Gender difference in rumination subtypes connecting to depression

Abstract

Purpose

The prevalence of depression is higher in women than in men. This may be because women are more prone to rumination. However, there is a lack of evidence about which rumination symptoms are most strongly associated with depression in women. According to the network theory of mental disorders, the complex interaction between different subtypes of rumination and depressive symptoms is confusing. We utilized the network analysis method to examine the connection between different subtypes of rumination and depressive symptoms across different gender groups and to investigate gender differences in rumination-depression networks.

Methods

798 Chinese university students (50.5% males and 49.5% females) completed The Rumination Response Scale (RRS) and The Beck Depression Inventory-Second Edition (BDI-II) scales via an online survey platform for this study. The networks were analyzed and built with scale dimensions as nodes.

Results

The line graph indicates that the B (Brooding) and N (Negative attitude) nodes had the highest BEI values in the female group, whereas the R (Reflection) and N nodes had the highest BEI values in the male group.

Conclusion

The findings revealed that males and females in the rumination-depression network had distinct rumination bridge nodes (male: Reflection; female: Brooding) but had a common depression bridge node (Negative attitude). The connection between different types of rumination and depressive symptoms was more pronounced in the female network. The findings enhance comprehension of gender disparities in the co-occurrence of rumination and depression, offering specific subtypes for targeted intervention in rumination.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00737-025-01569-y.

Highlights

Differences exist in the rumination subtypes connecting depression in males and females.

Female depression-related rumination is brooding, while male rumination is reflection.

Negative attitude was the depressive symptom most linked to rumination in male and female.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00737-025-01569-y.

Introduction

Depression is now a widespread and urgent mental health issue, contributing significantly to the global burden of disease (Monroe and Harkness 2022). According to the Global Burden of Disease (GBD) assessment report by the World Health Organization, it is projected that depression will become the leading cause of global disease burden by 2030 (WHO 2012). Current evidence suggests that there were 172 million cases in 1990 and 258 million cases in 2017, showing a 49.86% rise (Liu et al. 2020). Identifying risk factors like rumination may aid in early diagnosis and intervention of depression (Moulds and McEvoy 2025). Rumination is defined as the repetitive negative thinking process whereby an individual, following the occurrence of a negative life event, engages in contemplation regarding the underlying causes, circumstances, and consequences of that event (Nolen-Hoeksema 1991; Watkins 2008). It is regarded as a significant risk factor throughout the onset, persistence, and recovery phases of depression.

In accordance with the classification of depressive symptoms proposed by Osman et al. (1997) and Wu (2010), the depressive symptoms measured by the BDI can be divided into three components: negative attitudes, performance difficulties, and somatic elements. From the perspective of relevant theories on depression, these symptom variables can comprehensively reflect the level of depression (Ingram et al. 1983; LeMoult and Gotlib 2019; Monroe and Harkness 2022). Rumination also has different subtypes, which can be evaluated through the Ruminative Responses Scale. Of note, depressed-symptom rumination items on the scale appear to overlap with items on measures of depressive symptomatology (Roelofs et al. 2006; Treynor et al. 2003). Consequently, two distinct subtypes of rumination have been identified. Brooding is self-critical thinking that reflects a passive tendency to compare one’s current situation with some abstract or unachieved standard. Reflection is a cognitive tendency to actively attend to and analyze internal thoughts in a problem-solving-oriented way (Owens et al. 2021; Treynor et al. 2003).

The relationship between rumination and depression has long been a focus of researchers. Recent research continues to deepen the depth of exploration of the relationship. For example, a study used ecological transient assessment to explore the relationship and found that increased ruminative thoughts may represent early indicators of vulnerability to depression (Moretta and Messerotti Benvenuti 2022). Li et al. (2023) also discovered that heightened rumination was linked to higher total depression scores. However, these studies only identify a positive correlation between total rumination levels and total depression score levels, they don’t consider the relationship between different rumination subtypes and different depressive symptoms in accordance with the network theory of mental disorders (Moretta and Messerotti Benvenuti 2022; Li et al. 2023).

Currently, network theory and its associated analytical methods have been extensively applied in research within the mental health and psychological domains (Bringmann et al. 2022). According to the network theory of mental disorders, mental disorders arise from the causal interaction between symptoms in a network. In other words, instead of being effects of a common cause, psychiatric symptoms have been argued to cause each other (Borsboom 2017). Thus, there may be a complex interaction between different rumination subtypes and different depressive symptoms. Identifying this complex interaction effect contributes to a more precise intervention for depression. Of note, there is a statistically significant gender difference in the prevalence of depression, with a higher incidence among females than males (Potter and Yoon 2023; Sun et al. 2021; Zhou et al. 2023). Additionally, rumination is seen as an important factor moderating gender differences in the prevalence of depression (Li et al. 2023; Treynor et al. 2003). This suggests that there may be differences in rumination symptoms connecting to depression between males and females.

Previous studies have employed regression analysis and structural equation modeling to investigate the predictive relationship between rumination and depression (Moretta and Messerotti Benvenuti 2022; Li et al. 2023). These approaches are based on the traditional concept of the latent variable model, which posits that symptoms are the observed variable and mental illness is the latent variable. Symptoms are the observed manifestations of mental illness and are assumed to be independent of each other, thereby neglecting the interactions between different symptoms (Hofmann et al. 2016). Thus, these approaches are insufficient for attaining a comprehensive understanding of the concept of mental disorders. Conversely, based on data-driven concepts, the network analysis approach allows for a more integrated examination of the symptoms of mental disorders, as well as the interplay of the symptoms in relation to each other (Borsboom 2017). In this case, symptoms play an active role in activating and maintaining mental disorders rather than being passive indicators of mental disorders.

Therefore, the present study employed a network analysis approach to examine and visualize the complex interrelationships between the two subtypes of rumination (brooding and reflection) and depressive symptoms in male and female populations, which contributed to the identification of symptomatic targets for the prevention and treatment of rumination-depression comorbidity in patients of different genders.

Methods

Participants

A Chinese questionnaire online survey platform, wenjuanxing (https://www.wjx.cn/), was employed to gather data. Considering that college students are in a critical period characterized by a high incidence of mental disorders (Pedrelli et al. 2015), we selected them as the research subjects. From March 15th to April 15th, 2024, using the snowball sampling method, we recruited a total of 817 college students from three comprehensive universities in Shaanxi Province, China. Three universities were randomly sampled from Xi’an City, Shaanxi Province. Informed consent was obtained from all participants on the first page of the online questionnaire. All procedures in this study were in accordance with the Declaration of Helsinki and the study protocol was approved by Air Force Medical University Research Ethics Committee. A series of quality control measures were implemented, including restrictions on questionnaire submission, content filling, and the setting of attention check items. The questionnaire included one attention check item (“Please select the second option for this question”) to check whether the participants answered the question seriously. Following the exclusion of those who did not respond to the question in an appropriate manner, a total of 798 participants (mean age 21.40 ± 2.42) were recruited into the study, with a return rate of 97.7%. Among them, 403 were male (mean age 21.38 ± 2.36) and 395 were female (mean age 21.42 ± 2.49). The demographic characteristics of the participants are presented in Table 1.

Table 1.

Demographic characteristics of the participants

Gender	N	Age				Education years
		M	SD	t	p	M	SD	t	p
Male	403	21.38	2.36	-0.24	0.813	15.25	1.90	0.45	0.650
Female	395	21.42	2.49			15.18	1.89

Note. N: Number; M: Mean; SD: Standard Deviation

Measures

The Rumination Response Scale (RRS; Nolen-Hoeksema and Morrow 1991) was used to measure both subtypes of rumination. The scale consists of a total of 22 items, covering three components: brooding, reflection, and depressed-symptom rumination. Previous studies have demonstrated that this scale has good reliability and validity and is suitable for college students (Nolen-Hoeksema and Davis 1999; Treynor et al. 2003; Han and Yang 2009). The depressed-symptom rumination, which overlaps with the meaning of depression, includes 12 items and is not included in the research for analysis. The brooding subtype consisted of 5 items and the reflection subtype consisted of 5 items. Participants were asked to report on whether some self-report descriptions were in line with their own situations according to a four-point Likert scale, ranging from 1 (“never”) to 4 (“always”). Internal consistency in the current study was acceptable (Brooding: α = 0.85; Reflection: α = 0.80).

The Beck Depression Inventory-Second Edition (BDI-II; Beck et al. 1996) was used to measure different depressive symptoms. The inventory consists of 21 items assessing the severity of symptoms of depression and three components (negative attitudes, performance difficulties, and somatic elements). Each item has a four-point scale ranging from one to four. Each item describes a specific depressive symptom. Participants are required to rate each item based on their personal experiences, with “1” indicating no symptoms and “4” indicating very severe symptoms. Higher total scores reflect greater severity of depression. In this study, the internal consistency of the three dimensions was acceptable, with Cronbach’s alpha coefficients of 0.93, 0.91, and 0.86, respectively.

Data analysis

Descriptive statistics

The data was analyzed using SPSS 26.0, with basic demographic information and descriptive statistics being generated. Origin 2021 was used to create correlation heat maps. Based on the stated research objectives and the conditions of use for network analysis, this study utilized the mean scores of each dimension of the RRS (brooding, reflection) and BDI (negative attitudes, performance difficulties, somatic elements) scales as symptom nodes for analysis.

Network analysis

Using the R-packages “qgraph” and “bootnet” (Epskamp et al. 2018), the rumination-depression network for the male and female groups was constructed and visualized. Age was incorporated into the network to ascertain its influence on network characteristics. However, given that the age covariate was not the primary focus of the study and didn’t affect the characteristics of the target network, the covariate-network was not employed in this study (supplementary materials S10). The regularized partial correlation network consists of various nodes and edges. Each node signifies a distinct measurement symptom, and the edge connecting nodes indicates the relationship between the two nodes. Greater edge weight indicates a higher association between the two nodes. Greater thickness of edges signifies stronger edge weights. Blue edges represent a positive correlation between two nodes, while red edges represent a negative correlation. Regarding network construction, the graphical least absolute shrinkage and selection operator (graphical LASSO or glasso) was used to regularize the estimated network model (Friedman et al. 2008). The degree of regularization of the network model is determined by a tuning parameter, which is selected by minimizing the Extended Bayesian Information Criterion (EBIC) (Chen and Chen 2008). The current study utilized Spearman correlations as the foundation for network estimations, with the parameter value set to the default of 0.5 (Epskamp and Fried 2018).

The study selected Bridge Expected Influence (BEI) as the centrality index. Research has demonstrated that expected influence is a more precise way to describe the relationships between nodes in networks with negative edges compared to typical centrality metrics (Robinaugh et al. 2016). The importance of a node in a network is indicated by its expected influence (EI); the higher the EI, the more significant the node is in the network. Likewise, the bridge expected influence (1-step) can be expressed as the total of the values of all edges between node A and all other nodes that do not share a community. This is used to investigate the bridge symptoms that are crucial in linking two different symptom communities. The R-package “networktools” can calculate the BEI.

To ascertain the precision and stability of the network, we calculated nonparametric confidence intervals (CIs) for each edge in the network using bootstrapping with 2,000 samples. Small bootstrapped confidence intervals suggest that the predicted network was precise. Next, we performed a case-dropping bootstrap procedure with 2,000 bootstrapped samples to get the correlation stability (CS) coefficient for the expected influence of the bridge. Research indicates that CS coefficients beyond 0.25 are considered satisfactory (Epskamp et al. 2018). Finally, we conducted bootstrapped difference tests to compare two edge weights or two node bridge expected influences for significant differences. The “bootnet” package (Epskamp and Fried 2018) was used for the three operations.

Network comparison

The study classified participants into male and female subgroups according to their gender. A network comparison test (NCT) was employed utilizing the R-package “NetworkComparisonTest” (van Borkulo et al. 2016). As a resampling-based permutation test, the NCT divides the data into two groups at random after pooling the data (van Borkulo et al. 2022). This test enables the examination of differences between two networks in terms of their network topology and global network strength (van Borkulo et al. 2022). A total of 5,000 permutations were carried out for the current analysis.

Results

Descriptive statistics

The correlation matrix and correlation heat map between items and between dimensions are shown in the supplementary material (S1-S3). Table 2 presents the mean scores and standard deviations of all the variables for the male and female groups.

Table 2.

Mean scores and standard deviations for each variable selected in the rumination-depression network

	Male (N = 403)			Female (N = 395)
	M	SD	BEI	M	SD	BEI
Rumination
Brooding	2.30	0.73	0.07	2.46	0.71	0.28
Reflection	2.29	0.73	0.18	2.35	0.68	0.14
Depression
Negative Attitude	1.62	0.63	0.17	1.56	0.61	0.22
Performance Difficulty	1.65	0.65	0.03	1.60	0.62	0.11
Somatic Elements	1.77	0.69	0.04	1.81	0.68	0.09

Note. M: Mean; SD: Standard Deviation; BEI: Bridge Expected Influence

Network estimation

Both male and female network structures include five nodes, two of which are rumination and three of which are depression. In the male group network, 8 of 10 possible edges were non-zero (density = 80%). The mean weight was 0.185. In the female group network, 9 of 10 possible edges were non-zero (density = 90%). The mean weight was 0.179. Figure 1 illustrates the graphical representation of the male and female group networks, as well as the BEI centrality index for each node in both group networks. The BEI values for each node are shown in Table 2. The edge weight matrix is shown in Table 3.

Fig. 1.

Network structure comparison diagram of each subgroup. Note: Regularized partial correlation networks of the male group (left-top; N = 403) and female group (right-top; N = 395). In the network visualization, rumination subtypes are depicted in red, while depressive symptoms are shown in blue; Positive associations are indicated by blue lines, while negative associations are represented by red lines; The orange arrows indicate the presence of bridge nodes in disparate communities within the two networks; Comparable subgroup networks have the same network arrangement to compare their internal structural differences more clearly. The line chart (bottom) depicts the BEI centrality index for each node in the male and female networks, with scores being z-standardized. The two points with the highest BEI values in the subgroups are the bridge nodes in the rumination and depression communities, respectively

Table 3.

Edge weights in male and female networks

Variables	B	R	N	P	S
B	-	0.000[0.17]***	0.002[0.16]**	0.063[0.05]	0.265[0.01]
R	0.74(0.57)	-	0.006[0.11]**	0.533[0.03]	0.487[0.04]
N	0.07(0.22)	0.11(0.00)	-	0.022[0.12]*	0.401[0.06]
P	0.00(0.05)	0.03(0.06)	0.56(0.43)	-	0.603[0.03]
S	0.00(0.01)	0.04(0.08)	0.21(0.27)	0.56(0.54)	-

Note. B = Brooding, R = Reflection, N = Negative Attitude, P = Performance Difficulty, S = Somatic Elements. The numbers in the lower left corner of the table represent the edge weight values for the male and female groups, and the values in “()” represent those for the female group. The bold numbers in the upper right corner of the table represent the results of the edge invariance tests, where the values outside “[]” are p-values and those inside “[]” are standard error values. *p < 0.05, **p < 0.01, ***p < 0.001

Network comparison

All edges in Fig. 1 are blue, showing that the various rumination subtypes were positively associated with each depressive symptom in both male and female groups. The male and female networks’ graphical representations were compared and statistically analyzed for differences in edges using NCT. The male network exhibited the presence of the R-N edge, which had a significantly higher weight compared to the female network (p = 0.006). Conversely, the female network displayed the B-P edge and the B-S edge, with the B-P edge having a marginally higher weight than in the male network (p = 0.063), while the weight of the B-S edge did not significantly differ from that of the males (p = 0.265).

Only three edges, B-N edge, B-R edge, and N-P edge, show significant differences between male and female network structures. The weights of these edges are significantly higher in female networks compared to male networks (B-N: p = 0.002, B-R: p < 0.001, N-P: p = 0.022). None of the other edges exhibit significant differences between men and women, with p-values greater than 0.05. The network structure graph also shows that the B-N edge is thicker in the female network compared to the male network.

The line graph indicates that the B and N nodes had the highest BEI values in the female group, whereas the R and N nodes had the highest BEI values in the male group. The bridge node connecting female rumination is Brooding, while the bridge node connecting male rumination is Reflection. The bridge node in the depression community is Negative Attitude, which applies to both male and female groups.

Stability analysis

The accuracy and robustness of the current network are shown in Supplementary Materials. Figures S4 and S5 illustrate the bootstrapped 95% CIs for the male and female edge weights, respectively, and both figures have narrow bootstrapped CIs, which suggests that both networks are accurate. The CS coefficient for the BEI is 0.44 for the male network and 0.28 for the female network (shown in supplementary materials S6 and S7). According to the criteria of Epskamp et al. (2018), both male and female networks are considered acceptable and relatively stable. Results of the bootstrapped differences tests for BEI and edge weights are shown in the supplementary material (S8, S9).

Discussion

The study revealed a strong positive correlation between both rumination subtypes and all depressive symptoms, indicating a high positive correlation between rumination and depression, similar to previous research findings (Li et al. 2023; Watkins and Roberts 2020). In the male network, reflection and negative attitude were identified as bridge nodes connecting rumination and depression relationships. Conversely, in the female network, brooding and negative attitude served as bridge nodes connecting rumination and depressed connections, respectively.

The findings are significant for two reasons. First, the study identified differences in the rumination subtypes most significantly associated with depression in male and female groups by discriminating between various rumination subtypes. Brooding is a significant bridge factor in the rumination-depression network among females, while reflection serves as a bridge factor in the rumination-depression network among males. The results indicated that rumination, a significant risk factor for depression, varied between genders in terms of bridge-connecting symptoms, with brooding symptoms observed in females and reflection symptoms in males. This is consistent with the existing research results (Vergara-Lopez et al. 2024). Brooding involves a psychological process of self-criticism. Research indicates that self-critical personality can forecast the onset of depression, particularly in females who are considered a vulnerable category for self-critical personality (Kopala-Sibley et al. 2017). Therefore, compared to men, women’s tendency towards excessively self-critical thinking may be more likely to result in the development of depressive symptoms, which are considered the most crucial symptom associated with depression. Conversely, this study found that reflection is the key subtype linking depression in men, which is an innovative discovery. The possible reasons are as follows. Reflection is a highly analytical thought process focused on problem-solving rather than taking action. Although men exhibit a stronger tendency toward analytical thinking, this non-action-oriented resolution is a factor that can lead to the development of depressive symptoms, according to the social problem-solving model proposed by Bell and D’Zurilla (2009). An individual who engages in reflective rumination, characterized by over-analyzing a problem, may deplete excessive cognitive resources. This will impede the development of problem-solving strategies and undermine effective problem resolution, potentially leading to a vicious cycle characterized by depressive symptoms and rumination (Guo et al. 2024).

Second, in both male and female networks, rumination was most closely linked to the symptom node “negative attitude” rather than the depressive symptoms “performance difficulty” and “somatic elements”. The results align with previous research (Li et al. 2023) indicating that among all depressive symptoms, only the negative attitude symptom showed a strong correlation with rumination. Negative attitudes are typical symptoms of depression and include negative subjective experiences such as pessimism, feelings of failure, guilt feelings, punishment feelings, self-dislike, and worthlessness. Nolen-Hoeksema’s concept of rumination describes an individual’s repetitive abstract contemplation of negative events (Nolen-Hoeksema et al. 2008). This excessive abstract thinking leads to immersion in negative emotions and hinders the ability to shift attention away from these emotions, creating a cycle of bidirectional influence between analyzing negative events and generating negative emotions (Yang et al. 2017).

Additionally, it was discovered that the connections between brooding and reflection, as well as between negative attitude and performance difficulty, had considerably higher edge weight values in the female network compared to the male network. On the one hand, these findings indicate a stronger correlation between brooding and reflection and depressive symptoms in females compared to males. On the other hand, brooding acts as a crucial connecting point, indicating that the reflection rumination symptoms in women may worsen their brooding symptoms, thereby impacting depression more directly through brooding. This has also been supported by previous research results (Kim and Kang 2022). Intervening in brooding symptoms in females may successfully alleviate their negative attitude and performance difficulty symptoms simultaneously. In addition to the gender differences between the rumination-depression networks mentioned above, there are some commonalities between the two networks. For example, the relationship of somatic elements in the depression community to other depressive symptoms in the same community, as well as to the two rumination subtypes in the rumination community, was consistent across male and female groups. That is, the relationships of somatic elements with negative attitude, performance difficulty, brooding, and reflection did not differ between the male and female groups, suggesting that the somatization of depression induced by susceptibility symptoms is consistent in both males and females.

The study employed a network analysis method to investigate gender differences in rumination, depression, and their interconnectedness from a particular symptom-oriented viewpoint. The findings are conducive to promoting the application of the network theory of mental disorders in understanding the relationship between rumination and depression. They also identify specific targets for more accurate and efficient therapies for rumination and depressive symptoms. Nevertheless, the study does have limitations. Firstly, the cross-sectional data did not indicate the direction of the correlation or offer a clear explanation for the causal link between rumination and depression. Secondly, the network analysis approach relied on the specific questionnaire instrument utilized in the study, and employing various depressed symptom assessment instruments could have led to varied measurements. Thirdly, the demographic characteristics included in this study are limited. In future research, more demographic characteristics can be incorporated into the analysis when exploring the network relationship between rumination and depressive symptoms.

Conclusion

This study has identified distinct rumination bridge nodes for males and females: Reflection for males and Brooding for females. Both genders share the same depression bridge node, Negative Attitude. Targeted interventions on those identified core subtypes in different gender network patterns may be more effective in disrupting network structures and maximizing reductions in psychopathology.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization: Mingxuan Zou, Bin Liu, Xiuchao Wang, Hui Wang; Data acquisition: Mingxuan Zou, Jing Ji; Data analysis: Mingxuan Zou, Bin Liu, Lei Ren; Data interpretation: Mingxuan Zou, Bin Liu, Yuqing He, Huaihuai Wei, Mengxin Yin; Drafting: Bin Liu, Lei Ren; Revision: Xufeng Liu, Shengjun Wu, Jing Ji, Hui Wang, Xiuchao Wang.

Funding

This research was funded by the Air Force Military Medical University “Rapid Response” Project (2023KXKT060) and the Key Research and Development Program of Shaanxi (2024SF-YBXM-063).

Data availability

The datasets and R-codes used in this study are not publicly available but are available from the first author.

Declarations

Ethical statement

Informed consentwas obtained from all participants. All procedures in this study were in accordance with the Declaration of Helsinki and the study protocol was approved by Air Force Medical University Research Ethics Committee.

Competing interests

The authors declare that they have no conflict of interest.