While structured psychiatric interviews have generally succeeded in identifying higher rates of schizotypal personality disorder in relatives of schizophrenia versus control probands, self-report questionnaires designed to assess schizotypy have been less successful at distinguishing these groups of relatives. In the Roscommon Family Study, an epidemiologically based, case-controlled study conducted in the west of Ireland, relatives were administered a short form of Eysenck's Psychoticism scale and shortened and modified versions of the scales for magical ideation and social anhedonia developed by Chapman and colleagues. We compared, with relatives of matched controls, relatives of four proband groups: schizophrenia, other nonaffective psychoses (ONAP), psychotic affective illness (PAI), and nonpsychotic affective illness (NPAI). Only social anhedonia scores successfully differentiated, at modest levels of significance, relatives of schizophrenia versus control probands. Levels of magical ideation did not distinguish relatives of schizophrenia, ONAP, PAI, or NPAI probands from relatives of controls. Compared to controls, ONAP probands had significantly elevated psychoticism scores, but no such increase was seen in relatives of schizophrenia, PAI, or NPAI probands. Dimensions of schizotypy assessed at personal interview were significantly better at differentiating relatives of schizophrenia and control probands than our measures of social anhedonia, magical ideation, or psychoticism. Although psychiatric interviews in this sample have shown that clinically assessed schizotypal personality disorder and traits strongly aggregate in relatives of schizophrenia patients, of the three self-report instruments designed to assess schizotypy, only one even modestly identifies relatives of schizophrenia versus control probands. These results suggest that, compared with psychiatric interviews, self-report questionnaires are less successful at assessing underlying familial vulnerability to schizophrenia.