Clarification of the nature of the liability to schizophrenia transmitted within families is a major goal of family studies. In the Roscommon Family Study, the risk in relatives of schizophrenic vs. control probands was significantly increased for psychotic affective illness (PAI), but not for non-psychotic affective illness (NPAI) or alcoholism (ALC). We attempt to confirm these findings independently by examining, within families of schizophrenia spectrum probands, the relationship between PAI, NPAI and ALC in parents and the risk for schizophrenia spectrum disorders in siblings. A parental diagnosis of PAI predicted a significantly increased risk for schizophrenia, affective illness and anxiety disorders in siblings. In particular, schizophrenia in siblings was predicted by a parental diagnosis of mood-incongruent PAI. By contrast, neither a parental diagnosis of NPAI nor a parental diagnosis of ALC was associated with an increased risk for schizophrenia spectrum disorders in siblings. Consistent with our earlier findings, these results suggest that (a) the familial liability to schizophrenia is neither extremely non-specific nor extremely specific and (b) PAI, and particularly mood-incongruent PAI, may occur particularly in parents of schizophrenics because, although it reflects a significant familial liability to schizophrenia, it does not markedly impair the ability to marry and reproduce.