We found evidence previously of familial aggregation of psychotic symptoms in 65 bipolar disorder pedigrees. This finding, together with prior evidence from clinical, family, neurobiological, and linkage studies, suggested that psychotic bipolar disorder may delineate a valid subtype. We sought to replicate this finding in 69 new bipolar disorder pedigrees. The presence of psychotic symptoms, defined as hallucinations or delusions, during an affective episode was compared in families of 46 psychotic and 23 non-psychotic bipolar I probands ascertained at Johns Hopkins for the NIMH Bipolar Disorder Genetics Initiative. There were 198 first-degree relatives with major affective disorder including 90 with bipolar I disorder. Significantly more psychotic proband families than non-psychotic proband families (76% vs. 48%) contained at least one affected relative with psychotic symptoms. Psychotic symptoms occurred in 35% of relatives of psychotic probands and in 22% of relatives of non-psychotic probands (P = 0.10). Both psychotic affective disorder generally and psychotic bipolar I disorder clustered significantly in families. These results are consistent with our prior report although the magnitude of the predictive effect of a psychotic proband is less in the replication families. Our findings provide modest support for the validity of psychotic bipolar disorder as a subtype of bipolar disorder. This clinically defined subtype may prove more homogeneous than the disorder as a whole at the level of genetic etiology and of neuropathology/pathophysiology. Families with this subtype should be used to search for susceptibility genes common to bipolar disorder and schizophrenia, and for biological markers that may be shared with schizophrenia.