Eye-tracking deficits in schizophrenic patients and relatives have generated interest in using eye movements to mark schizophrenia liability. Efforts to develop specific quantitative measures have provided insight into the nature of the deficit and suggested what underlying neurophysiological mechanisms are involved. This study used receiver operating characteristic curve analyses to evaluate and compare the sensitivity, specificity, and overall accuracy of predicting disease liability using single and combined specific and global quantitative measures. Results indicate that measures of predictive pursuit and leading saccades significantly increased predictive accuracy compared with traditional global measures. Combining specific measures provided greater predictive accuracy compared with single measures. Implications for the use of specific eye movement measures to define schizophrenia-related phenotypes in genetic studies are discussed.