There is considerable evidence that schizophrenia spectrum personality (SSP) disorders mark genetic risk for schizophrenia. Use of the spectrum phenotype in genetic and neurophysiological studies may prove informative. However, the degree to which the current diagnostic criteria correspond with genetic risk is unclear. This can be assessed by observing how measures of liability among SSP subjects change as a function of diagnostic criteria. In this study the generalized estimating equation method was used to assess changes in eye-tracking performance among SSP and non-SSP family and community groups employing various diagnostic criteria. Eye-tracking deficits among SSP relatives remained statistically higher compared with the other groups across progressively more liberal SSP criteria. The results suggest that fewer traits than are used in clinical diagnoses can effectively identify the spectrum phenotype among relatives of schizophrenia patients. Thus, reduced criteria may be used in research to increase "high risk" sample size and power to detect neurophysiological and genetic differences. Our results provide suggestive evidence that the use of clinical criteria in research may, in fact, underidentify at-risk individuals--potentially distorting genetic and neurophysiological findings.