The methodology of a number of prognostic studies has been critically considered, with only incidental reference to the substantive content. The latter has validity only when principles of the scientific method have been applied in deriving and evaluating data. The extensive literature on psychological tests used in prognosis has not been reviewed, since that already has been done by Windle(32).Most of the research has been found to be so deficient in the use of research procedures as to place in doubt any of the findings. Together with the prognostic statements of the didactic literature, much of the content is of value chiefly because of the conceptual framework and the suggestions they provide of factors whose prognostic relevance should be investigated. These factors have appeared so often in the literature that it seems that sheer force of repetition may have given them an importance not always demonstrated by the specific research reviewed.Definitive research will be facilitated by a greater degree of interdisciplinary planning. The narrow orientation of particular investigators may have been reflected in the fact that definitions so frequently were not clarified, indicating perhaps the investigator's assumption that no other point of view existed. Little attention, for example, has been paid to contingency factors such as the association of particular furlough situations with outcome. Full coverage of biogenic, psychogenic, and sociogenic factors demands that research psychiatrists join forces with other social scientists, some of whom have been working for years on a related problem, the prediction of recidivism (21).Research design must be refined. The population must be identified, the sampling process indicated, and the criterion groups defined. All variables under study should be objectively described. Obtaining reliable and valid measures of factors is perhaps a greater problem than the analysis of these measures. Conclusions should be statistically evaluated for significance, and it should be recognized that generalizations apply only to the population sampled. In general, enough information should be reported so that replication would be possible. Prognostic research, which may involve establishing contact with a mobile group of former patients, is expensive. To produce results worth the cost in time and money, strict adherence to scientific procedure is necessary.Abstract Teaser