To determine the psychometric properties of the Toronto Alexithymia Scale (TAS-20), an established self-report measure of alexithymia, for a substance user sample participating in a clinical trial of outpatient cognitive–behavioral therapies (N=230).Confirmatory and exploratory factor analyses were used to determine the number and nature of the factors underlying the TAS-20 in a sample of substance users. Structural equation modeling was used to determine the predictive validity of the TAS-20.A factor structure comparable, but not identical, with TAS-20 psychometric results with other populations was found; alpha coefficients were .88 for the feelings factor, .62 for the external thinking factor, and .87 for the total score. Although, on average, the substance users did not appear to have elevated alexithymia scores compared with the undergraduate students, alexithymia predicted less treatment engagement, i.e., fewer sessions attended and weaker helping alliance. Alexithymia also predicted alcohol use outcomes but not drug use outcomes. The relation between alexithymia and drinking outcome was conditional on whether the patient was using solely alcohol at baseline.The TAS-20 has reasonably good psychometric properties in this sample, which might be improved by dropping several marginal questionnaire items. Alexithymia appears to attenuate substance abuse treatment engagement. More clinical and research experience with this construct and specific instrument in substance user samples is needed.