Working memory and its neural characteristics in male schizophrenia patients with or without a history of violent behavior: an exploratory fNIRS study

Abstract

Background

Evidence has suggested an association between working memory impairment and violence in people with schizophrenia. However, the neural characteristics of working memory deficit in this patient population remain unclear. This study aimed to examine the neurobiological differences in brain activity using the working memory task in schizophrenia patients with and without a history of violent behavior.

Method

A total of 194 patients, including 106 with a history of severe violent behavior and 88 without such a history, as well as 66 healthy individuals, were recruited. All the participants were required to complete the n-back task (0-, 1-, and 2-back) under functional near-infrared spectroscopy.

Results

Patients with a history of violent behavior, but not those without, exhibited hyperactivity in the left dorsolateral prefrontal gyrus (all p < 0.05) and hypoactivity in the triangular part of the inferior frontal gyrus (all p < 0.05), as compared to healthy individuals. Compared to patient without a violent history, they demonstrated higher accuracy in 0- and 1-back tasks (both p < 0.001), however, their response times were also significantly prolonged (p < 0.001; p = 0.591).

Conclusion

This study revealed severe damage to the overall working memory in schizophrenia patients with a history of violent behavior, with abnormal activation of brain regions related to working memory and semantic processing, accompanied by compensatory brain responses.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12888-025-07626-1.

Background

Schizophrenia is a severe mental disorder with a prevalence of 0.29% worldwide [1] and 0.5% in China [2]. A systematic review investigating violent behavior in Chinese inpatients with schizophrenia revealed a prevalence ranging from 15.3%-53.2% [3]. The core symptoms of schizophrenia include positive symptoms (e.g., hallucinations and delusions), negative symptoms (e.g., emotional blunting and poverty of speech), and cognitive impairment [4]. Beyond this core constellation, depressive symptom frequently occur across all disease stages and may contribute to adverse outcomes [5, 6]. Additionally, while most patients with schizophrenia may not exhibit aggressive behaviors, evidence has shown that this disorder can increase the risk of violent behavior [7]. A meta-analysis by Fazel et al. [8] has shown that the risk of violent behavior and crime in patients with schizophrenia is significantly higher than that in the general population with no history of psychiatric disorder. A cohort study [9] in Sweden found that the rate of crimes in patients with schizophrenia was 4.2 times higher than that in their unaffected siblings. Violence can impact the patients, their families, and society to varying degrees. Thus, studying the violent behaviors of patients with schizophrenia and its underlying mechanisms is of great significance to the treatment of schizophrenia, destigmatization, and public safety.

Studies have shown that impaired cognition may increase the risk of violence. Impaired working memory may reduce one’s ability to assess the risk of their potential behaviors and the consequences. Meanwhile, information retrieved from long-term memory may become deviated, resulting in inappropriate social behaviors, such as violent behaviors [10]. Impaired working memory may also contribute to violent behaviors in individuals with schizophrenia. However, the findings of relevant studies are inconsistent. For example, Kumari et al. [11] found that schizophrenia patients with violent behaviors demonstrated more pronounced working memory impairments compared to schizophrenia patients without such behaviors during the challenging n-back task. However, Silver et al. [12] suggested that both violent and non-violent patients with schizophrenia exhibited lower working memory capacities than healthy individuals, with no significant difference observed between the two patient groups.

Neuroimaging studies have suggested that violent behaviors in schizophrenia are associated with structural abnormalities in brain regions related to working memory. In a study on the MRI and clinical data of 49 patients with schizophrenia or schizoaffective disorder who were on long-term treatment, Hoptman et al. performed volumetric delineation on the structural images of the orbitofrontal cortex, which indicated that increased gray matter volume of the left orbitofrontal cortex was associated with an increased level of aggression [13]. A study by Puri et al. found that schizophrenia patients with a history of severe violent acts showed decreased cortical volume in the nearby supramarginal gyrus (BA39/40), which is integral to semantic working memory, as compared to non-violent patients [14]. Some brain regions are particularly important in working memory, such as the frontal lobe. Abnormal working memory suggests the malfunction of frontal lobe. Kumari et al. [11] found that schizophrenia patients with a history of violent behavior had reduced activation of bilateral frontal lobes during both simple and complex working memory tasks compared to healthy individuals. However, for schizophrenia patients with no history of violent behavior, the pattern and level of activation of the frontal lobe was similar to that of healthy controls.

In light of these results, abnormal activation of the frontal cortex may be related to working memory impairment in schizophrenia patients with a history of violent behavior. Severe damage to working memory may limit the ability of these patients to process new information, potentially leading to violent behaviors. Appraising the brain activity of schizophrenia patients with past violent behavior in working memory task may help to understand the neural mechanisms underlying violent behaviors within the population. However, existing studies have involved different definitions of violent behavior (from verbal to physical aggression), different time frames for the occurrence of violent behaviors, various neurocognitive tasks, and small sample sizes (fewer than 100 participants), which resulted in different or even conflicting findings. Therefore, this study aimed to expand the sample size, recruit schizophrenia patients with a history of severe violent behavior after onset, and evaluate the brain activity of the subjects using the n-back task under fNIRS.

Functional near-infrared spectroscopy (fNIRS), a newly developed imaging technique, can evaluate neural activity by detecting changes in oxygenated hemoglobin concentration without any invasion. With superior spatial resolution to electroencephalography (EEG) and greater temporal resolution than magnetic resonance imaging (MRI), fNIRS has been widely used in studies on schizophrenia [15]. Due to its relative insensitivity to motion artifacts, portability, and applicability in regular clinical settings [16, 17], fNIRS is particularly suitable for studying the violent behaviors of patients with schizophrenia. This makes the technology well-suited for administering cognitive tasks in naturalistic environments while monitoring brain activity under simulated real-world condition [15].For these reasons, this study used fNIRS to assess the activity of the prefrontal cortex using a block-designed numeric n-back (0-back, 1-back, and 2-back) task.

The prevalence of violent behaviors varies by gender. Previous studies have indicated that males are more likely than females to engage in physical violence and they also have a higher risk of committing violent crimes [18–20]. Studies also indicate that assessment tools have higher validity in males than in females [21] and that the impact of the same risk factors for violent behaviors differs between males and females [9]. These findings suggest that the mechanism underlying violent behaviors may differ by gender. Therefore, this study focuses solely on males.

The present study aimed to evaluate the neural characteristics of male schizophrenia patients with severe post-onset violent behaviors and those without lifetime violent behaviors using the n-back task, in the hope of facilitating the development of cognitive rehabilitation programs. We hypothesized that (1) male schizophrenia patients with a history of violent behavior might exhibit poorer n-back behavioral performance than those without such a history, and (2) activation of task-related brain regions might decrease more significantly in male schizophrenia patients with a history of violent behavior.

Methods

Participants

This study used a case-control design, and all participants were recruited from September 2021 to November 2022. The participants were divided into three groups: (i) male schizophrenia patients with a history of violence (the VS group), recruited from the Forensic Psychiatric Hospital of Hunan (China) and the forensic department in general psychiatric hospitals, including Pingjiang County Psychiatric Hospital, the Ninth Hospital of Changsha and Xiangtan Fifth People’s Hospital. All of the mentioned hospitals are located in Hunan Province, China, (ii) male schizophrenia patients without history of violence (the NVS group), recruited from other psychiatric hospitals in Hunan Province, and (iii) healthy males (the healthy control [HC] group), recruited from the community through advertisement.

The VS group consisted of patients who were engaged in violent crimes (e.g., intentional homicide or intentional physical assault) and were determined as incapable of criminal responsibility through jurisdictional identification of mental diseases. Participants in the VS group had evidence of severe aggressive behaviors after the onset of schizophrenia, and participants in the NVS group and HCs were not engaged in aggressive behaviors throughout their lifetime. Information on aggressive behaviors and crimes was obtained from reports of forensic psychiatric examination and medical record data within the electronic medical record system of the Forensic Psychiatric Hospital of Hunan (China) and the forensic department in general psychiatric hospitals.

All the enrolled participants met the following criteria: (i) aged between 18 and 60 years, of Han nationality, and were right-handed, (ii) voluntarily participating in this study by signing the informed consent form, and (iii) able to understand instructions and cooperate in all study tests. Participants in both the VS and NVS groups were diagnosed with schizophrenia according to the International Classification of Diseases and Related Issues, 10th Revision (ICD-10). Patients with any organic brain disease, history of severe head injury, history of alcohol and drug dependence, severe somatic diseases, history of electroconvulsive therapy in the past year before assessment, or auditory or visual impairments were excluded. Patients in the VS and NVS groups who were found to have comorbid mental disorders, including schizoaffective disorder, bipolar disorder, intellectual disability, anxiety spectrum disorder, and mental disorders caused by psychoactive substances were also excluded. A healthy individual would also be excluded from the study if they met any of the diagnostic criteria in ICD-10 or had a previous history of any mental illness.

Upon full understanding of the purpose, process, benefits, and potential risks of this study, all the participants provided written informed consent. The research has been reviewed and approved by the ethics committee of the Second Xiangya Hospital of Central South University.

Clinical assessment

Trained and experienced psychiatrists employed a comprehensive set of diagnostic and assessment tools, including Mini International Neuropsychiatric Interview (MINI), the Positive and Negative Symptom Scale (PANSS), the Clinical Global Impression Severity of Illness (CGI-SI), the Social and Occupational Functioning Assessment Scale (SOFAS), and the Modified Over Aggression Scale (MOAS), to evaluate the patients. In the study, MINI [22] as used to confirm the diagnosis of schizophrenia or other mental disorders. A self-designed questionnaire was used to collect sociodemographic data (e.g. age, gender, education level, ethnicity, registered residence, and marital status), and clinical information and past lifestyle information (e.g. duration of disease (months), duration of current hospitalization (months), history of smoking, history of drinking, history of substances, and family history of mental illness). In addition, this study evaluated the current clinical status of patients using CGI-SI [23] and assessed their current social functioning with SOFAS [24].

The severity of symptoms within the last month in patients with schizophrenia was rated by PANSS, compiled by Kay et al., is a standardized scale used to assess the severity of symptoms within the last month in patients with schizophrenia [25]. The Chinese version of PANSS has demonstrated good reliability and validity [26]. This scale consists of three subscales, i.e., the positive symptom scale (7 items), the negative symptom scale (7 items), and the general psychopathological symptom scale (16 items). Each item is rated on a 7-point Likert scale (1 = none to 7 = extremely severe). The total score of PANSS ranges from 30 to 210, with higher scores indicating greater severity of symptoms.

The aggressive behaviors in the month before the assessment was measured by using MOAS [24] was used to assess the aggressive behaviors in the month before the assessment. The scale includes four subscales: verbal aggression, aggression against property, physical aggression to themself, and physical aggression to other persons. Each item is rated on a 5-point Likert scale, with 0 indicating no aggressive behavior and 4 indicating the most serious aggressive behavior. The total score is calculated based on the following formula: the total score = (score of verbal aggression × 1) + (score of aggression against property × 2) + (score of physical aggression to themself × 3) + (score of physical aggression to other persons × 4).

N-back working memory paradigm

The procedure is shown in Fig. 1. Block-designed numerical n-back paradigm was programmed using E-prime software (3.0 version). The n-back task consisted of three subtasks with three difficulty levels (0-back, 1-back, and 2-back). The task was presented in the following order: 0-back → 1-back → 2-back. Each subtask consists of 3 equally difficult blocks, and each block contained n + 15 numbers, with the ratio of target to non-target stimuli being 1:2. All numbers were presented pseudo-randomly. The interval between tasks of different levels was 60 s, and the interval between different blocks of the same task was 30 s. The stimulus presentation time was 500 ms, and the maximum time allowed for participants to respond was 2000 ms. A fixation sign “+” on the computer screen (t = 0 ms) indicated the start of the task, which lasted for 500 ms. Participants were instructed to respond promptly by pressing a button. They were asked to press the “F” key with their left index finger when a target stimulus appeared, and they needed to press the “J” key with their right index finger when a non-target stimulus appeared. The procedure needed approximately 10 min to complete.

Fig. 1.

Procedure of n-back paradigm in the study

The task was completed in a separate and quiet room. Participants were asked to sit in a comfortable position in front of the computer screen, with their eyes 100 cm away from the screen. The task consisted of two parts, i.e., a practice test and a formal test, which had the same procedure. Participants with an accuracy rate lower than 90% were asked to perform the practice task again, and those with an accuracy rate higher than 90% in the practice test would be invited to complete the formal test. This threshold was set to ensure that participants thoroughly comprehend the task and can perform it correctly, ensuring consistency in performance during the formal experiment [27, 28].

Neuroimaging acquisition

Neuroimaging data were collected using NirSmart functional near-infrared spectroscope (Danyang Huichuang Medical Equipment Co., Ltd., China). The device, including the data acquisition and display systems, consists of 14 light-source probes and 14 detection probes, forming a total of 35 effective channels. The distance between probes was 3 centimeters, with the lowest probe placed along the Fpz, C3 and C4, according to the international 10–20 system. The main coverage area was the prefrontal lobe, with some channels covering the temporal and parietal lobes. The measured near-infrared wavelengths were 730 nm and 850 nm. The full channel sampling rate was ≥ 11 Hz. The arrangement of probes and channels is shown in Fig. 2.

Fig. 2.

Location of fNIRS channels in the study

Statistical analysis

Analysis of fNIRS data

NirSpark 1.7.2 software (Danyang Huichuang Medical Equipment Co., Ltd., China) was used for the preprocessing of fNIRS data. The preprocessing included (i) removal of motion artifacts by using moving standard deviation, (ii) filtration of instrumental and physiological noise with cut-off frequencies of 0.01–0.2 Hz, and (iii) calculation of oxygenated hemoglobin (Oxy Hb) and deoxygenated hemoglobin (Deoxy Hb) concentration changes using modified Beer-Lambert Law. Given that Oxy Hb generally demonstrated a higher signal-to-noise ratio than Deoxy Hb and were more reflective of task-related activation [29, 30], Oxy Hb was selected as the primary measure for subsequent analyses. Then, the General Linear Model (GLM) was used using the same software to assessed to estimate the temporal correlation between hemodynamic changes and task conditions. At the individual level, GLM was conducted to generate β-value for each condition. These β-values represent the magnitude of task-related neural activity. The β-values were then utilized for subsequent statistical analyses and the generation of activation maps.

This study adopted an analysis method based on regions of interest (ROI) [31, 32]. Before analyzing fNIRS data, ROIs were selected based on the current research purpose and previous research findings. According to prior studies, the dorsolateral frontal cortex (DLFC) [33] is an important brain region for working memory, and the left middle frontal gyrus (lMFG) [34] and inferior frontal gyrus (IFG) [35] also play an important part. Therefore, bilateral dorsolateral superior frontal gyri (CH7, CH10, CH24, CH31), bilateral inferior frontal gyri (CH4, CH12), and the left middle frontal gyrus (CH25, CH26, CH27, CH34) were selected as ROIs in this study (see Supplementary materials for details). Due to the distribution of fNIRS channels in this study, the inferior frontal gyrus only included the triangular part of the inferior frontal gyrus.

Analysis of variance was used to compare the Oxy Hb changes in five ROIs between the VS, NVS, and HC groups, with the use of Bonferroni correction. All the tests were two-tailed, with p < 0.05 indicating statistically significant differences.

Analysis of clinical data

The measurement data were presented as mean ± standard deviation and compared between groups using independent samples t-test or one-way analysis of variance. The least significant difference (LSD) method and the Games Howell method were used for pairwise comparison in the post hoc test. Enumeration data were presented as frequencies and composition ratios and analyzed using chi-square test or Fisher exact probability test. One-way analysis of covariance (ANCOVA) was used to examine whether there were differences in the adjusted means between two or more groups. Due to the significant differences in age and years of education between the VS group and the other two groups, age and years of education were used as covariates in the comparison of behavioral outcomes and fNIRS data between groups to minimize the impact of the two factors. All the above statistical analyses were performed using the SPSS (23.0) software.

Results

Sociodemographic and clinical data

A total of 194 patients with schizophrenia, including 106 in the VS group and 88 in the NVS group, as well as 66 HCs, were enrolled. Compared to the NVS group, patients in the VS group had a lower education level (p < 0.001). The mean age was 40.15 ± 7.45 years for the VS group, 42.90 ± 11.37 years for the NVS group, and 43.88 ± 11.22 years for HCs, with a significant difference found between the VS and HC groups only. There was no significant difference in age and marital status between the VS and NVS groups, and no significant difference was found in terms of total duration of disease and duration of current hospitalization between the two patient groups, either (Table 1).

Table 1.

Demographic and clinical information of all groups

Age (years)		VS (N = 106)	NVS (N = 88)	HC (N = 66)	p	t / χ2 / F
		40.15 ± 7.45	42.90 ± 11.37	43.88 ± 11.22	0.035b	3.39
Educational level (years)		9.17 ± 2.81	11.39 ± 3.38	10.92 ± 3.00	< 0.001 a, b	14.22
Marital status					< 0.001 b, c	119.28
	Unmarried	81 (76.4%)	70 (79.5%)	9 (13.6%)
	Married/cohabitating	10 (9.4%)	7 (8.0%)	50 (75.8%)
	Others	15 (14.2%)	11 (12.5%)	7 (10.6%)
Total duration of disease (months)		195.51 ± 88.03	219.47 ± 141.21	-	0.155	-1.43
History of smoking					< 0.001 a, c	105.51
	No	36 (34.0%)	87 (98.9%)	19 (28.8%)
	Yes	70 (66.0%)	1 (1.1%)	47 (71.2%)
History of drinking					< 0.001 a, b	50.38
	No	59 (55.7%)	83 (94.3%)	60 (90.9%)
	Yes	47 (44.3%)	5 (5.7%)	6 (9.1%)
History of drugs					< 0.001 a, b	16.69
	No	95 (89.6%)	88 (100.0%)	66 (100.0%)
	Yes	11 (10.4%)	0 (0.0%)	0 (0.0%)
Family history of mental illness					0.035 c	6.68
	No	88 (83.0%)	72 (81.8%)	62 (95.4%)
	Yes	18 (17.0%)	16 (18.2%)	3 (4.6%)

VS: schizophrenia patients with a history of violent behavior; NVS: schizophrenia patients without a history of violent behavior; HC: healthy control. a: significantly different between the VS and NVS groups; b: significantly different between the VS and HC groups; c: significantly different between the NVS and HC groups

Post-hoc comparisons were performed using LSD method or the Games Howell method to adjust for multiple comparison. p-value in year, educational level, marital status, history of smoking, history of drinking, history of drugs and family history of mental illness was adjusted

Clinical characteristics

Compared to the NVS group, the VS group had higher scores on the positive scale and general psychopathology scale in PANSS. There was no significant difference in the scores of the negative subscale in PANSS, ITAQ, CGI-SI, and SOFAS between the VS and NVS groups (Table 2). The VS group also scored higher than the NVS group regarding the physical aggression subscale in MOAS (Table 2).

Table 2.

Clinical characteristics of the VS and NVS groups

		VS (N = 106)	NVS (N = 88)	t	p
PANSS-Positive scale		13.34 ± 6.33	12.75 ± 6.29	0.648	0.518
PANSS-Negative scale		17.40 ± 7.58	17.77 ± 7.49	-0.346	0.729
PANSS-General psychopathological scale		27.19 ± 7.51	26.28 ± 7.31	0.845	0.399
PANSS-total score		57.92 ± 18.17	56.81 ± 17.76	0.431	0.667
CGI-SI		3.51 ± 1.59	3.68 ± 1.69	-0.728	0.467
ITAQ-total score		12.58 ± 8.23	13.01 ± 8.59	-0.352	0.725
SOFAS score		64.29 ± 16.48	64.25 ± 13.17	0.020	0.984
MOAS	Verbal aggression	1.00 ± 1.19	1.38 ± 1.24	-0.801	0.426
	Aggression against property	1.88 ± 0.99	1.64 ± 1.31	0.479	0.633
	Autoaggression	1.00 ± 1.85	0.86 ± 1.57	0.236	0.814
	Physical aggression towards others	3.00 ± 1.41	1.14 ± 1.12	4.250	< 0.001
	MOAS total score	19.75 ± 4.86	11.8 ± 7.58	2.870	0.006

VS: schizophrenia patients with a history of violent behavior; NVS: schizophrenia patients without a history of violent behavior; HC: healthy controls

Behavioral results

The accuracy of the VS group decreased with the increasing complexity of the n-back task. For the accuracy in the 0- and 1-back tasks, the NVS group exhibited the poorest performance, followed by the VS group and HCs, while for the accuracy in the 2-back task, the VS group showed the poorest performance, followed by the NVS group and HCs. The average accuracy of HCs was significantly higher than that of the VS group and the NVS group (both p < 0.001).

The response time of the VS group also increased with the increase in complexity of the task. In the 0- and 1-back tasks, the VS group had the longest response time, followed by the NVS group and HCs, but in the 1-back task, there was no significant difference between the VS and NVS groups. No significant difference in response time was found between the three groups in the 2-back task (Table 3).

Table 3.

Behavioral characteristics of the VS, NVS, and HC groups in the n-back tasks

		VS (N = 106)	NVS (N = 88)	HC (N = 66)	p	F Statistics
Accuracy
	0-back	0.766 ± 0.206	0.653 ± 0.237	0.919 ± 0.107	< 0.001 a, b,c	33.84
	1-back	0.750 ± 0.226	0.680 ± 0.242	0.911 ± 0.123	< 0.001 a, b,c	23.98
	2-back	0.666 ± 0.221	0.695 ± 0.245	0.828 ± 0.122	< 0.001 b, c	12.92
Response time
	0-back	665 ± 202	621 ± 177	537 ± 113	< 0.001 a, b,c	10.82
	1-back	697 ± 189	695 ± 188	588 ± 126	< 0.001 b, c	9.31
	2-back	740 ± 190	766 ± 224	714 ± 193	0.391	0.94

VS: schizophrenia patients with a history of violent behavior; NVS: schizophrenia patients without a history of violent behavior; HC: healthy controls. a: significantly different between the VS and NVS groups; b: significantly different between the VS and HC groups; c: significantly different between the NVS and HC groups

Post-hoc comparisons were performed using LSD method or the Games Howell method to adjust for multiple comparison. p-value was adjusted

fNIRS results

In the 0-, 1- and 2-back tasks, the VS group showed greater activation of the left dorsolateral superior frontal gyrus than the HC group (p = 0.039, 0.026, and 0.031, respectively). In the 1- and 2-back tasks, the VS group also showed greater activation of this region than the NVS group (all p < 0.05), and no significant difference was found between the NVS group and HCs. The VS group also showed greater activation of the right dorsolateral superior frontal gyrus than the NVS group (all p < 0.05) during the 1- and 2-back tasks, and no significant difference was observed between the NVS group and HCs.

In the 0-, 1- and 2-back tasks, the VS group showed reduced activation of the left triangular part of the inferior frontal gyrus, compared to the HC group (p = 0.001, 0.035, and 0.002, respectively). No significant difference was found between the VS and NVS groups, as well as between the NVS and HC groups in all three n-back tasks with regard to the activation of the left triangular part of the inferior frontal gyrus. Regarding the right triangular part of the inferior frontal gyrus, the VS group showed significantly reduced activation compared to the HC group in all three tasks (p = 0.007, 0.013, and 0.002, respectively). In addition, the VS group showed lower activation of this region than the NVS group in the 2-back task (p < 0.001). There was no significant difference in the activation of the right inferior frontal gyrus between the NVS and HC groups in all three tasks. In the left middle frontal gyrus, and there was no significant difference in the activation of the left middle frontal gyrus between the three groups in all three tasks (Fig. 3, more details are presented in the supplementary material).

Fig. 3.

Results assessing brain activation through fNIRS among VS, NVS, and HC. Note: fNIRS: functional near infrared spectroscopy; VS: schizophrenia patients with a history of violent behaviors; NVS: schizophrenia patients without a history of violent behaviors; HC: healthy controls

Discussion

In this study, we examined brain functional activity in n-back working memory tasks using fNIRS, which showed differences in the activation of brain regions between the VS, NVS, and HC groups under different task conditions. In the 1- and 2-back tasks, the VS group showed increased activation of the bilateral dorsolateral prefrontal gyrus compared to the NVS group. When compared to HCs, they exhibited increased activation of the left dorsolateral prefrontal gyrus and decreased activation of bilateral triangular part of inferior frontal gyrus in all three n-back tasks. This study also found that participants in the VS group had a higher accuracy than the NVS group but responded more slowly in a low-load n-back task (0-back). These findings may shed light on the development of rehabilitation programs for schizophrenia patients with a history of violent behavior.

We observed increased activation of the left dorsolateral superior frontal gyrus in the VS group compared to HCs in all three n-back tasks. Furthermore, compared to the NVS group, the VS group exhibited greater activation in bilateral dorsolateral superior frontal gyri in the 1- and 2-back tasks. Although previous studies have indicated a decrease in prefrontal function in patients with schizophrenia, some studies have reported increased activation in some prefrontal regions during cognitive tasks, which is shared by both schizophrenia patients and healthy individuals [36]. The increased activation may reflect a compensatory response to dorsolateral superior frontal gyrus dysfunction in the VS group. The dorsolateral superior frontal gyrus, essential for working memory, is extensively interconnected with other brain structures, such as hippocampus. Due to spontaneous synchronization, co-activation with amygdala (critical for emotional processing) is often observed during task performance. Previous studies have found that schizophrenia with a history of violent behavior exhibit reduced hippocampal and amygdala volumes [37–39]. Functional MRI studies have further revealed abnormal connectivity between the prefrontal cortex and amygdala in this subgroup [40]. Based on the findings above, we propose the individuals in VS group may exhibit structural and functional abnormalities in the dorsolateral superior frontal gyrus and its associated networks, or disruptions in the neutral connectivity between these regions. Our results suggest that the abnormal activation pattern observed in the dorsolateral superior frontal gyrus may be attributed impaired connectivity with other functional brain areas. Such aberrant connections could trigger compensatory activation mechanisms, allowing the brain to reallocate resources to maintain task performance. Moreover, previous study have reported reduced glucose metabolism in the prefrontal cortex in schizophrenia patients with a history of violent behavior exhibit during cognitive task performance [41]. This suggests that abnormal activation observed in the present study may reflect decrease neural efficiency in dorsolateral superior frontal gyrus, with the increased activation potentially resulting from compensatory mechanisms employed by the brain to preserve cognitive function. Additionally, according to the hypothesis of Naudts and Hodgins [42], the increased activation in the prefrontal lobe may be associated with antisocial traits that manifest since childhood. Further research is needed to explore the relationship between violent behaviors, antisocial traits, and activation of the dorsolateral superior frontal gyrus.

In this study, participants in the VS group showed diminished activation of the bilateral triangular part of the inferior frontal gyrus in all three n-back tasks, as compared to HCs, while the NVS group showed a similar level of activation to the healthy individuals. In Baddeley’s model, individuals increase their response to the initial visual stimuli continuously through phonetic circuits in the form of language, thereby completing the process of working memory [43]. This process may lead to the activation of the triangular part of the inferior frontal gyrus. In the present study, the VS group showed reduced activation of the triangular part of the inferior frontal gyrus, which might indicate the VS group relied less on verbal encoding to complete the task. The underlying cause of this phenomenon remains unclear and requires further discussion as to whether it results from a severe impairment of processing and maintenance of the phonological loop, or whether it indicates a shift toward the adoption of alternative non-verbal encoding strategies in the VS group? According to the findings reported by Kumari et al., reduced activation of the left sensorimotor cortex was observed in the schizophrenia patients with a history of violent behavior compared to the healthy controls during the 0-back task; however, this pattern was also present in the schizophrenia patients without a history of violent behavior [11]. Thus, based on the currently limited evidence available, it is difficult to conclude that the VS group exhibit a shift from verbal to non-verbal encoding strategies during working memory task. Taken together, the decreased activation of the triangular part of the inferior frontal gyrus observed in the present study is more likely attributable to a reduction in functional capacity within this region. In addition, studies have shown that the left inferior frontal gyrus plays an important role in integrating information, which helps to distinguish abstract concepts [44]. Prior studies also showed that the inferior frontal gyrus plays a crucial role in response inhibition, which is a key aspect of cognitive control [45, 46]. Thus, the decreased activation of this region might also indicate defects in the ability to understand abstract concepts and response inhibition in schizophrenia patients with a history of violence. However, there are few studies on the association between activation of the triangular part of the inferior frontal gyrus during working memory tasks and violence in patients with schizophrenia, and researchers only observed changes in task-related activity in patients with other mental illnesses. A study by Prehn et al. found enhanced activation of the inferior frontal gyrus in individuals with personality disorders during low-load working memory tasks (with background stimuli), and using working memory tasks with near-infrared technology. Zhu et al. found that both bipolar and unipolar depression patients showed decreased activation of the inferior frontal gyrus, indicating the role of this brain region in emotional disorders [47, 48]. To date, there has been no relevant research on the relationship between abnormal activation of the inferior frontal gyrus and violent behavior in schizophrenia. Although our finding of decreased activation of the inferior frontal gyrus in working memory tasks in schizophrenia patients with a history of violent behavior may shed some light on this relationship, this finding still needs to be verified in further studies.

We also found that in the low-load n-back tasks (0-back), the VS group demonstrated greater accuracy compared to the NVS group, while in the high-load task (2-back), the difference between the two groups became insignificant. The low-load task is commonly considered an attention task. Using this task, prior studies have yielded consistent results [49] that schizophrenia patients who engaged in violent crime performed relatively well in attention tests but exhibited poorer performance than patients without violent behaviors in tasks assessing working memory and processing speed. These findings suggest that schizophrenia patients who have engaged in violence may exhibit a higher level of attention in relevant tasks.

Our study also found differences in the accuracy and response time between the VS and NVS groups during the n-back tasks. Participants in the VS group tended to prioritize accuracy, resulting in a higher accuracy rate despite a longer response time. Conversely, participants in the NVS group prioritized quicker task completion, resulting in a shorter response time despite lower accuracy. Previous studies have demonstrated the relationship between schizophrenia and working memory deficits [50, 51]. However, a recent study found that patients with schizophrenia had intact encoding; thus, the decreased accuracy of patients in the VS group might be the result of disrupted maintenance of working memory [52]. These results may also suggest a different strategy to allocate resources in schizophrenia patients with a history of violent behavior [53, 54]. Another explanation is that patients in the VS group might have been motivated to perform better in cognitive tests, which might seem beneficial to their situation during compulsory treatment [55, 56].

There are some limitations in this study. Firstly, all the enrolled patients were chronic schizophrenia patients with a history of long-term medication use at the time of study. Although the influence of medications on brain activities has been controversial, a study found that atypical antipsychotics had little impact on the changes of oxy-Hb under fNIRS [57]. However, a systematic review [58] has highlighted that clozapine can lead to decreased metabolism in the basal ganglia and thalamus in patients with refractory schizophrenia. Thus, based on the current evidence, this study did not implement specific restrictions regarding medication use. And large-scale fNIRS studies on unmedicated patients with schizophrenia are needed to elucidate the relationship between psychotropic drugs and the functional activity of the brain. Secondly, this study evaluated participants only based on the accuracy and response time in the working memory tasks, and information decoding, maintenance, and retrieval of schizophrenia patients with and without a history of violent behaviors should be explored in future studies.

Conclusion

This study used near-infrared brain imaging technology based on the n-back paradigm to explore the working memory characteristics and neural activity of schizophrenia patients with and without a history of violent behavior. The study found that patients with a history of violent behavior might have difficulty adjusting their behavioral strategies, indicating severe damage to their working memory and functional coordination. In the working memory tasks, schizophrenia patients with a history of violent behaviors showed functional abnormalities in a range of brain regions, especially the dorsolateral superior frontal gyrus. Due to the scarcity of research in this field, future studies with a larger cohort are needed to verify the findings of the present study, which may help to identify schizophrenia patients who are at a higher risk of violent behaviors.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

fNIRS

functional near-infrared spectroscopy

EEG

Electroencephalography

MRI

Magnetic resonance imaging

VS

Male schizophrenia patients with a history of violence

NVS

Male schizophrenia patients without history of violence

HC

healthy males

ICD-10

The International Classification of Diseases and Related Issues, 10th Revision

MINI

Mini International Neuropsychiatric Interview

PANSS

The Positive and Negative Symptom Scale

CGI-SI

The Clinical Global Impression Severity of Illness

SOFAS

The Social and Occupational Functioning Assessment Scale

MOAS

The Modified Over Aggression Scale

ROI

Regions of interest

DLFC

The dorsolateral frontal cortex

MFG

The left middle frontal gyrus

IFG

Inferior frontal gyrus

ANOVA

Analysis of variance

ANCOVA

Analysis of covariance

Author contributions

YG and HJG contributed to data collection, analysis and interpretation, and drafted the manuscript. KL, YXL, GPL, JH and TJC participated in data collection, reviewed the manuscript and gave constructive comments for the manuscript. YG, SLZ and QLS generated the visualizations, and were responsible for the literature review and revised the manuscript. YG, YH, JZW, SYW and YXZ performed the data extraction, data analyses and interpretation of data. JSZ and XPW supervised the study, critically read the manuscript, and provided intellectual comments. HJG, QLS, JSZ and XPW acquired the funding. All authors conceptualized the study, contributed to the manuscript and approved the final version of the manuscript.

Funding

This work was supported by the STI2030-Major Projects-2021ZD0200700; National Natural Science Foundation of China (82171509); National Natural Science Foundation of China (82301741); Key Research and Development Program of Hunan Province (2023SK2028); and National Natural Science Foundation of China (82071543), the Postdoctoral Fellowship Program of CPSF (GZC20233170) for Huijuan Guo.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

The research has been reviewed and approved by the ethics committee of the Second Xiangya Hospital of Central South University. All participants were fully informed about the study and provided written consent to participate in this study. All procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and conducted in strict compliance with the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.