The authors examine somatization disorder in a community population, using grade of membership analysis, a new multivariate analytical technique for the analysis of medical classification. The technique is used to examine whether somatic symptoms will cluster into a clinical syndrome resembling somatization disorder, as defined by the Diagnostic and Statistical Manual of Mental Disorders (DSM-III), if no a priori assumptions are made about the interrelationship of somatic symptoms or their clustering into clinical syndromes. Grade of membershp analysis is applied to all respondents in the US National Institute of Mental Health Epidemiological Catchment Area Project of the Piedmont region of North Carolina reporting three of more somatic symptoms from the somatization disorder section of the Dagnostic Interview Schedule. The analysis indicates that seven ‘pure’ types, roughly analogous to clusters in cluster analysis, best describe the interrelationship of the symptoms included in the analysis. One ‘pure’ type in the analysis is nearly identical to DSM-III somatization disorder and is associated with demographic characteristics consistently found among patients with DSM-III somatization disorder. The present results demonstrate that symptoms associated with this disorder do cluster in a highly predictable fashion and represent a strong validation of the natural occurrence of an entity resembling somatization disorder.