The basic hypothesis of the literature on alexithymia, i.e. that alexithymia has a higher prevalence in psychosomatic than in neurotic (and delusional) patients, was empirically tested by means of the well-validated Toronto Alexithymia Scale (TAS). Surprisingly, neurotic and delusional patients (N = 71) had significantly higher mean total scores on the TAS, compared with the psychosomatic group (N = 150); the normal control sample (N = 224) was, as predicted, the lowest scorer. This hierarchical distribution was confirmed for the first two factors of alexithymia: (1) difficulty in distinguishing between feelings and bodily sensations, and (2) difficulty in expressing feelings. The psychiatric group was, instead, the lowest scorer on the third factor (lack of fantasy life). A substantial cross-validation of the above findings was achieved by comparing on the TAS three subgroups of the normal sample (symptom-free, somatizing and 'neurotic' normal controls). The postulate of the non-neurotic nature of alexithymia, along with its many psychopathological and technical corollaries, is completely contradicted by the present findings.