In light of current linkage studies in schizophrenia, research on the "schizophrenia spectrum" deserves increased attention for an exact determination of the affected phenotype: Those disorders that have a much higher prevalence among biological relatives of schizophrenia patients are supposed to share common etiological factors with "core" schizophrenia. However, there is controversy over which of the DSM-III-R personality disorders should be included in the spectrum. In a controlled family study of inpatients with a DSM-III-R diagnosis of schizophrenia (n = 101), schizophreniform and schizoaffective disorders (n = 69), and unipolar major depression (n = 160), familial rates of personality disorders were assessed through personal interviews and compared with prevalence rates in 109 control families from the community. As predicted, schizotypal personality disorder occurred more frequently in the nonpsychotic relatives of schizophrenia probands (2.1%) than in the families of unscreened controls (0.3%). Paranoid personality disorder was more frequent in relatives of probands with unipolar depression (2.9%) than in relatives of schizophrenia patients (1.7%), and controls revealed the lowest rate (0.9%). Schizoid personality disorder, however, was extremely rare in all sample groups (between 0.3% and 0.7%), providing no sufficient statistical power for detection of group differences. Further analysis of the DSM-III-R criterion symptoms of schizotypal personality disorder demonstrated that items describing "negative" symptomatology are the main source of familial aggregation, but "psychotic-like" personality features are also contributing factors.