Alexithymia represents an individual difference dimension characterized by difficulties identifying emotions, difficulties describing emotions, and a utilitarian approach to thinking. The Toronto Alexithymia Scale-20 (TAS-20; Bagby, Parker, & Taylor, 1994) is a frequently used inventory to measure alexithymia, and although several studies have examined the factor structure of the TAS-20, a number of issues remain unresolved. Specifically, the severely unbalanced item-keyed nature of the TAS-20 has been suggested to limit the interpretation of the substantive Externally Oriented Thinking (EOT) factor. Further, it has also been suggested that the EOT factor may be better represented by 2 oblique factors. A review of the TAS-20 confirmatory factory analysis (CFA) literature has suggested that some improvement in CFA strategies could possibly be afforded by using a nested factors modeling approach. Based on a sample of 355 participants, we demonstrated that the TAS-20 was better represented by a nested factors model with 5 substantive factors. A novel, latent variable approach to estimating internal consistency reliability revealed that the subscales within the TAS-20 were associated with unacceptably low levels of reliability independently of the global alexithymia factor. Although there was some CFA evidence to suggest the plausibility of a negatively keyed factor, a thorough examination of the items in question offered an alternative interpretation. Further development of the TAS-20's Externally Oriented Thinking subscale is encouraged.