Results of two earlier studies imply that schizophrenics could be assigned, according to their pattern of presenting symptomatology, to that phenothiazine on which they are likely to improve most. This paper reports an attempt to replicate these findings. The results are negative: prediction of better and worse drug assignment could not be made. Factors leading to our failure to predict as compared with successes in earlier work are discussed. The current "state of the art" in prediction of response to phenothiazines is equivocal; positive results in the literature preclude dismissing the idea totally, but are insufficient to establish the relationship as fact. For the practicing clinician no system is yet available, either from empirical research or accumulated clinical experience, to enable matching of particular phenothiazines with particular schizophrenic patients in terms of their symptom profiles.