Brief reportA confirmatory factor analysis of the Toronto Alexithymia Scale (TAS-20) in an alcohol-dependent sample
Introduction
Alexithymia refers to difficulty in identifying, differentiating and communicating feelings, a lack of imagination, and an externally oriented thinking style (Nemiah et al., 1976). The Toronto, Alexithymia Scale (TAS-20) (Bagby et al., 1994) is a reliable self-report measure of alexithymia. It consists of three factors: difficulties identifying feelings (DIF), difficulties describing feelings (DDF), and externally oriented thinking (EOT). Given the link between addiction and alexithymia, it is important to validate the TAS-20 in substance misuse samples (Thorberg et al., 2009). There have only been a few such studies. The TAS-20 confirmatory factor analyses (CFA) undertaken by Cleland et al. (2005) in a substance disorder sample indicated a marginal fit for the two-factor model (Loas et al., 1996, Kooiman et al., 2002) and the original three-factor model (Bagby et al., 1994). In a combined drug- and alcohol-dependent group, Besharat (2008) confirmed the three-factor model, but the psychometric properties of the TAS-20 were described as suboptimal. In a sample composed of a group with substance dependence and a group with eating disorders (Loas et al., 2001) the original three-factor structure was confirmed and was a superior fit than the two-factor model. In a psychiatric sample including alcohol abusers, a four-factor model consisting of DIF, DDF, pragmatic thinking, and lack of importance of emotions provided the best fit compared to the other models (Mueller et al., 2003). Given the lack of consistency in previous attempts to identify the TAS-20 factor structure in substance-dependent samples, and the lack of such research in groups with primary alcohol dependence, further studies of this type are warranted. This study aimed to establish the underlying factor structure of the TAS-20 in an alcohol-dependent sample.
Section snippets
Participants
Participants were undertaking outpatient cognitive behavioural therapy for alcohol dependence at a Brisbane hospital. A total of 210 participants (144 males) with a mean age of 38.17 years (S.D. = 10.82) diagnosed with alcohol dependence as assessed by Diagnostic and Statistical Manual of Mental Disorders, fourth edition, third revision (DSM-IV) criteria (American Psychiatric Association, 2000), Brief Michigan Alcoholism Screening Test (bMAST Pokorny et al., 1972), and the Alcohol Use Disorder
Means, S.D. and intercorrelations
The mean score for the TAS-20 was 53.82 (S.D. = 11.87), DIF 19.38 (S.D. = 6.23), DDF 14.11 (S.D. = 4.47), EOT 20.33 (S.D. = 4.42), bMAST 16.60 (S.D. = 8.27), and AUDIT 26.03 (S.D. = 9.80). Pearson's correlations between TAS-20 total score, DIF, DDF, and EOT were significant (r = 0.83; r = 0.87; and r = 0.63, respectively). The DIF scale was significantly correlated with DDF (r = 0.63) and EOT (r = 0.19) and DDF was significantly correlated with EOT (r = 0.43). The AUDIT was significantly correlated with TAS-20 (r =
Discussion
The original three-factor model provided the most plausible theoretical explanation of the data compared with the other models. It had a significantly lower χ2 scores as well as a significant chi-square difference indicating a superior global fit. This result supports previous findings (Loas et al., 2001, Mueller et al., 2003, Cleland et al., 2005). DIF and DDF factor loadings were significant and higher than the recommended cut-off (Stevens, 1996), with the exception of DDF item 4 (Table 1).
References (28)
- et al.
Alexithymia: relationship to personality disorders
Comprehensive Psychiatry
(1994) - et al.
The twenty-item Toronto Alexithymia Scale: I. Item selection and cross-validation of the factor structure
Journal of Psychosomatic Research
(1994) - et al.
Cross validation of the factor structure of the 20-item Toronto Alexithymia Scale: an Italian multicenter study
Journal of Psychosomatic Research
(1996) - et al.
Psychometric properties of the Toronto Alexithymia Scale (TAS-20) for substance users
Journal of Psychosomatic Research
(2005) - et al.
The assessment of alexithymia: a critical review of the literature and a psychometric study of the Toronto Alexithymia Scale-20
Journal of Psychosomatic Research
(2002) - et al.
Factorial structure of the 20-item Toronto Alexithymia Scale confirmatory factorial analyses in nonclinical and clinical samples
Journal of Psychosomatic Research
(2001) - et al.
Alexithymia and alcohol use disorders: a critical review
Addictive Behaviors
(2009) - American Psychiatric Association, 2000. Diagnostic and Statistical Manual of Mental Disorders 4th ed. text rev....
Comparative fit indexes in structural models
Psychological Bulletin
(1990)Assessing reliability and validity of the Farsi version of the Toronto Alexithymia Scale in a sample of substance-using patients
Psychological Reports
(2008)
The relationship between dimensions of alexithymia and dissociation
Psychotherapy and Psychosomatics
Imputing cross-sectional missing data: comparison of common techniques
Australian and New Zealand Journal of Psychiatry
Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives
Structural Equation Modelling
Paternal alcohol exposure and Turner syndrome
Alcohol and Alcoholism
Cited by (24)
The associations between alexithymia and both non-suicidal self-injury and risky drinking: A systematic review and meta-analysis
2020, Journal of Affective DisordersCitation Excerpt :Kopera et al. (2018) conducted one study with Polish male adults and Khosravani, Bastan, Avatefi, & Mofidi (2017) conducted another with Iranian adults. Thorberg, Lyvers, and colleagues conducted the remaining four studies (Lyvers et al., 2014b; Thorberg et al., 2010, 2011a,b) with Australian adults. Kopera et al. (2018) and Lyvers et al. (2014b) employed case-control designs; the remaining four studies were cross-sectional.
The psychometric assessment of alexithymia: Development and validation of the Perth Alexithymia Questionnaire
2018, Personality and Individual DifferencesCitation Excerpt :High levels of internal consistency reliability were also evident for all PAQ subscale and composite scores, and the excellent internal consistency reliability of the PAQ G-EOT subscale is of particular note given that some previous attempts to measure EOT via self-report have had low reliability. In a similar Australian sample to the two analysed here, for example, Preece, Becerra, Robinson, and Dandy (2017) found that the TAS-20 EOT subscale had a low Cronbach's alpha reliability coefficient of 0.64, and half the TAS-20 EOT items did not load meaningfully on the EOT factor in factor analysis (see also, Gignac et al., 2007; Thorberg et al., 2010). We think future research with the PAQ could consequently help enhance theoretical understanding of the alexithymia construct and its relationship to other variables.
Alexithymia and problematic alcohol use: A critical update
2018, Addictive BehaviorsCitation Excerpt :In contrast the three factor model of alexithymia was reported to be a superior fit among an AD population, although concerns were raised about the reliability of the EOT subscale (Thorberg et al., 2010b). Thorberg et al. (2010b) suggest caution when applying and interpreting the TAS-20 with AD populations and suggest a revision of the EOT items could be warranted. However, they also warned that changes could result in a modified construct that is not true to the original conceptualisation.
The association between depression and craving in alcohol dependency is moderated by gender and by alexithymia factors
2016, Psychiatry ResearchCitation Excerpt :Because the TAS-20 was more strongly associated with alcohol problems than the OAS, they concluded that it is preferable to use the TAS-20 over the OAS for this population. Finally, Thorberg et al. (2010a) investigated the factorial validity of the TAS-20 and examined several models in a group of an alcohol- dependent sample. Although the three-factor model provided a better fit across four out of five fit indices compared to the one and two-factor models, the CFI did not reach the recommended 0.90 cut-off and all models were rejected given their poor fit to the data.