Validation of a Greek adaptation of the 20-item Toronto Alexithymia Scale

Abstract

Background

The purpose of the current investigation was (1) to test whether the 3-factor structure of the 20-item Toronto Alexithymia Scale (TAS-20) corresponding to the theoretical conceptualization of the alexithymia construct could be recovered in a Greek translation of the scale (the TAS-20-G), (2) to assess if a 3-factor structure provides a better fit to the TAS-20-G compared with the recently proposed alternative factor structures, and (3) to evaluate the internal reliability of the TAS-20-G.

Methods

The English version of the TAS-20 was translated into Greek and then back-translated and modified until cross-language equivalence was established. The Greek version was then administered to 340 university students. Confirmatory factor analyses were conducted, and 4 different factor structure models were compared. Internal consistency and item-to-scale homogeneity of the TAS-20-G and its factor scales were also evaluated.

Results

The 3-factor model provided a good fit to the data and proved superior to alternative 1-, 2-, and 4-factor models. Apart from a coefficient α below the recommended range for the externally oriented thinking factor, the TAS-20-G and its factor scales demonstrated adequate internal consistency and homogeneity.

Conclusion

The TAS-20-G is a valid and reliable measure of alexithymia in university students and may be suitable for investigations of alexithymia in other Greek-speaking population samples.

Introduction

The term “alexithymia” has its linguistic roots in the Greek language; coined by Sifneos [1], this synthetic Greek term literally means having no words for emotions. The construct of alexithymia was formulated by Nemiah and Sifneos [2], [3] to encompass difficulties some people have in identifying and describing their own feelings and emotions as well as an impoverished fantasy life and an externally oriented mode of thinking. Rather than being a psychiatric diagnosis or a categorical phenomenon, alexithymia is conceptualized as a dimensional personality trait [4]; a taxometric investigation has provided support for this dimensional view of the construct [5].

A number of investigators have reported associations between alexithymia and a variety of medical and psychiatric disorders, including substance use disorders, eating disorders, posttraumatic stress disorder, functional gastrointestinal disorders, and essential hypertension [4], [6]. There is also accumulating evidence that alexithymia influences patients' responses to some medical and psychotherapeutic treatments and may even predict treatment response. Porcelli et al [7], for example, reported that alexithymia was a significant predictor of treatment outcome for patients with functional gastrointestinal disorder. Clinical researchers have also found alexithymia to be associated with a less favorable response to psychodynamic psychotherapy [8] and a higher level of residual symptoms than reported by nonalexithymic patients even when the psychotherapy achieves significant symptom reduction [9].

Bagby and colleagues developed the 20-item Toronto Alexithymia Scale (TAS-20) [10], [11], which has become the most widely and frequently used instrument to assess the alexithymia construct [12]. Based on factor analytic work [13], the 20 items that compose the TAS-20 are organized into 3 factor scales—Difficulty Identifying Feelings (DIF), Difficulty Describing Feelings (DDF), and Externally Oriented Thinking (EOT). The DIF factor scale assesses difficulties identifying feelings and distinguishing them from the somatic sensations that accompany emotional arousal; DDF assesses difficulties describing feelings to other people; and EOT assesses EOT. There is evidence that this factor also indirectly assesses fantasy and other imaginal processes [11], [14], [15].

Numerable investigators have examined the latent structure of the TAS-20. Using confirmatory factor analysis (CFA), the Toronto team of researchers demonstrated that the 3-factor structure recovered initially in the derivation sample [10] was replicable in both clinical and nonclinical English-speaking populations [16], [17]. This 3-factor model has also been replicated with translations of the TAS-20 administered to nonclinical and clinical samples in many different languages including Chinese [18], Danish [19], Dutch [19], Finnish [20], French [21], German [16], [22], Hebrew [19], Hungarian [23], Hindi [24], Farsi [25], Italian [26], Japanese [19], Korean [27], Lithuanian [28], Norwegian [19], Portuguese [28], Spanish [29], and Swedish [30].

Although the 3-factor model, identified initially in the English-speaking derivation sample, has been replicated in a variety of samples speaking different languages from diverse cultures (eg, Asian, European, and Middle-Eastern), some investigators have reported that alternative 2- and 4-factor models provide better statistical fit than the original 3-factor structure. For example, Loas et al [31] administered a French translation of the TAS-20 to a university student sample and obtained a 2-factor solution, with the DIF and the DDF items constituting a single factor and the EOT items comprising a second factor. A similar 2-factor solution was obtained by Erni et al [32], who had a sample of medical students complete a German translation of the TAS-20. Both of these studies, however, used exploratory factor analysis rather than CFA, to explore the factor structure of the translated versions of the TAS-20. When CFA was applied to the French data, a 3-factor model provided a better fit to the data than did the 2-factor model [33]. More recently, Müller et al [34] compared 5 different models with 1 to 4 factors in clinical and nonclinical (ie, university students) samples in Germany and found that a 4-factor model, in which the EOT items were divided into 2 distinct factors—“Pragmatic Thinking” and “Lack of subjective significance or importance of emotions”—provided a better fit to the data compared with 1-, 2- and 3-factor models. When the same models were compared in the validation study of the Chinese translation of the TAS-20, the 4-factor model provided a better fit than did the standard 3-factor model in a student sample, but not in a clinical sample [18].

In a review of 18 different language versions of the TAS-20, Taylor et al [19] refer to an unpublished Greek version of the scale that was translated by T. Anagnostopoulou and evaluated with a sample that comprised 104 normal adults and 96 asthmatic patients. Although results from CFA revealed that the 3-factor structure provided a “marginally” acceptable fit (χ²/df ratio, 1.68; Goodness-of-Fit Index [GFI], 0.88; Adjusted Goodness of Fit Index [AGFI], 0.85; Standardized Root Mean Square Residual [SRMR], 0.07), the goodness-of-fit indices did not include an incremental fit index, which is now regarded as an essential index in assessing goodness-of-fit. Moreover, the estimates of internal reliability for the DDF and EOT factor scales were below the generally acceptable standard. The mixed clinical and nonclinical sample used in that study and some disagreements over the way some items had been translated into Greek prompted us to develop a new Greek translation of the TAS-20 (the TAS-20-G) and to examine whether the theoretical structure that underlies the factor structure of the original TAS-20 could be recovered in this Greek adaptation of the scale. Furthermore, we were interested in investigating whether alternative factor structures, such as a 2-factor model [31], [32] or a 4-factor model [34], represent the latent structure of the TAS-20-G items better than the original 3-factor structure in a sample of Greek-speaking university students. Finally, we were interested in evaluating the internal reliability of this latest Greek translation of the scale.