Numerous studies have replicated the finding that schizophrenia patients make an increased number of errors on an antisaccade task. Some studies have reported that relatives of schizophrenia patients also make an increased number of antisaccade errors, a finding that has been interpreted to support the usefulness of compromised antisaccade performance as an index of genetic liability for schizophrenia. We examined performance on an antisaccade task in schizophrenia patients, nonpsychiatric controls, first-degree relatives of schizophrenia patients and first-degree relatives of nonpsychiatric controls. Schizophrenia patients made significantly more errors than did nonpsychiatric controls, but relatives of schizophrenia patients did not differ from relatives of controls or from all controls. Increased antisaccade errors on the standard version of the antisaccade task are associated with schizophrenia, but do not seem to be a co-familial trait for schizophrenia.