Neural correlates of impairments in music emotion processing in major depressive disorder: Evidenced from an event-related potential study

Abstract

BACKGROUND

Patients with major depressive disorder (MDD) commonly exhibit widespread cognitive impairments. Music, as a complex auditory stimulus with relatively high ecological validity, can be utilized to investigate brain information processing mechanisms. Event-related potentials (ERPs) are well-suited for capturing the temporal dynamics of neural processing across successive cognitive stages. However, the number of systematic ERPs studies examining multi-stage musical information processing in MDD patients remains relatively limited.

AIM

To elucidate the neural mechanisms underlying musical emotion processing deficits in MDD using a multi-stage ERPs framework, and to explore potential neurobiological markers associated with cognitive impairment.

METHODS

Thirty MDD patients (Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition diagnosis, 24-item Hamilton Depression Rating Scale ≥ 20) and twenty-nine demographically matched healthy controls (HCs) completed a category judgment task with neutral, negative, and positive musical stimuli (4-7 seconds each). Electroencephalogram was recorded using a 64-channel system, and core ERPs components (N100, P200, P300) from the left prefrontal, right prefrontal, and central regions were analyzed. Behavioral (accuracy, reaction time) and ERPs data were examined via repeated-measures ANOVA.

RESULTS

Behaviorally, MDD patients showed significantly lower overall accuracy (P = 0.008) and longer reaction times (P = 0.014) than HC. Both groups responded faster to positive music than neutral and negative music (P < 0.001). Neurophysiologically, significant “group × emotional condition” or “group × region” interactions emerged for N100 (button response: P = 0.007), P200 (onset: P = 0.012), and P300 (onset: P = 0.008). Key neural features of MDD included enhanced central N100 amplitude, failure to differentiate neutral from negative music at the P200 stage, and absent stimulus type-related modulation of P300, contrasting with HCs’ differentiated neural responses. Behaviorally, MDD patients showed significantly lower overall accuracy (P = 0.008) and longer reaction times (P = 0.014) than HC. Both groups responded faster to positive music than neutral and negative music (P < 0.001). Neurophysiologically, significant “group × emotional condition” or “group × region” interactions emerged for N100 (button response: P = 0.007), P200 (onset: P = 0.012), and P300 (onset: P = 0.008). Key neural features of MDD included enhanced central N100 amplitude, failure to differentiate neutral from negative music at the P200 stage, and absent stimulus type-related modulation of P300, contrasting with HCs’ differentiated neural responses.

CONCLUSION

MDD patients exhibit multi-stage neural functional abnormalities in musical emotional processing. ERPs abnormalities reflect deficits in early sensory-attentional allocation (N100), stimulus feature discrimination (P200), and late cognitive evaluation (P300). These stage-specific ERPs profiles can serve as potential neurobiological markers for cognitive impairments in MDD, highlighting the utility of musical paradigms in unraveling the brain functional mechanisms of depression.

Key Words: Major depressive disorder; Emotional processing; Musical emotional information; Event-related potentials; N100; P200; P300

Core Tip: Using ecologically valid traditional Chinese instrumental music as stimuli and multi-stage event-related potentials analysis, this study found that patients with major depressive disorder exhibit multi-stage abnormalities in musical emotion processing: Enhanced central N100 amplitude (early sensory hypervigilance), failure of P200 to distinguish neutral from negative music (feature discrimination deficit), and absence of P300 emotional modulation (impaired late cognitive evaluation), accompanied by lower behavioral accuracy and longer reaction times. These event-related potentials profiles may serve as potential neurobiological markers for cognitive impairments in major depressive disorder.

INTRODUCTION

Major depressive disorder (MDD) is a highly prevalent mental illness worldwide, often accompanied by severe functional impairment. In addition to abnormalities in emotion regulation, cognitive impairment is now recognized as one of its core pathological features[1-3]. These cognitive deficits span multiple dimensions including attentional allocation, sensory discrimination, and higher-order cognitive evaluation, and are closely associated with poor treatment response and unfavorable long-term functional recovery[3]. Despite its growing clinical significance, the neural mechanisms underlying multi-stage cognitive processing impairments in MDD patients remain unclear, particularly within the context of complex, realistic stimuli.

Auditory emotion processing, a core component of social cognition, is frequently impaired in MDD patients. However, existing research largely relies on simple stimuli with low ecological validity, such as pure tones or isolated emotional words[4,5]. Music, as a complex, naturalistic auditory stimulus integrating pitch, timbre, rhythm, and emotional valence, better simulates real-world information processing scenarios[6]. Nevertheless, critical gaps persist in research on musical emotion processing in MDD. It is particularly unclear whether impairments are pervasive across all processing stages, from early sensory encoding to mid-stage feature discrimination and late-stage cognitive evaluation, or whether they occur specifically in certain phases. Furthermore, while behavioral studies confirm difficulties in musical emotion categorization tasks in MDD patients, the precise neural origins of these deficits, whether from early sensory dysfunction, impaired mid-stage discriminative ability, or late-stage cognitive impairment, remain to be elucidated. Do neural responses to music classified by its emotional valence, where positive music evokes pleasant feelings, negative music evokes unpleasant feelings, and neutral music is neither strongly pleasant nor unpleasant, differ between MDD patients and healthy individuals? Compounding these issues, the current literature often lacks ecological validity or the temporal resolution necessary to pinpoint the exact stages of dysfunction within a nuanced, multi-layered cognitive process like musical emotion judgment.

Event-related potentials (ERPs), with their millisecond-level temporal resolution, provide a unique tool for parsing the temporal dynamics of neural processing[7,8]. Three core components are particularly relevant to emotional cognition: N100 (80-120 milliseconds)[9,10], responsible for early sensory input and attentional allocation; P200 (150-200 milliseconds)[9,11], involved in rapid stimulus feature discrimination and initial affective categorization; and P300 (500-800 milliseconds)[12], governing late cognitive evaluation and information integration. Although existing studies have reported abnormalities in these components in MDD patients, the findings contain contradictions (e.g., regarding P300 amplitude), which may be related to insufficient ecological validity of stimuli and a lack of systematic multi-stage analysis[13,14].

Another unresolved scientific issue is whether the neural abnormalities associated with MDD reflect generalized processing deficits or valence-specific impairments. Behavioral studies have confirmed difficulties in musical emotion categorization tasks in MDD patients, but it remains unclear whether these deficits originate from early sensory dysfunction, impaired mid-stage discriminative ability, or late-stage cognitive impairment. Additionally, the issue of cultural specificity in emotional stimuli has received scant attention most existing studies employ Western music, limiting the generalizability of their findings to non-Western populations.

To address these key scientific gaps, this study employs a multi-stage ERPs framework, utilizing ecologically valid stimuli from traditional Chinese instrumental music to systematically investigate: (1) The behavioral characteristics of musical emotion processing in MDD patients; (2) Stage-specific neural abnormalities in the N100, P200, and P300 components; and (3) Whether neural modulation in response to music of different emotional valences is impaired in MDD patients. By addressing these questions, this study aims to clarify the neural mechanisms of musical emotion processing impairments in MDD, identify potential neurobiological markers, and provide new perspectives for the clinical assessment and intervention of cognitive deficits in depression.

MATERIALS AND METHODS

Time and setting

This study was conducted between January 1, 2025, and August 1, 2025, at the Department of Psychiatry and the Department of Clinical Mental Rehabilitation, the Affiliated Mental Health Center of Jiangnan University. The study protocol received approval from the Ethics Committee of the Affiliated Mental Health Center of Jiangnan University, No. WXMHCIRB2025 LLky011, and was carried out in accordance with the principles outlined in the Declaration of Helsinki. This study has been registered with the Chinese Clinical Trial Registry under registration number ChiCTR2500113809. Prior to participation, all participants were fully informed about the experimental procedures and equipment and provided written informed consent.

Participants

This study initially enrolled 32 patients diagnosed with MDD and 30 healthy control (HC) participants matched on key demographic variables. All participants were recruited from the Department of Psychiatry at the Affiliated Mental Health Center of Jiangnan University. MDD diagnoses were confirmed by board-certified psychiatrists according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria.

The inclusion criteria for the MDD group were: (1) A score of 20 or higher on the 24-item Hamilton Depression Rating Scale (HAMD-24); (2) Age between 18 years and 65 years; (3) Absence of comorbid medical or neurological conditions, traumatic brain injury, or history of substance use disorder; and (4) Normal auditory, visual, verbal, and writing abilities. The exclusion criteria included: (1) A comorbid diagnosis of any other DSM-5 psychiatric disorder; and (2) A history of electroconvulsive therapy or modified electroconvulsive therapy within the six months prior to enrollment. All MDD patients were experiencing an acute depressive episode and completed the experimental procedures within one week of hospital admission.

During the study period, all MDD patients were undergoing antidepressant treatment. The treatment regimens included selective serotonin reuptake inhibitors, serotonin and norepinephrine reuptake inhibitors, and other commonly prescribed antidepressants. Specifically, eight patients were prescribed sertraline (mean dose: 185.0 ± 26.7 mg/day), five received escitalopram oxalate (mean dose: 18.4 ± 2.5 mg/day), seven were taking mirtazapine (mean dose: 34.6 ± 8.3 mg/day), four received paroxetine (mean dose: 25.0 ± 8.1 mg/day), and five were on venlafaxine (mean dose: 225.6 ± 24.1 mg/day). According to a previously established conversion method[15], the mean fluoxetine-equivalent dose for all patients was 32.5 ± 13.7 mg/day.

HC participants were recruited through community advertisements and were group-matched to MDD patients based on age, sex, education level, or handedness distribution. For HC individuals, exclusion criteria included not only any personal history of psychiatric disorder (as diagnosed by DSM-5 criteria) but also any history of major chronic physical illnesses (e.g., severe cardiovascular disease, neurological disorders, uncontrolled diabetes, or autoimmune diseases), significant head injury, or current use of medications known to affect central nervous system function or cognitive processing. Specifically, all HC participants were medically healthy and not undergoing any concurrent medical treatments that could confound the results.

An a priori power analysis was performed using G*Power software (version 3.1.9.7). With parameters set to a statistical power of 0.95 (1 - β = 0.95) and a two-tailed alpha level of 0.05 for F-tests, the results indicated that a minimum sample size of 23 participants per group was required to detect the anticipated effects. Following data collection, two MDD patients and one HC participant were excluded from the final analyses due to excessive artifacts in the electroencephalographic recordings, including poor electrode contact or ocular and muscle movements. Consequently, the final analytical sample comprised 30 MDD patients and 29 HC participants. As summarized in Table 1, the final MDD and HC groups showed no statistically significant differences in age, years of education, or handedness distribution (all P > 0.05).

Table 1 Demographic and clinical characteristics of participants, mean ± SD.

Variable	MDD (n = 30)	HC (n = 29)	t/χ²	P value
Age (years)	32.30 ± 10.63	30.93 ± 8.53	-0.55	0.588
Sex (male/female)	10/20	13/16	5.18	0.075
Education (years)	16.00 ± 2.58	14.20 ± 2.61	1.82	0.074
Handedness (R/M/L)	13/7/10	11/8/10	0.216	0.897
Age of onset (years)	28.93 ± 10.55
Duration of illness (months)	40.00 ± 45.39
HAMD-24	24.33 ± 3.65

MDD: Major depressive disorder; HC: Healthy controls; HAMD-24: 24-item Hamilton Depression Rating Scale; R: Right; M: Mixed; L: Left.

Stimulus materials

The musical stimuli for this study were selected and prepared in collaboration with a registered music therapist and a professional composer. Candidate pieces were chosen from traditional Chinese instrumental works recommended by mainstream music platforms nationally and internationally, adhering strictly to three selection criteria: (1) Exclusion of pieces containing vocal singing; (2) Avoidance of widely known film or animation soundtracks to minimize familiarity effects; and (3) Requirement for each piece to possess a distinct and consistent emotional tone. Based on emotional valence and energy level, three representative works were selected as emotional prototypes: The lively and uplifting “Xi Yang Yang” (played on the yangqin) for positive emotion; the serene and flowing “Chun Jiang Hua Yue Ye” (played on the guzheng) as a neutral reference; and the melancholic and sorrowful “Er Quan Ying Yue” (played on the erhu) for negative emotion.

All musical stimuli were edited into 4-7 seconds segments using Cool Edit Pro 2.1 software (Syntrillium Software Corporation, Phoenix, AZ, United States). This classic professional digital audio workstation was utilized for precise trimming of the musical pieces to achieve the desired segment lengths, following recommendations from music psychology research. This duration is considered sufficient to elicit basic emotional responses while avoiding emotional fluctuations or auditory fatigue due to excessive length[16]. Additionally, all segments underwent uniform sound quality optimization and loudness normalization (calibrated to 70 dB) using Cool Edit Pro 2.1’s powerful audio processing capabilities, specifically its normalize and noise reduction tools. This ensured that all musical stimuli presented to participants maintained consistent acoustic characteristics and loudness levels, thereby controlling for the potential confounding effects of acoustic intensity on electroencephalogram (EEG) data.

To validate the emotional elicitation effectiveness of the stimuli, behavioral assessments were conducted in stages. First, an initial pool of music segments underwent behavioral evaluation. Twenty graduate students specializing in emotion research rated the emotional valence of all segments using a 5-point Likert scale (1 = very sad, 5 = very happy). Based on the rating results, only segments where the target emotion dimension score was significantly dominant were retained. This screening process ultimately yielded 84 segments with clear emotional specificity (28 positive, 28 neutral, and 28 negative) for inclusion in the formal experiment.

Subsequently, another 20 independent raters (10 male, 10 female) with no professional music training were recruited for a secondary validation of a randomly selected subset of 19 core segments (8 positive, 3 neutral, 8 negative) from the final set of 84 stimuli. They rated these segments on emotional valence (5-point scale) and energy level (9-point scale, 1 = very low energy, 9 = very high energy). Statistical analysis of the ratings for these 19 segments further confirmed the validity of the stimulus materials. All data management, descriptive statistics, and inferential analyses, including the calculation of Cronbach’s α coefficients, execution of one-sample t-tests, and performance of ANOVA, were conducted using IBM SPSS Statistics software, Version 26.0 (IBM Corp., Armonk, NY, United States). Cronbach’s α coefficients were 0.89 for emotional valence ratings and 0.85 for energy level ratings, indicating good inter-rater reliability and stable results. One-sample t-tests (test value = 3, the neutral point) revealed that the mean valence of positive segments (4.16 ± 0.52) was significantly higher than the neutral point [t (19) = 10.32, P < 0.001], the mean valence of negative segments (1.81 ± 0.15) was significantly lower than the neutral point [t (19) = -32.67, P < 0.001], and the mean valence of neutral segments (2.95 ± 0.10) did not differ significantly from the neutral point [t (19) = -1.83, P = 0.085]. Energy level ratings showed: Positive segments (7.82 ± 0.61) > neutral segments (4.95 ± 0.48) > negative segments (2.33 ± 0.57). ANOVA indicated significant main effects of stimulus category on both emotional valence [F (2, 38) = 689.45, P < 0.001, η² = 0.97] and energy level [F (2, 38) = 346.72, P < 0.001, η² = 0.95]. Post-hoc tests (Bonferroni corrected) further confirmed that positive segments had significantly higher valence and energy levels than both neutral and negative segments (all P < 0.001), and neutral segments were significantly higher than negative segments on both measures (all P < 0.001). These results collectively demonstrate that the selected musical stimulus materials (including the full set of 84 segments) can effectively and specifically induce the target emotions, with clear differentiation in valence and energy levels, thus meeting the requirements of the experimental design.

Emotional task

The experiment was conducted in a sound-attenuated laboratory. Auditory stimuli were presented at a pre-set moderate volume through speakers. Participants were seated approximately 60 cm from a computer screen with their head position stabilized using a chin rest. During the task, participants placed the index finger of their right hand on designated response keys (“J”, “B”, and “F”) in the alphanumeric section of the keyboard, while their left hand rested naturally on their thigh.

Each trial began with a fixation cross (“+”) displayed at the center of the screen for 15000 milliseconds, during which a music stimulus was played simultaneously. Participants were instructed to remain relaxed and focus on listening to the music. Immediately following the offset of the music stimulus, the screen advanced automatically to an emotional rating interface. Participants were required to rate the emotional valence of the music excerpt by pressing the corresponding key: “J” for positive, “B” for neutral, and “F” for negative. A maximum response window of 15000 milliseconds was allowed for each rating, and participants were asked to respond based on their immediate subjective feeling.

Upon completion of each rating, the experiment proceeded automatically to the next trial. A total of 84 music excerpts were presented in a fully randomized order. To minimize potential influences of emotional carry-over effects and to ensure that participants fully understood the task requirements, all participants received a detailed explanation of the procedure and the response key assignments prior to the formal experiment. Additionally, practice trials using excerpts independent of the formal stimulus set were administered to familiarize participants with the task. A visual illustration of the above-described emotional task flow is provided in Figure 1.

Figure 1 Schematic diagram of the emotional task.

EEG recording and analysis

EEG data were recorded continuously during the emotion task using a 64-channel EasyCap system (Brain Products GmbH, Germany) connected to a BrainAmp Standard DC amplifier (Brain Products GmbH, Gilching, Germany). Signals were sampled at 500 Hz with an online band-pass filter set to 0.1-100 Hz. The reference electrode was placed at FCz on the forehead, and the ground electrode was positioned 1-2 cm below the left clavicle. Horizontal electrooculogram electrodes were attached 1 cm lateral to the outer canthi of both eyes, and a vertical electrooculogram electrode was placed 1 cm below the pupil of the left eye to monitor ocular artifacts. All electrodes were maintained with impedances below 5 kΩ. Stimuli presentation was precisely controlled by E-Prime 3.0 software (Psychology Software Tools, Sharpsburg, PA, United States) running on a dedicated computer with a Lenovo display. EEG data acquisition occurred in an electrically shielded and sound-attenuated room.

Offline preprocessing was conducted using EEGLAB 2021 (Delorme and Makeig, 2004) running in MATLAB 2020b (The MathWorks, Inc., Natick, MA, United States). Continuous data were band-pass filtered between 0.1 Hz and 30 Hz (24 dB/octave slope), and re-referenced to the average of the left and right mastoids. Independent component analysis using the runica algorithm was applied to identify and remove artifacts related to ocular, muscular, and cardiac activity. Components identified as artifactual were visually inspected and removed. Epochs containing amplitudes exceeding ± 100 μV were excluded from further analysis.

Analysis focused on two types of event-locked epochs: Those aligned to the onset of emotional stimuli (neutral, negative, positive) and those aligned to the participant’s button-press response. For stimulus-locked ERPs, epochs were extracted from -200 milliseconds to 1000 milliseconds post-stimulus, with a -200 millisecond to 0 milliseconds pre-stimulus baseline used for amplitude correction. and mean amplitudes were computed for the N100 (80-120 milliseconds), P200 (150-200 milliseconds), and P300 (500-800 milliseconds) components across regions of interest: Left prefrontal (F3, F5, FC3), right prefrontal (F4, F6, FC4), and central (Fz, Cz, FCz). For response-locked ERPs, data were segmented from -200 milliseconds to +800 milliseconds relative to the button press to examine brain activity associated with motor responses, with a -200 millisecond to 0 milliseconds pre-response baseline.

For each emotional condition, separate repeated-measures ANOVAs were performed to assess the effects of emotion and electrode scalp region on the ERP components. The Greenhouse-Geisser correction was applied when sphericity was violated, and post-hoc analyses with Bonferroni correction were conducted where appropriate to further investigate significant effects.

Assessment for the severity of MDD

The HAMD-24 was employed to evaluate and quantify the severity of depressive symptoms in patients with MDD. This scale represents one of the most widely used international instruments for clinical assessment of depression. According to the standardized scoring criteria, total scores below 20 are considered within the normal range, scores ≥ 20 indicate depressive states, and scores exceeding 35 reflect severe depression. In this study, all enrolled MDD patients (n = 30) had HAMD-24 scores meeting the diagnostic threshold for depression (24.33 ± 3.65), confirming that the study sample was in a clinically depressed state. The use of this extensively validated scale enabled accurate characterization of depression severity in the patient cohort and provided crucial clinical reference for subsequent analyses examining relationships between depressive symptoms and cognitive measures.

Statistical analysis

All statistical analyses were conducted using IBM SPSS Statistics (Version 22; IBM Corp., Armonk, NY, United States). Demographic and clinical continuous variables were compared between groups using independent samples t-tests, while categorical variables were assessed using χ² tests. Behavioral performance, including emotional valence judgment accuracy and reaction time and ERP data were analyzed using repeated-measures ANOVA. When the sphericity assumption was violated, the Greenhouse-Geisser correction was applied. For significant interaction effects, simple effect analyses were performed, followed by post-hoc comparisons using the Bonferroni method. Relationships between HAMD-24 scores and behavioral and ERP measures were examined using two-tailed Pearson correlation analysis. Effect sizes are reported as partial eta-squared (η²_p) for ANOVA and Cohen’s d for t-tests, with the statistical significance threshold set at P < 0.05.

Emotional valence judgment accuracy was defined as the proportion of trials in which the musical emotion type was correctly identified; reaction time referred to the interval from the onset of the rating interface to a correct keypress response. Behavioral data were analyzed using a 3 (emotional condition: Neutral, negative, positive) × 2 (group: HC, MDD) mixed-design ANOVA, applied separately to accuracy rates and reaction times.

For stimulus-locked ERP analyses, mean amplitudes of the N100 (80-120 milliseconds), P200 (150-200 milliseconds), and P300 (500-800 milliseconds) components were evaluated using a 3 (emotional condition) × 2 (group) × 3 [region: Left prefrontal (F3, F5, FC3), right prefrontal (F4, F6, FC4), central (Fz, Cz, FCz)] mixed-design ANOVA. For response-locked analyses, EEG data spanning from -200 milliseconds to +800 milliseconds relative to the button press were extracted and analyzed using the same factorial design. All ERP analyses were performed separately for each emotional valence condition (neutral, negative, positive). Repeated-measures ANOVAs were used to examine the effects of group, emotional condition, and electrode region on ERP components.

RESULTS

Participant characteristics

The results of independent samples t-tests and χ² tests showed no significant differences between the MDD and HC groups in terms of age, sex, education level, or handedness distribution (all P > 0.05, Table 1).

Behavioral results

Task accuracy: A 3 (emotional condition) × 2 (group) repeated-measures ANOVA on accuracy rates revealed a significant violation of sphericity (Mauchly’s W = 0.654, P < 0.001), therefore the Greenhouse-Geisser correction was applied. The analysis showed no significant main effect of emotional condition [F (1.49, 83.22) = 1.33, P = 0.265, η²_p = 0.023], and no significant condition × group interaction [F (1.49, 83.22) = 0.45, P = 0.583, η²_p = 0.008]. However, a significant main effect of group was found [F (1, 56) = 7.605, P = 0.008, η²_p = 0.232]. Pairwise comparisons indicated that the HC group showed significantly higher overall accuracy rates compared to the MDD group [mean difference (MD) = 0.126, P = 0.008]. Specific accuracy values across emotional conditions and groups are presented in Table 2.

Table 2 Behavioral results (accuracy and reaction times), mean ± SD.

	MDD (n = 30)			HC (n = 29)
	Neutral	Negative	Positive	Neutral	Negative	Positive
ACC	0.75 ± 0.27	0.84 ± 0.28	0.81 ± 0.32	0.90 ± 0.13	0.92 ± 0.21	0.96 ± 0.19
RTs	846.22 ± 306.88	771.50 ± 280.46	649.20 ± 255.55	676.47 ± 270.25	591.95 ± 254.38	504.39 ± 220.23

ACC: Accuracy; RTs: Reaction times; MDD: Major depressive disorder; HC: Healthy controls.

Reaction time: A 3 (emotional condition) × 2 (group) repeated-measures ANOVA on reaction times showed no violation of sphericity (Mauchly’s W = 0.926, P = 0.113), yet the Greenhouse-Geisser correction was applied for consistency. The analysis revealed a significant main effect of emotional condition [F (1.86, 108.03) = 43.08, P < 0.001, η²_p = 0.426]. Multivariate tests also showed a significant main effect of emotional condition [Wilks’ Λ = 0.331, F (2, 57) = 57.649, P < 0.001, η²_p = 0.669].

Post-hoc tests indicated that all participants responded significantly faster to positive music compared to both neutral (MD = 184.550 milliseconds, P < 0.001) and negative music (MD = 104.929 milliseconds, P < 0.001). Additionally, responses to neutral music were significantly faster than those to negative music (MD = 79.622 milliseconds, P = 0.001). A significant main effect of group was also found [F (1, 58) = 6.443, P = 0.014, η²_p = 0.100]. Pairwise comparisons demonstrated that MDD patients had significantly longer reaction times compared to HCs (MD = 164.703 milliseconds, P = 0.014). The condition × group interaction was not significant [F (1.86, 108.03) = 0.043, P = 0.655, η²_p = 0.064]. Detailed reaction time data and accuracy data across all conditions and groups are shown in Table 2, and the full results of the repeated-measures ANOVA for accuracy and reaction time are summarized in Table 3.

Table 3 Repeated-measures ANOVA on accuracy and reaction time.

	F	P value	η2p	Df (num)	Df (den)
ACC
Group	7.605	0.008	0.232	1	56
Condition	1.33	0.265	0.023	1.49	83.22
Condition × group	0.45	0.583	0.008	1.49	83.22
RTs
Group	6.443	0.014	0.100	1	58
Condition	43.076	< 0.001	0.426	1.86	108.03
Condition × group	0.043	0.655	0.064	1.86	108.03

ACC: Accuracy; RTs: Reaction times; η²_p: Partial eta-squared; df: Degrees of Freedom; MDD: Major depressive disorder; HC: Healthy controls.

ERP results

During the ERP analysis of EEG data, the left forehead, right forehead and central region were selected for analysis and comparison of the differences between MDD and HC under three different emotional conditions (neutral, negative and positive). Statistical analyses were performed on the amplitudes of core ERP components (N100, P200, P300). Components that exhibited significant interactions of group, emotion, or brain region are visually presented in Figures 2, 3, and 4. As Supplementary Tables 1 and 2 are too large, they have been moved to the Supplementary material to maintain a streamlined flow of the main text. Key findings are summarized in the main body, and complete data can be found in the them.

Figure 2 Amplitude differences of the N100 event-related potential during stimulus onset and button-response phases. A-C: These panels illustrate the amplitude distribution of the N100 component across various electrode locations (left prefrontal, right prefrontal, central regions) and stimulus types (neutral, negative, positive) during the stimulus onset phase; D-F: Correspondingly, these panels present the N100 amplitude distribution across the same electrode locations and stimulus types during the button-response phase. Green bars represent the patient group, while dark blue bars represent the healthy control group. Error bars indicate standard errors. HC: Healthy control; MDD: Major depressive disorder.

Figure 3 Amplitude distribution of the P200 event-related potential across stimulus onset and button-response phases. A-C: Display the P200 amplitude across different electrode locations (left prefrontal, right prefrontal, central regions) and three stimulus types (neutral, negative, positive) during the stimulus onset phase; D-F: Illustrate the P200 amplitude at corresponding electrode locations and stimulus types during the button-response phase. Green bars represent the patient group, and dark blue bars represent the healthy control group; error bars indicate standard errors. MDD: Major depressive disorder; HC: Healthy control.

Figure 4 Amplitude distribution of the P300 event-related potential during stimulus onset and button-response phases. A-C: Present the P300 amplitude across different electrode locations (left prefrontal, right prefrontal, central regions) and three stimulus types (neutral, negative, positive) in the stimulus onset phase; D-F: Illustrate the P300 amplitude at corresponding electrode locations and stimulus types in the button-response phase. Green bars represent the patient group, and dark blue bars represent the healthy control group; error bars indicate standard errors. MDD: Major depressive disorder; HC: Healthy control.

N100: Analysis of the stimulus-onset phase revealed a significant main effect of emotional condition on N100 amplitude [F (2, 114) = 9.554, P < 0.001, η²_p = 0.144], with neutral stimuli eliciting smaller amplitudes than both negative (P < 0.001) and positive (P = 0.001) stimuli. A significant condition × region interaction [F (2.522, 143.754) = 2.844, P = 0.049, η²_p = 0.048] was also observed.

During the button-response phase, significant main effects of emotional condition [F (2, 114) = 7.648, P = 0.001, η²_p = 0.118] and brain region [F (1.793, 102.208) = 4.872, P = 0.012, η²_p = 0.079] emerged, along with a significant region × group interaction [F (1.793, 102.208) = 5.573, P = 0.007, η²_p = 0.089]. Post-hoc analysis indicated that MDD patients demonstrated significantly enhanced N100 amplitudes in the central region compared to other areas, a pattern not observed in HCs.

N100 latency analysis revealed a significant main effect of brain region specifically during the button-response phase [F (1.792, 102.142) = 5.290, P = 0.008, η²_p = 0.085], with prolonged latency in the left forehead region compared to both right forehead (P = 0.008) and central areas (P = 0.013). These group differences in waveform morphology and spatial distribution characteristics are illustrated in Figure 5.

Figure 5 Grand-avg N100 (40-120 milliseconds) event-related potential waveforms/topo maps of major depressive disorder and healthy control under neutral/negative/positive emotional music. Waveforms (left prefrontal: F3, F5, FC3; right prefrontal: F4, F6, FC4; central: FZ, CZ, FCZ) show N100 latency/amplitude group differences. Topo maps show spatial differences in early music emotion processing. MDD: Major depressive disorder; HC: Healthy control.

P200: Analysis of P200 amplitude during the stimulus-onset phase revealed significant main effects of emotional condition [F (1.971, 112.336) = 10.50, P < 0.001, η²_p = 0.156] and brain region [F (1.994, 113.673) = 22.01, P < 0.001, η²_p = 0.279]. Most crucially, a significant emotional condition × group interaction was observed [F (1.971, 112.336) = 4.65, P = 0.012, η²_p = 0.075]. Simple effect analysis delineated two distinct emotional processing patterns between the groups: HCs exhibited finely differentiated neural responses across all emotional conditions (all pairwise comparisons P < 0.05), whereas patients with MDD specifically failed to effectively discriminate between neutral and negative stimuli (P = 1.000). Between-group comparisons further confirmed that MDD patients exhibited significantly larger P200 amplitudes specifically in response to negative stimuli compared to HC [central region: t (57) = 2.03, P = 0.047; right prefrontal region: t (57) = 2.01, P = 0.049; left prefrontal region: t (57) = 1.98, P = 0.052]. Although a significant main effect of brain region was present, the critical emotional condition × group interaction did not yield a significant three-way interaction with brain region (P = 0.362), indicating that the impairment in differentiating neutral from negative stimuli in MDD patients represents a generalized characteristic across brain regions, rather than being specific to any particular area. These specific response patterns associated with the P200 component and their corresponding brain activation topography are illustrated in Figure 6.

Figure 6 Grand-avg P200 (150-250 milliseconds) event-related potential waveforms/topo maps of major depressive disorder and healthy control under neutral/negative/positive emotional music. Waveforms (left prefrontal: F3, F5, FC3; right prefrontal: F4, F6, FC4; central: FZ, CZ, FCZ) show P200 latency/amplitude group differences. Topo maps show spatial differences in early music emotion processing. MDD: Major depressive disorder; HC: Healthy control.

During the button-response phase, the analysis of P200 amplitude showed that only the main effect of brain region remained significant [F (1.730, 98.582) = 6.970, P = 0.002, η²_p = 0.109], manifested as enhanced amplitudes over central regions. In contrast, neither the main effect of emotional condition nor the emotional condition × group interaction reached significance (all P > 0.05). Analysis of P200 latency further revealed differences in processing speed. Significant main effects of emotional condition were found during both the stimulus-onset phase [F (2, 114) = 13.932, P < 0.001, η²_p = 0.196] and the button-response phase [F (1.939, 110.515) = 4.916, P = 0.010, η²_p = 0.079], with positive stimuli consistently eliciting the shortest latencies.

P300: Analysis of the stimulus-onset phase revealed significant main effects of brain region [F (1.905, 108.560) = 13.468, P < 0.001, η²_p = 0.191] and significant condition × group [F (1.993, 113.588) = 5.003, P = 0.008, η²_p = 0.081] and condition × region interactions [F (3.232, 184.234) = 4.362, P = 0.004, η²_p = 0.071]. HCs demonstrated significantly larger P300 amplitudes for neutral compared to negative stimuli, particularly in central regions, while MDD patients showed no significant amplitude modulation across emotional conditions.

Notably, no significant emotional modulation or group differences were observed in P300 amplitudes or latencies during the button-response phase. All analyses incorporated Greenhouse-Geisser corrections where appropriate, and effect sizes are reported as η²_p. The characteristic waveform patterns and topographic distribution of the P300 component are collectively presented in Figure 7.

Figure 7 Grand-avg P300 (250-400 milliseconds) event-related potential waveforms/topo maps of major depressive disorder and healthy control under neutral/negative/positive emotional music. Waveforms (left prefrontal: F3, F5, FC3; right prefrontal: F4, F6, FC4; central: FZ, CZ, FCZ) show P300 latency/amplitude group differences. Topo maps show spatial differences in early music emotion processing. MDD: Major depressive disorder; HC: Healthy control.

Correlation

To examine the relationship between depression severity and emotional information processing, we performed separate correlation analyses between behavioral measures and ERP data. The behavioral correlations revealed several key findings. First, reaction times across all three emotional conditions were strongly and positively correlated (all r > 0.75, P < 0.001), indicating a consistent pattern of response speed irrespective of emotional valence. Second, a strong positive correlation was observed between accuracy rates for positive and negative emotions (r = 0.690, P < 0.001), suggesting a common emotion discrimination ability.

Critically, a speed-accuracy trade-off was evident: Reaction times under all emotional conditions were positively correlated with accuracy for neutral stimuli (all r approximately was 0.39, P < 0.05), indicating that slower responses were associated with more accurate identification of neutral music. Regarding clinical correlations, HAMD scores were significantly positively correlated with accuracy for neutral stimuli (r = 0.380, P < 0.05), but were not associated with reaction times or with accuracy for positive or negative emotions. Neurophysiologically, no significant correlations were found between HAMD scores and the mean amplitudes of any ERP components across brain regions or emotional conditions (Figure 8).

Figure 8 This correlation matrix heatmap illustrates relationships among reaction times, accuracy in neutral/negative/positive emotional tasks, and Hamilton Depression Rating Scale scores in major depressive disorder patients. Color intensity and dot size reflect correlation strength (dark blue = strong negative, light = positive). Positive correlations exist within reaction times and within accuracy across emotions; Hamilton Depression Rating Scale shows negative correlations with several emotional processing indicators, quantifying links between emotional cognition and depressive severity in major depressive disorder. RTs: Reaction times; ACC: Accuracy; HAMD: Hamilton Depression Rating Scale.

DISCUSSION

The present study employed a multi-stage ERP approach with ecologically valid musical stimuli to delineate the temporal dynamics of emotional processing impairments in MDD. While confirming broader behavioral deficits and late-stage positive processing diminution (P300), our most salient finding reveals a specific and early neural dysfunction in MDD: A failure to neurally discriminate between neutral and negative emotional stimuli at the P200 stage. This impairment in early feature discrimination suggests a fundamental blurring in the initial affective classification system, providing a novel mechanistic account for the emergence of negative cognitive bias in depression.

The core finding: A blurred affective classification system at the P200 stage

The P200 component is widely recognized as an index of early feature discrimination and the initial affective evaluation of a stimulus’s salience[17,18]. Our data demonstrate that while HCs exhibit a finely tuned P200 response that clearly differentiates between neutral, negative, and positive music, MDD patients display a crippled discriminatory capacity. Specifically, their neural responses fail to distinguish neutral from negative stimuli.

This finding moves beyond the conventional narrative of a simple “negativity bias”. It posits that the pathology in MDD is not merely an amplification of negative processing but, more fundamentally, a failure of early affective categorization. The brain’s initial “sorting mechanism” appears over-generalized, mis-categorizing non-threatening, neutral information as possessing a negative valence. This could stem from dysfunction in a prefrontal-parietal network responsible for rapid stimulus evaluation[18] and/or disrupted amygdala-prefrontal circuitry that assigns emotional significance[19]. An overactive amygdala might tag ambiguous or neutral stimuli as potentially threatening, while a hypoactive prefrontal cortex fails to provide the top-down regulatory signal needed for precise differentiation. This early misclassification provides a distorted foundation upon which all subsequent cognitive operations are built, effectively “priming” the brain for a negative interpretive bias.

A point warranting specific discussion is whether the similarity in lower arousal levels between neutral and negative music constitutes an acoustic confound leading to the indistinguishable P200 responses. However, several lines of evidence lead us to favor the “blurred affective classification” pathophysiological explanation. First, behavioral data reveal a difference in classification accuracy between neutral and negative music in MDD patients (neutral: 0.75 vs negative: 0.84), indicating that they retain a certain capacity for behavioral discrimination. This highlights a discrepancy between their early neural classification (P200) and the eventual behavioral outcome. Furthermore, and most compellingly, HCs exhibited clearly differentiated P200 responses to the identical set of musical stimuli. This demonstrates that the acoustic properties of the stimulus materials themselves are discernible and effectively encoded by a healthy brain. Therefore, the P200 blurring in MDD patients more likely reflects an intrinsic neural processing abnormality rather than being attributable to the physical properties of the external stimuli.

Neural mechanisms of N100 abnormality in MDD

N100, as an early sensory-attentional component (80-120 milliseconds), primarily reflects sensory input processing and attention resource allocation[17,20]. Our results show that MDD patients exhibit significantly enhanced N100 amplitude in the central region, which differs from the spatial distribution pattern of HCs. This enhancement is not limited to negative stimuli but is observed across all emotional conditions, suggesting a generalized hypervigilance rather than selective attention bias toward negative information.

Traditionally, N100 enhancement in MDD has been attributed to early sensory hyperarousal or negative attention bias[17]. However, our findings suggest a more specific interpretation: N100 abnormality is an independent early processing deficit parallel to the P200 classification defect. It may originate from congenital functional impairments in the emotional processing network, leading to over-sensitivity of the sensory-attentional system. This hypervigilance increases the cognitive load for subsequent stimulus discrimination, as the brain receives excessive undifferentiated sensory signals, further exacerbating the P200 classification deficit. Unlike a one-way causal relationship, N100 hypervigilance and P200 classification failure interact dynamically, forming the initial source of abnormal emotional processing in MDD.

Neural mechanisms of P300 abnormality in MDD

P300 (250-400 milliseconds) is closely related to late cognitive assessment, value judgment and information integration[14,20]. This study reproduced the finding that P300 in MDD patients lacks effective emotion modulation: Unlike the HC group, which showed differentiated P300 amplitudes under different emotional conditions, MDD patients presented flattened responses to neutral, negative and positive music.

This P300 passivation is not a primary defect of late cognitive function, but a secondary result of abnormal early processing: On the one hand, P200 cannot distinguish between neutral and negative stimuli, causing the brain to continuously process vague “quasi-negative” signals and resulting in excessive consumption of early cognitive resources; on the other hand, the persistent excessive alertness of N100, as another independent early processing anomaly, will simultaneously disrupt the rational allocation of attention resources. Under the combined effect of the two, the P300 stage, due to the lack of sufficient cognitive and attention resources, is difficult to conduct in-depth processing and value assessment of positive stimuli. Therefore, the P300 anomaly reflects a systematic collapse of the entire emotional processing chain, marking the cumulative endpoint of processing damage from the early stage to the late stage.

Our multi-stage ERP findings, which delineate early (N100, P200) and late (P300) temporal dynamics of emotional processing deficits in MDD, offer crucial insights that are highly complementary to those obtained from other neuroimaging modalities. For instance, functional magnetic resonance imaging (fMRI) studies, with their superior spatial resolution, have consistently identified dysfunctional patterns in limbic and prefrontal regions during emotional music processing in depression. Aust et al[21] reported altered neural correlates of emotional experience in remitted depression using fMRI, underscoring persistent neural vulnerabilities even after symptomatic remission. Similarly, research by Mitterschiffthaler et al[22] utilized fMRI to elucidate brain regions associated with happy and sad affective states induced by classical music, highlighting the role of distributed emotional networks. While these fMRI studies effectively localize affected brain regions, our ERP approach precisely delineates the timing of these dysfunctions, providing critical temporal resolution to pinpoint when specific processing impairments (e.g., N100 hypervigilance, P200 classification failure, and P300 passivation) occur. As comprehensively reviewed by Sun et al[23] integrating both neuroimaging and neuroelectrophysiological features is essential for a complete and mechanistic understanding of music’s effects on anhedonia in MDD. Therefore, our study, by identifying distinct early and late ERP markers, complements these fMRI findings by adding a layer of temporal precision to the understanding of regional brain activity changes and their impact on the entire emotional processing chain in MDD.

Clinical implications

Our findings provide critical insights into the pathophysiology of MDD that hold significant clinical implications. The identified ERP markers, particularly the early N100 hypervigilance and the P200 deficit in neutral-negative discrimination, offer objective neurophysiological correlates of emotional processing dysfunction. These markers could serve as valuable tools for guiding treatment and aiding in prognosis.

Specifically, these ERP components hold promise as objective biomarkers for identifying individuals with distinct emotional processing deficits, potentially leading to more targeted interventions. For instance, patients exhibiting prominent P200 classification deficits might benefit from therapies specifically designed to refine early affective categorization, such as certain cognitive behavioral techniques or even music-based interventions focused on enhancing emotional differentiation. Furthermore, the ability of these ERP markers to objectively quantify improvements could facilitate the monitoring of treatment efficacy, providing early neurophysiological indicators of an intervention’s impact on underlying emotional processing mechanisms, beyond subjective symptom reports.

Regarding prognosis, baseline ERP patterns or changes in these markers over time could offer valuable predictive insights. For example, the severity of P200 blurring or N100 hypervigilance at diagnosis might correlate with treatment resistance or the long-term course of MDD. Such prognostic indicators could enable clinicians to stratify patients into different risk groups, allowing for more personalized and proactive long-term management strategies, potentially mitigating relapse risk or guiding the selection of more intensive interventions for those identified as less responsive to standard care.

Limitations and future directions

Several limitations should be considered. First, all MDD patients were undergoing antidepressant treatment, and while the specific P200 neutral-negative discrimination deficit is unlikely to be solely medication-driven, the potential influence of drugs on N100 and P300 components cannot be fully ruled out. Specifically, the classes of antidepressants used (selective serotonin reuptake inhibitors, norepinephrine reuptake inhibitors) are known to modulate neural excitability and ERP amplitudes, and their specific, differential efficacy at various dosages on particular cognitive functions remains complex and largely undetermined. Thus, while our study reflects a real-world clinical population, this makes it challenging to definitively disentangle the effects of MDD pathology from the effects of antidepressant treatment on certain ERP components. Future studies should recruit medication-free patients to validate the specificity of these ERP abnormalities. Second, the use of Chinese musical stimuli enhances ecological validity for the sample but may limit cross-cultural generalizability, as emotional perception of music is partially culturally shaped. Third, the sample size, though adequate for primary analyses, may lack power to detect subtle correlations between specific ERP components and clinical symptom dimensions. For example, a larger sample size might facilitate a more robust analysis of the relationship between specific ERP parameters and varying degrees of anxiety or anhedonia. Third, the sample size (n = 30 MDD, n = 29 HC), though adequate for primary analyses, may lack power to detect subtle correlations between specific ERP components and clinical symptom dimensions. For example, a larger sample size might facilitate a more robust analysis of the relationship between specific ERP parameters and varying degrees of anxiety or anhedonia.

Future research directions include: (1) Employing longitudinal designs to determine whether N100 hypervigilance, P200 classification deficit, and P300 passivation are state or trait markers of MDD by tracking ERP changes during treatment and remission; (2) Recruiting medication-free patients to validate the specificity of these ERP abnormalities and to investigate medication-naive or remitted populations; (3) Utilizing source localization techniques (e.g., fMRI, high-density EEG) to identify the neural generators of N100, P200, and P300 abnormalities, clarifying their underlying brain circuitry mechanisms; (4) Conducting cross-cultural replications with diverse musical stimuli and participant populations to assess the generalizability of our findings beyond a specific cultural context; (5) Exploring the clinical utility of these ERP metrics, such as whether they can predict treatment response or serve as targets for neuromodulation interventions tailored to different processing stages; and (6) Future studies should also consider incorporating a broader range of clinical assessments, including specific measures of anhedonia, anxiety, and cognitive symptoms, to better characterize the heterogeneity within MDD populations and allow for more nuanced correlations with ERP findings.

CONCLUSION

This study explored the neural mechanisms of musical emotion processing impairments in MDD using ERPs and ecologically valid Chinese musical stimuli. The results confirmed multi-stage emotional processing deficits, providing novel insights into MDD’s neurobiological basis. Behaviorally, MDD patients showed lower accuracy and longer reaction times in musical emotion judgment than HCs, indicating generalized deficits in processing complex emotional auditory stimuli.

Neurophysiologically, stage-specific ERP abnormalities were identified: Enhanced central N100 (generalized hypervigilance), blunted P200 (failure to differentiate neutral-negative music), and flattened P300 (lack of emotional modulation). These reflect intrinsic neural abnormalities, with early deficits initiating subsequent processing collapse. This study highlights musical paradigms’ utility for MDD research, offering culturally appropriate assessment perspectives and laying a foundation for clinical applications like treatment prediction.