Methods for improving gaze tracking are presented. These methods eliminate the need for a separate initial calibration step to determine whether the subject is properly centrally fixated during testing. In one embodiment, whether the subject is centrally fixated is determined by identifying inlier and outlier gaze descriptors; the inliers are assumed to be associated with central fixation. This reduces the total time required for testing. These methods also improve the accuracy of gaze tracking during testing without the need for individual subject calibration. Instead, a database of reference eyes is used. The subject's eye can be compared to the reference eyes, and the mapping function for one or more reference matches can be used to estimate the subject's gaze direction. In another embodiment, previous calibration data from the same subject can be used to determine whether the subject is properly centrally fixated and/or the direction of the subject's gaze.