Neural correlates of attentional switching dysfunction in major depressive disorder: Evidence from an event-related potential study with a dual-task paradigm

Abstract

BACKGROUND

Research has consistently demonstrated that patients with major depressive disorder (MDD) exhibit attentional switching dysfunction, and the dual-task paradigm has emerged as a valuable tool for probing cognitive deficits. However, the neuroelectrophysiological mechanism underlying this deficit has not been clarified.

AIM

To investigate the event-related potential (ERP) characteristics of attentional switching dysfunction and further explore the neuroelectrophysiological mechanism of the cognitive processing deficits underlying attentional switching dysfunction in MDD.

METHODS

The participants included 29 MDD patients and 29 healthy controls (HCs). The ERPs of the participants were measured while they performed the dual-task paradigm. The behavioral and ERP N100, P200, P300, and late positive potential (LPP) data were analyzed.

RESULTS

This study revealed greater accuracy in HCs and slower reaction times (RTs) in MDD patients. Angry facial pictures led to lower accuracy. The results also revealed shorter RTs for happy facial pictures and the longest RTs for the 500-ms stimulus onset asynchrony. With respect to ERP characteristics, happy facial pictures and neutral facial pictures evoked higher amplitudes. The N100, P200, P300, and LPP amplitudes at Pz were the highest. MDD patients had lower P200 mean amplitudes and LPP amplitudes than HCs did.

CONCLUSION

In conclusion, MDD patients exhibited abnormal ERP characteristics evoked by the dual-task paradigm, which could be the neural correlates of the known abnormalities in attentional switching in patients with MDD. These results provide valuable insights into the understanding of the neural mechanisms of attentional switching function and may guide targeted interventions in patients with MDD.

Key Words: Major depressive disorder; Attentional switching function; Dual-task paradigm; Event-related potential; Cognition

Core Tip: We have developed and designed a novel dual-task paradigm to investigate attentional switching function in patients with major depressive disorder. This paradigm integrates event-related potential technology, enabling the precise capture of brain activity during attentional switching tasks. This study revealed neurophysiological characteristics associated with attentional switching abnormalities in patients with major depressive disorder, offering valuable insights into the underlying neural mechanisms and potentially guiding future targeted interventions for depression.

INTRODUCTION

Major depressive disorder (MDD) is a prevalent and debilitating psychiatric condition characterized primarily by persistent low mood, anhedonia, and a range of cognitive and somatic symptoms[1]. Despite the availability of pharmacological and psychotherapeutic interventions, a substantial proportion of patients with MDD experience treatment resistance or relapse, underscoring the need for an improved understanding of its pathophysiology and the development of more personalized treatment strategies. Research into the cognitive aspects of MDD is extensive and continually evolving, revealing critical insights pivotal for both diagnosis and treatment strategies. Cognitive deficits in MDD involve multiple functions, including attention, executive function, memory, and processing speed[2]. Research consistently indicates that those with MDD suffer from notable cognitive deficits in these domains[3].

The attentional switching function refers to a key executive function that enables individuals to flexibly shift their focus of attention between different tasks, mental sets, or stimulus features in response to changing environmental demands[4,5]. It is a fundamental aspect of cognitive control and adaptability, allowing for efficient regulation of goal-directed behavior in dynamic contexts[6]. The dual-task paradigm is a widely used experimental approach in cognitive psychology and neuroscience that assesses an individual’s capacity to perform two tasks simultaneously, thereby probing cognitive resource allocation, executive control, and attentional processing[7]. In particular, the dual-task paradigm has been instrumental in studying the mechanisms underlying attentional control and switching[8,9].

In recent years, the dual-task paradigm has emerged as a valuable tool for probing cognitive deficits, offering insights into the interplay between attentional control[10], working memory[11], and cognitive resource allocation in patients with MDD[12]. Many studies have reported that MDD patients exhibit impairments in attentional switching[13], such as exhibiting larger P3 difference wave amplitudes and longer P2 difference wave latencies in cognitive flexibility tasks[14]. MDD patients have difficulty disengaging from negative stimuli, as indicated by higher P300 amplitudes in response to angry faces compared to neutral faces[15]. Previous studies have highlighted cognitive control deficits in MDD, specifically impaired cognitive switching and backward inhibition, with rumination playing a key role in exacerbating these impairments[16]. Deficits in this domain have been linked to the characteristic cognitive rigidity and maladaptive emotional processing observed in MDD. Studies employing dual-task paradigms in MDD populations have consistently demonstrated that patients exhibit greater performance decreases under dual-task conditions than healthy controls (HCs) do[17]. These decreases manifest as increased reaction times (RTs), higher error rates, or reduced task efficiency in one or both tasks. This dual-task cost is interpreted as evidence of impaired cognitive resource allocation and limited attentional control capacity, both of which are core features of cognitive dysfunction in MDD[18]. Moreover, research suggests that dual-task performance deficits in patients with MDD are not merely a consequence of psychomotor slowing but also reflect more fundamental impairments in executive control, including attentional switching, task coordination, and inhibitory processes[19].

Recently, functional magnetic resonance imaging (fMRI) studies have consistently revealed that MDD patients exhibit hypoactivation of key components of the frontoparietal control network, including the dorsolateral prefrontal cortex, anterior cingulate cortex, and posterior parietal cortex[20], regions known to support attentional control, cognitive flexibility, and working memory. These neural deficits are accompanied by impaired coordination between task-relevant networks, such as reduced functional connectivity between the default mode network and executive control networks, suggesting disrupted task-set reconfiguration and interference control. Although fMRI has better spatial resolution, its temporal resolution is limited; moreover, fMRI does not directly record neuronal firing; rather, it infers activity through neurovascular coupling, which may be modulated by various nonneuronal factors (e.g., vascular reactivity and medication)[21].

Event-related potentials (ERPs) are electrophysiological responses to sensory, cognitive, or motor stimuli that are time-locked, obtained through electroencephalography (EEG), and processed using signal averaging methods. As noninvasive, direct measurements of neural activity that are sensitive to subtle cognitive processes and well-characterized component structures and use temporally precise methods, ERPs have become fundamental tools in cognitive neuroscience, psychophysiology, and clinical research[22]. To date, no studies on the ERP characteristics of the attentional switching function with the dual-task paradigm in patients with MDD have been reported. Further exploration of ERP characteristics related to attentional switching in MDD patients could aid in understanding the neural mechanisms behind their attentional impairments. Additionally, insights into the neuroelectrophysiological basis of these deficits might reveal new targets for pharmacological and neuromodulatory interventions. In this study, the participants included MDD patients and HCs, and measurements of ERPs during the dual-task paradigm were used to investigate the neural process underlying the attentional switching function. The aim of this study was to examine the ERP characteristics of attentional switching in MDD and further investigate the neuroelectrophysiological mechanisms underlying the cognitive processing deficits associated with attentional switching dysfunction in this disorder.

MATERIALS AND METHODS

Time and setting

This study was conducted between January 1 and May 31, 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 LLky018) and was carried out in accordance with the principles outlined in the Declaration of Helsinki. The trial was prospectively registered with the Chinese Clinical Trial Registry, with the unique identifier No. ChiCTR2500103379. Prior to participation, all participants were fully informed about the experimental procedures and equipment and provided written informed consent.

Participants

The inclusion criteria for the MDD group were as follows: (1) Met the criteria for MDD in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5); (2) Had Hamilton Depression Rating Scale (24-item edition) (HAMD-24) scores ≥ 21[23]; (3) Were aged 18-65 years; (4) Had no physical/neurological illnesses, traumatic brain injury, or substance abuse history; and (5) Had intact auditory/visual, verbal, and writing abilities. The exclusion criteria were as follows: (1) Met the criteria for any other mental disorder according to the DSM-5; and (2) Were treated with electroconvulsive therapy or modified electroconvulsive therapy within 6 months before recruitment. This study included MDD patients who were in the acute phase of depression and who participated in the experiment within one week of admission. Patients in remission or with significant comorbidities were excluded. The acute phase of depression significantly impacts cognitive performance, and the aim of this study was to assess the cognitive impairments associated with MDD during this critical period.

Healthy people whose gender, age, and years of education were matched were selected as the HC group. HCs were recruited through local advertisements in the residential communities of Wuxi, China. Sample size estimation was performed using G*Power software (version 3.1.9.7; Heinrich Heine University, Düsseldorf, Germany), with the parameters set to achieve a statistical power of 0.95 (1-β = 0.95) and an alpha level of 0.05 for the F tests. The calculation indicated that a minimum of 23 participants per group was required. In the end, 33 MDD patients and 32 HCs were enrolled, though four MDD patients and three HCs were excluded due to poor electrode contact or excessive blink and muscle artifacts.

Clinical assessments

A semistructured interview, adapted from the Structured Clinical Interview for DSM-5 Disorders, was employed to collect sociodemographic and clinical information from all participants[24]. The HAMD-24 was used to rate the severity of depressive symptoms in MDD patients by a resident psychiatrist and an attending psychiatrist. Additionally, the Annett handedness scale was used to rate the participants’ handedness[25].

Experimental paradigm

Stimulus materials: As shown in Figure 1, fifty-four facial images from the Chinese Affective Picture System were selected[26], representing three emotional expressions (happy, angry, neutral) from 54 distinct actors (27 male, 27 female). Nine images were included for each emotion-sex combination. To minimize practice effects, additional facial stimuli were incorporated into the practice trials preceding the experimental session. The emotional images used in this study were uniformly adjusted to dimensions of 15.5 cm × 11.6 cm and pixel resolution of 260 × 300, with Picture Manager being employed for resizing. To ensure consistency, the images’ background, contrast, and brightness were calibrated to a standard value. The stimuli were then displayed on a 19-inch monitor at a resolution of 1280 × 1024 pixels, with a 60 Hertz (Hz) refresh rate. Auditory stimuli comprised 100 millisecond (ms) continuous tones at two frequencies: 300 Hz (low) and 1000 Hz (high). The digital audio parameters included a 176 kbps bitrate, 8-bit sample depth, and monaural channel configuration.

Figure 1 Schematic illustration of dual task. A: Illustration of the dual task, there are six blocks in total, each containing 54 trials; B: The emotional facial stimuli employed in the dual task. Including pictures of angry, happy, and neutral emotions. ms: Milliseconds; Hz: Hertz; SOA: Stimulus onset asynchrony; ITI: Inter-trial interval. The permission for the photos is shown in the Supplementary material.

Dual task: The study involved two stages: The practice phase and the formal testing phase. Participants were seated approximately 60 cm from the computer screen in a moderately lit, soundproof room. They were instructed to maintain a fixed gaze on the monitor and to minimize eye blinks and body movements throughout the task. A practice session was conducted prior to the formal experiment to ensure that all the participants were adequately familiarized with the procedure. The experiment consisted of six blocks, each comprising 54 trials (324 total trials). Each trial began with the random presentation of a 100-ms tone, either low (300 Hz) or high (1000 Hz). The participants categorized the tone via a keypress (“Z” for low, “X” for high), with responses recorded within a 1500 ms window; responses exceeding this time limit were coded as missed trials. Following the tone offset, one of three stimulus onset asynchronies (SOAs: 500, 1000, or 1500 ms) was presented as a blank interval. A facial stimulus (happy, angry, or neutral) subsequently appeared centered on a white background for 500 ms. The participants then categorized the emotion using keypresses (“V” = happy, “B” = angry, “N” = neutral).

We designated the auditory task as task 1 and the emotional image recognition task as task 2. In both tasks, participants were required to respond as quickly as possible while maintaining accuracy. They were encouraged to take a break after completing each block. In line with typical psychological refractory period paradigms, participants were told to first respond to task 1 and then to task 2, avoiding the grouping of both responses. RTs and accuracy were recorded to assess attentional switching ability. RTs were measured as the time taken to respond to each stimulus, whereas accuracy was defined as the percentage of correct responses. Both RTs and accuracy are commonly used behavioral indicators of attentional switching, which refers to the ability to shift attention flexibly between stimuli in response to changes in the environment. We hypothesized that MDD participants, owing to the cognitive impairments associated with the disorder, would exhibit slower RTs and reduced accuracy.

EEG recording and analysis

During the dual-task experiment, EEG data were continuously recorded using a 64-channel EasyCap system (Brain Products GmbH, Wörthsee, Bavaria, Germany) and a BrainAmp Standard amplifier (Brain Products GmbH, Wörthsee, Bavaria, Germany). EEG signals were continuously sampled at a rate of 500 Hz and amplified with a 0.1-100 Hz bandpass filter for subsequent offline analysis. The reference electrode was placed in the center of the forehead, and the ground electrode was positioned 1-2 cm below the left clavicle. To monitor ocular artifacts, two horizontal electrooculogram electrodes were positioned approximately 1 cm lateral to the outer canthi of both eyes, and one vertical electrooculogram electrode was placed approximately 1 cm below the center of the lower margin of the left eye. Electrode impedance was maintained below 5 kilohms throughout the recording session.

EEG data preprocessing was performed using the EEGLAB 2021 toolbox within the MATLAB 2020b environment (MathWorks, Inc., Natick, MA, United States). The data were bandpass filtered between 0.1 and 30 Hz and re-referenced to the average of the bilateral mastoid electrodes. Artifacts arising from eye movements, muscle activity, or cardiac signals were identified and removed using independent component analysis. Independent component analysis is effective in isolating and removing components related to artifacts while preserving the brain-related signal[27]. Trials containing residual artifacts exceeding ± 100 μV were excluded from further analysis.

During the research phase, two types of experiments were conducted: Those related to stimulus onset asynchrony (SOA: 500, 1000, and 1500 ms) and those involving emotional image stimuli (angry, happy, and neutral). On the basis of the grand average of ERPs and previous studies[28], the ERP components analyzed included the N100 (40-120 ms), P200 (150-250 ms), P300 (250-400 ms), and late positive potential (LPP) (400-600 ms), examined over various duration windows. In this study, the average amplitudes of the ERP components were measured at various electrode sites, and the average amplitudes across three electrode sites were then calculated.

Statistical analysis

Data analysis was conducted using IBM SPSS Statistics version 25 (IBM Corp., Armonk, NY, United States). The independent t-test (two-tailed) and Pearson χ² test were used to compare quantitative data and qualitative data, respectively, between the MDD group and the HC group. A mixed-design analysis of variance (ANOVA) was conducted on both behavioral and ERP data. When appropriate, statistical results were adjusted using the Greenhouse-Geisser correction method. Pairwise comparisons of interactions were corrected using the Bonferroni method. A two-tailed Pearson correlation was calculated for the HAMD-24 scores. Effect sizes were reported as partial eta squared (η²P) and Cohen’s d, with a significance threshold set at 0.05.

RESULTS

Demographics and clinical characteristics

According to Table 1, there were no significant differences between the MDD and HC groups for age, sex distribution, educational background, or handedness. All MDD patients were undergoing antidepressant treatment, with eight prescribed sertraline (185.0 ± 26.7 mg/day), five given escitalopram oxalate (18.4 ± 2.5 mg/day), seven taking mirtazapine (34.6 ± 8.3 mg/day), four on paroxetine (25.0 ± 8.1 mg/day), and five on venlafaxine (225.6 ± 24.1 mg/day). The mean fluoxetine-equivalent dose across patients was 32.5 ± 13.7 mg/day, which was calculated according to a previously established conversion method[29].

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

Variables	MDD (n = 29)	HC (n = 29)	Test statistic
Age (years)	25.0 (3.20)	30.31 (6.40)	P = 0.063
Sex (M/F)	11/18	14/15	χ2 = 0.633, P = 0.426
Education (years)	15.02 (4.50)	15.45 (2.50)	P = 0.063
Age at onset (years)	24.17 (7.83)	-	-
Duration of illness (month)	36.32 (6.51)	-	-
Handedness (R/M/L)	11/9/9	10/10/9	χ2 = 0.809, P = 0.35
HAMD-24	34.24 (4.59)	-	-

MDD: Patients with major depressive disorder; HC: Healthy controls; M: Male; F: Female; R: Right; M: Mixed; L: Left; HAMD: Hamilton Depression Rating Scale.

Analysis of behavioral data

For the behavioral data, accuracy and RTs were analyzed. RT was recorded as the duration from when the image appeared to the pressing of the keyboard keys in response. Trials with response times for auditory tones or facial expressions below 200 ms or above 3000 ms were excluded, resulting in a 12.02% loss of trials in the dual-task condition. Trial loss was greater for MDD patients than for HCs [W = 663.00, P = 0.003, r = 0.39 (M_MDD = 14.4%, M_HC = 9.6%)]. Task 2 analyses were performed using a mixed-design ANOVA with group (MDD vs HC) as a between-groups factor and emotion (angry vs happy vs neutral) and SOA (500 ms vs 1000 ms vs 1500 ms) as within-subjects factors.

As shown in Tables 2 and 3, for accuracy, a main effect of group was observed. MDD patients had lower accuracy than HCs did. A main effect of emotion was observed, whereas a main effect of SOA was not observed. Angry facial pictures led to lower accuracy than both happy facial pictures and neutral facial pictures did (all P < 0.001). There was no significant difference in accuracy between happy facial pictures and neutral facial pictures. For RTs, an emotion main effect was observed. The RTs of happy facial pictures were shorter than those of angry facial pictures and neutral facial pictures (all P < 0.001). There was no significant difference in RTs between angry facial pictures and neutral facial pictures. A main effect of SOA was observed. The RTs of the 500 ms SOA were longer than those of the 1000 ms SOA and 1500 ms SOA (all P < 0.001). Additionally, the RTs of the 1000 ms SOA were longer than those of the 1500 ms SOA (P < 0.001).

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

Dual task	Task 2 expression	SOA	MDD	HC
Task 2 RTs1	Angry	500	1160 (250)	1020 (254)
	Angry	1000	1073 (227)	934 (212)
	Angry	1500	1037 (242)	905 (182)
	Happy	500	1054 (238)	973 (229)
	Happy	1000	938 (196)	889 (196)
	Happy	1500	909 (174)	852 (196)
	Neutral	500	1111 (220)	1018 (179)
	Neutral	1000	993 (158)	948 (177)
	Neutral	1500	962 (144)	936 (194)
Task 2 accuracy (%)2	Angry	500	80.3 (12.2)	88.1 (10.0)
	Angry	1000	81.3 (11.7)	88.1 (11.7)
	Angry	1500	81.6 (12.9)	89.2 (10.6)
	Happy	500	91.2 (7.3)	92.1 (7.7)
	Happy	1000	89.8 (6.9)	90.9 (8.8)
	Happy	1500	89.0 (8.2)	92.8 (10.2)
	Neutral	500	87.3 (8.2)	92.8 (8.0)
	Neutral	1000	88.3 (8.7)	93.3 (6.7)
	Neutral	1500	88.8 (6.4)	93.0 (6.5)

¹Reactive time presented in milliseconds (ms).

²Accuracy presented in percentage correct.

SOA: Stimulus onset asynchrony; MDD: Patients with major depressive disorder; HC: Healthy controls; RTs: Reaction times.

Table 3 Analysis of variances outcome of the behavioral data (dual task).

	Df (num)	Df (den)	F	P value	η2p
Dual task accuracy
Group	1	561	11.851	0.001	0.1751
Emotion	2	551	10.0961	< 0.001	0.2691
SOA	2	55	0.382	0.685	0.014
Group × emotion	2	55	1.664	0.199	0.057
Group × SOA	2	55	0.271	0.764	0.01
Emotion × SOA	4	53	0.89	0.477	0.063
Group × emotion × SOA	4	53	0.868	0.489	0.061
Dual task reaction time
Group	1	56	3.116	0.083	0.053
Emotion	2	551	12.6171	< 0.001	0.3141
SOA	2	551	101.331	< 0.001	0.6441
Group × emotion	2	55	2.601	0.083	0.086
Group × SOA	2	55	1.126	0.332	0.039
Emotion × SOA	4	53	0.464	0.751	0.008
Group × emotion × SOA	4	53	0.878	0.483	0.062

¹Represent that probability values were corrected for departures from compound symmetry using the Greenhouse-Geisser correction.

SOA: Stimulus onset asynchrony.

Analysis of ERP data

Given that ERP activity associated with task-switching is typically more pronounced at midline electrodes, we selected frontal (Fz), central (Cz), and parietal (Pz) sites for our analysis[30]. A mixed-design ANOVA with group (MDD vs HC) as a between-groups factor and electrode (Fz vs Cz vs Pz) and SOA (500 ms vs 1000 ms vs 1500 ms) as within-subjects factors was conducted separately for N100, P200, P300, and LPP. For the ERP characteristics of task 2, a mixed-design ANOVA with group (MDD vs HC) as a between-groups factor and electrode (Fz vs Cz vs Pz) and emotion (angry vs neutral vs happy) as within-subjects factors was conducted separately for N100, P200, P300, and LPP. Post hoc analyses were conducted for all interaction effects. The ERP data and analysis outcomes are shown in Tables 4, 5, and 6, respectively. The ERP waveform and topographic maps for emotion and SOA are shown in Figures 2 and 3, respectively. Figure 4 shows the mean amplitudes of the N100, P200, P300, and LPP evoked by task 2.

Figure 2 Grand averaged event-related potentials in frontal, central, and parietal locations and topographical distribution of grand averaged image-evoked N100, P200, P300 and late positive potential of both major depressive disorder and healthy controls groups. The emotional picture-evoked N100 time window is 40-120 milliseconds (ms); the emotional picture-evoked P200 time window is 150-250 ms, the emotional picture-evoked P300 time window is 250-400 ms, the emotional picture-evoked late positive potential time window is 400-600 ms. MDD: Patients with major depressive disorder; HC: Healthy controls; LPP: Late positive potential.

Figure 3 Grand averaged ERPs in frontal, central, and parietal locations and topographical distribution of grand averaged stimulus onset asynchrony-evoked N100, P200, P300 and late positive potential of both major depressive disorder and healthy controls groups. The stimulus onset asynchrony (SOA)-evoked N100 time window is 40-120 milliseconds (ms); the SOA-evoked P200 time window is 150-250 ms, the SOA-evoked P300 time window is 250-400 ms, the SOA-evoked late positive potential time window is 400-600 ms. SOA: Stimulus onset asynchrony; MDD: Patients with major depressive disorder; HC: Healthy controls; LPP: Late positive potential.

Figure 4 The mean amplitudes of N100, P200, P300 and late positive potential evoked by task 2. Fz: Frontal; Cz: Central; Pz: Parietal; MDD: Patients with major depressive disorder; HC: Healthy controls; LPP: Late positive potential.

Table 4 The mean amplitude and latencies of event-related potential components.

Variables1	MDD (n = 29)						HC (n = 29)
	Fz		Cz		Pz		Fz		Cz		Pz
	Amplitude	Latency	Amplitude	Latency	Amplitude	Latency	Amplitude	Latency	Amplitude	Latency	Amplitude	Latency
N100
500	-4.4 (3.1)	71.6 (14.3)	-2.5 (3.0)	61.4 (14.9)	-0.2 (2.5)	57.2 (22.9)	-5.3 (2.5)	73.7 (13.4)	-3.2 (2.0)	65.2 (14.8)	-0.6 (1.6)	55.2 (14.6)
1000	-4.5 (3.1)	70.6 (12.2)	-2.5 (2.8)	61.5 (13.8)	-0.2 (2.0)	58.9 (22.2)	-5.2 (2.5)	76.1 (16.5)	-3.1 (2.1)	64.7 (15.2)	-0.1 (2.3)	56.4 (20.5)
1500	-4.6 (2.9)	73.1 (14.5)	-2.6 (2.8)	62.9 (15.3)	-0.1 (2.3)	57.4 (19.5)	-5.2 (2.7)	74.3 (14.2)	-3.1 (2.3)	66.3 (15.7)	-0.5 (1.9)	57.7 (21.0)
P200
500	-4.4 (5.9)	178.6 (40.7)	-0.9 (4.9)	181.3 (41.7)	1.9 (4.0)	199.9 (47.2)	-4.0 (4.5)	173.0 (35.7)	-0.5 (3.5)	176.8 (37.0)	1.3 (2.7)	190.3 (45.9)
1000	-4.6 (6.1)	177.7 (41.8)	-1.1 (4.8)	179.0 (41.5)	1.7 (3.7)	204.5 (46.5)	-3.3 (4.6)	171.7 (32.5)	-0.2 (3.7)	178.0 (37.2)	1.6 (3.0)	186.7 (44.3)
1500	-4.5 (5.2)	178.3 (41.3)	-1.0 (4.5)	177.4 (38.9)	1.9 (3.7)	200.2 (46.3)	-3.5 (4.4)	177.7 (38.5)	0.1 (3.7)	186.7 (42.7)	1.7 (3.0)	192.3 (45.5)
P300
500	-7.1 (9.5)	278.8 (40.5)	-3.3 (6.3)	276.1 (42.4)	1.3 (4.7)	291.2 (51.4)	-9.6 (9.3)	274.6 (44.1)	-3.6 (6.2)	293.0 (57.4)	0.9 (4.9)	311.0 (59.0)
1000	-7.0 (9.4)	270.8 (34.2)	-3.3 (6.3)	282.6 (49.7)	1.3 (4.6)	295.5 (51.3)	-9.2 (9.7)	271.5 (42.3)	-3.5 (6.3)	293.0 (57.6)	1.0 (5.0)	302.6 (60.5)
1500	-7.3 (9.1)	277.2 (41.5)	-3.5 (6.2)	275.0 (43.5)	1.2 (4.6)	285.2 (47.6)	-9.6 (10.1)	265.4 (28.8)	-3.3 (6.5)	284.1 (54.1)	1.0 (5.0)	304.3 (54.7)
LPP
500	-9.3 (10.6)	435.8 (64.3)	-5.2 (6.6)	410.3 (52.9)	-0.3 (4.7)	405.1 (46.8)	-13.7 (11.5)	407.3 (56.5)	-5.8 (7.6)	419.8 (60.8)	-0.1 (5.8)	411.1 (53.3)
1000	-9.6 (10.8)	425.3 (60.6)	-5.3 (7.0)	412.8 (54.2)	-0.2 (4.9)	413.5 (53.2)	-14.0 (12.7)	413.6 (63.0)	-6.0 (8.0)	422.6 (57.3)	-0.3 (6.2)	423.4 (59.6)
1500	-9.5 (10.1)	423.0 (63.3)	-5.3 (6.8)	410.3 (55.2)	-0.2 (4.8)	409.2 (50.6)	-14.2 (11.8)	407.7 (58.1)	-5.7 (7.5)	425.8 (62.5)	-0.1 (5.4)	424.2 (61.8)
N100
Angry	-2.8 (2.6)	70.1 (16.3)	-1.7 (2.2)	61.9 (15.2)	0.1 (1.7)	59.0 (22.1)	-3.1 (2.2)	71.9 (14.8)	-2.0 (1.6)	63.2 (16.3)	-0.5 (1.3)	58.1 (20.1)
Happy	0.7 (2.7)	75.9 (24.0)	-0.5 (1.9)	66.6 (20.3)	0.3 (1.5)	65.4 (28.1)	-0.9 (2.5)	71.7 (21.6)	-0.9 (1.6)	66.3 (22.9)	-0.3 (1.2)	69.8 (30.7)
Neutral	-1.0 (2.4)	74.1 (23.7)	-1.0 (1.6)	72.8 (25.8)	-0.1 (1.2)	65.0 (28.1)	-1.0 (2.4)	68.7 (20.2)	-1.0 (1.4)	66.9 (23.2)	-0.6 (1.2)	65.9 (29.0)
P200
Angry	-1.9 (3.4)	179.1 (40.8)	0.2 (3.2)	185.0 (40.8)	2.0 (2.7)	214.0 (42.7)	-1.1 (2.6)	177.3 (34.6)	0.8 (2.3)	188.9 (40.3)	1.6 (1.9)	204.3 (42.4)
Happy	2.1 (3.2)	192.8 (27.1)	2.5 (2.9)	200.5 (32.4)	3.6 (2.4)	224.3 (30.0)	2.5 (3.6)	192.2 (23.2)	2.8 (2.7)	196.6 (23.0)	2.7 (2.0)	225.3 (27.3)
Neutral	1.8 (3.1)	194.7 (28.0)	2.0 (2.9)	196.9 (29.7)	3.2 (2.3)	222.0 (29.5)	2.3 (3.7)	194.6 (24.7)	2.4 (2.6)	201.3 (26.1)	2.3 (2.0)	221.7 (28.3)
P300
Angry	-3.7 (5.6)	274.1 (39.7)	-1.5 (4.1)	279.9 (47.2)	1.9 (3.1)	289.3 (49.3)	-5.2 (5.8)	276.6 (44.6)	-1.3 (4.0)	294.8 (58.5)	1.6 (3.4)	308.5 (58.6)
Happy	0.7 (4.1)	290.1 (57.9)	1.3 (3.9)	295.9 (62.3)	4.0 (2.7)	319.5 (61.6)	0.2 (3.7)	289.4 (50.6)	1.7 (3.2)	306.4 (58.3)	3.8 (3.5)	317.6 (51.2)
Neutral	0.4 (3.6)	292.5 (56.9)	1.0 (3.6)	301.2 (60.9)	3.9 (2.5)	324.9 (56.6)	-0.2 (3.7)	297.2 (59.9)	1.5 (3.1)	311.0 (63.1)	3.6 (3.0)	337.7 (51.8)
LPP
Angry	-5.3 (6.8)	423.0 (63.6)	-2.9 (4.6)	410.7 (56.2)	0.7 (3.3)	411.1 (57.3)	-8.6 (7.7)	404.4 (57.0)	-3.1 (5.2)	413.9 (59.8)	0.7 (4.0)	405.6 (54.6)
Happy	-0.1 (5.0)	440.7 (68.1)	0.7 (4.4)	445.3 (67.0)	3.1 (2.9)	416.9 (59.8)	-2.2 (4.5)	410.9 (60.9)	0.5 (3.4)	426.6 (71.0)	2.8 (3.2)	416.2 (66.5)
Neutral	0.0 (4.1)	450.3 (66.4)	0.7 (4.1)	441.8 (64.4)	3.3 (2.9)	424.7 (61.2)	-2.2 (4.3)	401.1 (49.8)	0.7 (3.7)	424.9 (66.6)	3.0 (3.4)	405.3 (52.0)

¹All values presented in microvolts (μV).

MDD: Patients with major depressive disorder; HC: Healthy controls; Fz: Frontal; Cz: Central; Pz: Parietal.

Table 5 Analysis of variances outcome of event-related potential mean amplitudes.

	Df (num)	Df (den)	F	P value	η2p
Dual task: Task 2 N100
Group	1	56	0.637	0.428	0.011
Emotion	2	112	28.5031	< 0.001	0.3371
Electrode	2	112	39.2671	< 0.001	0.4121
Group × emotion	2	112	0.234	0.791	0.004
Group × electrode	2	112	0.909	0.406	0.016
Emotion × electrode	4	224	110.7951	< 0.001	0.6641
Group × emotion × electrode	4	224	0.688	0.601	0.012
Dual task: Task 2 P200
Group	1	56	0.024	0.877	0.000
Emotion	2	112	37.5591	< 0.001	0.4011
Electrode	2	112	29.2621	< 0.001	0.3431
Group × emotion	2	112	0.287	0.751	0.005
Group × electrode	2	112	5.9631	0.010	0.0961
Emotion × electrode	4	224	42.4771	< 0.001	0.4311
Group × emotion × electrode	4	224	0.092	0.888	0.002
Dual task: Task 2 P300
Group	1	56	0.104	0.748	0.002
Emotion	2	112	36.7951	< 0.001	0.3961
Electrode	2	112	125.8531	< 0.001	0.6921
Group × emotion	2	112	0.175	0.725	0.003
Group × electrode	2	112	2.615	0.094	0.045
Emotion × electrode	4	224	21.6391	< 0.001	0.2791
Group × emotion × electrode	4	224	0.703	0.467	0.012
Dual task: Task 2 LPP
Group	1	56	1.030	0.315	0.018
Emotion	2	112	55.4121	< 0.001	0.4971
Electrode	2	112	34.6301	< 0.001	0.7001
Group × emotion	2	112	0.098	0.830	0.002
Group × electrode	2	112	8.172	0.002	0.127
Emotion × electrode	4	224	32.1731	< 0.001	0.3651
Group × emotion × electrode	4	224	1.514	0.225	0.026

¹Represent that probability values were corrected for departures from compound symmetry using the Greenhouse-Geisser correction.

LPP: Late positive potential.

Table 6 Group differences between major depressive disorder and healthy controls in late positive potential amplitudes across electrodes.

Electrode	Group	Mean	SE	P value	η2p
Fz	HC	-4.35	0.88	0.047	0.068
	MDD	-1.83
Cz	HC	-0.66	0.72	0.857	0.001
	MDD	-0.47
Pz	HC	2.15	0.57	0.777	0.001
	MDD	2.38

SE: Standard error; Fz: Frontal; Cz: Central; Pz: Parietal; HC: Healthy controls; MDD: Patients with major depressive disorder.

N100: An emotion × electrode interaction was observed. Happy facial pictures and neutral facial pictures evoked higher amplitudes than angry facial pictures did at Fz and Cz (all P < 0.001). A main effect of emotion was observed. Happy facial pictures and neutral facial pictures evoked higher amplitudes than angry facial pictures did (all P < 0.001). A main effect of the electrode was observed. N100 amplitudes at Pz were greater than those at Fz and Cz (all P < 0.001); N100 amplitudes at Cz were greater than those at Fz (P < 0.05). There was no significant difference in latency between the two groups.

P200: A group × electrode interaction was observed. The mean amplitudes (Fz, Cz, and Pz) of the HCs were greater than those of the MDD patients (P < 0.05). An emotion × electrode interaction was observed. Happy facial pictures and neutral facial pictures evoked higher amplitudes than angry facial pictures did at Fz, Cz, and Pz (all P < 0.001). A main effect of emotion was observed. Happy facial pictures and neutral facial pictures evoked higher amplitudes than angry facial pictures did (all P < 0.001). A main effect of the electrode was observed. P200 amplitudes at Pz were greater than those at Fz and Cz (all P < 0.001); P200 amplitudes at Cz were greater than those at Fz (P < 0.001). There were no significant differences in latency between the two groups.

P300: An emotion × electrode interaction was observed. Happy facial pictures and neutral facial pictures evoked higher amplitudes than angry facial pictures did at Fz, Cz, and Pz (all P < 0.001). A main effect of emotion was observed. Happy facial pictures and neutral facial pictures evoked higher amplitudes than angry facial pictures did (all P < 0.001). A main effect of the electrode was observed. P300 amplitudes at Pz were greater than those at Fz and Cz (all P < 0.001); P300 amplitudes at Cz were greater than those at Fz (P < 0.001). There was no significant difference in latency between the two groups.

LPP: A group × electrode interaction was observed. MDD patients had lower LPP amplitudes than HCs did at Fz (P < 0.05). An emotion × electrode interaction was observed. Happy facial pictures and neutral facial pictures evoked higher amplitudes than angry facial pictures did at Fz, Cz, and Pz (all P < 0.001). A main effect of emotion was observed. Happy facial pictures and neutral facial pictures evoked higher amplitudes than angry facial pictures did (all P < 0.001). A main effect of the electrode was observed. LPP amplitudes at Pz were greater than those at Fz and Cz (all P < 0.001). LPP amplitudes at Cz were greater than those at Fz (P < 0.001). There were no significant differences in latency between the two groups.

Correlations

As shown in Figure 5, there was a moderately robust negative correlation between HAMD-24 scores and the RTs of task 2 and a significant negative correlation between LPP latencies and the accuracy of task 2 for angry face stimuli. However, the HAMD-24 scores were not correlated with the other ERP components.

Figure 5 The correlation analysis of late positive potential mean amplitudes and latencies, reaction times, accuracy and Hamilton Depression Rating Scale (24-item edition) scores under different emotional faces in the major depressive disorder group. ^aP < 0.05; ^bP < 0.01. HAMD-24: Hamilton Depression Rating Scale (24-item edition); A: Amplitude; L: Latency; ACC: Accuracy; RT: Reaction time.

DISCUSSION

This study investigated the ERP characteristics of the attentional switching function by measuring ERPs during the dual-task paradigm and further explored the neuroelectrophysiological mechanism of the cognitive processing underlying the attentional switching dysfunction in MDD. This study revealed greater accuracy in HCs and slower RTs in MDD patients. Angry facial pictures led to lower accuracy. The results also revealed shorter RTs for happy facial pictures and the longest RTs for the 500 ms SOA. With respect to ERP characteristics, happy facial pictures and neutral facial pictures evoked higher amplitudes. The N100, P200, P300, and LPP amplitudes at Pz were the highest. MDD patients had lower P200 mean amplitudes and LPP amplitudes than HCs did.

The dual-task paradigm is a widely utilized experimental framework in cognitive psychology and neuroscience that is designed to investigate the allocation and limitations of attentional resources[31,32]. This paradigm has been instrumental in elucidating the mechanisms of divided attention, cognitive load, and interference across various populations, including aging individuals and those with psychiatric or neurological disorders[33-35]. Closely related is the concept of the attentional switching function, which refers to the cognitive ability to flexibly reallocate attention between different tasks, goals, or mental sets. Deficits in attentional switching have been associated with a range of neurocognitive conditions, including depression[36], schizophrenia[37,38], and frontal lobe dysfunction[39,40], highlighting its critical role in cognitive flexibility and control.

The behavioral outcomes of this study indicated that MDD patients have impaired cognitive resource allocation and limited attentional control capacity, which is similar to the findings of previous studies, i.e., excessive internal focus or ruminative processing is occurring in MDD patients that disrupts the integration of task-relevant information[17-19]. In this study, happy facial pictures and neutral facial pictures evoked higher amplitudes, suggesting that nonnegative emotional stimulation might lead to early cognitive processing. MDD patients had lower P200 mean amplitudes, suggesting that MDD patients might allocate fewer attentional resources during early emotional processing, especially when encountering negative emotional stimuli. However, MDD patients had lower LPP amplitudes, indicating deficits in later-stage attention allocation and stimulus evaluation processes. We hypothesize that the observed ERP deficits may not only reflect impairments in attentional switching but also difficulties in regulating emotional responses during the task. In this context, MDD patients may struggle to modulate emotional reactions to stimuli, which could further exacerbate difficulties in shifting attention. For example, deficits in emotion regulation could result in heightened emotional reactivity to stimuli (such as angry faces), which might interfere with the ability to effectively switch attention and process information. ERP findings from MDD patients in the dual-task paradigm showed a distinct pattern of disrupted neural processing across task phases, indicating attentional switching abnormalities.

In summary, this study revealed greater accuracy in HCs and slower RTs in MDD patients, and for all the participants, angry facial pictures led to lower accuracy. The higher behavioral exclusion rate in MDD may be associated with depressive symptoms, such as cognitive slowing and inattention. From a behavioral perspective, these findings indicate that patients with MDD have impaired attention shift function. MDD patients, while not necessarily demonstrating a greater attentional bias toward angry faces, may still exhibit difficulty in regulating attention when emotional stimuli are present. Most importantly, MDD patients exhibited abnormal ERP characteristics evoked by the dual-task paradigm, which may be the neural correlates of abnormalities in attentional switching in patients with MDD. In this study, we observed significant differences in ERP components (P200 and LPP) between MDD patients and HCs, particularly in response to emotional stimuli such as angry faces. These ERP markers reflect impairments in attentional switching, offering important insights into the cognitive deficits present in MDD.

Our study revealed that while HAMD-24 scores were correlated with RTs and LPP latencies in response to angry faces, there was no correlation with P200 or LPP amplitude differences. It suggests that these ERP markers might reflect stable, trait-like deficits in MDD rather than being dependent on the severity of depressive symptoms. We suggest that these ERP markers could serve as potential biomarkers for attentional dysfunction in MDD, providing critical clues for intervention. Given that attentional switching dysfunction in MDD may be influenced by difficulties in regulating emotional responses, combining emotion regulation training with attention control techniques may offer a more comprehensive approach to treatment. Our results suggest that the reduced accuracy when angry faces are being processed is more likely related to cognitive flexibility and task-switching demands rather than being solely attributable to the traditional notion of negative attentional bias in MDD. Future research could explore how emotion regulation and cognitive flexibility interact with attention to negative stimuli in MDD patients under different experimental paradigms, which may yield a more nuanced understanding of attentional dysfunction in this disorder[15,41,42].

Our research has two limitations. First, as this study has a relatively small sample size, our research results are preliminary. A small sample size increases the risk of Type II errors, meaning that subtle or complex group differences may be missed. Future studies with larger sample sizes using consistent parameters are required to replicate these findings. Second, ERP equipment has low spatial resolution, and fMRI provides excellent spatial resolution, enabling researchers to identify the brain regions activated during emotional processing and attentional tasks. Additionally, magnetoencephalography also strikes an effective balance between temporal and spatial resolution. Future studies should utilize a combination of ERP and high-spatial-resolution fMRI or magnetoencephalography to further elucidate the neuroelectrophysiological mechanism of the deficits in cognitive processing underlying the attentional switching dysfunction, as well as impairment in cognitive flexibility in patients with MDD.

CONCLUSION

In conclusion, MDD patients exhibit attentional switching dysfunction and abnormal ERP characteristics evoked by the dual-task paradigm, which may reflect the neural correlates of the abnormalities in attentional switching in patients with MDD. The correlations between ERP components and clinical symptoms suggest ERP markers might reflect stable, trait-like deficits in MDD. These results provide valuable insights into the understanding of the neural mechanisms underlying attentional switching function and may guide targeted interventions in patients with MDD.

ACKNOWLEDGEMENTS

We thank the subjects and their families who participated in this study, and we would like to acknowledge everyone who helped us in this project.