An impairment in abstracting ability has frequently been proposed as a reason for schizophrenic thought disorder. The performance of hospitalized chronic paranoid schizophrenics and non-paranoid schizophrenics were compared to a normal control group on two types of abstraction; a traditional conceptual abstraction task (similarities, Trunnell, 1964) and an inferential abstraction task (relational abstraction, Bransford, Barclay & Franks, 1972). These two measures allowed a differential interpretation of the nature of the abstraction impairment in schizophrenia. The two clinical groups did not significantly differ on the traditional hierarchical measure of abstraction. Performance of both schizophrenic groups, however, differed significantly from that of controls in that schizophrenic subjects employed less abstract concepts to classify items in this task. On the second measure of abstraction no significant differences were found between schizophrenic subjects and the control group. Differences between paranoid and non-paranoid subjects did not reach significance on this task but there was some indication that each of these schizophrenic sub-groups used different cognitive strategies on this measure. Paranoid schizophrenics appeared not to elaborate information beyond its original form. The non-paranoids, on the other hand, appeared to elaborate stimulus material but were confused between inferential and original information. The present results indicate that chronic paranoid schizophrenics have a different type of abstraction impairment to chronic non-paranoid schizophrenics on the inferential conceptual abstraction task. These findings indicate the utility of using two indices of abstraction and the importance of not treating schizophrenics as a homogeneous group.