Is Richard Baxter’s Melancholia Still Relevant Today?: Comparative Network Structures of Depressive Symptoms in Asian Individuals With Guilt-Rich and Guilt-Free Depression
Article information
, Junghyung Kim2, Jungsoo Park3, Kiwon Kim4, Eunkyung Kim5, Jaechang Lee1, Tae Young Choi6, Sung-Won Jung7, Hyung-Jun Yoon8, Hyun Soo Kim9, Eunsoo Moon10, Hyun-Ju Yang11, Narae Jeong12, Heechan Yoon13, Tian-Mei Si14, Roy Abraham Kallivayalil15, Andi J. Tanra16, Amir Hossein Jalali Nadoushan17, Kok Yoon Chee18, Afzal Javed19, Kang Sim20, Pornjira Pariwatcharakul21, Takahiro A. Kato22, Shih-Ku Lin23,24, Naotaka Shinfuku25, Norman Sartorius26, Seon-Cheol Park,27,28,29
Abstract
Objective
Historical theological frameworks, particularly those rooted in Christianity, have conceptualized guilt as a defining component of melancholy. This study compared the network structures of depressive symptoms in guilt-rich and guilt-free presentations of depression using a large-scale cross-national Asian dataset.
Methods
We analyzed data from the Research on Asian Psychotropic Prescription Patterns for Antidepressants, Phase 3 (REAP-AD3), which included patients with depressive disorders from 11 Asian countries. Network analysis was used to estimate symptom interactions and identify the central symptoms in each group. Participants were dichotomized according to the presence or absence of prominent guilt symptoms, as determined by the “feeling bad about yourself” item of the Patient Health Questionnaire-9 (PHQ-9).
Results
The network structures of the two groups differed substantially. In the guilt-rich group, depressed mood was the central node that was closely linked to self-blame. By contrast, the guilt-free group exhibited centralized networks with a loss of energy.
Conclusion
Variations in symptom network structures can be partly explained as broadly consonant with early ideas of Richard Baxter and Robert Burton on melancholia. Clinically, symptoms related to moral judgment and self-criticism may indicate potentially different specific therapeutic attention from individuals with significant guilt. Conversely, for those without pronounced guilt, interventions targeting physical and motivational symptoms such as fatigue might inform hypothesis-driven treatment strategies.
INTRODUCTION
In Western Christianity, guilt traditionally functions as a key moral guide influencing the self-evaluation and emotional responses of individuals through the lens of perceived divine scrutiny. These configurations of symptoms may indicate culturally distinct ways of internalizing moral obligation, largely influenced by religious teachings surrounding sin, personal responsibility, and the pursuit of redemption [1]. According to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) [2], the operational diagnostic criteria for major depression comprise: 1) at least one of the core symptoms of depressive mood and loss of interest or pleasure, 2) the presence of five or more depressive symptoms, and 3) experience of the presented symptoms for at least 2 weeks. Excessive feelings of guilt are a significant element in the diagnostic criteria for depressive disorders in both the DSM-5 and the International Classification of Diseases, 10th revision (ICD-10) [3]. Thus, guilt is not only an indicator of severe depressive episodes, but also a potential behavioral marker for certain types [4]. Brain imaging research indicates that moral emotions, including guilt, are linked to neural activation in areas such as the medial prefrontal cortex and anterior insula [5]. A neurobiological perspective suggests that guilt is associated with 5-HT1A receptors [6]. However, modeling guilt-like behavior in animal models remains a substantial challenge [7].
Christian cultural contexts [8] has historically emphasized that guilt is a diagnostic criterion for depressive disorder. Richard Baxter, a 17th-century Puritan theologian and clergyman, considered guilt both as a possible cause and as a symptom of melancholia. He recognized that excessive guilt could contribute to the onset of depressive symptoms, while also acknowledging that guilt itself may arise as a manifestation of the disorder. In “A Christian Dictionary,” [9] he attributed depressed states to religious anxiety, guilt about sin, fear of divine punishment, grief over worldly loss, and intellectual overexertion. He noted that individuals with melancholia experienced magnified feelings of guilt, danger, and helplessness. His account mainly focused on psychological symptoms, with less attention paid to physical symptoms. Baxter’s intellectual and theological orientations may have led him to focus more on the mind than the body in his discussions of melancholia. The common symptoms he noted were exaggerated fear, persistent gloom, frequent weeping, guilt over the inability to feel joy, self-blame, despair over salvation, identification with others’ suffering, belief in the uniqueness of one’s case, rejection of comfort, punitive conscience, loss of faith and religious habits, irritability, social withdrawal, and the inability to control one’s thoughts. Occasionally observed symptoms include ruminations, confused and circular thinking, inactivity and staying in bed, obsessive scruples and rituals, accusatory hallucinations, urges toward profanity or blasphemy, intense guilt over blasphemous thoughts, delusions of demonic possession, visual hallucinations, suicidal tendencies, mutism, loss of interest, and relapse or resistance to improvement. Baxter emphasized that distorted images of God—particularly viewing God solely as a judge rather than as a loving and forgiving presence—could exacerbate feelings of guilt and despair. As part of the cure, he encouraged individuals to correct their understanding of divine grace and forgiveness.
Baxter [9] portrayed melancholia as a state marked by profound guilt, hopelessness, and self-reproach shaped by religious concerns, attributing a central role to spiritual distress and maladaptive interpretations of divine judgment. By contrast, Burton [10] understood melancholia primarily in terms of fear, sadness, and bodily exhaustion, placing far less emphasis on guilt as a defining feature. Later in 1921, Emil Kraepelin recognized the significance of guilt in the symptoms of melancholia [11]. In 1928, Sigmund Freud theorized that guilt was a central mechanism in the development of melancholia [11]. Spiritual and religious aspects were minimally reflected in the diagnostic criteria of the DSM-III and DSM-IV, as these frameworks prioritized objective, biologically grounded approaches that tended to disregard subjective dimensions such as spirituality [12]. However, since the publication of symptom-based criteria in the DSM-III, excessive guilt has been considered as one of the nine criteria for major depression in operational diagnostic frameworks, including the DSM and ICD. The prevalence and expression of guilt can depend on the religious and cultural backgrounds of individuals with major depression [13]. Guilt has been reported more frequently among melancholic individuals from Western Christian societies than those from non-Christian cultures [14]. Americans with major depression exhibit higher levels of guilt and sadness, whereas Koreans with major depression exhibit higher levels of suicidal behaviors and health-related anxiety. A recent cohort study reported significant differences in the prevalence of guilt between non-Hispanic Caucasians and Hispanic individuals [15]. Guilt can arise from secular ethical systems, interpersonal norms, or culturally embedded standards independent of formal religious belief. Although Baxter’s account arose from a Christian context, guilt as a psychological experience is not confined to religion. Cross-cultural research shows that concerns about moral failure and social obligation appear across societies, including in Asian settings, where they are often expressed in interpersonal rather than theological terms [16]. Accordingly, Baxter’s description is treated as a historical example of a broader moral–emotional pattern that can vary by culture.
The considerable heterogeneity of depressive disorders challenges the foundational assumptions of the “disease essentialism paradigm,” which asserts that categorical, criteria-based diagnoses reflect a singular neurobiological core underlying psychiatric symptoms [17]. Wittgenstein [18] used the example of “games”—including board games, card games, and sports—to show that no single feature is shared by all, but we understand them as belonging to the same category due to their interconnected traits. Wittgenstein questioned whether clarity is always preferable, and suggested that concepts with blurred boundaries offer more appropriate representations. From this perspective, the various presentations of depressive disorders do not necessarily share an essential core, but are connected through partial symptom similarities, following Wittgenstein’s analogy to describe the diagnostic entity for depressive disorders. This conceptual shift has led to the development of network psychiatry, which employs computational methods to explore the complex interconnections between depressive disorder symptoms [19]. Rather than assuming a common underlying cause, network analysis focuses on how symptoms influence each other and collectively constitute a disorder [20]. Within such networks, centrality refers to the degree of a symptom’s connection with others. Symptoms with a higher centrality are likely to exert greater influence and may serve as the main drivers of symptom clusters, suggesting that they are therapeutic targets. This approach reframes depressive disorders as systems of interacting symptoms, moving away from the traditional model that views symptoms as mere reflections of an underlying disease [21]. Network analysis provides a bottom-up perspective that prioritizes symptom-level interactions [22]. This method allows researchers to identify symptoms central to the experience of depressive disorders [23]. Additionally, the prognosis and development of depression are significantly influenced by the connectivity and prominence of specific symptoms within the network [24].
The significance of guilt within the diagnostic framework of depressive disorders can be studied using network analysis. In this study, we compared the network structures of depressive symptoms in individuals with guilt-rich and guilt-free depressive disorders using data from the Research on Asian Psychotropic Prescription Patterns for Antidepressants, Phase 3 (REAP-AD3) study [25,26].
METHODS
Study overview and participants
As described elsewhere [25,26], the REAP is the largest international collaborative psychiatric study conducted in Asia. The REAP-AD3 was used to evaluate antidepressant use and clinical correlates among psychiatric patients in 11 Asian countries: China, India, Indonesia, Iran, Japan, Malaysia, Pakistan, Singapore, Korea, Taiwan, and Thailand. Convenience sampling was used to recruit psychiatric patients taking antidepressants between August 2023 and February 2024. The inclusion criteria were as follows: 1) confirmed diagnosis of any psychiatric disorder by the research center clinical psychiatrists according to the (ICD-10) [27], 2) use of at least one antidepressant, defined by the Anatomical Therapeutic Chemical classification system developed by the World Health Organization Collaborating Centre for Drug Statistics Methodology [28]. The exclusion criterion was medical or neurological illness severe enough to interfere with the study evaluations and interviews. Bipolar affective disorder (F31) was not included because the REAPAD3 protocol focused on unipolar depressive disorders. All Institutional Review Boards of the participating survey centers, including Hanyang University Guri Hospital, Guri, Republic of Korea (receipt number: 2023-05-021-006), approved the study protocols and provided written informed consent. Written informed consent for the use of their data for research purposes was obtained from all study participants. Regular consensus meetings were held before beginning the survey to ensure that data collection and entry methods remained consistent. The data collection center was located at the Taipei City Psychiatric Center, Taipei, Taiwan.
To estimate the network structures of depressive symptoms, excluding guilt, in patients with guilt-rich and guilt-free depressive disorders, the selection criteria were as follows: 1) any diagnosis of depressive disorders, including depressive episode (F32), recurrent depressive disorder (F33), and persistent depressive disorder (F34), according to the ICD-10; 2) completion of the Patient Health Questionnaire-9 (PHQ-9) [29]; and 3) age of 18–80 years. Ultimately, 2,488 participants, comprising 1,766 with guilt-rich depressive disorder and 722 with guilt-free depressive disorder, were included in this study.
Evaluation of depressive symptoms
The PHQ-9 is a self-report questionnaire with high reliability (α=0.89) and validity to evaluate nine depressive symptoms in respondents over the period of 2 weeks. The PHQ-9 scored the nine items on a 0–3 Likert scale, with higher scores indicating greater severity [29]. The PHQ-9 was formally translated into different Asian languages, and the translated versions were standardized [30]. The PHQ-9 was used to evaluate depressive symptoms and define guilt-rich and guilt-free depressive disorders. Guilt-rich depressive disorders were defined as scores ≥1 on a guilt item of the PHQ-9, and guilt-free depressive disorders were defined as scores of 0 on the guilt items of the PHQ-9. The abbreviations used for the PHQ-9 items are as follows: APE, poor appetite or overeating; CNC, trouble concentrating on things such as reading the newspaper or watching television; DEP, feeling down, depressed, or hopeless; ENG, feeling tired or having little energy; HUR, thoughts that you would be better off dead or hurting yourself in some way; INS, trouble falling or staying asleep, or sleeping too much; PLE, little interest or pleasure in doing things; SLW, moving or speaking so slowly that others could have noticed or being so fidgety or restless that you have been moving around much more than usual.
Statistical analyses
The network structures of the PHQ-9 items, excluding guilt, were estimated for patients with guilt-rich and guilt-free depressive disorders using the qgraph package in R [31]. All items were categorized as ordered categorical variables ranging from 0–3. Polychoric correlations were used to estimate the network structures [32], where the nodes and edges indicate symptoms and associations, respectively. All network structures were estimated to be unidirectional, because cross-sectional data cannot be used to make causal inferences. The Least Absolute Shrinkage and Selection Operator (LASSO) method was used to control for false-positive edges, where very small edges were set to exactly zero. Graphical LASSO (GLASSO) was used to estimate the mean edges to evaluate the relationship between the two symptoms while controlling for all other relationships within the network. Therefore, the edges represent partial correlation coefficients. The Bayesian information criterion [33] was used to minimize the shrinkage parameter and the underlying network structure was accurately recovered. Closely connected nodes were positioned together, and the network was visualized using the Fruchterman–Reingold algorithm [34]. A modularity-based community detection algorithm was used to cluster nodes within the network structure. The spin-glass algorithm was used to determine whether the number and strength of the edges within each cluster exceeded those within other clusters in the network communities. The spring-class community function from the igraph R package was applied to the GLASSO network (weights=null, vertex=null, parupdate=false, gamma=0.5, start temperature=1, stop temperature=0.01, cooling factor=0.99, and spins=17).
Using bootstrapping techniques, as recommended by Epskamp et al. [32], the stability and accuracy of the networks were evaluated to ensure robustness. The bootstrapped confidence intervals for edge weights demonstrated that relatively high precision was represented by stronger connections, whereas estimation uncertainty was indicated by weaker edges with wide confidence intervals. Case-dropping bootstrapping was used to determine the stability of the centrality indices, despite the removal of a substantial portion of the sample. The reliability of the centrality estimate was supported by a correlation stability (CS) coefficient for node strength centrality >0.25. Furthermore, a CS coefficient >0.5 was considered preferable [35]. The centrality indices of the depressive symptoms, excluding guilt, were estimated as follows: 1) the sum of all connections of a given node with all other nodes to denote node strength centrality, which is a stable and widely used central metric in network analysis [36]; 2) the shortest path connecting any two nodes to denote betweenness centrality; and 3) the close connection level of a symptom with all other symptoms in the network to denote closeness centrality. As node strength, betweenness, and closeness centralities were strongly correlated, the network structures of the PHQ-9 items, excluding guilt, in patients with guilt-rich depressive disorders and guilt-free depressive disorders were mainly based on node strength centrality.
Following the statistical approaches recommended by Fried et al. [37], the two network structures of the PHQ-9 items, excluding guilt, in patients with guilt-rich depressive disorders and those with guilt-free depressive disorders were compared as follows: 1) node strength centralities and edge weights of the two network structures were compared by Spearman’s rank correlations; 2) differences in network structure, edge strength, and global strength (two-tailed permutation test) were examined by the R package NetworkComparisonTest (NCT) [38]; and 3) a cross-sample network combining the two groups was estimated to understand the network structure characteristics of the entire sample. Statistical significance was set at p<0.05 (two-tailed). All statistical analyses were performed using R 4.3.3 (R Foundation for Statistical Computing).
RESULTS
General characteristics of patients with guilt-rich and guilt-free depressive disorders
As shown in Table 1, patients with guilt-rich depressive disorders were more likely to be older, female, outpatients, and South or West Asians than those with guilt-free depressive disorders. In addition, those with guilt-rich depressive disorders were more likely to use adjunctive antipsychotics and mood stabilizers and less likely to use anxiolytic and hypnotic drugs than those with guilt-free depressive disorders. As shown in Table 1, those with guilt-rich depressive disorders presented with a greater severity of all depressive symptoms, excluding guilt, than those with guilt-free depressive disorders.
General descriptions and frequency distributions of PHQ-9 items (excluding guilt) of patients with guilt-rich and guilt-free depressive disorders
Network structure of depressive symptoms among individuals with guilt-rich depressive disorders
Among the 1,766 individuals with guilt-rich depressive disorders, a network analysis of the PHQ-9 items, excluding guilt, revealed that 24 (85.7%) of the 28 possible edges were estimated to be >0 (Figure 1A). Several strong interconnections were identified in the network, including the following pairs of items: PLE-DEP (weight=0.456), DEP-HUR (0.294), CNCSLW (0.287), SLW-HUR (0.285), INS-APE (0.266), and INS-ENG (0.249). Edge statistics revealed preferred interpretability, with a CS coefficient of 0.750. As shown in Figure 1B, the node strength centrality of the PHQ-9 items excluding guilt indicated that DEP was the most central symptom, followed by PLE and ENG. By contrast, APE was the least interconnected symptom in the network, followed by HUR and INS. Node strength centrality revealed preferred interpretability, with a CS of 0.672. The community detection method demonstrated only one meaningful cluster, including the PHQ-9 items, excluding guilt.
Network analysis of depressive symptom profiles, excluding guilt, among individuals with guilt-rich depressive disorders. A: Network structure of depressive symptom profiles. B: Node-strength centrality. The green lines represent positive associations between connecting nodes. The line thickness reflects the edge strength. Nodes represent depressive symptom profiles evaluated using the Patient Health Questionnaire-9. APE, poor appetite or overeating; CNC, trouble concentrating on activities such as reading the newspaper or watching television; DEP, feeling disheartened, depressed, or hopeless; ENG, feeling tired or having little energy; HUR, thoughts that you should be dead or hurting yourself in some way; INS, trouble falling, staying asleep, or sleeping excessively; PLE, little interest or pleasure in activities; SLW, moving or speaking so slowly that others may have noticed or the opposite, being so fidgety or restless that you have been moving around more than usual.
Network structure of depressive symptoms among individuals with guilt-free depressive disorders
Among the 722 individuals with guilt-free depressive disorders, network analysis of the PHQ-9 items revealed that 27 (96.4%) of the 28 possible edges were >0 (Figure 2A). Several strong interconnections were identified in the network, including the following pairs of items: PLE-DEP (weight=0.443), SLW-HUR (0.385), DEP-HUR (0.314), ENG-APE (0.279), INS-ENG (0.278), and CNC-SLW (0.277). The edge statistics revealed preferred interpretability, with a CS of 0.672. As shown in Figure 2B, the node strength centrality of the PHQ-9 items indicated that ENG was the most central symptom, followed by DEP and HUR. By contrast, CNC was the least interconnected symptom in the network, followed by INS and APE. Node strength centrality revealed reliable interpretability, with a CS of 0.361. The community detection method demonstrated only one meaningful cluster, including the PHQ-9 items, excluding guilt.
Network analysis of depressive symptom profiles, excluding guilt, among individuals with guilt-free depressive disorders. A: Network structure of depressive symptom profiles. B: Node-strength centrality. The green lines represent positive associations between connecting nodes, whereas the red lines represent negative associations. The line thickness reflects the edge strength. Nodes represent depressive symptom profiles evaluated using the Patient Health Questionnaire-9. APE, poor appetite or overeating; CNC, trouble concentrating on activities such as reading the newspaper or watching television; DEP, feeling disheartened, depressed, or hopeless; ENG, feeling tired or having little energy; HUR, thoughts that you should be dead or hurting yourself in some way; INS, trouble falling, staying asleep, or sleeping excessively; PLE, little interest or pleasure in activities; SLW, moving or speaking so slowly that others may have noticed or the opposite, being so fidgety or restless that you have been moving around more than usual.
Comparing the network structures of depressive symptoms between individuals with guilt-rich and guilt-free depressive disorders
As shown in Table 2, a network comparison between guiltrich (Network 1) and guilt-free disorders (Network 2) revealed several key findings. NCT revealed significant differences in network structure invariance (p=0.01) and network structure invariance-equal sample sizes (p=0.01). However, the results for the global strength invariance (p=0.12) and global strength invariance with equal sample sizes (p=0.35) showed no significant differences. Individual edge-invariance tests showed that two of the total edges (7.1%) differed significantly between the two networks. The Spearman correlation coefficient for node strength centrality between Networks 1 and 2 was 0.48, indicating a moderate degree of similarity in the importance of individual symptoms. In addition, the correlation coefficient for the edges was 0.89, suggesting similar specific connections between symptoms. Only 1 (3.6%) rank-order difference was observed.
Pairwise comparison of the two networks of PHQ-9 items between patients with guilt-rich and guilt-free depressive disorders
Finally, the network structure of the PHQ-9 items, excluding guilt, was estimated using a combined sample of the two groups to visualize the similarities and differences in edge connections across the groups. As shown in Figure 3A, network analysis of depressive symptoms revealed that 26 of the 28 possible edges were >0. Several strong interconnections were identified in the network, including the following pairs of items: PLE-DEP (weight=0.454), SLW-HUR (0.307), DEP-HUR (0.304), CNC-SLW (0.296), and INS-ENG (0.252). The edge statistics revealed preferred interpretability, with a CS of 0.75. As shown in Figure 3B, the node strength centrality of the PHQ-9 items indicated that DEP was the most central symptom, followed by PLE and ENG. In contrast, INS was the least interconnected symptom in the network, followed by APE, and HUR. Node strength centrality revealed preferred interpretability, with a CS of 0.672. The community detection method demonstrated only one meaningful cluster, including the PHQ-9 items, excluding guilt.
Network analysis of depressive symptom profiles, excluding guilt, in the cross-sample data set combining both guilt rich and guilt free depressive disorder groups. A: Network structure of depressive symptom profiles. B: Node-strength centrality. The green lines represent positive associations between connecting nodes. The line thickness reflects the edge strength. Nodes represent depressive symptom profiles evaluated using the Patient Health Questionnaire-9. APE, poor appetite or overeating; CNC, trouble concentrating on activities such as reading the newspaper or watching television; DEP, feeling disheartened, depressed, or hopeless; ENG, feeling tired or having little energy; HUR, thoughts that you should be dead or hurting yourself in some way; INS, trouble falling, staying asleep, or sleeping excessively; PLE, little interest or pleasure in activities; SLW, moving or speaking so slowly that others may have noticed or the opposite, being so fidgety or restless that you have been moving around more than usual.
DISCUSSION
Our findings revealed that the network structures of depressive symptoms were significantly different between individuals with guilt-rich and guilt-free depressive disorders in a large, multi-country sample of Asian psychiatric patients, although the guilt-rich and guilt-free groups did not differ in global strength invariance. Specifically, depressed mood (DEP) was the most central symptom within the network structure of depressive symptoms in the guilt-rich group, whereas loss of energy (ENG) was most centrally situated within the network in the guilt-free group. These findings suggest that guilt alters the network architecture of depression, offering a new perspective for understanding the heterogeneity of depressive disorders.
These findings may reflect the differences between Baxter’s and Burton’s early understandings of melancholia [9,10]. As noted earlier, Baxter wrote that persistent sadness and despair, guilt and self-condemnation, fear of condemnation and death, religious doubt, and lack of belief were the main clinical features of melancholia. He regarded melancholia as a complex interplay of physiological vulnerability and spiritual stress. Thus, the identification of mood-centered networks of depressive symptoms in the guilt-rich group is partly consistent with Baxter’s idea of melancholia. Lundy [39] has argued that Baxter’s pastoral methods anticipate principles of modern cognitive behavioral therapy, especially in challenging distorted cognitions related to divine judgment. This may indicate that guilt reinforces the centrality of affective symptoms, such as depressed mood and suicidal ideation, creating tighter feedback loops. In contrast, Burton stated that loss of energy, loss of interest, sadness, despair, anxiety, obsession, and somatic symptoms were the main characteristics of melancholia, recognizing it to be rooted in both the mind and body. Thus, the energy loss-centered network of depressive symptoms in the guiltfree group is partly consistent with Burton’s idea of melancholia. In our study, although DEP (depressive mood) appeared to be the most central symptom within the guilt-rich group, its prominence was similarly observed in the combined network encompassing both groups. This pattern implies that depressive mood likely represents a fundamental aspect of depressive disorders in general rather than a feature specific to the guilt-rich subgroup. In addition, the present results align with earlier network-based research indicating that the relative importance of depressive symptoms is not uniform but varies according to clinical features and sociocultural settings. Prior studies have reported that affective symptoms [40], reduced energy [41], and suicidal thoughts [42] can assume a central position in symptom networks depending on the characteristics of the sample and contextual stressors. These findings support the view that depression lacks a single, stable hierarchy of symptoms and instead reflects configurations that are sensitive to contextual influences. Building on this work, our study highlights guilt as a possible organizing factor that may shape these symptom patterns within Asian populations. Moreover, cultural psychiatry has long observed that depression in many Asian populations is more often conveyed through somatic or psychomotor symptoms than through explicitly stated guilt or self-blame. These patterns are understood as culturally shaped expressions of distress, in which emotional suffering is communicated in bodily or energy-related terms [43]. Accordingly, the prominence of fatigue and behavioral symptoms in the guilt-free network may reflect differences in how moral distress is experienced and expressed, rather than its absence, within sociocultural settings that emphasize relational harmony and perseverance.
These differing symptom centralities can be supported by distinct underlying neurobiological mechanisms. Individuals with remitted major depressive disorder present with symptoms of guilt accompanied by abnormal functional magnetic resonance imaging connectivity between the right superior anterior temporal lobe and the subgenual cingulate cortex and adjacent septal region. These findings were observed as a recurrence risk in remitted major depression [42], suggesting that symptom targeted treatment in specific individuals with major depression is effective. Another case-control study also reported a similar finding, an association between feelings of guilt and temporo-fronto-limbic decoupling from functional imaging, suggesting vulnerability to major depression by case-control study [44]. Maladaptive coping style, represented as guilt, was associated with vulnerability to major depression [45]. In addition, previous studies have reported that guilt-related depressive symptoms are associated with altered activity in the anterior cingulate and medial frontal cortices, which are mainly involved in moral reasoning and self-referential processing [46]. However, energy-related depressive symptoms, including those in the ventral striatum and mesolimbic dopamine pathway, are associated with dysregulation of reward and motivation circuits [47]. The most central symptom within a network structure is a potential therapeutic target [24]. From a clinical perspective, these findings suggest that individuals with depression who exhibit feelings of guilt respond well to treatments focusing on cognitive distortions and maladaptive moral evaluations [44]. Patients with guilt-related depressive symptoms are also recommended to be vulnerable to relapse. Conversely, patients without guilt symptoms benefit more from interventions that address low energy and anhedonia, including behavioral activation and pharmacological treatments that enhance dopaminergic activity [48]. Future studies should use subjective reports related to guilt and objective tests (neuroimaging with cognitive tasks) to follow the treatment response.
This differentiation may offer preliminary implications for symptom-informed approaches to individualized care, which require confirmation in longitudinal and mechanistic studies. These findings align with philosophical critiques of categorical psychiatric diagnoses. The concept of “family resemblance” by Ludwig Wittgenstein suggests that categories such as depression may consist of overlapping symptom patterns rather than a shared essential core [18]. This notion supports the shift in network theory from categorical models to a dimensional understanding of psychopathology. Our results empirically underscore that different symptom networks may be valid forms of depressive disorder without necessitating a single latent construct.
Statistical tests of global strength invariance (p=0.12) and the adjusted test, accounting for equal group sizes (p=0.34), demonstrated no significant differences between individuals with guilt-rich and guilt-free depression. This outcome indicated that the total magnitude of the association between depressive symptoms was largely consistent between the two groups. Statistically, the overall architecture of the networks did not appear to differ significantly. From a clinical standpoint, this similarity suggests that the degree of symptom connectivity across depressive syndromes may not provide a basis for distinguishing between the subtypes. Although the presence or absence of guilt defines symptom profiles, the interconnectedness of depressive symptoms remains stable. As a result, broad treatment strategies targeting the overall network activation could potentially be used for both groups. However, the lack of global differences highlights the need to explore finer-grained network features such as highly connected symptoms or key symptom relationships, which may be clinically relevant for personalized interventions [49].
Our study had some limitations. First, the study design was cross-sectional, which limits inferences regarding the causality or temporal stability of the network structures. Longitudinal network analyses are required to explore how symptom centrality evolves over the course of illness or treatment. Second, the classification of guilt-rich and guilt-free depression was based on a single self-report item, which can be affected by cultural, linguistic, or interpretive differences across diverse Asian samples, suggesting the need for more comprehensive guilt assessment tools or multifaceted measurement approaches in future research. Third, while symptom network analysis offers an innovative perspective on psychopathology, it remains a correlational method and does not infer latent constructs or external influences, such as trauma history, medication use, or comorbidity. Fourth, because the initial study design did not account for religious factors, it was not possible to investigate how the network structure of depression varied based on religious affiliation. Fifth, although the network structure of the guilt-rich group was estimated by excluding guilt items, it was impossible to account for the extent to which this exclusion influenced the resulting network structure. Sixth, the narrow response range of the PHQ-9 (0–4) may have influenced the accuracy of the centrality estimates, as symptoms with lower average scores and limited variability may be less likely to demonstrate strong associations with other items. While such constraints could introduce bias due to the potential link between variability and centrality, the correlation between standard deviations and node strength was minimal (r=0.157), indicating that this effect was likely negligible [50].
Nonetheless, these findings raise the possibility of guilt-informed subtyping of individuals with depressive disorders. Differentiating depressive profiles based on symptom centrality, particularly guilt versus somatic-motivational features, may support more personalized treatment algorithms and encourage the refinement of diagnostic criteria beyond the current nosological conventions. Despite these limitations, the present study demonstrates that guilt-related depressive features are not only clinically significant, but also structurally influential within depressive symptom networks. In conclusion, our results suggest that the presence or absence of guilt alters the network structure of depressive symptoms, findings that are consistent with the historical conceptions of melancholia by Baxter and Burton and supported by recent neurobiological evidence. Overall, the results indicate that, while depressive mood consistently emerges as a central feature across various forms of depression, the symptom network specific to the guilt-free group, characterized by the prominence of loss of energy and suicidal ideation, may reveal distinct somatic and existential aspects of depression that are not shaped by moral self-criticism. This distinction highlights the value of symptom network analysis in elucidating diverse clinical profiles of depressive disorders.
Notes
Availability of Data and Material
The datasets generated or analyzed during the study are not publicly available because of institutional ownership but are available from the corresponding author upon reasonable request.
Conflicts of Interest
Seon-Cheol Park and Norman Sartorius, a contributing editor of the Psychiatry Investigation, were not involved in the editorial evaluation or decision to publish this article. All remaining authors have declared no conflicts of interest.
Author Contributions
Conceptualization: Seon-Cheol Park, Afzal Javed, Kang Sim, Shih-Ku Lin, Takahiro A. Kato, Naotaka Shinfuku, Norman Sartorius. Data curation: Seon-Cheol Park, Hong Seok Oh, Tae Young Choi, Sung-Won Jung, Hyung-Jun Yoon, Hyun Soo Kim, Eunsoo Moon, Hyun-Ju Yang, Narae Jeong, Heechan Yoon, Tian-Mei Si, Roy Abraham Kallivayalil, Andi J. Tanra, Amir Hossein Jalali Nadoushan, Kok Yoon Chee, Afzal Javed, Kang Sim, Pornjira Pariwatcharakul, Shih-Ku Lin, Takahiro A. Kato. Formal analysis: Seon-Cheol Park, Jung Hyung Kim, Kiwon Kim, Jeongsoo Park, Eunkyung Kim. Investigation: Seon-Cheol Park, Hong Seok Oh, Jaechang Lee. Resources: Seon-Cheol Park, Tian-Mei Si, Roy Abraham Kallivayalil, Andi J. Tanra, Amir Hossein Jalali Nadoushan, Kok Yoon Chee, Afzal Javed, Kang Sim, Pornjira Pariwatcharakul, Shih-Ku Lin, Takahiro A. Kato. Supervision: Afzal Javed, Kang Sim, Shih-Ku Lin, Takahiro A. Kato, Naotaka Shinfuku, Norman Sartorius. Validation: Seon-Cheol Park, Hong Seok Oh, Jaechang Lee. Visualization: Seon-Cheol Park, Hong Seok Oh, Jaechang Lee. Writing—original draft: Seon-Cheol Park, Hong Seok Oh. Writing—review & editing: all authors.
Funding Statement
This work was supported by Jisan Cultural Psychiatry Research Fund (2025) of the Korean Foundation of Neuropsychiatry Research.
Acknowledgments
We would like to thank Dr. Jinseob Kim for his valuable help with the network analysis.