An Attempt at Validation of Traditional Psychiatric Syndromes by Cluster Analysis

Abstract

During the last twenty years countless attempts have been made to refine, replace or validate our intuitive clinical classifications of mental illness by the statistical analysis of symptom ratings. By and large these attempts have been unsuccessful. In spite of all the elegant and complex manipulations to which innumerable sets of data have been subjected, the classifications used by psychiatrists remain much as they were before the computer was invented. The reasons for this failure—for as such it must be seen at least by clinicians—are complex, but we would attribute it mainly to two factors: a failure to appreciate how unreliable and open to bias are the clinical ratings on which most analyses have been based, and too exclusive a reliance on factor analysis at the expense of other forms of multivariate analysis.