Neural correlates of aggressive impulsive behavior differ between schizophrenia and bipolar disorder: An event-related potential study with a go/no-go task

Abstract

BACKGROUND

Although schizophrenia (SCZ) and bipolar disorder (BD) may exhibit some superficial behavioral similarities in terms of aggression and impulsivity, the underlying mechanisms distinguishing these two conditions remain poorly understood. Further elucidation of event-related potential (ERP) characteristics associated with aggressive and impulsive behaviors in SCZ and BD patients may provide reliable evidence for the different neural mechanisms of aggressive and impulsive behaviors in these two disorders. Our hypothesis is that the neural correlates of aggressive and impulsive behaviors in SCZ patients may differ from those in BD patients.

AIM

To investigate the differences in the ERP characteristics of SCZ and BD patients with aggressive behavior and explore the neural mechanism of cognitive processing.

METHODS

This cross-sectional, controlled study used questionnaires to assess aggression and ERPs to examine neural activation when 47 SCZ patients, 44 BD patients and 56 healthy controls (HCs) performed the emotional go/no-go task. A 3 × 2 repeated measures analysis of variance was used to compare the response time and accuracy of the three groups under the two neutral and angry emotional stimuli. The cluster-based permutation test in FieldTrip was used to analyze the differences in the average amplitudes of the P100, N170, N200, P300 and late positive potential components under the two emotional conditions.

RESULTS

Greater motor impulsivity and selfdirected aggression were observed in patients with BD. ERP data revealed that under the go condition, the P100 evoked by angry faces was greater than that evoked by neutral faces in HCs, and the N170 evoked by angry faces was greater than that evoked by neutral faces in BD patients; under the no-go condition, the N200, and P300 evoked by angry faces or neutral faces in SCZ patients were greater than those evoked by BD patients, and the late positive potential evoked by angry faces was greater than that evoked by neutral faces in HCs.

CONCLUSION

This study utilized the emotional go/no-go task to support the possibility that patients with schizophrenia and BD who exhibit aggressive impulsive behaviors have differences in their behavioral performance and neural mechanisms.

Key Words: Schizophrenia; Bipolar disorder; Event-related potential; The emotional go/no-go task; Aggressive impulsive behavior

Core Tip: This study utilized the emotional go/no-go task to examine behavioral performance, event-related potential characteristics, and aggressive impulse scale data, aiming to reveal the possible differences in the neural mechanisms of aggressive behavior in patients with schizophrenia (SCZ) and bipolar disorder (BD), and to further deepen the research on the cognitive function of patients with SCZ and BD, to provide new therapeutic targets and targeted interventions for aggressive behavior in patients with SCZ and BD.

INTRODUCTION

Schizophrenia (SCZ) and bipolar disorder (BD) are significant psychiatric disorders that affect a substantial portion of individuals worldwide[1-3]. Although the symptoms of the two disorders vary greatly, aggressive impulsive behavior is a clinical feature of SCZ and BD[4,5].

Aggressive impulsive behavior is associated with inhibitory control dysfunction and is moderated by contextual or environmental factors[6-8]. The go/no-go task is a widely used paradigm for probing mechanisms of inhibitory control function. Additionally, inhibitory control function is closely related to emotion regulation ability. It has been suggested that negative emotions overactivate the limbic system and inhibit the inhibitory control function of the prefrontal lobe, resulting in increased negative emotions, and decreased inhibition of action choice and response[9].

Studies using functional magnetic resonance imaging (fMRI) have shown that aggressive impulsive behavior leads to increased activation levels in the prefrontal lobe, whereas individuals with dyscontrol experience inadequate regulation of the prefrontal region, which in turn affects the limbic system and insula involved in emotions, abnormal mood swings, and rapid reactions and behavioral inhibitory functions[10,11]. Further research on the interaction between negative emotion processing and response inhibition revealed that the anterior insula, dorsal anterior cingulate cortex (dACC), and dorsolateral prefrontal cortex (DLPFC) are all subregions of response inhibition and emotion regulation, and the connectivity between the anterior insula and amygdala, the DLPFC and lateral orbitofrontal cortex, and the anterior insula and dACC is altered when processing emotion-related go/no-go tasks in patients with SCZ[12]. Because fMRI has a low temporal resolution, the findings might not be entirely reliable.

Event-related potentials (ERPs) provide a direct measure of neural activity associated with the processing of these events, providing a unique temporal resolution in milliseconds[13-15]. An increasing number of studies have shown that compared with healthy controls (HCs), patients with SCZ and BD, have N200, and P300 abnormalities, which are evoked by related experimental paradigms, and reflect deficits in dyscontrol, which can be used as biomarkers of inhibitory control function[16,17].

Clinically, there are differences in aggressive impulsive behavior between SCZ patients and BD patients. In SCZ patients, aggression is often reactive, and rooted in psychotic symptoms such as persecutory delusions and command hallucinations[18-20]. This behavior tends to stem from impulse control and deficits in judgment, insight, and social cognition[21,22]. However, in BD, aggression or impulsivity is more goal-directed, and linked to elevated mood, irritability, and heightened energy[4,23]. Behavioral dyscontrol often aligns with impulsive activities such as reckless spending during mood episodes[24,25]. Self-reported impulsivity [via the Barratt Impulsiveness Scale-11 (BIS-11)] is greater in BD patients than in SCZ patients[26]. SCZ patients, in contrast, exhibited greater risk aversion than BD patients did on performance-based tasks such as the Balloon Analogue Risk Task[27]. Additionally, in SCZ patients, comorbid substance use disorder or antisocial personality disorder multiplies the risk of violence-by up to eightfold compared with a twofold increase in patients without comorbidities[28,29]. In BD, impulsivity often persists between episodes, framing it as a trait vulnerability contributing to substance abuse and risky conduct[30,31].

To date, although SCZ and BD may exhibit some superficial behavioral similarities in terms of aggression and impulsivity, the underlying mechanisms distinguishing these two conditions remain poorly understood. Further elucidation of ERP characteristics associated with dyscontrol in SCZ and BD patients may enhance our understanding of the neurophysiological processes (i.e., neural correlates) underlying aggressive and impulsive behaviors in these disorders, thereby providing reliable neuroelectrophysiological evidence for the different neural mechanisms of aggressive and impulsive behaviors in the two disorders.

In this study, participants included SCZ and BD patients, as well as HCs. ERPs were recorded during the emotional go/no-go task to examine the neural correlates of inhibitory control. This study aimed to investigate the ERP characteristics of aggressive and impulsive behaviors in SCZ patients and BD patients, clarify the differences in the neural correlations of inhibitory control dysfunction between SCZ patients and BD patients, and provide reliable evidence for further exploration of the different neural mechanisms of cognitive processing deficits in inhibitory control in these two disorders.

MATERIALS AND METHODS

Time and setting

This study was conducted from February 1, 2025 to July 31, 2025, at the Department of Psychology, Affiliated Mental Health Center of Jiangnan University. It received ethical approval from the Research Ethics Committee (approval No. WXMHCIRB2025 LLky014) of the Wuxi Mental Health Center in China and complies with the Helsinki Declaration. The trial was registered on the Chinese Clinical Trial Registry (chictr.org.cn) on June 19, 2025 (ChiCTR-ICR-2500104625) and followed the CONSORT guidelines. Before the start of the experiment, all subjects signed informed consent after the experimental procedures were fully explained. In accordance with hospital ethics management requirements, researchers also contacted the families of inpatients with SCZ and BD patients and explained the study content in detail to obtain informed consent from the families.

Participants

Participants, including SCZ patients, BD patients and HCs, and SCZ patients and BD patients were selected from the Department of Psychiatry, the Affiliated Mental Health Center of Jiangnan University, China. HCs were recruited from local residential community advertisements in Wuxi city, Jiangsu Province, China.

The inclusion criteria for the SCZ and BD groups were: (1) Met the criteria of SCZ and BD (currently a manic episode) according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth edition (DSM-5), respectively; (2) Were aged 18 years to 65 years; (3) Had Modified Overt Aggression Scale (MOAS) scores > 4[32]; (4) Had the ability to complete the entire experimental procedure with the care of medical staff; and (5) Had no current or history of neurological illness or any other kind of severe physical illness that would affect his or her cognitive function. The exclusion criteria were: (1) Met the criteria for any other mental disorder according to the DSM-5; (2) Treated with electroconvulsive therapy or modified electroconvulsive therapy within 6 months before recruitment; and (3) Had nicotine/other substance misuse or dependence within the past 6 months. The inclusion criteria for the HCs were: (1) No diagnosis of psychiatric disorders according to the DSM-5 criteria; (2) No family history of any mental disorders; and (3) Age ranging from 18 years to 65 years.

G-power (version 3.1.9.7; Heinrich-Heine-Universität Düsseldorf, Düsseldorf, Germany) software was used to calculate the sample size, targeting a statistical power of 0.95 (1-β = 0.95) with an alpha of 0.05 for F tests, resulting in a minimum requirement of 22 participants per group. Considering potential data loss in clinical sample recruitment [such as excessive electroencephalograph (EEG) artifacts, device failures, and poor participant cooperation], we adopted a more conservative recruitment strategy, planning to recruit 50 subjects per group. A total of 153 subjects, including 49 SCZ patients, 46 BD patients, and 58 HCs, were included in this study; however, owing to poor electrode contact, frequent blinking, muscle artifacts, or device recording flaws, 2 SCZ patients, 2 BD patients, and 2 HCs were excluded from the subsequent analysis, and 147 subjects were ultimately analyzed.

Clinical assessment

All clinical data were assessed by a psychiatrist and a psychiatric resident. For SCZ patients, the Positive and Negative Symptom Scale (PANSS) was used to assess psychopathology[33]. For BD patients, the Young Mania Rating Scale (YMRS) was used to assess manic symptoms and severity[34].

Aggression was assessed in a multifaceted manner using different questionnaires. The Reactive-Proactive Aggression Questionnaire (RPAQ) was used to assess the motivation and causes of the aggression which included both the reactive and proactive aggression subscales[35,36]. The BIS-11 was used to assess psychological characteristics on the basis of three dimensions: Attentional impulsiveness, motor impulsiveness and no plan impulsiveness[37]. The MOAS was used to assess the types of aggression, including verbal aggression, aggression against property, autoaggression, and physical aggression[32]. The Buss-Perry Aggression Questionnaire (BPAQ) was used to assess impulsive behavior and personality traits, including physical aggression, verbal aggression, anger, hostility, and self-directed aggression[38].

The emotional go/no-go task

The emotional go/no-go task with emotional facial cues included angry and neutral pictures from the Chinese Facial Affective Picture System[39]. E-Prime 3.0 software (Psychology Software Tools Incorporated, Pittsburgh, PA, United States) was used to run the task and program on computer. As shown in Figure 1, angry faces and neutral faces were used as target stimuli. The emotional go/no-go task include go stimulus (angry face) vs no-go stimulus (neutral face) to investigate neural inhibition, and a go stimulus (neutral face) vs a no-go stimulus (angry face) to investigate emotional inhibition. First, a 150-milliseconds fixation “+” was presented, followed by 1-5 go stimuli which included a 500-millisecond go stimulus and a 1300-1500 milliseconds stimulation interval; finally, the no-go stimulus presentation consisted of three parts: 300-500 milliseconds no-go stimulus, which was similar to the go stimulus faces; a 300-500 milliseconds no-go stimulus, which was dissimilar to go stimulus faces; and a 2000-millisecond blank interval. The subjects were asked to respond to the target stimulus as quickly as possible (go) and stop responding to the nontarget stimulus (no-go) without two consecutive no-go stimuli. All the stimuli were presented in pseudorandom order. The presented ratio of the go stimulation to the no-go stimulation was 7:3 (350 trials for target stimulation and 120 trials for nontarget stimulation), and a total of 940 trials were performed in the emotional go/no-go task. There are 5 practical phases of the trial before the formal experiment.

Figure 1 Flowchart of emotional go/no-go task.

Electroencephalogram recording and preprocessing

Continuous EEG recordings (bandpass filtered between 0.05 Hz and 100 Hz, digitized at 500 Hz) were obtained using a 64-channel Ag-AgCl elastic cap adhering to the International 10/20 System, via the BioSemi Active Two system (https://www.biosemi.com/Products_ActiveTwo.htm). Vertical electrooculograms were recorded with electrodes placed above and below the left eye, and horizontal electrooculograms were recorded with electrodes placed in the outer canthi of both eyes. The reference electrodes were the left and right mastoids, and the ground electrodes were placed under the left clavicle site. All interelectrode impedances were maintained below 5 kΩ.

In this study, the ERP waveforms were time-locked to the onset of each facial picture. The epoch for ERP, including a-500 millisecond baseline, was 1500 milliseconds. After recording, the data were processed offline using MATLAB version 2020b (MathWorks, Inc., Natick, MA, United States) with the EEGLAB toolbox. EEG data filtering including rejection and correction of atypical artifacts from eye movement, muscle, or cardiac components was performed using independent component analysis[40,41].

To examine how the brain maintains stable neural responses, we employed a spatiotemporal clustering technique for the EEG/ERP data. In this process, the Microstate EEGlab toolbox was used, as it offers both traditional microstate analysis and k means clustering adapted for ERP data[42]. For subsequent statistical comparisons, we used FieldTrip - a versatile MATLAB toolbox equipped with a variety of EEG analysis modules, including preprocessing, ERP computation, spectral estimation (both parametric and nonparametric), forward/inverse source modeling, functional connectivity evaluation, real-time processing capabilities, and statistical inference. More specifically, we conducted group-level comparisons of mean ERP amplitudes under different experimental conditions using the cluster-based permutation test available in FieldTrip. On the basis of the total average ERP and previous studies, the following ERP components were analyzed: The P100 amplitude was taken as the average amplitude of the O1, O2, and OZ electrode locations between 90-150 milliseconds[43]; the N170 amplitude was taken as the average amplitude at the P6 and P8 electrode locations between 160-200 milliseconds[43,44]; the N200 amplitude is taken as the average amplitude of the AF3, AF4, and AFZ electrode locations between 200-350 milliseconds[43]; the P300 amplitude time window was the same as that of N200, using the average amplitudes of the PO3, PO4, and POZ electrode locations[43]; the late positive potential (LPP) amplitude was taken as the average amplitude of the PO3, PO4, POZ, O1, and O2, and OZ electrode locations between 300-600 milliseconds[45,46].

Statistical analysis

Data were analyzed using the Statistical Program for Social Sciences software version 22.0 (SPSS, IBM Corp, Armonk, NY, United States). Normality was assessed using the Shapiro-Wilk test, and the criteria recommended by Finney and DiStefano[47] for approximate normality were applied to evaluate distribution characteristics (skewness < |2|, and kurtosis < |7| were considered approximately normal). The results revealed that the vast majority of the variables followed a normal or nearly normal distribution, with only active aggression [RPAQ-proactive aggression (RPAQ-PA)] significantly deviating from the normal distribution (skewness = 3.710, kurtosis = 14.620).

One-way analysis of variance of demographic and clinical characteristics was used to compare the differences between the SCZ, BD and HC groups, followed by post hoc Bonferroni or Tamhane’s T2 test. For RPAQ-PA data that deviated from the normal distribution, the descriptive statistic was changed to the median (quartile interval), and the analysis of variance for between-group comparisons was changed to the Kruskal-Wallis H test. The reaction times (RTs) and accuracy rates (ACCs) of the stimuli were compared using mixed-model repeated measures analysis of variance with a 3-group (SCZ vs BD vs HC) × face type (angry vs neutral) design. The degrees of freedom were corrected using the Greenhouse-Geisser method if the assumption of sphericity was violated. Effect sizes were further estimated using partial eta squared (η_p²). Post hoc analyses were performed when there was a significant interaction, and Bonferroni correction was employed to manage the possible type I errors caused by multiple comparisons. Alpha values of 0.05 were regarded as significant.

The cluster-based permutation test in FieldTrip was used to compare the differences in the mean amplitudes between the three groups on the ERP components under the emotional go/no-go task. We used the clustering parameter maxsum, the number of iterations was 3000, and the statistical significance threshold was set at P < 0.05. All resulting P values were further adjusted for multiple using false discovery rate correction.

Pearson correlation analysis was used to investigate potential correlations between aggression scores, PANSS scores, YMRS scores and ERP components. The correlations between the RPAQ-PA and the other variables were analyzed using Spearman’s rank correlation.

RESULTS

Demographic characteristics and clinical data of the participants

The demographic characteristics and clinical data of all the participants are shown in Table 1. There were no significant differences in age, sex, body mass index, or years of education between the SCZ group, BD group, or HC group. However, there were differences in disease duration and total PANSS, PANSS positive subscale, PANSS negative subscale, PANSS general psychopathology subscale and YMRS scores between the BD group and SCZ group.

Table 1 Demographics characteristics and clinical assessment (mean ± SD).

Variables	HC (n = 56)	BD (n = 44)	SCZ (n = 47)	Statistics	Bonferroni/T2
Age (years)	32.20 ± 7.41	31.91 ± 7.68	31.55 ± 8.57	F = 0.086, P = 0.918
Sex (M/F)	32/24	23/21	29/18	χ2 = 0.825, P = 0.662
BMI (kg/m2)	23.87 ± 4.70	25.49 ± 5.06	25.23 ± 5.09	F = 1.609, P = 0.204
Education (years)	14.14 ± 2.45	13.23 ± 2.77	12.96 ± 3.18	F = 2.717, P = 0.071
Duration of illness (months)	-	120.6 ± 92.5	74.2 ± 90.6	t = 2.417, P = 0.018	BD > SCZ
PANSS	-	55.11 ± 16.34	71.00 ± 12.71	t = -5.194, P = 0.000	BD < SCZ
PANSS-P	-	12.80 ± 4.58	15.57 ± 5.75	t = -2.539, P = 0.013	BD < SCZ
PANSS-N	-	12.39 ± 6.55	18.68 ± 5.79	t = -4.863, P = 0.000	BD < SCZ
PANSS-G	-	29.93 ± 10.13	36.74 ± 7.03	t = -3.747, P = 0.000	BD < SCZ
YMRS	-	8.48 ± 3.00	6.70 ± 2.98	t =2.828, P = 0.006	BD > SCZ
BIS-11	70.46 ± 16.50	89.82 ± 14.60	82.87 ± 23.59	F = 19.406, P = 0.000	BD = SCZ > HC
BIS-11--MI	18.93 ± 5.20	29.43 ± 6.57	22.68 ± 8.84	F = 37.388, P = 0.000	BD > SCZ > HC
BIS-11-AI	26.25 ± 8.15	29.34 ± 7.50	29.66 ± 9.62	F = 2.577, P = 0.080
BIS-11-NPI	25.29 ± 8.39	31.05 ± 7.56	30.53 ± 10.02	F = 6.920, P = 0.001	BD = SCZ > HC
RPAQ	33.29 ± 12.17	44.14 ± 13.33	41.04 ± 10.86	F = 10.818, P = 0.000	BD = SCZ > HC
RPAQ-RA	21.41 ± 9.16	32.36 ± 11.25	29.45 ± 10.02	F = 16.132, P = 0.000	BD = SCZ > HC
RPAQ-PA, median (interquartile range)	10 (10, 11)	10 (10, 13)	10 (10, 12)	F = 0.647, P = 0.724
BPAQ	52.29 ± 17.82	68.11 ± 21.81	63.02 ± 19.35	F = 8.656, P = 0.000	BD = SCZ > HC
BPAQ-PA	12.09 ± 4.85	13.70 ± 6.71	12.68 ± 6.52	F = 0.899, P = 0.409
BPAQ-VA	9.30 ± 3.62	11.91 ± 4.75	11.13 ± 3.94	F = 5.443, P = 0.005	BD = SCZ > HC
BPAQ-A	11.00 ± 4.39	15.82 ± 5.62	13.72±5.51	F = 11.024, P = 0.000	BD = SCZ > HC
BPAQ-H	11.98 ± 4.17	16.30 ± 6.35	17.40 ± 7.34	F = 14.267, P = 0.000	BD = SCZ > HC
BPAQ-SDA	7.91 ± 2.66	10.39 ± 5.37	8.09 ± 3.71	F = 3.949, P = 0.023	BD > SCZ = HC
MOAS	0.18 ± 0.61	15 ± 8.25	11.87 ± 7.89	F = 120.290, P = 0.000	BD = SCZ > HC
MOAS-VA	0.11 ± 0.31	2.02 ± 0.85	2.02 ± 0.92	F = 175.977, P = 0.000	BD = SCZ > HC
MOAS-AP	0.04 ± 0.19	2.00 ± 1.51	1.66 ± 1.52	F = 61.785, P = 0.000	BD = SCZ > HC
MOAS-AA	0.00 ± 0.00	1.30 ± 1.75	0.53 ± 1.30	t = 2.353, P = 0.021	BD > SCZ
MOAS-PA	0.00 ± 0.00	1.27 ± 1.48	1.23 ± 1.39	t = 0.129, P = 0.898

The distribution of Reactive-Proactive Aggression Questionnaire-proactive aggression scores was skewed. Therefore: Data are presented as median (interquartile range). BD: Bipolar disorder group; HC: Healthy control group; SCZ: Schizophrenia group; Bonferroni: Bonferroni Correction; T2: Tamhane’s T2 test; F: Female; M: Male; BMI: Body mass index; PANSS: Positive and Negative Syndrome Scale; PANSS-P: Positive and Negative Syndrome-Positive Scale; PANSS-N: Positive and Negative Syndrome-Negative Scale; PANSS-G: Positive and Negative Syndrome-General Psychopathology Scale; YMRS: Young Manic Rating Scale; BIS-11: Barratt Impulsiveness Scale-11; BIS-11-MI: Barratt Impulsiveness Scale-11-motor impulsiveness; BIS-11-AI: Barratt Impulsiveness Scale-11-attentional impulsiveness; BIS-11-NPI: Barratt Impulsiveness Scale-11-no plan impulsiveness; RPAQ: Reactive-Proactive Aggression Questionnaire; RPAQ-PA: Reactive-Proactive Aggression Questionnaire-proactive aggression; RPAQ-RA: Reactive-Proactive Aggression Questionnaire-reactive aggression; BPAQ: The Buss-Perry Aggression Questionnaire; BPAQ-PA: The Buss-Perry Aggression Questionnaire-physical aggression; BPAQ-VA: The Buss-Perry Aggression Questionnaire-verbal aggression; BPAQ-A: The Buss-Perry Aggression Questionnaire-angry; BPAQ-H: The Buss-Perry Aggression Questionnaire-hostility; BPAQ-SDA: The Buss-Perry Aggression Questionnaire-self-directed aggression; MOAS: Modified Overt Aggression Scale; MOAS-VA: Modified Overt Aggression Scale-verbal aggression; MOAS-AP: Modified Overt Aggression Scale-aggression against property; MOAS-AA: Modified Overt Aggression Scale-auto-aggression; MOAS-PA: Modified Overt Aggression Scale-physical aggression.

All the SCZ patients were treated with the following antipsychotics: Eight with risperidone (mean dosage 4.1 ± 1.2 mg/day), eleven with olanzapine (mean dosage 13.55 ± 2.15 mg/day), six with quetiapine (mean dosage 515.0 ± 100.35 mg/day), seven with aripiprazole (mean dosage 15.5 ± 9.23 mg/day), eight with amisulpride (mean dosage 255.15 ± 23.24 mg/day), and seven with paliperidone (mean dosage 7.55 ± 2.50 mg/day). All BD patients were treated with stabilizers (lithium carbonate) combined with antipsychotics. The mean dosage of lithium carbonate was 1000.0 ± 125.8 mg/day. Twenty-one received olanzapine (mean dosage 10.25 ± 3.10 mg/day), nineteen received quetiapine (mean dosage 400.35 ± 100.50 mg/day), and four received aripiprazole (mean dosage 10.0 ± 5.38 mg/day). As calculated according to a previous report[48], for SCZ patients, the mean chlorpromazine-equivalent dose was 250.35 ± 15.13 mg/day; for BD patients, the mean chlorpromazine-equivalent dose was 225.65 ± 10.85 mg/day.

Comparisons of the aggression scores of participants

As shown in Table 1, the BIS-11 total scores and factor scores, total RPAQ scores and factor scores, total BPAQ scores and factor scores and total MOAS scores and factor scores were higher in the BD group and SCZ group than in the HC group. The motor impulsiveness scores on the BIS-11, self-directed aggression scores on the BPAQ and self-directed aggression scores on the MOAS were higher in the BD group than in the SCZ group; however, no differences were observed in the other BIS-11 scores, other BPAQ scores or other MOAS scores between the BD and SCZ groups. No differences were observed in the RPAQ scores between the BD and SCZ groups.

Behavioral data analysis

ACC: As shown in Table 2 and Figure 2, for the ACC of No-go trials, the interaction effect for group (SCZ vs BD vs HC) × face type (angry vs neutral) was not significant [F (2, 143) = 0.325, P = 0.723, η_p² = 0.005]. The main effects of face type [F (1, 143) = 0.717, P = 0.399, η_p² = 0.005] were not significant. The main effect of group [F (2, 143) = 23.048, P = 0.000, η_p² = 0.244] was significant. The ACCs of the no-go trials in the SCZ and BD groups were lower than those in the HC group (all P < 0.001). However, there were no differences in the ACC between the BD group and the SCZ group.

Figure 2 Comparisons of behavioral data. A and B: The accuracy of no-go trials of angry faces or neutral faces between schizophrenia group, bipolar disorder group and healthy control group; C and D: The accuracy of reaction time of go trials of angry faces or neutral faces between schizophrenia group, bipolar disorder group and healthy control group. RTs: Reaction times; ACC: Accuracy rate; HC: Healthy control group; BD: Bipolar disorder group; SCZ: Schizophrenia group.

Table 2 Behavior data of participants (mean ± SD).

Variables	Go RTs		No-go ACC
	Neutral-go	Angry-go	Angry-no-go	Neutral-no-go
HC	840.11 ± 169.05	840.42 ± 180.73	0.81 ± 0.24	0.82 ± 0.24
BD	608.29 ± 227.56	625.03 ± 202.62	0.45 ± 0.33	0.48 ± 0.32
SCZ	646.78 ± 214.86	649.72 ± 210.29	0.50 ± 0.33	0.50 ± 0.34
Statistics	F = 21.436c	F = 18.494c	F = 25.196c	F = 23.123c
T2	BD = SCZ < HC	BD = SCZ < HC	BD = SCZ < HC	BD = SCZ < HC

^cP ≤ 0.001.

RTs: Reaction times; ACC: Accuracy rate; HC: Healthy control group; BD: Bipolar disorder group; SCZ: Schizophrenia group; T2: Tamhane’s T2 test.

RTs: As shown in Table 2 and Figure 2, for the RTs in the go trials, the interaction effect for group (SCZ vs BD vs HC) × face type (angry vs neutral) was not significant [F (2, 144) = 0.371, P = 0.691, η_p² = 0.005]. The main effects of face type [F (1, 144) = 0.654, P = 0.420, η_p² = 0.005] were not significant. The main effect of group [F (2, 144) = 20.151, P = 0.000, η_p² = 0.219] was significant. The RTs of the go trials in the SCZ and BD groups were shorter than those in the HC group (all P < 0.001). However, there were no differences in RTs between the BD group and the SCZ group.

ERP data analysis

P100 evoked by the go stimuli: As shown in Figure 3A-D[49], under the go stimulus condition, the P100 amplitudes evoked by angry faces were significantly greater than those evoked by neutral faces in the HC group (P < 0.05, marked by the red dots in Figure 3); however, no differences were observed in either the SCZ group or the BD group (all P > 0.05). There were no differences in the P100 amplitudes evoked by angry faces or neutral faces among the three groups (all P > 0.05).

Figure 3 The significant differences of P100 and N170 in different emotion types between three groups. A: For the go stimulus condition, the P100 mean amplitudes in the 90-150 milliseconds time window over occipital regions (O1, O2, Oz) across group schizophrenia group (SCZ) vs bipolar disorder group (BD) vs healthy control group (HC) × face type (angry vs neutral); B-D: For the P100, the difference topographic maps of emotion types within SCZ, BD and HC group; E: For the go stimulus condition, the N170 mean amplitudes in the 160-200 milliseconds time window over right temporal regions (P6, P8) across group (SCZ vs BD vs HC) × face type (angry vs neutral); F-H: For the N170, the difference topographic maps of emotion types within SCZ, BD and HC group. Electrodes marked with red color indicate a significant group-difference. HC: Healthy control group; BD: Bipolar disorder group; SCZ: Schizophrenia group.

N170 evoked by the go stimuli: As shown in Figure 3E-H[50], under the go stimulus condition, the N170 amplitudes evoked by angry faces were significantly greater than those evoked by neutral faces in the BD group (P < 0.05, marked by the red dots in Figure 3), however, no differences were observed in either the SCZ group or the HC group (all P > 0.05). There were no differences in the N170 amplitudes evoked by angry faces or neutral faces among the three groups (all P > 0.05).

N200 evoked by the no-go stimuli: As shown in Figure 4A-C[51], under the no-go stimulus condition, the N200 amplitudes evoked by neutral faces were greater than those evoked by angry faces in the HC group (P < 0.05). There were differences in the N200 amplitudes evoked by angry faces or neutral faces between the three groups (the SCZ group > the HC group > the BD group) (P < 0.05).

Figure 4 The significant differences of N200 and P300 in different emotion types between three groups. A: For the no-go stimulus condition, the N200 mean amplitudes in the 200-350 milliseconds time window over prefrontal region (AF3, AF4, AFZ) across group [schizophrenia group (SCZ) vs bipolar disorder group (BD) vs healthy control group (HC)] × face type (angry vs neutral); B and C: For the N200 and P300, the difference topographic maps of emotion types within the healthy control group as well as the difference topographic maps of both emotion types within group (SCZ vs BD vs HC); D: For the no-go stimulus condition, the P300 mean amplitudes in the 200-350 milliseconds time window over occipital region (PO3, PO4, POz) across group (SCZ vs BD vs HC) × face type (angry vs neutral). Electrodes marked with red color indicate a significant group-difference. HC: Healthy control group; BD: Bipolar disorder group; SCZ: Schizophrenia group.

P300 evoked by the no-go stimuli: As shown in Figure 4B-D[52], under the no-go stimulus condition, the P300 amplitudes evoked by neutral faces were greater than those evoked by angry faces in the HC group (P < 0.05). There were differences in the P300 amplitudes evoked by angry faces or neutral faces between the three groups (the SCZ group > the HC group > the BD group) (P < 0.05).

LPP evoked by the no-go stimuli: As shown in Figure 5[53,54], under the no-go stimulus condition, the LPP amplitudes evoked by angry faces were greater than those evoked by neutral faces in the HC group (P < 0.05). There were differences in the LPP amplitudes evoked by angry faces or neutral faces between the three groups (the HC group > the SCZ group = the BD group) (P < 0.05).

Figure 5 The significant differences of late positive potential in different emotion types between three groups. A: For the no-go stimulus condition, the late positive potential mean amplitudes in the 300-600 milliseconds time window over occipital regions (O1, O2, Oz, PO3, PO4, POZ) across group (schizophrenia group vs bipolar disorder group vs healthy control group) × face type (angry vs neutral); B: For the late positive potential, the difference topographic maps of emotion types within the healthy control group as well as the difference topographic maps of both emotion types within group (schizophrenia group vs bipolar disorder group vs healthy control group). Electrodes marked with red color indicate a significant group-difference. HC: Healthy control group; BD: Bipolar disorder group; SCZ: Schizophrenia group; LPP: Late positive potential.

Analysis of correlations

As shown in Figure 6, the Pearson correlation was used to analyze the correlations between aggression scales, PANSS and YMRS scores and ERP components. In the BD group, the N170 amplitudes in the go trials in both the neutral and angry stimulus conditions were negatively correlated with the attentional impulsiveness scores in the BIS-11, whereas the N170 amplitudes in the go trials in both the neutral and angry stimulus conditions were positively correlated with the motor impulsiveness scores in the BIS-11 and YMRS. The N200 amplitudes of the no-go trials under neutral stimulus conditions were negatively correlated with YMRS scores.

Figure 6 Correlation analysis of event-related potential data and clinical data within bipolar disorder group and schizophrenia group. A: Bipolar disorder group; B: Schizophrenia group. ^aP ≤ 0.050, ^bP ≤ 0.010. BD: Bipolar disorder group; SCZ: Schizophrenia group; BIS-11: Barratt impulsiveness Scale-11; BIS-11-MI: Barratt impulsiveness Scale-11-motor impulsiveness; BIS-11-AI: Barratt impulsiveness Scale-11-attentional impulsiveness; BIS-11-NPI: Barratt impulsiveness Scale-11-no plan impulsiveness; RPAQ: Reactive-Proactive Aggression Questionnaire; RPAQ-PA: Reactive-Proactive Aggression Questionnaire-proactive aggression; RPAQ-RA: Reactive-Proactive Aggression Questionnaire-reactive aggression; BPAQ: The Buss-Perry Aggression Questionnaire; BPAQ-PA: The Buss-Perry Aggression Questionnaire-physical aggression; BPAQ-VA: The Buss-Perry Aggression Questionnaire-verbal aggression; BPAQ-A: The Buss-Perry Aggression Questionnaire-angry; BPAQ-H: The Buss-Perry Aggression Questionnaire-hostility; BPAQ-SDA: The Buss-Perry Aggression Questionnaire-self-directed aggression; MOAS: Modified Overt Aggression Scale; MOAS-VA: Modified Overt Aggression Scale-verbal aggression; MOAS-AP: Modified Overt Aggression Scale-aggression against property; MOAS-AA: Modified Overt Aggression Scale-auto-aggression; MOAS-PA: Modified Overt Aggression Scale-physical aggression; PANSS: Positive and Negative Syndrome Scale; PANSS-P: PANSS-positive scale; PANSS-N: PANSS-negative scale; PANSS-G: PANSS-general psychopathology scale; YMRS: Young Manic Rating Scale.

In the SCZ group, the P100 amplitudes in the go trials under neutral stimulus conditions were positively correlated with the BIS-11 total scores and the motor impulsiveness scores in the BIS-11, whereas the P100 amplitudes in the go trials under angry stimulus conditions were positively correlated with the verbal aggression scores in the MOAS. The P100 amplitudes of the go trials and the P300 amplitudes as well as the LPP amplitudes of the no-go trials in both the neutral and angry stimulus conditions were positively correlated with the aggression against the properties of the MOAS. The N170 amplitudes of the go trials of angry stimulus conditions were negatively correlated with the hostility of the BPAQ. The N200 amplitudes of the no-go trials in angry stimulus conditions were negatively correlated with the aggression against the properties of the MOAS. The P300 amplitudes of the no-go trials in the neutral stimulus condition were negatively correlated with the negative scale scores of the PANSS.

DISCUSSION

This study investigated the possible differences in the mechanism of aggressive impulsive behavior between patients with SCZ and patients with BD in response to the emotional go/no-go task. Our findings revealed that, on the basis of psychological scales, BD patients and SCZ patients exhibited obvious impulsive aggressive behaviors. The motor impulsiveness of the BIS-11, self-directed aggression according to the BPAQ scores and self-directed aggression according to the MOAS scores were higher in BD patients than in SCZ patients; however, no differences were observed in the other psychological scale measurements between BD patients and SCZ patients. The greater motor impulsivity and self-directed aggression observed in BD patients than in SCZ patients likely arises from the interplay of emotion-triggered impulsivity[55,56], hyperactive reward and limbic circuits[57-59], dysregulated prefrontal inhibitory control[60-62], and elevated risk of self-harm[63,64]. In contrast, the characteristic cognitive disorganization, executive deficits, and diminished affective drive of patients with SCZ result in a different behavioral phenotype - one less prone to overt impulsive or self-directed aggression[21,65,66].

Behavioral data analysis revealed that the ACCs of the go trials and the RTs of the no-go trials in both the SCZ group and the BD group were lower than those in the HC group; however, there were no differences between the BD group and the SCZ group. Consistent with the findings of previous studies, compared with HCs, SCZ and BD patients responded significantly shorter to the go task. These results may reflect increased impulsivity, but they may also be related to the subject’s motivation, level of effort, or processing strategy[67,68]. In terms of performance on no-go tasks, SCZ and BD patients had significantly lower accuracy than HCs did, possibly because emotional situations increase the difficulty of response suppression in individuals with attention deficit, resulting in behavioral inhibition failure and reduced task completion accuracy[69]. Moreover, both have deficiencies in recognizing emotional faces; thus, there are no significant differences in disease groups in response to different categories of emotional stimuli[70,71]. These findings suggest that both SCZ and BD patients lack a complete cognitive processing process, possibly reflecting a more fundamental neurological dysfunction behind them.

ERP results revealed that under the go condition, the P100 evoked by angry faces was greater than that evoked by neutral faces in the HC group, and the N170 evoked by angry faces was greater than that evoked by neutral faces in the BD group; under the no-go condition, the N200 and P300 evoked by angry faces or neutral faces in the SCZ group were greater than those in the BD group and HC group, and the LPP evoked by angry faces or neutral faces in the HC group was greater than those in the BD group and SCZ group. P100 is involved primarily in visual attention, whereas this component is affected by the face[72]. Analysis of P100 indicated that owing to attention impairment or overall deficits in early visual processing stages in SCZ and BD patients, poor emotional discrimination and widespread perception deficits within SCZ patients led to reduced P100 amplitude[73]. In BD patients, deficits in early sensory processing and attention allocation led to reduced P100 amplitudes, which is consistent with the findings of previous studies[74]. Analysis of N170 indicated that BD patients might have negative biases in the face of emotional stimuli and are more sensitive to angry faces[70], whereas SCZ patients are insensitive to differences between two emotional faces and tend to misinterpret neutral expressions as negative emotions (such as anger) and misjudge anger as neutral[75]. These findings indicate that SCZ patients have subtle and unique emotional cognitive deficits. Additionally, N170 is regulated mainly by emotional content and attention levels and is influenced by task demands rather than solely reflecting emotion recognition processing[76]. Therefore, these results may also be influenced by early attention bias or task processing strategies.

Previous research has shown that the no-go N200 component is indicative of the early stages of response inhibition, whereas the no-go P300 component may represent the latter stages, linked to the actual suppression of motor activity[77]. In this study, the N200 and P300 components were the greatest in SCZ patients, which indicates that SCZ patients with aggressive impulsive behavior have facial recognition impairment and need effort to recognize pictures and emotions to inhibit nontarget stimuli. However, in this study, compared with those in SCZ patients and HCs, the N200 and P300 components in BD patients were the smallest, which suggests that BD patients who are more sensitive to angry faces exhibit behavioral impulsivity, indicating potential deficits in suppressing control functions. Moreover, one study reported that emotional stimuli interfere with the response inhibition process in individuals with cognitive impairments, possibly because emotional information is associated with increased attention resources, thereby interfering with response inhibition[78]. This may explain the abnormal phenomena in SCZ and BD patients with cognitive deficits in the N200 and P300 components, which represent inhibitory control. However, the ERP component reflects the comprehensive output of multilevel cognitive processing[79], and both N200 and P300 are affected by factors such as task difficulty, response strategies, and motor skills. Therefore, the observed amplitude anomalies cannot yet be directly attributed.

LPP begins approximately 300 milliseconds after emotional stimuli appear and continues until the stimulus ends[80]. LPP reflects not only the automatic allocation of attention resources to emotional stimuli[81] but also the degree of involvement in emotional regulation[82]. LPP analysis revealed that compared with neutral faces, HCs paid more sustained attention to angry facial stimuli. This finding is consistent with that of a previous study and shows a tendency to prioritize threatening stimuli and allocate attention resources[83]. Our results indicated that SCZ patients have difficulty continuously assigning attention to emotional stimuli because of their insensitivity with them, however, compared to SCZ patients, BD patients may pay less attention to facial stimuli for a long time because of the impairments associated with assigning attention. In addition to difficulty maintaining attention, it may also reflect difficulty using emotional regulation strategies or poor regulation effects; currently, it is impossible to distinguish between these two explanations.

Analysis of correlations revealed that the P100, N170, N200, P300 and LPP amplitudes were correlated with the impulsive aggressive behaviors measured by psychological scales, suggesting that ERP biomarkers might reflect the characteristics of impulsive aggressive behaviors for both SCZ patients and BD patients. Additionally, ERP components are correlated with the PANSS and YMRS, which suggests that abnormal ERP components might be traits in SCZ and BD.

In conclusion, the ERP components evoked by the emotional go/no-go task confirm the neural correlates of aggressive impulsive behavior in patients with SCZ and BD. The differences in neural correlation between the two disorders further suggest that the neural mechanisms of BD and SCZ may be fundamentally different. Our study enhances the understanding of the neurophysiological processes underlying aggressive and impulsive behaviors in patients with BD and SCZ, thereby providing reliable neuroelectrophysiological evidence for the different neural mechanisms of aggressive and impulsive behaviors in these two disorders. Moreover, clarifying these neural mechanisms could offer novel targets for pharmacological and neuromodulatory interventions aimed at mitigating such behavioral impairments.

Limitations

The current research has the following limitations. First, owing to the small sample size and the inclusion of specific groups, such as SCZ and BD with aggressive impulse behaviors, the conclusions cannot be directly generalized to a broader population. Future research should validate the findings in larger sample sizes and broader populations. Second, while the use of ERPs in this study offered advantageous temporal resolution, it is important to note the disadvantages in spatial resolution. Future investigations could benefit from the integration of high spatial resolution methods, such as fMRI and magnetoencephalography, to further verify our outcomes. Finally, this study did not strictly control for variables related to patients' medications, which may have led to bias in the results and requires a more rigorous design in subsequent studies.

CONCLUSION

This study utilized the emotional go/no-go task to support the possibility that patients with SCZ and BD who exhibit aggressive impulsive behaviors have differences in their behavioral performance and neural mechanisms.

ACKNOWLEDGEMENTS

We are grateful to all the people who took part in this study.