Schizophrenia is a mental disorder affecting approximately 1% of the global population, leading to significant functional impairments (1). Accumulated evidence suggests that cognitive symptoms are the mainstay factor for functional impairments in patients. Difficulties in acquiring, encoding, and retrieving new information—along with challenges in integrating previously learned material—are common cognitive impairments that significantly contribute to functional disability (2). It is thought that the cognitive symptoms stem from disrupted neuroplasticity, hindering the brain’s ability to process environmental stimuli essential for daily functioning (3). One major challenge in studying neuroplasticity in the in vivo brain is the limited availability of applicable tools. However, resting-state functional MRI (rs-fMRI), which captures changes in the functional coupling between brain regions following learning, has emerged as a promising method for investigating neuroplasticity-related alterations (4).

The Serial Reaction Time Task (SRTT) is commonly used to assess implicit learning—the unconscious acquisition of skills or knowledge—which is often impaired in individuals with schizophrenia. Reaction times (RTs) typically decrease as they become familiar with the predetermined sequence of cues but show a rebound when random stimuli are introduced. (5–7). The subject learns the predetermined sequence implicitly during the task, so, neuroplasticity is expected to occur in related brain regions. Neuroimaging studies revealed that SRTT primarily engages the motor cortex, premotor cortex (PMC), parietal cortex, basal ganglia, and medial temporal lobe (8–18). Indeed, theoretical models and experimental findings highlight the roles of parallel cortico-cerebellar and cortico-striatal circuits in procedural implicit learning (19–21). In the early stages of learning, the cerebellum facilitates motor adaptation and rapid skill acquisition, while the striatum plays a more prominent role in consolidating learned skills over time. Functional connectivity studies suggest a shift in connectivity during learning, decreased cerebellar and increased striatal connectivity to the motor cortex, reflecting the transition from adaptation to consolidation (19).

In the current study, we incorporated a reward-enhanced SRTT protocol between two rs-fMRI sessions to assess neuroplasticity deficits in the cortico-striato-cerebellar circuit and evaluate differences between schizophrenia patients and healthy controls. Based on the literature, we propose that the altered neuroplasticity in schizophrenia patients compared to healthy controls can be assessed by observing functional connectivity changes after SRTT. We included a reward component in SRTT to investigate its modulatory effect on learning and neuroplasticity, given that it has been shown to modulate both implicit and explicit learning (22) and enhance striatal-prefrontal connectivity in motor tasks (23). The use of a reward-enhanced SRTT may help to alleviate motivational deficits associated with negative symptoms and, in turn, enhance cortico-striatal activation during task performance. Thus, this study also highlights the potential role of motivation and reward mechanisms in mitigating these deficits.

In this study, we examined the relationship between implicit learning of schizophrenia patients with neuroplasticity using a reward-enhanced SRTT via rs-fMRI. Our findings showed patients had a lower proportion of correct responses compared to controls, suggesting implicit learning deficits in schizophrenia. The observed baseline hyperconnectivity in cortico-striato-cerebellar regions became even more robust and widespread following task completion in patients. However, increased connectivity was not associated with the patients’ performance. It was further observed that hyperconnectivity after the task was a differentiating feature that could classify individuals into different groups.

A dissociation in behavioral performance was observed in schizophrenia patients. While their reduction in reaction times (RTs) across sequential compared to random blocks was comparable to that of healthy controls—suggesting partially preserved implicit learning—they demonstrated significantly lower accuracy in the sequential blocks. This discrepancy points to potential deficits in motor execution, attentional regulation, or the integration of learned sequences. Additionally, the inclusion of a reward component in the task, together with the group’s relatively low ESRS scores, may have influenced reaction time patterns positively in patients. However, the lack of sequence-specific improvement in the post-test condition, along with a lower proportion of correct responses, further supports the presence of subtle impairments in implicit learning among schizophrenia patients. Our findings are also consistent with the meta-analysis conducted by Siegert et al. (27). In reviewing the studies included in this meta-analysis, for example, Green et al. (29) reported that patients with chronic schizophrenia exhibited significantly lower levels of sequence learning compared to healthy controls. Similarly, Marvel et al. (30) found impaired implicit learning in a group of psychosis, although the inclusion of individuals with schizoaffective disorder was noted as a limitation. To eliminate the effects of medication, Exner et al. (31) investigated drug-naïve first-episode patients and reported that impairments in implicit learning may be related to the nature of the disorder. These findings were supported by Zedkova et al. (32), who also found impaired implicit learning in patients with chronic schizophrenia. Consistent with these findings, our study also showed that patients had lower accuracy compared to controls, suggesting implicit learning deficits.

Our finding of the increased baseline cortico-striato-cerebellar hyperconnectivity in schizophrenia patients has been reported previously from the onset of the disease and has been proposed as a potential soft neurological sign of the disorder (33–37). The observation of hyperconnectivity between the thalamus and cortical regions in the current study supports the idea that the excitatory-inhibitory imbalance between glutamatergic and GABAergic neurons leads to dysfunction in thalamo-cortical neurocircuits (38). Northoff et al. proposed that increased dopaminergic and decreased serotonergic signaling in psychotic disorders, including schizophrenia, may contribute to increased sensorimotor-thalamic functional connectivity (39).

In previous studies, healthy subjects have increased cortico-striato-cerebellar connectivity after SRTT and one study reported healthy controls have more cortico-striato-cerebellar connectivity compared to schizophrenia patients (40). However, we could not observed connectivity changes in the healthy controls. On the other hand, our schizophrenia patients had more widespread increased cortico-striato-cerebellar connectivity after SRTT. The reason for the discrepancy between our and the previous study might be due to the type antipsychotics that the patients were on. In the previous study, those who remained on typical antipsychotics continued to show a lack of procedural learning, while patients who switched to atypical antipsychotics demonstrated significant improvements in procedural learning, accompanied by increased activation in the superior-middle frontal gyrus, anterior cingulate, and striatum (41). In our study, the fact that all patients were using atypical antipsychotics may have contributed to the observed results, potentially due to the relatively weaker nigrostriatal D2 blockade associated with these medications.

Our results show that although cortico-striato-cerebellar hyperconnectivity present in schizophrenia patients, its lack of correlation with behavioral performance supports the dysfunctional brain networks hypothesis in schizophrenia (42). At the cellular level, the disturbances in the integration of electrical stimuli impair the consolidation of incoming information, resulting in excessive information overload in the brain. Consequently, even if connectivity within certain brain networks increases, functional impairments may still occur (43). Additionally, studies using psychedelic-induced psychosis models suggest that abnormal synaptogenesis leads to hyperconnectivity and increased neural entropy (44). However, this hyperconnectivity might not be functionally beneficial as showed in a previous study by Wu et al. (45), the finding that sensorimotor-thalamic hyperconnectivity was associated with poorer task performance further supports the dysfunctional brain networks hypothesis in schizophrenia patients.

In their study, Chang et al. (46) demonstrated the presence of microstructural abnormalities in the cerebellar peduncles of patients with schizophrenia. These abnormalities were specific to the middle and inferior cerebellar peduncles and were characterized by reduced fractional anisotropy and increased radial diffusivity. Notably, these changes were significantly associated with cognitive performance in healthy controls, but not in patients. This suggests that impaired cerebellar white matter integrity may underlie cognitive impairments in schizophrenia. Damage to the cerebellar pathways has been shown to impair efficient information processing, which, in turn, leads to increased functional connectivity within the circuit as a compensatory mechanism to maintain basic task performance.

Recent evidence from several neuropsychiatric disorders suggests that hyperconnectivity within large-scale brain networks may reflect a compensatory response to underlying structural or functional disturbances. For example, patients with mild traumatic brain injury showed increased functional connectivity between the inferior fronto-occipital fasciculus and primary sensorimotor along with cerebellar networks, which was associated with impaired information processing speed, suggesting a maladaptive compensatory mechanism (47). Similarly, patients with generalized tonic-clonic seizures showed increased variability in functional connectivity, particularly in the default mode, attention and cognitive control networks, reflecting a dynamic reorganization that may be related to both epileptic activity and cognitive dysfunction (48). Consistent with these findings, our study in schizophrenia revealed hyperconnectivity within the cortico-striato-cerebellar circuit, which we interpret as a compensatory adaptation to widespread disruptions in brain function. Taken together, these findings support the notion that increased network connectivity may serve as a general compensatory mechanism in neuropsychiatric disorders, albeit potentially associated with inefficient cognitive processing.

In this study, we employed a reward-related SRTT, which is expected to modulate neural activity in emotion-related regions. Our findings revealed reduced connectivity between the insula and the posterior cingulate cortex (PCC), a key region of the default mode network (DMN). This result aligns with Ebisch et al. (49), who reported similar disruptions in first-episode schizophrenia patients, suggesting that schizophrenia may impair the insula’s role in emotional processing (50).

Additionally, we observed increased insula-thalamus connectivity following the task, consistent with previous findings that connectivity in this pathway strengthens with feedback (51). This suggests that schizophrenia patients remain responsive to feedback mechanisms. Furthermore, we found enhanced connectivity between the insula, a core component of the salience network, and the cerebellum in schizophrenia. This may reflect a compensatory mechanism for executive dysfunction, potentially engaging the reward system.

While Chrobak et al. (52) reported no relationship between the sensorimotor network and SRTT performance, our findings suggest that using a reward-enhanced SRTT may have contributed to increased connectivity between the premotor cortex (PMC), a key sensorimotor region, and the nucleus accumbens. This indicates that despite neuroplasticity deficits, schizophrenia patients may still process rewards, albeit through alternative neural pathways.

Finally, contrary to the meta-analysis by Radua et al. (53), we found increased connectivity between the nucleus accumbens and both the thalamus and PMC. This discrepancy may stem from the fact that all participants in our study were treated with atypical antipsychotics, which are known to normalize striatal activity. Overall, our findings suggest that implicit learning deficits in schizophrenia are not solely attributable to motivational factors or negative symptoms, as often debated, but are likely rooted in fundamental neural dysfunctions, including impaired neuroplasticity.

Our findings demonstrate that SRTT-induced changes can predict clinical groups with high accuracy (AUC: 0.95–1). Key predictors included the PMC, bilateral insula, thalamus, and cerebellum ( Figure 6 ). To assess the necessity of neuroimaging, behavioral metrics (reaction times, correct response rates) were included as predictors. However, their weak contributions to PC1, with significant weights (>0.2) only in the post-task fMRI condition, highlight the primacy of neural changes in classification. This suggests that fMRI-based functional connectivity provides deeper insights into neuroplasticity than behavioral measures alone, particularly in task-induced resting-state alterations. Further studies are needed to explore these interactions in greater detail and clarify their role in reward processing and its dysregulation in schizophrenia.

This study has limitations, both in sample size and study design. One limitation is the inability to perform fMRI during the task. This was mainly due to the task’s total duration of approximately 75 minutes, making it challenging for patients to minimize their head movements. Relatively small sample size, the exclusion of left-handed participants, and the cross-sectional design decrease the generalizability of the findings. Additionally, the small sample size may contribute to an overestimation of the SVM results; however, the baseline AUC values appear sufficiently robust to support the findings. Similarly, external validation is needed further to support the effectiveness of SRTT in group distinction. Future research should use longitudinal designs and multimodal imaging to better understand neuroplasticity dynamics and implicit learning over time. Additionally, studying unmedicated patients or those at different illness stages may provide further insights.

This study provides evidence of altered neuroplasticity in schizophrenia, characterized by hyperconnectivity within cortico-striato-cerebellar circuits during implicit learning. While this adaptation may partly compensate for neuroplasticity deficits, it appears insufficient to fully support task-specific learning. These findings emphasize the importance of targeting neuroplasticity in the development of interventions, such as non-invasive brain stimulation techniques (TMS, tDCS) or pharmacological modulation of glutamate and dopamine systems, to help improve cognitive impairments in schizophrenia. These findings contribute to a growing understanding of the neural basis of implicit learning deficits in this population.