Validation of the Hungarian Version of the Cognitive Fusion Questionnaire (CFQ-7)

Cognitive fusion is a key aspect of psychological inflexibility in the framework of ACT. It can be described as the process by which thoughts about an event merge with the actual event (Assaz et al., 2023; Hayes et al., 1999). Cognitive fusion implies that people react to thoughts as if they were literal reality and thus may act in a way that is inconsistent with their actual environment and available choices (Gillanders et al., 2014a, b). In this sense, cognitive fusion represents the phenomenon by which individuals believe the literal meaning of their thoughts instead of viewing them as transient internal states (e.g., the thought “I am a loser”, is equivalent to the psychological experience of failure) (Greco et al., 2008, p. 20). When fused with their thoughts, individuals tend to respond to them as if they were absolute facts (“I am a loser’’), triggering experiential avoidance strategies (attempts to avoid, escape, modify or control the experience) (e.g.: not apply for a job) and turning these internal experiences more painful (Hayes et al., 2006).

Various studies have shown that higher levels of cognitive fusion are associated with higher levels of depression, anxiety, burnout, lower levels of quality of life and life satisfaction, and maladaptive coping strategies (e.g., experiential avoidance, frequency of automatic thoughts) (Fergus et al., 2012; Gillanders et al., 2014a, b; Shi et al., 2024; Zucchelli et al., 2020). Coyne and Wilson (2004) described how parents’ fusion with their negative thoughts (e.g. “I am a bad parent.”, “I cannot tolerate my child’s behavior.”) may lead to dysfunctional parenting behaviors such as anger outbursts, withdrawal, or overcontrol to avoid sadness and despair. Gillanders and his colleagues (2015) conducted a quantitative cross-sectional study among adults with various cancer diagnoses which proved that cognitive fusion was the strongest predictor of anxiety symptoms, while cancer- related cognitions and avoidant coping were the strongest predictors of depressive symptoms and quality of life.

Several scales and questionnaires have been developed to measure cognitive fusion and defusion. Most of them were designed to measure fusion in specific populations: the Believability of Anxious Feelings and Thoughts Questionnaire (BAFT; Herzberg et al., 2012) was developed for non-clinical undergraduates and anxious community samples; the Automatic Thoughts Questionnaire (ATQ; Hollon & Kendall, 1980) measures depressogenic thought frequency and is widely used in depression studies to assess the impact of cognitive therapy; ATQ was subsequently altered to assess the believability of depressogenic thoughts (ATQ-B; Zettle & Hayes, 1986); the Cognitive Fusion Questionnaire - Body Image (CFQ-BI; Lucena-Santos et al., 2017) assesses body image-related cognitive fusion; the Cognitive Fusion Questionnaire– Food Craving (CFQ-FC; Duarte et al., 2016) was designed to assess cognitive fusion with undesirable thoughts regarding food craving and urges to eat; the Stigmatizing Attitudes Believability Scale (Hayes et al., 2004) measures substance abuse therapists´ stigmatizing thoughts about their clients; the Avoidance and Fusion Questionnaire for Youth (AFQ-Y; Greco et al., 2008) was developed for children and adolescents; the AFQ-Y was recently adapted for nonclinical adults (Schmalz & Murrell, 2010) and for adults with anxiety disorders; Drexel Defusion Scale (DDS; Forman et al., 2012) measures ability to achieve psychological distance from 10 unpleasant internal experiences. There is a special available instrument includes cognitive fusion items as part of measurement of the psychological inflexibility construct, in which other ACT processes are also present: the Psychological Inflexibility in Pain Scale —Cognitive Fusion subscale (PIPS; Wicksell et al., 2008) measures psychological flexibility, including cognitive fusion, specifically related to pain.

Cognitive Fusion Questionnaire (CFQ; Gillanders et al., 2014a, b) was designed to assess fusion with cognition in general, rather than with particular forms of cognition (e.g., anxious thoughts), and was tested in a series of studies designed to generate evidence of validity and reliability for the measure involving over 1,800 people across diverse samples (e.g., community-dwelling healthy adults, people with multiple sclerosis, sample of different mental health difficulties, people with major depressive disorder, and caregivers of people with dementia). The CFQ started from a pool of 42 items and was progressively improved by reducing it to a 7-items final version (Gillanders et al., 2014a, b). The original studies provide good evidence of the CFQ’s unifactorial structure, reliability, temporal stability, validity, discriminant validity, and sensitivity to treatment effects. The seven-item scale (CFQ-7) has become the most widely used self-report instrument for assessing cognitive fusion in both clinical and research settings (Bolderston et al., 2019).

The CFQ-7 has been translated and validated in numerous languages, including Brazilian Portuguese (Lucena-Santos et al., 2017), Catalan (Solé et al., 2016), Chinese (Zhang et al., 2014), French (Dionne et al., 2016), German (China et al., 2018), Greek (Zacharia et al., 2021), Italian (Donati et al., 2021; Policardo et al., 2023), Japanese (Shima et al., 2016), Korean (Kim & Cho, 2015), Persian (Soltani et al., 2022), Spanish (Romero-Moreno et al., 2014; Ruiz et al., 2017) and Turkish (Kervancioğlu et al., 2023). Studies usually support the one-factor structure, but some studies achieve an adequate-fit of the one-factor structure with correlations between the items’ error terms (Kim & Cho, 2015; Lucena-Santos et al., 2017). Lucena-Santos and colleagues (2017) allow for correlations between items 1 and 2, and also items 2 and 3; while Kim and Cho (2015) place a correlation between items 6 and 7. The internal consistency and test-retest stability of a single-dimensional structure are usually adequate. Configural, scalar and metric invariance can also be confirmed on different samples (Gillanders et al., 2014a, b). In Latin-speaking countries Ruiz and colleagues (2017) confirmed gender invariance. In addition to gender invariance, Ruiz and colleagues (2017) found a lower cognitive fusion mean score in men sample of undergraduates, and participants without a psychiatric history also had lower scores. In Romero-Moreno and colleagues’ (2014) and Losada and colleagues’ (2006) suggested that clinical depressive symptoms are positively associated with scores on cognitive fusion, but only in the case of female dementia caregivers. Similar differences were found in a previous study in which female caregivers with scores close to clinical depressive symptoms presented higher frequencies of dysfunctional thoughts. According to Nolen-Hoeksema and Jackson (2001), rumination as a coping strategy is used more frequently by women than men. It can contribute to their greater tendency toward cognitive fusion. Ruiz and colleagues (2017) also supported the invariance of the measuring instrument in certain clinical conditions, but there is no information on whether the invariance of CFQ-7 can be established regardless of age and education levels. Thus, in the present study, we examined the invariance of the Cognitive Fusion Questionnaire-7 by gender, clinical condition, age and education.

Additionally, some studies have examined the external validity of the CFQ-7 in terms of psychological flexibility, mindfulness, quality of life, life satisfaction, anxiety, and depression (e.g. Dionne et al., 2016; Donati et al., 2021; Gillanders et al., 2014a, b; Kim & Cho, 2015). The phenomenon of cognitive fusion occurs more prevalent in states of anxiety and depression. However, it can also be observed in life situations characterized by low well-being or stress. Cognitive fusion appears during certain psychological challenges or in connection with certain clinical conditions (e.g. China et al., 2018; Donati et al., 2021; Kim & Cho, 2015). The second aim of this paper is to examine the external validity of the CFQ-7.

The present study focuses on validating the Hungarian version of the Cognitive Fusion Questionnaire in a community sample. The study examined (a) the factor structure of the Hungarian version of the CFQ-7, (b) respectively the measurement invariance across age, gender, education, and psychological treatment history of the participants (c) evaluate its psychometric properties (external validity) in relation to other constructs, such as psychological inflexibility, general tendency to suppress thoughts, satisfaction with life and distress.

The sample consisted of Hungarian-speaking participants recruited from community samples via email and social media. The present study was part of a larger online investigation about psychological flexibility and its elements, such as mindfulness, experiential avoidance and value-driven life (The ethics license was approved by the Károli Gáspár University of Reformed Church, ethical permission number: 369/2016/2/P and BTK-PI/8323-1/2024). After explaining the purpose of the study (i.e. to better understand their personal experiences), the participants provided informed consent, understanding that participation was voluntary and anonymous. A total of 1231 participants completed the questionnaires with a men age of 29.35 years old (SD = 11.10 years, minimum 18, maximum 60 years). The majority of respondents were women (n = 805, 75.56%), high school graduates (n = 714, 58.0%). The respondents were mainly single (n = 494, 41.1%) or in a relationship (n = 324, 26,3%). Most of the participants did not have children (n = 892, 72.5%). 77.7 (n = 956) percent of the respondents had no previous psychological treatment history, and 22.3 (n = 275) percent of the respondents had. We examined the temporal stability of the CFQ-7 over a 1-month interval with part of the sample. This subsample consisted mainly of women (n = 152, 91.6%), married people (n = 59, 35.5%), and people with secondary school degrees (n = 103, 62.0%), and people who had no children (n = 97, 58.4%). The average age was 30.57 years (SD = 9.93 years), and a significant percentage of this subsample had previous psychological treatment history (n = 54, 32.5%) (see detailed information - Table 1).

First, we tested the factor structure of the CFQ-7 using confirmatory factor analysis with the robust maximum likelihood (MLR) estimator because violations of univariate and multivariate normality were observed. Considering the excellent sample size of over 1000 people, we used the following fit indices during confirmatory factor analysis (Kyriazos, 2018). The χ2-test compares the covariance matrix of the factor model and the covariance matrix of the observed model. In case of a large sample, the probability of Type II error increases at this index, so it is difficult to find a properly fitting model. Thus, several authors recommend its distribution with degrees of freedom (χ2/df), the value of which is accepted below 5. In case of larger samples, the use of the Comparative Fit and Tucker-Lewis indices is also an excellent choice, the values of which are considered acceptable above 0.900 and excellent above 0.950. Comparative Fit Index (CFI) evaluates the fit of the observed model on a continuum between the worst-fitting null model and the perfectly fitting saturated model. The Tucker-Lewis Index (TLI) is also a comparative goodness-of-fit index that shows how well the covariance matrix of the observed variables fits the hypothetical factor model. Among the error indicators, we considered SRMR and RMSEA, which calculate the difference between the observed correlations and the correlations assumed by the factor model. The smaller their value, the better the model fits the assumed model, so in the case of SRMR and RMSEA, a value below 0.080 is acceptable, and in the case of RMSEA, a value below 0.050 is excellent (Browne & Cudeck, 1992; Hu & Bentler, 1998; Schreiber et al., 2006).

To check for factorial invariance, multigroup confirmatory factor analysis (MCFA) was conducted across age groups, sex (male/female), education (secondary school/university degree), and psychological treatment history (yes/no) in the entire sample. Multigroup factor analysis is suitable for analyzing whether the structural characteristics of the measuring instrument are the same in the compared groups and whether it is not distorted by group-specific characteristics (Gregorich, 2006). Three nested models were adopted: (1) a configural model in which all factor parameters were freely estimated; (2) a weak factorial invariance model (metric), in which item loadings were constrained to be equal across groups; and (3) a scalar invariance model, in which the equality of thresholds or intercepts of the items were constrained to be equal across groups. We took into account Kline ‘s(2016) recommendation that at least 100 people per group be compared during multigroup confirmatory factor analysis. Successively restrictive models were compared with the Satorra-Bentler scaled χ2 difference tests using scaling correction factors and with CFI, RMSEA, and SRMR difference scores, the latter serving as indices for final decisions. As the sample size was large, changes of ≤ −.010 for CFI, ≤.015 for RMSEA and ≤.030 / ≤.010 for SRMR (depending on the level of testing) showed invariance (Chen, 2007). Finally, the external validity and test-retest reliability were tested with a Kendall-tau correlation. Taking into account Cicchetti’s (1994) criteria, the test-retest reliability between 0.4 and 0.59 is fair, between 0.60 and 0.74 is good, and above 0.75 is excellent. For the procedures used, the significance level was set at ≤.005. Analyses were performed using SPSS 29.0 and Mplus (Muthén & Muthén, 2017).

The unidimensional structure of the CFQ-7 was tested using CFA and showed an inadequate fit of RMSEA (χ2 = 240.85, df = 14, p <.001, χ2/df = 17.20, CFI = 0.941, TLI = 0.911, RMSEA = 0.115 [90% CI 0.102, 0.128], SRMR = 0.036). Modification indices were checked for suggestions to improve the model: error covariance was added for items 1 (“My thoughts cause me distress or emotional pain.”) and 2 (“I get so caught up in my thoughts that I am unable to do the things that I most want to do.”) (with MI = 98.45), and for items 4 (“I struggle with my thoughts”) and 5 (“I get upset with myself for having certain thoughts”) (with MI = 85.85). Each item focuses on negative events that can be linked to thoughts (struggle, sadness, helplessness). After implementing these modifications, an adequate fit was found (χ2 = 64.24, df = 12, p <.001, χ2/df = 5.35, CFI = 0.985, TLI = 0.974, RMSEA = 0.062 [90% CI 0.048, 0.077], SRMR = 0.020.) (see Table 2). In addition, no standardized factor loading was below 0.695, indicating a proper fit of individual items. The internal consistency (Cronbach’s alpha) was 0.918. We also examined the temporal stability of the instrument over a month interval. The test-retest reliability of the CFQ-7 was r =.665.

Configural, metric and scalar invariance were tested among different age groups (18–25, 26–40, 41–60 years), gender, educational level and psychological treatment history. In all multi-group analyses based on Chen’s (2007) criteria, configural and metric invariance held (see Table 3). The change in CFI was between − 0.001 and − 0.011, the change in RMSEA was between − 0.004 and 0.013, and the change in SRMR was between 0.016 and − 0.008. The fit indices are also in the desired range (CFI between 0.976 and 0.986; RMSEA between 0.049 and 0.063; SRMR between 0.021 and 0.052). These results indicate that not only are all factor loadings seem to be equal across groups, but also item intercepts and residuals as well, showing that psychological meaning of the items, and the concept of cognitive fusion are the same regardless of the age, gender, education level and psychological treatment history of the participants. We also calculated the group means and standard deviations of the CFQ-7 according to gender, age, education level, and psychiatric treatment history. We found that cognitive fusion decreases with advancing age (F(2, 1231) = 3.19, p <.001, η² = 0.082), higher education level (F(2, 1231) = 32.10, p <.001, η² = 0.044); and that those treated with a psychiatric diagnosis show a significantly higher value compared to those not treated (F(2, 1231) = 55.99, p <.001, η² = 0.044). The internal consistency of the questionnaire was good in all studied groups (between 0.891 and 0.924). In the entire sample, the measuring instrument had a Cronbach’s alpha value of 0.918 (see Table 4).

As shown in Table 5, the correlation between the CFQ-7 total score and the acceptance and action (AAQ-II) and suppression (WBSI) total scores was strong (r =.79 and r =.62). It shows a moderate relationship with the other constructs measuring external validity: general psychological distress scale of DASS-21 (r =.59), subscale of depression (r =.57), anxiety (r =.59) and stress (r =.61); and also satisfaction with life (SWLS, r = −.38).

Our study has several limitations, the consideration of which may allow for a more accurate research design in the future. An important disadvantage of the study is the self-reported questionnaire form, since in this situation we could not control the environment in which the questionnaire was filled out. Thus, the quality of the responses can depend on how much time the individuals take to think about the questions, their level of self-awarness, in what state of mind, or how undisturbed they answer them. The scores of both the total sample and the compared subsamples were also relatively low. Out of the possible 49 points, the subsample with a psychological treatment history scored only 25.94 points, which means that the questions related to cognitive fusion mostly characterize them moderately. It would be more balanced in terms of CFQ-7 if the sample included more men and older people. It would also be useful if not only those with a psychiatric treatment history but also those with a current psychiatric diagnosis were included in the validity and reliability analysis of the questionnaire, because the one-factor structure could have been created as an artifact of the sample selection. Stratified sampling rather than convenience sampling would improve the paper’s quality. It would also be advisable for future research to conduct cross-cultural studies to test the psychometric properties of the CFQ-7 in different cultural, socioeconomic and linguistic contexts. The factorial invariance of the CFQ-7 among these subgroups would further increase the validity of the questionnaire.

In summary, the one-factor structure of the CFQ-7 can be confirmed, and the retention of the 7 items may be justified from a therapeutic planning perspective, however, actual therapeutic experiences can only be obtained in cases where the measuring instrument is validated in a therapeutic setting and on a clinical sample.