in depression. The P600 has been considered an index of controlled, attention-demanding cognitive operations. Individuals with MDD often show deficits in executive functions, including reduced attentional control, slowed processing speed, and impaired cognitive flexibility, all of which may limit the resources available for task-relevant controlled processing and consequently attenuate P600 amplitudes[64]. Furthermore, Vissers et al[65] demonstrated that induced sadness in healthy individuals can attenuate the P600 effect, suggesting that negative affect alone may reduce P600 amplitude. Accordingly, the reduced P600 observed in the MDD group is likely a cross-task, non-specific effect, reflecting diminished cognitive resources and the impact of affective state on controlled processing, rather than a selective abnormality in stereotype processing. This interpretation is consistent with resource depletion accounts of depression.

Finally, we analyzed the correlation between clinical features and ERP data in patients with MDD. The results showed that the fluoxetine-equivalent dose was positively correlated with N400 amplitude (male stereotype-male face), and the duration of illness was positively correlated with N400 amplitude (female stereotype-male face). These findings indicate that patients with MDD who have a longer duration of illness may allocate greater cognitive resources for semantic integration when processing stereotype-incongruent information, suggesting a more pronounced stereotype activation effect. However, we found no significant correlation between total HDRS-24 scores and N170 or N400 amplitudes. There could be several reasons for this. For one thing, the MDD group was relatively small. While the sample size was sufficient for the main group comparisons, it offered limited power for within-group correlational analyses. Most of our patients had moderate to severe depression, and the restricted range of symptom scores may have further weakened the chance of detecting meaningful correlations. Another possibility is that the HDRS-24 captures multiple dimensions—emotional, somatic, and cognitive. It might be that impairments in stereotype processing are more closely tied to cognitive symptoms, such as poor attentional control, cognitive rigidity, or executive dysfunction, rather than to emotional or somatic complaints. Using total scores may obscure the specific contribution of the cognitive domain. Future studies should examine the relationships between ERP data and individual HDRS-24 subscales, particularly the cognitive dimension, to more precisely capture the link between symptom characteristics and neural responses. A third factor is medication. All patients in the current study were receiving antidepressant treatment. Exploratory analyses indicated that disease duration, rather than symptom severity, was significantly associated with N400 amplitude, suggesting that medication may attenuate the direct influence of depressive symptoms on ERP components. The chronicity of the disorder may exert a stronger effect on stereotype processing than the severity of current symptoms. Future research with larger samples should consider HDRS-24 subdomains, medication status, disease duration, and social cognitive measures to clarify the specific associations and potential moderating mechanisms between depression severity and ERP components.

Several limitations of this study should be pointed out. One limitation was that our sample was relatively small, which might affect both the statistical power and the extent to which the findings can be generalized. Future studies should include larger and more varied groups of participants to enhance the reliability of the results. Another point is that the current work was based on a cross-sectional design. We plan to implement longitudinal follow-up methods in future research to explore how these indicators shift as time goes on and to evaluate their potential application value in diagnosis, prognosis prediction, and treatment guidance. In addition, it should be noted that some patients with MDD in the present study exhibited mild anxiety symptoms, which may introduce potential confounding effects on the results to some extent. Although individuals meeting diagnostic criteria for independent anxiety disorders were excluded, and efforts were made to control for anxiety severity in the statistical analyses, the potential modulatory effects of subclinical anxiety symptoms cannot be completely ruled out. Future studies are therefore warranted to further refine sample stratification, clearly distinguish between “pure MDD” and “MDD with comorbid anxiety symptoms”, and systematically compare the distinct neural mechanisms of these subgroups in larger samples.

CONCLUSION

This study was rarely studied in clinical samples to combine the GST paradigm with ERP technology to explore whether there are different neural mechanisms between patients with MDD and healthy individuals when processing stereotypes. The behavioral data did not show any differences in reaction times between the two groups. However, we observed that both groups had significantly longer reaction times when processing stereotype-incongruent information compared to stereotype-congruent information. This indicates that processing counter-stereotype information requires more cognitive resources for decision-making and encoding, which leads to prolonged reaction times. Based on the ERP results, we summarize the following conclusions: First of all, both patients with MDD and healthy controls exhibited a significant enhancement in N170 amplitude under conditions of gender-stereotype incongruence. This finding indicates that when individuals are exposed to information that contradicts existing stereotypes, the cognitive processing mechanisms of the brain are correspondingly adjusted and enhanced. This phenomenon may be closely related to the generation of cognitive conflict, where individuals need to allocate additional cognitive resources to process information that violates stereotypes. It is noteworthy that this effect was more pronounced in male participants, suggesting that males may exhibit greater neural sensitivity when processing stereotype-incongruent information. Furthermore, compared with healthy controls, patients with MDD showed a significantly enhanced N400 amplitude when processing stereotype-incongruent information. This finding suggests that patients with MDD may exhibit stronger activation of stereotypes and require more cognitive resources to integrate incongruent semantic information. Correlation analyses further revealed that this effect was positively associated with disease duration, suggesting that a longer illness course may be accompanied by a more pronounced N400 effect.

Stereotypes represent one of the most important research topics in social cognition. As a typical form of social thinking, its cognitive characteristics are prominently manifested in the automatic categorization of social groups and the categorical representation of those groups. In the process of forming impressions of others, individuals often tend to use cognitive shortcuts by engaging in stereotype processing based on the social categories of the target (e.g., gender, age, and race), which influences the development and maintenance of different associations and attitudes (i.e., prejudice) toward members of these groups[66]. Stereotypes affect both the perceiver and the actor, which in turn changes subjective perceptions of that person or their behaviors[67]. It is particularly noteworthy that this stereotypemediated interaction impairment is more pronounced in MDD. Not only are these patients more likely to become targets of social prejudice, but they also tend to interpret others' behavior in a negative and stereotyped manner, thereby accelerating the erosion of social support and the accumulation of loneliness. Ultimately, this constitutes an important psychosocial mechanism underlying the persistence and exacerbation of depressive symptoms. Therefore, our study has significant practical implications. By identifying characteristic electrophysiological biomarkers, we introduced neural mechanisms into the traditional framework of social cognition research, thereby integrating and establishing connections among “behavior and experience”, “individual and social context”, “cognition and information processing”, and “neural mechanisms”. This approach aims to provide neurophysiological evidence for understanding cognitive biases in MDD and lays a foundation for developing intervention strategies for MDD based on cognitive neural mechanisms. Of course, current research on stereotypes in MDD is still at an early stage, and there are still some issues that need further improvement, refinement, and in-depth exploration in future studies.

ACKNOWLEDGEMENTS

We are grateful to all the participants who took part in the current study, as well as everyone who assisted with this research.