Automatic Thought Systems: A Multinational Network Analysis Study

Background and Objectives

The study presented by Ramos-Vera and colleagues (2026) addresses a fundamental topic in cognitive psychology and cognitive-behavioral therapy: the structure and nature of negative automatic thoughts. Automatic thoughts constitute a central component in cognitive models of depression and other psychological disorders. Although previous research has established the importance of these cognitive processes, a significant gap has existed in understanding how these thoughts interact and mutually reinforce one another within complex systems, especially in multicultural contexts.

The Automatic Thoughts Questionnaire (ATQ; Hollon & Kendall, 1980) has been widely used since its development to assess depressive negative thoughts. However, most research on psychometric validity has concentrated on Western samples, which raises questions about its cross-cultural applicability. The present study seeks to overcome these limitations through network analysis of the abbreviated 8-item version (ATQ-8) in a multinational sample.

The primary objective of the study is to represent gender-based network models within a multinational sample to enhance understanding of the interaction and clustering of automatic thought indicators, and to advance the cross-cultural validity of the ATQ-8. The researchers hypothesized that network analysis would allow identification of the most central and influential thoughts in different cultures and genders, revealing the systemic structure of these cognitive processes beyond what traditional factorial models could offer.

Method

Participants

The study included a total of 3,964 participants from six countries, with ages ranging from 18 to 72 years (M = 25.0; SD = 9.20; 63.8% female). The composition by country was as follows:

Saudi Arabia: 440 participants (M = 19.6; SD = 0.90; 70.9% female)
China: 955 participants (M = 19.9; SD = 1.21; 59.4% female)
Spain: 1,147 participants (M = 32.9; SD = 11.6; 74.5% female)
Georgia: 359 participants (M = 26.3; SD = 8.91; 61.3% female)
Turkey: 680 participants (M = 22.0; SD = 2.41; 48.2% female)
Peru: 383 participants (M = 24.7; SD = 9.19; 64.8% female)

The sample was predominantly university-based (n = 2,760; 69.62%), with 1,202 employed participants (30.38%). A non-probabilistic convenience sampling method was employed based on specific eligibility criteria: being of legal age, having completed secondary education, residing in one of the participating countries, completely filling out assessment instruments, and providing informed consent.

Instruments

The primary instrument was the Automatic Thoughts Questionnaire abbreviated version (ATQ-8; Netemeyer et al., 2002). This version consists of 8 items developed as a more practical alternative to the original 30-item ATQ while maintaining psychometric validity. The items include:

ATQ1: "I'm no good"
ATQ2: "I'm so disappointed in myself"
ATQ3: "What's wrong with me?"
ATQ4: "I'm worthless"
ATQ5: "I feel so helpless"
ATQ6: "Something has to change"
ATQ7: "My future is bleak"
ATQ8: "I can't finish anything"

Each item is rated on a 5-point Likert scale (1 = Strongly disagree to 5 = Strongly agree), with total scores ranging from 8 to 40. Researchers culturally adapted the items in each participating country and verified adequate psychometric properties in each national context.

Procedure

Data were obtained from researchers who had previously used the ATQ in various contexts. Data published from 2020 onward were included, and authors who responded positively to requests were contacted. The Peruvian sample consisted of university students from a private university. The study was ethically approved by the Chair of Medical Psychology of the Faculty of Medical Sciences of the National University of Asunción, Paraguay. Informed consent was obtained from all participants in compliance with the Declaration of Helsinki.

Data Analysis

All analyses were conducted using R software version 4.1.3 with specialized packages (qgraph, bootnet, mgm, psych, networktools). Analysis was structured in several stages:

First, descriptive statistics were obtained for each ATQ-8 item in all countries, including mean, standard deviation, skewness, and kurtosis. Normality criteria were applied by establishing an acceptable range of ±1.5 for skewness and kurtosis.

Second, the validity of the unidimensional internal structure of the ATQ-8 was evaluated through confirmatory factor analysis (CFA) using the WLSMV estimator (Weighted Least Squares Mean and Variance adjusted). Good fit criteria were established as: χ²/df < 4, CFI > 0.90, TLI > 0.90, RMSEA < 0.08, and SRMR < 0.080. Factor loadings (λ) were required to exceed 0.30.

Reliability was evaluated using ordinal alpha and McDonald's omega, with values above 0.70 considered satisfactory.

Third, to analyze network structure, a Gaussian Graphical Model (GGM) was used, representing the network based on regularized partial correlations. Each item was represented as a node, and connections between them (edges) represent conditional dependency relationships after controlling for all other associations in the network. The network was estimated using the graphical LASSO method with extended Bayesian information criterion.

Network stability was evaluated using the correlation stability coefficient (CS), with minimum acceptable values of 0.25, although values above 0.50 are recommended. Predictability indices (R²) were calculated to estimate the proportion of variance explained in each node.

A nonparametric resampling procedure with 1,500 resamples was applied to evaluate the precision of edge weight estimates. Finally, the most central and statistically significant nodes of the network were identified.

Network structures were compared across countries and genders using the NetworkComparisonTest package, with 500 iterations and Benjamini-Hochberg correction for multiple comparisons. Comparisons focused on edge weights, node strength, and bridge connectivity.

Results

Factorial Validity

Results confirmed the unidimensional structure of the ATQ-8 in all six participating countries with satisfactory fit indices. Saudi Arabia: χ² = 96.35, df = 20, CFI = 0.983, TLI = 0.977, RMSEA = 0.062, SRMR = 0.048. China: χ² = 432.65, df = 20, CFI = 0.963, TLI = 0.948, RMSEA = 0.091, SRMR = 0.055. Spain: χ² = 280.03, df = 20, CFI = 0.976, TLI = 0.967, RMSEA = 0.073, SRMR = 0.048. Georgia: χ² = 86.08, df = 20, CFI = 0.970, TLI = 0.958, RMSEA = 0.096, SRMR = 0.051. Turkey: χ² = 62.97, df = 20, CFI = 0.991, TLI = 0.987, RMSEA = 0.056, SRMR = 0.033. Peru: χ² = 67.94, df = 20, CFI = 0.987, TLI = 0.982, RMSEA = 0.079, SRMR = 0.041.

All countries demonstrated satisfactory reliability with ordinal alpha ≥ 0.90 and omega ≥ 0.87. Factor loadings in all countries exceeded the minimum threshold of 0.30, demonstrating that each item contributes adequately to the underlying latent factor.

Network Analysis

Network analysis revealed that automatic thoughts do not operate in isolation but form a highly interconnected system. The strongest connections in the general network were observed between ATQ4 ("I'm worthless") and ATQ7 ("My future is bleak"; r = 0.29), as well as between ATQ3 ("What's wrong with me?") and ATQ6 ("Something has to change"; r = 0.26).

ATQ4 showed the highest strength centrality (S = 1.31), indicating that it is the most central and influential thought in the network. In contrast, ATQ6 presented the lowest value (-1.42), suggesting it has lower connection and influence within the automatic thought system. Average predictability across the network was 43.8%, with ATQ4 being the most predictable (53.1%), indicating that its variations can be predicted more precisely by its neighboring thoughts in the network.

Network stability was excellent (CS = 0.75), far exceeding the recommended threshold of 0.50. Edge precision was confirmed through resampling, with confidence intervals excluding zero in most cases.

Gender Differences

Gender analysis revealed distinct patterns in network structure between men and women. The connection between ATQ4 and ATQ7 was significantly stronger in women (r = 0.32) than in men (r = 0.25). The male network presented stronger unique associations between several item pairs, while the female network showed more robust general interconnection.

Women showed higher strength centrality for ATQ4, ATQ2 ("I'm so disappointed in myself"), and ATQ7, while in men, only ATQ4 showed high centrality value. This pattern suggests that in women, concerns about self-esteem, disappointment with oneself, and future hopelessness are more interconnected, reflecting greater cognitive rigidity. In men, the automatic thought system is more isolated and compartmentalized.

Connections exclusive to women included the association between ATQ5 ("I feel so helpless") and ATQ6 ("Something has to change"; r = 0.11). In men, unique associations existed between ATQ2-ATQ7 (r = 0.11) and ATQ1-ATQ5 (r = 0.15).

Country Variations

Network density varied significantly across countries, reflecting differences in the overall interconnection of automatic thoughts:

Saudi Arabia: 0.57 (16/28 edges)
China: 0.61 (17/28 edges)
Spain: 0.68 (19/28 edges)
Georgia: 0.57 (16/28 edges)
Turkey: 0.71 (20/28 edges)
Peru: 0.61 (17/28 edges)

The strongest correlations varied by country. In China, the strongest association was between ATQ4 and ATQ7 (ρ = 0.54), reflecting a particular connection between feelings of worthlessness and bleak future. In Georgia, the strongest connection was between ATQ4 and ATQ1 (ρ = 0.40), both related to negative self-esteem. In Saudi Arabia, ATQ5 and ATQ4 showed the strongest association (ρ = 0.36). In Spain and Peru, ATQ6 and ATQ3 were most strongly associated (ρ = 0.36 and 0.34 respectively).

Regarding strength centrality, ATQ4 was the most central thought in Georgia (S = 1.322), Peru (S = 1.666), and China (S = 1.407). ATQ5 showed highest centrality in Turkey (S = 1.319) and Saudi Arabia (S = 1.590). ATQ2 achieved highest centrality value in Saudi Arabia (S = 1.337) and Spain (S = 1.448).

A culturally significant finding was observed in China, where ATQ3 ("What's wrong with me?") had the lowest mean score, unlike all other countries where ATQ4 was lowest. This suggests that in Chinese culture, self-doubt may be a more socially acceptable form of negative self-reflection than absolute belief in one's worthlessness.

Discussion and Conclusions

Interpretation of Main Findings

The study confirmed that negative automatic thoughts do not function in isolation but are organized in highly interconnected systems operating according to network principles. ATQ4 ("I'm worthless") emerged as the most important central "hub" or node, being the most central, most predictable, and most influential on other automatic thoughts across all studied contexts.

The centrality and predictability of ATQ4 suggest it constitutes the core of negative thought patterns, reinforcing other negative thoughts in cascading fashion. Despite ATQ4 having the lowest response frequency (lowest mean score), its systemic role was predominant. This suggests that if a weighting system were applied, this item would deserve greater weight within the total ATQ score, given its structural impact on the entire automatic thought network.

The particularly strong connection between ATQ4 ("I'm worthless") and ATQ7 ("My future is bleak") indicates that feelings of worthlessness directly connect with bleak future perspectives, creating a reinforcing cognitive cycle that can maintain and exacerbate depressive symptoms. Similarly, the connection between ATQ3 ("What's wrong with me?") and ATQ6 ("Something has to change") suggests that concerns about self-esteem link with perception that change is required, although the mechanisms of how this translates into change behaviors require additional research.

Cross-Cultural Analysis

The factorial structure was consistent across six countries, demonstrating measurement invariance and cross-cultural validity. However, notable differences emerged in how these thoughts organize in different cultures. The general pattern of the automatic thought network was nearly identical across all nations, suggesting that symptom relationships are broadly consistent worldwide. This evidence of measurement invariance supports ATQ-8 applicability in multinational research.

Nevertheless, findings reveal significant cross-cultural differences. Specifically, in China, the dominant thought pattern was the association between worthlessness and bleak future (ATQ4-ATQ7), while self-doubt ("What's wrong with me?") showed the lowest mean value. This suggests that in Chinese culture, critical self-reflection on the self may be more socially acceptable than belief in absolute personal worthlessness. This finding has important implications for clinical case formulation and cross-cultural intervention design.

Previous research in Western populations has highlighted negative self-esteem as fundamental in depression. Present study findings extend this understanding, suggesting that while low self-esteem processes may be universal, their expression and cognitive organization may vary according to cultural contexts. Future research should identify cultural factors leading to automatic thought differences and attempt to evaluate these factors directly.

Gender Analysis

Gender analysis was particularly revealing. Women's networks demonstrated greater cognitive rigidity, with more interconnected and reinforcing thought patterns. Stronger connections in women between thoughts related to negative self-esteem (ATQ4, ATQ2, ATQ7) suggest that in women, these thoughts tend to activate and reinforce each other with greater likelihood.

In contrast, men's networks were less dense, with more isolated thoughts. The particular association between disappointment with oneself (ATQ2) and bleak future perspectives (ATQ7) was unique to men, suggesting a distinct cognitive route. In men, the relationship between feeling "ineffective" (ATQ1) and helplessness (ATQ5) was more prominent, which may reflect cognitive vulnerability toward incompetence and loss of control.

These gender-based findings on automatic thought network structure are novel in the literature. They suggest that therapeutic interventions could benefit from gender-sensitive approaches. For women, interventions targeting reduced negative self-esteem (ATQ4) could deactivate multiple connections in the network. For men, approaches addressing disappointment with oneself and future hopelessness, as well as competence sensations, could be particularly effective.

Practical Implications

For clinicians and therapists, these findings suggest ATQ4 ("I'm worthless") should be a priority in cognitive-behavioral interventions, given its central role as a network "hub." By weakening this core belief, multiple associated negative thoughts can be reduced in cascading fashion.

ATQ5 ("I feel so helpless") and ATQ7 ("My future is bleak") also emerged as highly predictable elements, suggesting that reducing hopelessness and improving sense of agency can help decrease emotional distress and future despair.

In contrast, ATQ6 ("Something has to change"), while theoretically important from a cognitive-behavioral therapy perspective, showed lowest network centrality. This suggests an interesting discrepancy: while cognitive-behavioral therapy typically emphasizes the need for change as a central intervention mechanism, in the real experience of automatic symptoms, perception of the need for change is less interconnected with other negative thoughts. This raises a reflective question: could cognitive-behavioral therapy tend to overestimate the importance of change relative to other interventions?

Interventions can be gender-sensitive. For women, approaches targeting self-esteem may be particularly effective, given multiple negative thoughts interconnect around this domain. For men, approaches addressing disappointment and incompetence may be more effective. Moreover, in Western societies, interventions targeting self-esteem may be more effective in reducing negative automatic thoughts, while in China, interventions may be more effective by focusing on self-doubt.

Limitations

The study presents several limitations to consider. First, the cross-sectional design does not allow inferring causality or direction of associations among automatic thoughts. Future research should incorporate longitudinal network analyses to examine temporality of these relationships.

Second, a non-probabilistic convenience sampling strategy was employed, limiting generalization of findings to populations of participating countries. Future research should employ probabilistic sampling to reduce potential bias and increase representativeness.

Third, although the ATQ-8 is a psychometrically robust instrument, the abbreviated version provides less detail than longer questionnaires. Future investigations should consider using scales such as the 30-item ATQ-R to capture more comprehensive understanding of automatic thoughts in clinical interventions. Additionally, exclusive reliance on self-reports may introduce response bias and prevents identification of atypical automatic thought features.

Fourth, RMSEA values in some countries (Turkey and Georgia) were slightly elevated, suggesting the synthesized model may deviate slightly from perfect fit and operate with some nuances in this sample. However, it is important to consider that no single fit index in isolation is sufficient for adequate interpretation.

Fifth, participant age ranges differed across countries. While some countries included only young adults, others encompassed broader adult age ranges. These differences could have influenced study findings, as previous research reported that younger individuals (ages 16-25) are less likely to experience negative automatic thoughts. Future studies should include samples with more comparable age ranges to corroborate or refute these findings.

Finally, findings regarding centrality within the total sample and across countries should be interpreted cautiously, as additional empirical evidence is needed to validate these results.

Significance and contribution

This study contributes significantly to the field of cognitive psychology and psychometrics through a multinational network analysis approach that reveals how negative automatic thoughts operate as interconnected systems rather than isolated processes. By demonstrating the cross-cultural validity of the ATQ-8 across six countries and the measurement invariance of the instrument, it supports its universal applicability in psychopathology research. Network analysis provides deeper understanding of the systemic structure of automatic thoughts, identifying central nodes and connectivity patterns that vary by gender and cultural context—information that enriches both theoretical knowledge about cognitive processes and gender-sensitive clinical case formulation in cognitive-behavioral therapy.