Curve Fitting
For each participant, we calculated the proportion of the eight trials at each direction of gaze on which the model was judged to be looking directly toward the participant, to the left, and to the right (see Fig. 2). To quantify sensitivity to eye contact, we fit logistic functions relating each participant’s proportion of left and right responses to the directions of gaze. All fits were carried out using the glmfit routines from the Statistics Toolbox in Matlab (R2008a). The sum of the left and right fitted functions was then subtracted from 1 to define a third function fitting the proportion of direct responses. Goodness of fit was within acceptable parameters described in Vida and Maurer for all fits. we calculated the width of the cone of gaze as the difference (in pixels) between the points of intersection between the fitted ‘‘direct’’ function and the left and right functions. These points of intersection correspond to the directions of gaze where the participant was equally likely to judge that the model was making eye contact or looking away. The distance between the right and left points of intersection provides a measure of the width of the horizontal cone of gaze. Three individuals in the ASD group were replaced because their response curves were too broad to allow confident estimation of the width of the cone of gaze in at least one condition.