The authors adopt a construct validity approach to examine the structural validity of diagnoses of paranoid, schizoid, and schizotypal personality disorders. Systematic descriptions of these diagnoses were developed based on features identified from the literature that were organized using the clinicians' ratings. Each diagnosis was described in terms of a series of behavioral dimensions. Sets of behavioral exemplars were developed to assess each dimension. Dimensions and diagnoses were found to exhibit satisfactory levels of internal consistency that were cross-validated in a general population sample of 274 subjects and a clinical sample of 133 patients with a diagnosis of personality disorder. Some dimensions did not correlate highly with the total diagnosis score. Diagnoses could probably be refined by eliminating these dimensions. The structural relationships between dimensions delineating each diagnosis were explored using factor analysis. Diagnoses were not unifactorial. Instead, each diagnosis was found to be composed of two factors. The factorial structure of each diagnosis was found to be similar in the general population and clinical sample. It is suggested that the results support a dimensional representation of these diagnoses using three dimensions: paranoid behaviors, social avoidance, and perceptual cognitive distortion.