The authors found no cohort effect. For their statistical analysis, they assumed that if a cohort effect were present, the three first-cohort medications (i.e., haloperidol, risperidone, and clozapine) should have all fared better in the second cohort. We question this assumption. If the second cohort consisted of patients with a better prognosis for second-generation antipsychotics, we would have expected the following: haloperidol should not have been effective in either cohort because both cohorts were selected to be resistant to neuroleptics, clozapine should have performed well in both cohorts, and risperidone should have been inferior to clozapine in the first cohort (1) and comparable in the second cohort (2, 3). The reported results fit these assumptions fairly well. Haloperidol was indeed ineffective in both cohorts. Clozapine did moderately well in both cohorts, with scores on the Positive and Negative Syndrome Scale increasing only a small amount in the second cohort (6.48 versus 7.05, respectively). The risperidone group’s improvement scores increased from –0.03 in the first cohort (N=25) to 7.93 in the second cohort (N=16). The latter appears comparable to the improvement with olanzapine (9.1, N=39). In summary, the cohort results appear too different to be validly combined.