I n recent discussions of the reliability and validity of assessments of personality, it has all too often been assumed that high reliability of judgments by a group of trained observers has guaranteed high validity. This assumption is tenable only under rather special conditions, and several studies-the Kelly-Fiske report being the best known-have shown that these conditions are often not present in assessment procedures.
Kelly, in a review article(') comments that "The curious state of affairs wherein the most widely (and confidently) used techniques are those for which there is little or no evidence is indeed a phenomenon appropriate for study by social psychologists. This reviewer can only assume that in the absence of evidence of their predictive validity, such techniques must serve an important function other than assessment". The purpose of this note is to suggest that much of the research work carried out by social psychologists on group consensus is relevant to the proceedings in assessment groups. A n examination of the factors that seem to be related to high degrees of agreement shows that many of them are irrelevant to the goal of valid assessment. The principal factors are: