The study of mental representations on projective tests such as the Rorschach has made substantial contributions to our understanding of psychopathology. It is not clear, however, whether the representations of paranoid schizophrenics can be differentiated from those of normals and whether the Rorschach is the best test for such comparison. Object representations on the Rorschach and a role-playing (Johnson & Quinlan, 1980, 1985) test were studied in groups of normal (n = 31) and schizophrenic subjects (divided into paranoid [n = 16], intermediate [n = 11], and nonparanoid groups [n = 16]). Developmental levels of representation on both tests were measured with the widely used system of Blatt, Brenneis, Schimek, and Glick (1976), derived from Werner's concepts of differentiation, articulation, and integration. Generally the groups were not differentiated on these measures on the Rorschach. The role-playing test showed greater discriminatory power than the Rorschach, possibly due to its explicit demand to produce representations of humans in interaction. Results of the role-playing test showed differentiation and integration scores were negatively correlated with measures of psychotic symptoms. Non-paranoids differed from the other three groups on differentiation measures, and paranoid and normal groups scored higher on integration measures than did the intermediate and nonparanoid groups. Paranoid subjects scored higher than nonparanoids on functional articulation and higher than intermediates on perceptual articulation; however, paranoid subjects were not differentiated from normals on any developmental measure. Differences between the schizophrenic sample and normals were explained entirely by the non-paranoid subgroup, supporting the paranoid-nonparanoid distinction and raising questions regarding the nature of the deficit in paranoid schizophrenia.