Assessing Measurement and Predictive Invariance of the Toronto Alexithymia Scale–20 in U.S. Anglo and U.S. Hispanic Student Samples

Abstract

We collected data from a predominately Anglo American student sample in the Southeastern United States and a predominately Hispanic student sample in the Southwestern United States. Along with an assessment of internal consistency reliability, we examined measurement invariance of the Toronto Alexithymia Scale–20 (TAS–20) using confirmatory factor analysis. We also assessed the predictive invariance of the TAS–20. Results indicate that 2 of the 3 TAS–20 subscales demonstrated satisfactory internal consistency reliability across samples. Items from the TAS–20 subscales demonstrated measurement invariance of the latent means. The relationship between 2 measures of emotional dysfunction and the TAS–20 also demonstrated slope and intercept invariance, indicating equivalent validity.

Acknowledgments

This research was supported in part by Grant 2 R24 MH47167–11 from the National Institutes of Mental Health awarded to O. F. Morera. Portions of this research were presented at the 2008 meeting of the Association for Psychological Science in Chicago, Illinois.

Notes

¹Data reported in a prior study (Culhane, Morera, & Hosch, 2004) indicates that Hispanic students in the United States were generally acculturated to United States norms. We did not collect acculturation data for this study. However, we reexamined data from the earlier project (Culhane, Morera, & Hosch, 2003) that used identical criteria for participant selection. The results were that 68.8% spent their entire life in the United States, and only 7.7% spent more of their lifetime in Hispanic countries (e.g., Mexico, Peru, etc.) than in the United States. Also, the majority of participants (67.7%) were second generation or later Americans (Culhane et al., 2003). There was no reason to believe that these participants were different. It is possible that results obtained in this project were a result of cultural similarities rather than differences.

²Data from the 29 participants who indicated mixed or other and had at least one identified Hispanic parent were also categorized as Hispanic. To test the question of whether the results differed when these participants were not included, we performed a series of simple one-way analyses of variance with and without them.

^a n = 214.

^b n = 317.

^a n = 214.

^b n = 317.

∗ p <.01.

∗∗ p <.001.

^a Items have factor pattern coefficients that were set to 1.00.

³We thank an anonymous reviewer for raising this point and bringing to our attention the Marsh et al. (2004) manuscript.

∗p <.05.

∗∗p <.01.