Abstract
We explore the relative importance of coding internal feature (e.g., eye) position relative to other internal features versus external head contour, and the possible role of head contour in coding feature position view-invariantly. All experiments employ the logic that size of a face aftereffect for a fixed physical deviation of adaptor gives a measure of sensitivity of coding along that axis in face-space, reflecting the slope of underlying two-pool neural response functions (see Susilo, McKone & Edwards, 2010, Vision Research). In Experiment Series 1, the baseline aftereffect was for +50 pixels vertical manipulation of eyeheight in a full head with all internal features and head contour present. Results were: (a) the aftereffect from a +50 pixels eyeheight adaptor in an internal-features-only stimulus was as large as baseline, arguing eye position is coded very accurately relative to only internal features; (b) the aftereffect was as large in an external-features-only stimulus (i.e., face and hair outline retained but nose and mouth ablated), arguing eye position is coded equally accurately relative to only external features; (c) the aftereffect from a more extreme eyeheight adaptor in an internal-features-only stimulus was larger, arguing linear 2-pool coding continues, in the absence of external contour, for eye positions that would normally be outside the head; and (d) the aftereffect from the extreme adaptor went to zero when head contour was added back in, arguing external contour trumps internal features and neural responses drop immediately to zero when the eyes are outside the face. In Series 2, we find mouth position aftereffects are as large from several manipulations breaking the first-order configuration of a front-view face (e.g., mouth left, mouth down-and-right) as from manipulations that do not (e.g., mouth up). This argues the purpose of the coding relative to head outline is to allow for 3D head rotation.
Australian Research Council DP0984558.