Inspection of Fig. 6 indicates that thresholds were lower for directions of gaze that diverged more from straight ahead. This result indicates that for larger shifts of gaze, there was more information available to discriminate between leftward and rightward gaze. This result is expected because, as shown in Fig. 1, a larger shift of gaze in a particular direction causes the eyes to appear less similar to eyes in which gaze is shifted in the opposite direction. Critically, Fig. 6 also indicates that angry and neutral faces produced identical thresholds, whereas fearful faces produced lower thresholds. Note that this will be true whether the stimulus is upright or inverted. This result suggests that there is more information available to perform the task in fearful faces than in angry or neutral faces, with no difference between the latter two. This pattern is different from our finding that in both the typical and ASD groups, the cone of gaze was wider for angry faces than for fearful or neutral faces, with no difference between the latter two, and that the effect of expression was limited to upright faces. Hence, our ideal observer analysis suggests that the wider cone of gaze for upright angry faces does not arise from a lack of low-level visual information available for discriminating the direction of gaze.