The face aftereffect spreads over changes in position, orientation and size in retinotopic, not space- or object-based coordinates

We examined the coordinate frame of face aftereffects (FAE) by measuring the FAE following eye movement, head movement, head rotation, or stimulus movement. 1) Following adaptation to a face at one location, subjects made a saccade. A test face was then presented either at the same location as the adaptor on the screen, the same retinal location as the adaptor or a different location with the same eccentricity as the adaptor. 2) Subjects tilted their head 45 deg to the right during adaptation, then tilted their head 45 deg to the left to view the test. The test was displayed with various orientations including the same retinal angle as the adaptor or the same angle on the screen as the adaptor. 3) Following adaptation, subjects moved their head and halved their distance to the monitor. The test face was then presented at various sizes including the same screen size and the same retinal size (half of the screen size) as the adaptor. 4) The adapting face turned around after adaptation, showing the blank back of the head, and moved to a new location. The test face was then presented at various locations including the original and the new location of the adaptor (where only the back of the head had been presented). In all experiments, the face aftereffect was strongest at the retinal position/angle/ size of the adaptor. There was substantial spread of the FAE across location, size and orientation but no additional benefit was found for test locations, sizes or orientations fixed in display-based or object-based coordinates (spatiotopy or object-otopy) rather than retinal coordinates. Our findings suggest that face analysis is grounded in a retinotopic coordinate frame and that spatiotopy across head and eye movements is not constructed at the level of visual feature/object analysis.