Alexithymia is a multidimensional trait comprised of difficulties identifying feelings, difficulties describing feelings, and externally orientated thinking. It is most commonly assessed using the 20-item Toronto Alexithymia Scale (TAS-20). However, over the past two decades, some authors have suggested that the TAS-20 might assess current levels of distress rather than levels of alexithymia. We explored this possibility by administering three alexithymia questionnaires, including the TAS-20, and a measure of psychological distress to five samples (Ns=300, 128, 216, 148, 103). In each sample, we used second-order exploratory factor analysis to test whether the alexithymia subscales loaded, as intended, on a separate factor to the distress subscales. All subscales of the Perth Alexithymia Questionnaire (PAQ) and Bermond-Vorst Alexithymia Questionnaire (BVAQ) demonstrated good discriminant validity against the distress factor. In contrast, the difficulties identifying feelings subscale of the TAS-20 consistently cross-loaded onto the distress factor, indicating that much of its variance reflected current levels of distress, not alexithymia. This could limit the utility of the TAS-20, and the results of studies using it may need to be interpreted with this in mind. The PAQ and BVAQ appear to be good self-report alternatives for measuring alexithymia as a construct that is separable from distress.