Although genetic influences on schizophrenia are well established, localization of the genes responsible for this illness has proven extremely difficult. Given evidence that genes predisposing to schizophrenia may be transmitted without expression of the clinical phenotype, efforts have focused on developing endophenotypes. While several neuropsychological measures have been proposed to be endophenotypes, few studies have systematically assessed batteries of neurocognitive tests to determine which tests are most sensitive to liability for the illness. Two hundred sixty-nine Latino individuals were administered a standard neuropsychological battery. Two hundred fourteen of these were members of families with at least two siblings diagnosed with schizophrenia or schizoaffective disorder. The remaining were community controls without history of major psychiatric illness. Neurocognitive measures found to be heritable were entered into analyses designed to determine which tests covary with the degree of genetic relationship to affected individuals. Although five measures were found to uniquely model genetic liability for schizophrenia, digit symbol coding was the most sensitive. To assess the specificity of these endophenotypes, performance on these measures were compared to family members with bipolar and unipolar affective disorders. These markers clearly distinguished between individuals with psychotic illnesses and those with major depression. As measures contributed uniquely to discriminate individuals at varying risk for schizophrenia, our findings imply multiple independently inherited elements to the liability for the illness. We present a practical model for adjudicating endophenotypes and determining which measures are best suited for use in linkage analyses. © 2006 Wiley-Liss, Inc.