This series of studies evaluated (1) hypotheses that poor antisaccade performance is attributable to confounding variables (e.g., visual attention deficits, incomplete understanding of task demands) and (2) the specificity of poor antisaccade performance to schizophrenia. In addition to self-correcting errors before being cued to do so, schizophrenia patients also showed the expected saccadic reaction time changes to fixation condition manipulations: decreased latencies for gap and increased latencies for overlap trials. These data suggest that schizophrenia patients are adequately engaged in and understand the antisaccade task. Schizophrenia patients made fewer correct antisaccade responses than other psychiatric patients (obsessive-compulsive and bipolar disorder) and normal subjects. The first-degree relatives of schizophrenia patients also generated a decreased proportion of correct antisaccade responses compared with normal subjects. For schizophrenia patients who performed below the range of normal subjects, 26% of their relatives also performed below the normal range. Conversely, patients who performed normally did not have a single poor-performing relative. These data suggest that increased antisaccade error rates may index a liability for schizophrenia within a subset of families.