Blatt and Shahar Reply

September 15, 2004. We appreciate Jonathan Rosenfeld's careful reading of our article and his pointing out the importance of examining pretreatment differences between psychoanalysis (PSA) and support-ive-expressive psychotherapy (SEP) in our further analyses of data from the Menninger Psychotherapy Research Project. It should be noted, however, that this had been done in an earlier analysis of this data set (Blatt 1992) and no signif icant pretreatment differences were found, either between the two treatment groups (PSA and SEP) or between the two groups of patients (anaclitic and introjective) on the Health-Sickness Rating Scale (Luborsky 1962). Nor was there a significant difference in the length of treatment, though by design patients in PSA were seen significantly more frequently each week (M = 4.67) than patients in SEP (M = 2.72). Additionally, no significant differences were found between the two treatment groups (PSA and SEP) on any of the outcome measures used in our analyses of these data, though indications suggest that introjective patients at pretreatment seemed more disturbed than anaclitic patients (they had significantly more thought disorder and greater investment in unrealistic [inaccurately perceived] object representations).

While the findings in this earlier report appear to address Rosenfeld's concern about the possibility of confounds in this non-randomized study, we believe that his observation has a more general implication for the often poorly understood nature of psychotherapy trials. Specifically, he observes that “in a study that is not randomized, it is possible that multiple factors may be distributed unequally among the groups being compared.” What many research consumers tend to overlook is the fact that even in a randomized trial, multiple factors might differentiate the groups. In other words, the success of randomization is probabilistic in nature, and should never be taken for granted. Further, even if randomization is successful