The Černis Felt Sense of Anomaly (ČEFSA) Scale
Psychometric Properties and Validity of a German Version
Abstract
Abstract: Past research on dissociation has predominantly relied on broad questionnaires, presenting ambiguous and overlapping themes in describing dissociative experiences. A new measure, the Černis Felt Sense of Anomaly (ČEFSA) scale (Černis et al., 2021), assesses a subgroup of dissociative experiences where felt sense of anomaly is inherent. In this article, we present a German translation of the 35-item ČEFSA scale. In Study 1 (N = 130), we investigated the factor structure of the initial translation. We describe how necessary revisions were implemented. In Study 2 (N = 243), we show that the seven ČEFSA subscales had an appropriate factor structure and sufficient reliability. In terms of validity, the convergent, discriminant, and criterion correlations were mostly as expected. Some subscales showed high intercorrelations, making it difficult to distinguish them. We conclude that the German version of the ČEFSA scale provides a reliable and valid tool for the detailed assessment of dissociation.
Dissociation is a ubiquitous phenomenon defined as “disruption of and/or discontinuity in the normal integration of consciousness, memory, identity, emotion, perception, body representation, motor control, and behavior” (American Psychiatric Association, 2022, p. 329). Dissociation comprises multiple distinct experiences occurring on a continuum from mild forms that minimally affect activities to pathological pervasive forms that can significantly interfere with daily functioning (Holmes et al., 2005). Current questionnaires encompass a broad spectrum of experiences identified as dissociative, yet they present ambiguous, overlapping themes in their characterization of these phenomena. Addressing this shortcoming, Černis et al. (2021) developed the Černis Felt Sense of Anomaly (ČEFSA) scale, an English-language questionnaire that assess dissociative experiences with a common theme of felt sense of anomaly. These experiences have in common “a subjective sense of ‘strangeness’ or anomaly,” including feelings of (unexpected) unreality, unfamiliarity, automaticity, disconnection, or absence in relation to one’s body, mind, mood, perception, identity, behaviour, or external environment” (Černis et al., 2024, p. 2). For example, feeling disconnected from one’s body, or finding one’s surroundings somehow unfamiliar. The ČEFSA scale was developed to assess individual differences in these various domains and types of dissociative experiences among adults for research purposes (Černis et al., 2021). The ČEFSA scale distinguishes three domains of dissociative experiences: anomalous experiences of the self, body, and emotion, as well as four types of anomalies: altered senses of familiarity, connection, agency, and reality. This factor structure was established in the initial validation study and reproduced in adolescents and patients with psychosis (Černis et al., 2021). Thus, unlike current questionnaires, the ČEFSA scale distinguishes between various domains and types of dissociative experience sharing a common subjective sense of anomaly, enabling a more detailed exploration of the prevalence and characteristics of different dissociative experiences. In a recent revision of the measure, most of the items in the ČEFSA scale (25 out of 35) were invariant across age and gender, which is a prerequisite of reliable comparisons of dissociation scores and to rule out the possibility that potential age and gender differences are due to differences in item comprehension (e.g., Spitzer et al., 2003). The ČEFSA scale was revised using invariant items into a 14-item brief form (ČEFSA-14; Černis et al., 2024). A sum score-based scoring guide for the ČEFSA-14 is available, which assigns severity categories (average, elevated, moderately severe, severe) and is recommended for use in tandem with clinical judgment (Černis et al., 2024). A clinical validation against severity ratings based on structured clinical interviews is still pending.
The goal of this research project was to develop and validate a German version of the ČEFSA scale. To achieve this goal, we conducted two studies. In the first study, we translated the scale to German and performed necessary revisions based on factor analysis results. In the second study, we investigated the psychometric properties and validity of the revised German version of the scale.
We expected dissociative experiences across different subscales to show significant intercorrelations, given their shared underlying theme. Nevertheless, we reasoned that these subscales should remain distinguishable (r < .90). To investigate convergent validity, we investigated correlations with established measures of dissociation. For the English-language version, the correlation of the ČEFSA and Dissociative Experience Scale II (DES-II; Carlson & Putnam, 1993) total scores was estimated at r = .80 (Černis et al., 2021). To investigate discriminant validity, we used a personality assessment. We expected to observe non-zero correlations between ČEFSA subscale scores and personality because dissociation as measured by the DES-II has shown correlations with personality traits, particularly neuroticism (Kwapil et al., 2002). However, we expected these correlations to be modest (r < .50) and significantly smaller than correlations with the DES-II (convergent measure), suggesting that the ČEFSA subscales capture a significant degree of unique variance.
Dissociation has been described as a multidetermined construct linked to trauma or stress, psychological distress (e.g., anxiety, depression), sleep disturbance, fantasy proneness (tendency to engage in fantasy and imagination), alexithymia (difficulties in identifying, labeling, and elaboration on emotions), as well as cognitive and affective self-regulation deficits (Lynn et al., 2022). A meta-analysis focusing on various forms of childhood interpersonal maltreatment and trait dissociation revealed associations such as higher retrospectively assessed childhood abuse and neglect being linked to increased levels of recurrent dissociation later in life (Vonderlin et al., 2018). Another meta-analysis revealed that trait dissociation levels were highest in individuals who fulfill criteria of disorders associated with past trauma such as borderline personality disorder, posttraumatic stress disorder, and dissociative disorders (Lyssenko et al., 2018). In line with this, the English-language based ČEFSA total score demonstrated a large positive correlation with current posttraumatic stress disorder symptoms (r = .71; Černis et al., 2021) and a moderate positive correlation with negative affectivity (r = .43; Lofthouse et al., 2023). A study using a network approach found negative (residual) correlations between the ČEFSA total score and both general self-efficacy, which refers to an individual’s belief in their ability to accomplish specific goals or tasks, and alexithymia (Černis et al., 2022). The negative correlation with alexithymia was particularly unexpected, as earlier research had generally indicated a positive link between dissociation and alexithymia. In addition, one study using partial correlation suggested that dissociation relates to personality dysfunction beyond its shared variance with trauma (Schulze et al., 2024). Studies on the relationships between the ČEFSA and other influential variables such as fantasy proneness are currently missing. Based on the reviewed research, we expected higher scores of the German ČEFSA in individuals reporting traumatic events (e.g., physical assault, child abuse, severe accident). Additionally, we anticipated scores to positively correlate with depression, anxiety, personality dysfunction, alexithymia, and fantasy proneness, and negatively with self-efficacy.
Study 1: Initial Translation and Psychometrics
In Study 1, we translated the ČEFSA scale to German, investigated the factor structure of the translation, and suggested necessary revisions. Data are available at https://doi.org/10.5281/zenodo.13169632.
Method
Power Analysis
Sample size requirements were determined using Monte Carlo simulation in Mplus (Muthén & Muthén, 2002). Parameter estimates (e.g., factor loadings, correlations between factors, error variances) were based on simple confirmatory factor analysis (CFA) results of the original scale (Černis et al., 2021, third subsample). We specified that the most discrepant mean parameter estimate in the models must be within 10% of the population value. Estimated power had to be 80% or greater (with α = .05) for all factor loadings and residual variances, and the analysis must not yield any errors (i.e., improper solutions or failures to converge; as recommended by Wolf et al., 2013). The power analysis indicated that a sample size of 130 participants would be required. Statistical code and full results are available at https://doi.org/10.17605/OSF.IO/3FK94.
Sample and Procedure
A sample of 130 participants was recruited on Clickworkers and paid €1.00 as compensation. We included adult native German speakers who live in Germany, Austria, or Switzerland at the time of data collection, and who passed all our quality checks, including assessments of whether participants read the instructions and questions carefully (see Measures section for details). One hundred sixty-eight participants clicked on the study link, of whom 38 were rejected because they did not pass our quality checks.
Participants’ Mage was 44.99 years (SD = 13.03, range = 19–75), with 55 (42.3%) men and 74 (56.9%) women, and two identifying as other. Among the participants, 25 (19.2%) reported having a migration background, with themselves or at least one parent born outside of Germany, Austria, or Switzerland. A majority (102) had a high school diploma with college prep or equivalent, 23 had a secondary school certificate, three had a school-leaving certificate, one was a current student, and one provided no information on education.
Measures
Participants completed the German translations of the ČEFSA scale (Černis et al., 2021), which includes seven factors with 35 items in total (see the Supplemental Materials S1 and S2 for details). Due to a technical error, the responses for Item 15 are missing from the first 28 participants. The scales use a 0 (never) to 4 (always) answer format. In addition, participants were asked to rate whether they understood each item (yes/no). The translation was developed using a forward-backward approach involving two native German-speaking psychologists with advanced English language skills and expertise in the field of dissociation (JH, AF; Beaton et al., 2000). We asked a convenience sample of 18 individuals with borderline personality disorder and/or posttraumatic stress disorder, as well as 10 nonclinical individuals to rate whether each statement makes sense and to suggest changes. This feedback was instrumental in enhancing readability and ensuring that each item was as comprehensible as possible (see the Supplemental Material S3 for details).
Our quality check items comprised one instructional manipulation check (Oppenheimer et al., 2009). We included the following text in the ČEFSA scale instruction text: “remember the answer ‘blue’ to show that you have read this.” After completing the scale, participants were asked to “please choose the correct color to show that you have read the instructions on the previous page.” The answer options were “yellow,” “blue,” “green,” and “red.” In addition, we included one instructed response item shown in between regular ČEFSA items (“please select the agree response option to indicate that you have read this”; Ward & Meade, 2023). Participants who failed a quality check were not forwarded to the next page and excluded from all analyses.
Statistical Analysis
We ran confirmatory factor analysis (CFA) to investigate the factor structure of the initial German translation of the ČEFSA scale. Specifically, we calculated a model with seven (latent) factors that converge on a single second-order factor (Model A; Černis et al., 2021) and a model with a single first-order factor that includes all 35 indicators (Model B). In addition, we calculated a CFA for each of the seven ČEFSA subscales to identify potentially necessary revisions. We used the following cutoff values: CFI > .95, RMSEA < .06, SRMR < .08 (Hu & Bentler, 1999). All CFA models were estimated using maximum likelihood robust (MLR) estimator. In addition, we estimated results for Models A and B using the mean- and variance-adjusted weighted least squares (WLSMV) estimator, which is more appropriate for handling the strong floor effects present in our indicators (see Table 1). The analysis used Mplus version 8.11 (Muthén & Muthén, 1998–2024). Statistical code and full results are available at https://doi.org/10.17605/OSF.IO/3FK94.
Results
Factor Structure
Using the MLR estimator, the seven-factor solution converging on a single higher-order factor (Model A) resulted in an error message. This issue was attributed to the high intercorrelations among the first-order factors, indicating that adjustments are necessary. Additionally, model fit parameters indicated an inappropriate solution for the model with a single first-order factor that includes all 35 indicators (Model B), χ2(560, N = 130) = 1,450.31, p < .001, CFI = .69, RMSEA = .11, 95% CI [0.10, 0.12], SRMR = .08. Using the WLMSV estimator, model fit parameters indicated appropriate fit for Model A, χ2(553, N = 130) = 682.36, p < .001, CFI = .99, RMSEA = .04, 95% CI [0.03, 0.05], SRMR = .06, but not Model B, χ2(560, N = 130) = 912.24, p < .001, CFI = .97, RMSEA = .07, 95% CI [0.06, 0.08], SRMR = .08. The chi-squared difference test for the WLMSV estimator was significant, diff χ2(7, N = 130) = 150.61, p < .001, indicating that Model A fits the data significantly better than Model B. Model fit parameters for subscales are shown in Table 2. Results indicated adequate fits for all ČEFSA subscales except Anomalous Experience of the Self, Altered Sense of Familiarity, and Altered Sense of Agency, indicating that revisions are necessary in these subscales.
Reliability
As shown in Table 2, all latent (subscale) factors had acceptable reliability estimates (ω ≥ .86; cutoff: > .70). The median of the part-whole corrected item-total correlations in relation to the respective subscale was .74 (range: .48–.87), indicating sufficient correlation of individual items with the sum of the other items in the ČEFSA subscales (cutoff: .30, Streiner et al., 2015; see Supplemental Material S4 for item-level estimates).
Subscale Descriptives
Subscale Ms and SDs, as well as medians, skew, and kurtosis are displayed in Table 2. The median of the subscale means was 0.53 (range: 0.33–0.84) on a 0–4 scale. This shows that the ČEFSA subscales were strongly skewed towards zero, indicating that most participants reported few or no dissociative experiences in the past two weeks, particularly in the Anomalous Experience of the Body (Mdn = 0.00) and Altered Sense of Familiarity (Mdn = 0.00) subscales. The median percentage of participants who selected the “never” option (score of 0) across all items was 67%, with a range of 31% to 86% (see Supplemental Material S4 for item-level estimates). We show density plots of subscale score distributions at https://doi.org/10.17605/OSF.IO/3FK94.
The estimated median of the absolute values of the correlations for the revised subscales was .81 (range: .33–.84) with a median absolute deviation of .11 (see Supplemental Material S5 for details). For eight subscale correlations, the confidence intervals included .90, suggesting insufficient distinguishability between these subscale pairs. Specifically, the Altered Sense of Familiarity subscale showed such correlations with all other subscales, except for Anomalous Experience of Emotion.
Item Comprehension and Revision of the German Translation
The median percentage of participants who said they do not comprehend a statement across all items was 24%, with a range of 17%–31% (see Supplement Material S4 for item-level estimates). To improve the quality of the translated scale, we revised items for which modification indices indicate substantial error correlations in subscale CFA models (Item 6 and Item 31), as well as items with low item loadings and a high proportion of “0” responses (Item 29 and Item 35).
Study 2: Psychometric Properties and Validity
In Study 2, we investigated the psychometric properties of the revised German version of the ČEFSA scale. In addition, we examined the properties of a subset of this scale, the 14-item brief ČEFSA scale. Data are available at https://doi.org/10.5281/zenodo.13169632.
Method
Power Analysis
Sample size was determined following the procedure described in Study 1. For Study 2, we estimated a multi group solution for Model A. Parameter estimates, including factor loadings, intercepts, error variances, were based on the CFA results of the original scale (Černis et al., 2021).
Sample and Procedure
A sample of 243 participants was recruited on Clickworkers and paid €2.40 as compensation. We included adult native German speakers who live in Germany, Austria, or Switzerland at the time of data collection, and who passed all our quality checks, including assessments of whether participants read the instructions and questions carefully (see Measures section for details). Two hundred ninety-four participants clicked on the study link, of whom 51 were rejected because they did not pass our quality checks.
Participants’ Mage was 44.04 years (SD = 11.96, range = 20–73), with 127 (52.3%) men and 115 (47.3%) women, and one participant identifying as other. Among the participants, 35 (14.4%) reported having a migration background, with themselves or at least one parent born outside of Germany, Austria, or Switzerland. A majority (179) had a high school diploma with college prep or equivalent, 53 had a secondary school certificate, 10 had a school-leaving certificate, and one person provided no information on education.
Measures
Participants completed the revised German version of the ČEFSA scale. In addition, participants completed the following questionnaires in this order.
Dissociative Experience Scale (DES)
We used the German short version comprising 20 items each rated from 0% to 100% (Spitzer et al., 2004). Omega was estimated at .98, 95% CI [.97, 98].
Posttraumatic Diagnostic Scale for DSM-5 (PDS-5)
We used the trauma screening questions from the German version to assess trauma history across eight categories (e.g., physical assault, childhood abuse) using a binary answer format (Wittmann et al., 2021).
Generalized Anxiety Disorder Screener (GAD-7)
We used the German version comprising seven items assessing anxiety symptoms (Hinz et al., 2017). The questionnaire covered the past two weeks, with responses ranging from 0 (not at all) to 3 (nearly every day). Omega was estimated at .90, 95% CI [.87, 92].
Patient Health Questionnaire-9 (PHQ-9)
We used the German version comprising nine assessing depressive symptoms (Löwe et al., 2004). The questionnaire covered the past two weeks, with responses ranging from 0 (not at all) to 3 (nearly every day). Omega was estimated at .88, 95% CI [.85, 90].
Ten-Item Personality Inventory (TIPI-G)
We used the German version comprising 10 items and using a 1 (strongly disagree) to 7 (strongly agree) answer format (Muck et al., 2007). Three subscales had acceptable reliability estimates: conscientiousness (ω = .63, 95% CI [.51, .74]), agreeableness (ω = .67, 95% CI [.37, .77]), and neuroticism (ω = .69, 95% CI [.58, .78]). Two subscales had insufficient estimates: openness (ω = .43, 95% CI [.26, .56]) and extraversion (ω = .57, 95% CI [.39, .72). For simplicity, we reported the aggregated results in the main text. Item-level analyses are described in Supplemental Material S6.
DSM-5 Level of Personality Functioning Scale Brief Form (LPFS)
We used the German version comprising 12 items assessing self-reported severity of personality pathology (Spitzer et al., 2021). The two subscales had acceptable reliability estimates: self (ω = .86, 95% CI [.82, .88]) and interpersonal (ω = .82, 95% CI [.78, .85]).
Creative Experiences Questionnaire (CEQ)
We used the German version comprising 25 items assessing fantasy proneness using a binary answer format (Merckelbach et al., 2001). The CEQ has been translated to German by Michael Hengartner and Vladeta Ajdacic-Gross (personal communication with Harald Merckelbach, November 29, 2023). When the items were grouped into five separate parcels and then combined into a single factor to address the binary answer format of the indicators, the scale demonstrated a sufficient reliability estimate (ω = .98, 95% CI [.97, .99]).
Perth Alexithymia Questionnaire Short Form (PAQ-S)
We used the German version comprising six items and using a 1 (strongly disagree) to 7 (strongly agree) answer format (Kaemmerer et al., 2021; Preece et al., 2023). The German version of the PAQ-S has not yet undergone validation. In this study, we conducted a detailed analysis of its factor structure. Consistent with the findings of the original English short version (Preece et al., 2023), the fit indices for a single-factor model without correlated error terms indicated an inappropriate fit, χ2(9, N = 243) = 46.36, p < .001, CFI = .89, RMSEA = 0.13, 95% CI [0.10, 0.17], SRMR = .06. However, fit was significantly improved after introducing theoretically informed correlated error terms as outlined in the original publication, χ2(9, N = 243) = 20.86, p < .001, CFI = .96, RMSEA = 0.09, 95% CI [0.05, 0.14], SRMR = .03. Omega was estimated at .84, 95% CI [.81, 87].
General Self-Efficacy Short Scale (ASKU)
We used the German version assessing general self-efficacy, comprising three items each rated from 1 (not at all) to 5 (absolutely; Beierlein et al., 2017). Omega was estimated at .92, 95% CI [.90, .94].
The psychometric properties of all covariates were evaluated (see https://doi.org/10.17605/OSF.IO/3FK94 for details). The quality check items in the ČEFSA scale matched those used in Study 1. Additionally, we included an instructional manipulation check in the CEQ instructions: “remember the answer ‘yellow’ to show that you have read this.” After completing the CEQ, participants were asked to “please choose the correct color to show that you have read the instructions on the previous page” (options: “yellow,” “blue,” “green,” “red”). We also included an instructed response item among the regular CEQ items: “select the yes response option to indicate that you have read this.”
Statistical Analysis
We repeated the analysis conducted in Study 1 using the data set from Study 2. In addition, we investigated bivariate latent correlations between ČEFSA subscale scores with all covariates to determine convergent and discriminant validity, as well as to explore relations with influential variables (concurrent validity). Finally, we calculated multiple group CFA to test invariance across the translated and original versions. The analysis used Mplus version 8.11 (Muthén & Muthén, 1998–2024). Statistical code and full results are available at https://doi.org/10.17605/OSF.IO/3FK94.
Results
Factor Structure
Model fit parameters indicated inappropriate solutions for the revised model with seven (latent) factors that converge on a single second-order factor (Model A), χ2(553, N = 243) = 1,080.99, p < .001, CFI = .88, RMSEA = 0.06, 95% CI [0.06, 0.07], SRMR = .06, and for the model with a single first-order factor that includes all 35 indicators (Model B), χ2(560, N = 243) = 1,422.42, p < .001, CFI = .81, RMSEA = 0.08, 95% CI [0.07, 0.08], SRMR = .06. The Satorra–Bentler scaled chi-squared difference test for the MLR estimator (Satorra & Bentler, 2010) was significant, diff χ2(7, N = 243) = 167.08, p < .001, indicating that Model A fits the data significantly better than Model B. Consistent with this result, the estimates for the Bayesian Information Criterion (BIC) were 15,930.61 for Model A and 16,393.29 for Model B (ΔBIC = 462.68), further suggesting that Model A is a better overall fit. See Figure 1 for an illustration of Model A. Using the WLMSV estimator, model fit parameters indicated appropriate fit for Model A, χ2(553, N = 243) = 940.26, p < .001, CFI = .98, RMSEA = 0.05, 95% CI [0.05, 0.06], SRMR = .05, but not Model B, χ2(560, N = 243) = 1,197.62, p < .001, CFI = .96, RMSEA = 0.07, 95% CI [0.06, 0.07], SRMR = .06. The chi-squared difference test for the WLMSV estimator was significant, diff χ2(7, N = 243) = 170.93, p < .001, again indicating that Model A fits the data significantly better than Model B. Overall, we conclude that Model A is an appropriate fit to the data and is more adequate than Model B. Model fit parameters for subscales in Table 2 suggest appropriate fit for all ČEFSA subscales.
Reliability
Results in Table 2 demonstrate that all revised subscale factors had at least acceptable reliability (ω ≥ .85). The median of the part-whole corrected item-total correlations in relation to the respective subscale was .71 (range: .53–.86). Item-level estimates are shown in Supplemental Material S4.
Subscale Descriptives
Subscale Ms and SDs, as well as medians, skew, and kurtosis are displayed in Table 2. The median of the subscale means was 0.63 (range: 0.42–0.88) on a 0–4 scale. This shows that the ČEFSA subscales were strongly skewed towards zero, indicating that most participants reported few or no dissociative experiences in the past two weeks. The median percentage of participants who selected the “never” option (score of 0) across all items was 57%, with a range of 30% to 73% (see Supplemental Material S4 for item-level estimates). We show density plots of subscale score distributions at https://doi.org/10.17605/OSF.IO/3FK94.
The estimated median of the absolute values of the correlations for the revised subscales was .81 (range: .69–.98) with a median absolute deviation of .07 (see Supplemental Material S5 for details). For nine subscale correlations, the confidence intervals included .90, suggesting insufficient distinguishability between these subscale pairs. Specifically, the Altered Sense of Familiarity subscale showed such correlations with all other subscales, except for Anomalous Experience of Emotion and Altered Sense of Agency. The Altered Sense of Connection subscale showed such correlations with all other subscales, except for Anomalous Experience of the Self and Anomalous Experience of the Body.
Measurement Invariance
We evaluated the equivalence of the measurement model structure of the German version of the PMERQ against the original English version using multiple group analysis. Contrary to our initial plan, we were unable to compare the results with the sample from the original validation study due to restricted access to the data arising from data protection concerns (Černis et al., 2021). Instead, the sample for the English version was n = 2,354 (Mage = 20.61 years, SD = 2.97, range = 16–25) with 1,569 women and 2072 persons who identified as White (see Černis et al., 2024 for more details). All participants were living in the UK at the time of the study. Model fit parameters indicated inappropriate for Model A in the sample for the English version using the MLR estimator, χ2(553, N = 2,384) = 5,232.12, p < .001, CFI = .91, RMSEA = 0.06, 95% CI [0.06, 0.06], SRMR = .05, and the WLMSV estimator, χ2(553, N = 2,384) = 8,319.92, p < .001, CFI = .94, RMSEA = 0.08, 95% CI [0.08, 0.08], SRMR = .05. We proceeded with the measurement invariance testing despite minor deviations from our pre-determined cut-offs for adequate model fit. For instance, CFI values between .90 and .95 are often deemed acceptable by many researchers (Putnick & Bornstein, 2016). We focused on comparing the models to identify differences in the measurement structure between the English and German versions, rather than overall model fit, as fit indices were impacted by the inadequate solution of the English version.
As a first step, we modeled the same factor structure (Model A) in both versions. The model fit indices for configural invariance were χ2(1,076, N = 2,627) = 5,471.30, p < .001, CFI = .91, RMSEA = 0.06, 95% CI [0.05, 0.06], SRMR = .05. In the next step, by equating the unstandardized factor loadings, the model fit indices for metric invariance were χ2(1,113, N = 2,627) = 5,749.67, p < .001, CFI = .91, RMSEA = 0.06, 95% CI [0.05, 0.06], SRMR = .08. The decrease in CFI (ΔCFI = −.005) and RMSEA (ΔRMSEA = .000) between the models was below recommended cutoff criteria (ΔCFI ≥ −.010, ΔRMSEA ≥ .015), indicating invariant factor loadings across scales (Chen, 2007). Finally, by equating the unstandardized intercepts, the model fit indices for scalar invariance were χ2(1,148, N = 2,627) = 7,723.36, p < .001, CFI = .87, RMSEA = 0.07, 95% CI [0.04, 0.04], SRMR = .14. The decrease in the CFI (ΔCFI = −.039) was above recommended cutoff criteria, indicating a lack of scalar invariance. Measurement invariance test results for subscales are presented in Supplemental Material S7.
Modification indices indicated non-invariance in the intercepts of the 17 items (see Supplemental Material S8 for details). For example, the intercepts for Item 1 in Anomalous Experience of the Body (original = 2.57 vs. translation = 1.32) and Item 2 in Anomalous Experience of the Body (original = 2.53 vs. translation = 1.24) differed across the two versions (see Table 2 for the English and German items). Fit indices indicated appropriate fit when allowing the intercepts of the 17 items to vary, while assuming invariant intercepts for all other items (partial metric invariance), χ2(1,131, N = 2,627) = 6,268.49, p < .001, CFI = .90, RMSEA = 0.06, 95% CI [0.06, 0.06], SRMR = .11. The change in the CFI was below recommended cutoff criteria (ΔCFI ≥ −.010), indicating invariant intercepts across scales for 18/35 items.
Results from the measurement invariance tests for subscales are presented in Supplemental Material S8. While all subscales achieved metric invariance, four subscales – Anomalous Experience of the Self, Anomalous Experience of the Body, Anomalous Experience of Emotion, and Altered Sense of Connection – did not meet scalar invariance criteria between the original and translated versions.
Validity
Convergent
We examined the convergent validity between ČEFSA subscales and other measures of dissociation. However, since the ČEFSA scale assesses a specific subgroup of dissociative experiences, we did not anticipate very high correlations. In line with this expectation, results in Table 3 show that the ČEFSA subscales strongly (>.50) relate to dissociation assessed using the German short form of the Dissociative Experience Scale (DES). The median of the correlation estimates was .65 (range: .60–.84). We conclude the ČEFSA scale has sufficient convergent validity with dissociation assessed using the DES.
Discriminant
We examined the discriminant validity between the ČEFSA subscales and personality measures, which should be weakly to moderately related. As shown in Table 3, we found zero correlations between the ČEFSA subscales and extraversion (range: −.15 to .12), as well as openness (range: −.11 to .14). Additionally, we found mostly moderate negative correlations between the ČEFSA subscales and agreeableness (median: −.40, range: −.41 to −.08), as well as conscientiousness (median: −.34, range: −.28 to −.43), and moderate positive correlations between the ČEFSA subscales and neuroticism (median: .38, range: .27–.44). For most ČEFSA subscales, the correlations with the convergent measure (DES) were significantly stronger than those with personality facets, as evidenced by non-overlapping 95% confidence intervals of the correlation estimates. The ČEFSA subscales Altered Sense of Familiarity and Altered Sense of Connection showed correlations with agreeableness and neuroticism that were as strong as their correlations with DES scores. We conclude the ČEFSA scale has sufficient discriminant validity from the Big Five personality factors.
Concurrent
We examined the relation between the ČEFSA subscales and indicators of psychopathology, which we expected to be positive. Results shown in Table 4 indicate all ČEFSA subscales strongly relate to PHQ-9 depression scores (median: .60, range: .51–.66) and LPFS personality dysfunction scores (median for self: 0.67, range for self: 0.53–.73, median for interpersonal: 0.67, range for interpersonal: 0.57–.72). Additionally, we found moderate to strong negative correlations with GAD-7 anxiety scores (Mdn: .52, range: .43–.58) and weak positive correlations with PAQ-S alexithymia scores (Mdn: −.21, range: −.25 to −.12). As expected, participants who reported specific traumatic events assessed using the PDS-5 had higher scores on several ČEFSA subscales. Results in Table 1 indicate that participants who reported experiencing a serious accident or childhood abuse had significantly higher scores on all ČEFSA subscales, except for Anomalous Experience of the Self. Additionally, other traumatic experiences were significantly related to ČEFSA subscale scores. Notably, significant relationships were found between reporting sexual assault and higher scores in Altered Sense of Familiarity, Altered Sense of Connection, and Altered Sense of Agency. The median difference in latent ČEFSA subscale scores between participants with and without trauma was 0.54 (range: 0.14–1.67), with 24 out of 56 differences being statistically significant. Refer to Supplemental Material S9 for the frequencies of various trauma categories. Finally, the ČEFSA subscales negatively related to generalized self-efficacy scores (Mdn: −.21, range: −.22 to −.12), and positively to CEQ fantasy proneness scores (median: .48, range: .44–.60).
Properties of the Brief Form
We examined psychometric properties of the 14-item ČEFSA short form that includes two items for each of the seven subscales (see Table 2). We replicated the evaluation steps from Study 1 for the brief form using the data set from Study 2.
Factor Structure
Model fit parameters indicated an appropriate solution for the model with seven (latent) factors, each with two indicators, that converge on a single second-order factor (Model A), χ2(70, N = 243) = 133.88, p < .001, CFI = .95, RMSEA = 0.06, 95% CI [0.05, 0.07], SRMR = .05, but not for the model with a single first-order factor that includes all 14 indicators (Model B), χ2(77, N = 243) = 213.28, p < .001, CFI = .89, RMSEA = 0.09, 95% CI [0.07, 0.10], SRMR = .06. The Satorra–Bentler scaled chi-squared difference test was significant, diff χ2(7, N = 243) = 73.11, p < .001, indicating that Model A fits the data significantly better than Model B. Consistent with this result, the estimates for the Bayesian Information Criterion (BIC) were 6,673.86 for Model A and 6,583.78 for Model B (ΔBIC = 90.08), further suggesting that Model A is a better overall fit. Subscale results are reported in Supplemental Material S10. Contrary to our original plan, we report measurement models for each subscale for the sake of comprehensiveness.
Reliability
Coefficient ω for the 14-item brief form was estimated at .94, 95% CI [.92, .95], suggesting an excellent composite reliability estimate. The median of the part-whole corrected item-total correlations in relation to the respective subscale was .73 (range: .69–.76). Item-level estimates are shown in Supplemental Material S4.
Descriptives
The scale mean of the 14-item brief form was estimated at 0.65 (SD = 0.21). The median percentage of participants who selected the “never” option (score of 0) across all items was 58% (range: 34%–73%) for the brief form compared with 57% (range: 30%–73%) for the long form.
Validity
The latent correlation between the total score of the 14-item brief form and dissociation assessed using the DES was estimated at .77, 95% CI [.67, .84], suggesting convergent validity. Contrary to our original plan, we report latent correlations instead of manifest correlations, as this approach has the advantage of accounting for measurement error in the subscales. We found a non-significant correlation between the 14-item brief form and openness, −.11, 95% CI [−.26, .03]. Moderate negative correlations were observed with agreeableness, −.42, 95% CI [−.60, −.26], and conscientiousness, −.32, 95% CI [−.48, −.16]. There was a weak positive correlation with extraversion, .17, 95% CI [.02, .31], and a moderate positive correlation with neuroticism, .40, 95% CI [.26, .54]. These findings indicate modest discriminant validity from the Big Five personality factors. Finally, the brief form positively related to PHQ-9 depression scores, .64, 95% CI [.51, .73], GAD-7 anxiety scores, .55, 95% CI [.43, .65], LPFS personality dysfunction scores in the domain self, .71, 95% CI [0.63, 0.79], and others, .72, 95% CI [0.63, 0.80], as well as PAQ-S alexithymia scores, .48, 95% CI [.36, .60]. The ČEFSA brief form negatively related to generalized self-efficacy scores, −.23, 95% CI [-0.35, −0.10], and positively to CEQ fantasy proneness scores, .52, 95% CI [0.33, 0.61]. Participants who reported specific traumatic events assessed using the PDS-5 had higher scores on the items of the brief form. See Supplemental Material S10 for further details. Contrary to our original plan, we also report correlations using the second order factor score of the brief version to inform the potential use of a general indicator for felt sense of anomaly.
Discussion
In previous publications, evidence was provided for the reliability and validity of the English version of the 35-item ČEFSA scale and its 14-item brief form in both general population and clinical samples, including adults and adolescents (Černis et al., 2021 ). The ČEFSA scale is a new measure designed to assess individual differences in dissociative experiences centered around the common theme of “felt sense of anomaly.” It identifies three domains of dissociative experiences – self, body, and emotion – and four types of anomalies: altered senses of familiarity, connection, agency, and reality. We presented the development of a German version (Study 1) and examined its psychometric properties (Studies 1 and 2) in general population samples.
Psychometric Properties
As in the original, we found that a CFA model with seven (latent) factors that converge on a single second-order factor was appropriate for the German ČEFSA scale. More so, this model was a better fit than a model with a single first-order factor that includes all 35 indicators, providing evidence that the ČEFSA scale comprises distinguishable subscales. In addition, for the final version, all subscales had acceptable model fit, indicating that the translated items assessed a single underlying construct for each individual ČEFSA subscale. However, we found high bivariate latent correlations (> .90) between nine of the 21 subscale pairs, suggesting limited distinguishability between these pairs, particularly when Altered Sense of Familiarity or Altered Sense of Connection were involved. One explanation for the high subscale correlations is that the large number of zero responses in our general population samples may have inflated these correlations. Future studies should explore whether improved differentiation can be achieved in clinical samples. If this is not the case, another perspective is that the ČEFSA scale may over-specify the felt sense of anomaly construct, suggesting that fewer subfactors could provide a more efficient yet comprehensive approach to measuring it. Future research should then focus on reducing the number of subscales. Such studies should also investigate whether the subscales are effective in differentiating subtypes of patients with dissociative symptoms, an objective that has proven difficult to achieve in previous research using other dissociation measures, such as the Dissociative Experiences Scale and the Clinician Administered Dissociative States Scale (Bremner et al., 2024; Černis et al., 2021).
Results in Studies 1 and 2 indicate sufficient reliability estimates of the German version of the ČEFSA subscales and sufficient discrimination of the items. In addition, we found high (r > .50) correlations between the ČEFSA subscales and an existing measure of dissociation, indicating convergent validity. The ČEFSA subscales demonstrated moderate correlations (r < .50) with agreeableness, conscientiousness, and neuroticism, as anticipated. We observed no significant correlations between the ČEFSA subscales and extraversion or openness, supporting the scales' discriminant validity. However, the confidence intervals for the latent correlations of three subscales (Anomalous Experience of Emotion, Altered Sense of Connection, and Altered Sense of Agency) with neuroticism overlapped with the confidence intervals for the latent correlations between these subscales and the Dissociative Experience Scale, our convergent measure. This finding highlights the close interrelation between dissociation and a person’s tendency to experience negative emotions, which aligns with ecological momentary assessment studies demonstrating a strong link between current negative affect and levels of dissociation (e.g., Heekerens et al., 2024; Stiglmayr et al., 2007), as well as studies linking dissociation with personality dysfunction (Spitzer et al., 2006). As expected, all ČEFSA subscales positively related to levels of depression, anxiety, personality dysfunction, and alexithymia. Individuals who reported having experienced physical or sexual violence, military combat, child abuse, and/or or involvement in a serious accident scored higher on several ČEFSA subscales. In addition, we found negative correlations with general self-efficacy and positive correlations with alexithymia (Černis et al., 2022). Finally, in line with previous research on dissociation, we found moderate to large correlations between the ČEFSA subscales and fantasy proneness (Merckelbach et al., 2022). Overall, these findings support the concurrent validity of the German version of the ČEFSA scale.
Contrary to expectations, we found only partial scalar invariance between the English and German versions of the ČEFSA scale. Our findings reveal that the intercepts of 17 out of 35 ČEFSA items were not equivalent between the two versions. As a result, comparisons across the affected subscales may lack validity and meaningfulness when these items are included (Millsap, 2011). However, latent means derived from a correctly specified partial scalar invariance multigroup model can still be used for valid group comparisons. For example, an item on the ČEFSA Anomalous Experience of the Body subscale – “I feel detached from my physical body (or parts of it)” in English versus “Ich fühle mich losgelöst von meinem Körper (oder Teilen davon)” in German – exhibited differing intercepts. This discrepancy might be attributed to differences in the samples used for the English and German versions. Specifically, the English sample was considerably younger, and previous studies have reported significantly higher ČEFSA scores in adolescent populations (Černis et al., 2024).
The psychometric properties of the 14-item version of the German ČEFSA scale were comparable to those of the 35-item version, suggesting that the brief version can be effectively used in settings where participant burden is a concern.
Application of the ČEFSA Scale
The German version of the ČEFSA scale has been validated for use in adult general population samples for research purposes. Following recommendations for the original English version and based on the findings from our studies presented here, we suggest using subscale scores if the full version of the ČEFSA scale is administered (Černis et al., 2021). A key strength of the ČEFSA scale is its ability to differentiate various aspects of dissociation, which is why we do not recommend using the second-order factor (composite) score. If one chooses to assess a general indicator of the felt sense of anomaly, we recommend using the second-order factor (composite) score of the brief version, as it should be sufficient and more economical to administer (Černis et al., 2024).
Subscales of the short version may be used in specific contexts, such as ecological momentary assessment studies, but our data do not provide information on the psychometric properties of the subscales in these contexts. Additionally, the sum score of the German brief form may be used to inform judgments about the severity of current dissociative experiences centered around a felt sense of anomaly (see Černis et al., 2024, for interpretive score ranges). However, we advise caution when using the German version in this way, as validation in clinical samples is still pending.
Limitations and Future Research
While we provide evidence supporting the German version of the ČEFSA scale, several limitations should be noted. First, we used general population samples in our study, which limits the generalizability of our findings. As our data show, dissociation is relatively rare in the general population. However, it is more frequent in individuals with mental disorders (Lyssenko et al., 2018). Our findings do not generalize to clinical samples, and a priority for future research should be to evaluate the German version of the ČEFSA scale in individuals diagnosed with mental disorders. We suggest selecting diagnostic groups in which dissociation is particularly frequent, such as borderline personality disorder, post-traumatic stress disorder, and dissociative disorders. However, it would also be valuable to test the German version of the ČEFSA scale and compare dissociation levels across a broader range of diagnostic categories. These studies should employ gold-standard assessment methods for mental disorders, such as widely used structured clinical interviews conducted by qualified professionals, with an emphasis on accurate differential diagnosis. Second, the ČEFSA scale relies on retrospective self-assessments, which are susceptible to memory biases and may be influenced by current emotional states. Future research could address this limitation by employing experience sampling methods that repeatedly assess dissociative experiences in real-time (Heekerens et al., 2024; Soffer-Dudek et al., 2017; Vannikov-Lugassi & Soffer-Dudek, 2018). For instance, administering a rotating set of one or two shortened items per subscale every four hours could be an effective approach (Carlson et al., 2016). Third, the ČEFSA scale focuses on a finite set of specific dissociative experiences, all of which revolve around the theme of a “felt sense of anomaly.” As a result, it does not capture the full spectrum of dissociative experiences, such as dissociative amnesia, which has been discussed in prior research (see Černis et al., 2021).
Conclusion
Across two studies, we described the development of a German version of the ČEFSA scale and investigations of its psychometric properties and validity in general population samples. Our results show the seven ČEFSA subscales of the German version had an appropriate factor structure and sufficient reliability estimates. Regarding validity, the convergent, discriminant, and criterion correlations largely aligned with expectations. However, some ČEFSA subscales could not be well distinguished from one another, and three subscales showed considerable overlap with neuroticism. Despite these challenges, we conclude that the German version of the ČEFSA is a reliable and valid tool for assessing dissociative experiences revolving around the theme of a “felt sense of anomaly.”
References
(2022). Diagnostic and statistical manual of mental disorders: DSM-5-TR.
(2000). Guidelines for the process of cross-cultural adaptation of self-report measures. Spine, 25(24), 3186–3191. 10.1097/00007632-200012150-00014
(2017). Short scale for measuring General Self-efficacy beliefs (ASKU). Methoden, Daten, Analysen, 7(2), 251–278. 10.12758/mda.2013.014
(2024). Psychometric properties of the 23-Item Clinician Administered Dissociative States Scale (CADSS) in a psychological trauma population. Journal of Affective Disorders, 364(1), 249–258. 10.1016/j.jad.2024.08.050
(2021). A new perspective and assessment measure for common dissociative experiences: ‘Felt Sense of Anomaly’. PLoS ONE, 16(2), Article
e0247037 . 10.1371/journal.pone.0247037(2022). Psychological mechanisms connected to dissociation: Generating hypotheses using network analyses. Journal of Psychiatric Research, 148(1), 165–173. 10.1016/j.jpsychires.2022.01.049
(2024). Measuring dissociation across adolescence and adulthood: Developing the short-form Černis Felt Sense of Anomaly scale (ČEFSA-14). Behavioural and Cognitive Psychotherapy, 52(2), 163–177. 10.1017/s1352465823000498
(2016). Advantages and psychometric validation of proximal intensive assessments of patient-reported outcomes collected in daily life. Quality of Life Research: An International Journal of Quality of Life Aspects of Treatment, Care and Rehabilitation, 25(3), 507–516. 10.1007/S10136-015-1170-9
(1993). An update on the Dissociative Experiences Scale. Dissociation: Progress in the Dissociative Disorders, 6(1), 16–27.
(2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modeling: A Multidisciplinary Journal, 14(3), 464–504. 10.1080/10705510701301834
(2025). Data sets for evaluation of the psychometric properties and validity of the German version of the Černis Felt Sense of Anomaly (ČEFSA) [Open data]. 10.5281/zenodo.13169632
(2025). ČEFSA – Translation (German) – Electronic Supplemental Material and Statistical Code [Open Data, Material, Code]. 10.17605/OSF.IO/3FK94
(2024). Affective arousal temporally precedes dissociation in patients with borderline personality disorder: A preliminary experience sampling study. Psychological Trauma: Theory, Research, Practice, and Policy, 16(7), 1129–1138. 10.1037/tra0001516
(2017). Psychometric evaluation of the Generalized Anxiety Disorder Screener GAD-7, based on a large German general population sample. Journal of Affective Disorders, 210(1), 338–344. 10.1016/j.jad.2016.12.012
(2005). Are there two qualitatively distinct forms of dissociation? A review and some clinical implications. Clinical Psychology Review, 25(1), 1–23. 10.1016/j.cpr.2004.08.006
(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6(1), 1–55. 10.1080/10705519909540118
(2021). German translation of the Perth Alexithymia Questionnaire. Retrieved August 22, 2024 from https://osf.io/preprints/psyarxiv/gj5e3/download
(2002). The five-factor personality structure of dissociative experiences. Personality and Individual Differences, 32(3), 431–443. 10.1016/s0191-8869(01)00035-6
(2023). Developing an understanding of the relationship between anxiety and dissociation in adolescence. Psychiatry Research, 324(1), Article
115219 . 10.1016/j.psychres.2023.115219(2004). Measuring depression outcome with a brief self-report instrument: sensitivity to change of the Patient Health Questionnaire (PHQ-9). Journal of Affective Disorders, 81(1), 61–66. 10.1016/s0165-0327(03)00198-8
(2022). Dissociation and dissociative disorders reconsidered: Beyond sociocognitive and trauma models toward a transtheoretical framework. Annual Review of Clinical Psychology, 18(1), 259–289. 10.1146/annurev-clinpsy-081219-102424
(2018). Dissociation in psychiatric disorders: A meta-analysis of studies using the dissociative experiences scale. The American Journal of Psychiatry, 175(1), 37–46. 10.1176/appi.ajp.2017.17010025
(2001). The Creative Experiences Questionnaire (CEQ): A brief self-report measure of fantasy proneness. Personality and Individual Differences, 31(6), 987–995. 10.1016/S0191-8869(00)00201-4
(2022). Empirical research on fantasy proneness and its correlates 2000–2018: A meta-analysis. Psychological Consciousness: Theory Research Practice, 9(1), 2–26. 10.1037/cns0000272
(2011). Statistical approaches to measurement invariance. Routledge/Taylor & Francis Group.
(2007). Construct validation of a short five-factor model instrument. European Journal of Psychological Assessment, 23(3), 166–175. 10.1027/1015-5759.23.3.166
(2002). How to use a Monte Carlo study to decide on sample size and determine power. Structural Equation Modeling: A Multidisciplinary Journal, 9(4), 599–620. 10.1207/s15328007sem0904_8
(1998-2024). Mplus User's Guide. Eight Edition.
(2009): Instructional manipulation checks: Detecting satisficing to increase statistical power. Journal of Experimental Social Psychology, 45(4), 867–872. 10.1016/j.jesp.2009.03.009
(2023). The Perth Alexithymia Questionnaire-Short Form (PAQ-S): A 6-item measure of alexithymia. Journal of Affective Disorders, 325(1), 493–501. 10.1016/j.jad.2023.01.036
(2016). Measurement invariance conventions and reporting: The state of the art and future directions for psychological research. Developmental Review, 41(2), 71–90. 10.1016/j.dr.2016.06.004
(2010). Ensuring positiveness of the scaled difference chi-square test statistic. Psychometrika, 75(2), 238–243. 10.1007/S10336-009-9135-y
(2024). Dissociative experiences, borderline personality disorder features, and childhood trauma: Generating hypotheses from data-driven network analysis in an international sample. Journal of Trauma & Dissociation, 25(4), 436–455. 10.1080/15299732.2024.2323974
(2017). Absorbed in sleep: Dissociative absorption as a predictor of sleepiness following sleep deprivation in two high-functioning samples. Consciousness and Cognition, 48(1), 161–170. 10.1016/j.concog.2016.11.009
(2006). Borderline personality organization and dissociation. Bulletin of the Menninger Clinic, 70(3), 210–221. 10.1521/bumc.2006.70.3.210
(2003). Gender differences in dissociation. Psychopathology, 36(2), 65–70. 10.1159/000070360
(2021). Die deutsche Version der Level of Personality Functioning Scale-Brief Form 2.0 (LPFS-BF): Faktorenstruktur, konvergente Validität und Normwerte in der Allgemeinbevölkerung [
The German version of the Level of Personality Functioning Scale-Brief Form 2.0 (LPFS-BF): Latent structure, convergent validity and norm values in the general population ]. Psychotherapie - Psychosomatik - Medizinische Psychologie, 71(1), 284–293. 10.1055/a-1343-2396(2004). Screening und Veränderungsmessung dissoziativer Psychopathologie: Psychometrische Charakteristika der Kurzform des Fragebogens zu dissoziativen Symptomen (FDS-20) [
Screening and change measurement of dissociative psychopathology: Psychometric characteristics of the short form of the Dissociative Experience Questionnaire (DES-20) ]. Psychotherapie, Psychosomatik, medizinische Psychologie, 54(3), 165–172. 10.1055/s-2003-814783(2007). Dissociative symptoms are positively related to stress in borderline personality disorder. Acta Psychiatrica Scandinavica, 117(2), 139–147. 10.1111/j.1600-0447.2007.01126.x
(2015). Introduction to health measurement scales. Oxford University Press.
(2018). No time like the present: Thinking about the past and the future is related to state dissociation among individuals with high levels of psychopathological symptoms. Frontiers in Psychology, 9, Article
2465 . 10.3389/fpsyg.2018.02465(2018). Dissociation in victims of childhood abuse or neglect: A meta-analytic review. Psychological Medicine, 48(15), 2467–2476. 10.1017/s0033291718000740
(2021). Psychometric properties and validity of the German version of the Post-Traumatic Diagnostic Scale for DSM-5 (PDS-5). European Journal of Psychotraumatology, 12(1), Article
1965339 . 10.1080/20008198.2021.1965339(2013). Sample size requirements for structural equation models: An evaluation of power, bias, and solution propriety. Educational and Psychological Measurement, 73(6), 913–934. 10.1177/0013164413495237
