Fig. 11

Overview and results from Experiment 4. (a) We showed the participants an image depicting a realistic scene with a salient, red ellipsoid (see Fig. 10) with a slant \(\theta\) w.r.t. the viewer. We required participants to fixate on a point with eccentricity \(\xi\). (b) We fitted one psychometric curve per angle \(\alpha\) between the target and the fixation, for six different angles (color-coded). The closer the fixation is to the target (i.e., the smaller the angle), the more accurately people detect the target’s width (and the closer the PSE is to 1.0). Larger angles (yellow-ish colored curves) present a clear overestimation of the target’s width (i.e., curves are shifted towards the left) (c) Point of subjective equality (PSE) for each angle \(\alpha\) (i.e., the width of an ellipsoid that would be perceived as a sphere), together with the lower (25%) and upper (75%) detection thresholds in dotted lines. As in (b), for smaller angles (i.e., lower \(\alpha\)), people accurately estimate the target being perfectly spherical, yet for larger angles they overestimate its width (e.g., for \(\alpha = 50^{\circ }\), viewers perceive an ellipsoid of width 0.85 as being perfectly spherical).