Abstract
A recent report (Goffaux and Rossion, 2006) found that, while horizontal shifts of eye position are easily discriminated in inverted faces, vertical shifts are more difficult to perceive. Along with the demonstrated difficulty in perceiving vertical shifts of the mouth, this has suggested that the perception of vertical feature position may be relatively more vulnerable to changes in orientation. The reasons for this are unclear. One possibility is that vertical eye shifts differ from horizontal eye shifts in that horizontal shifts can be appreciated via ‘short-range’ relations to nearby features (e.g. the other eye), whereas vertical shifts might be indexed to more distant facial features. To test this hypothesis we contrasted shifts in vertical mouth position, horizontal eye position, and two types of shifts in vertical eye position. In one, both the eyes and brows were shifted en bloc, as in the prior study. In the second, only the eyes were shifted vertically, so that the unchanged brows acted as a short-range relation to index this type of shift. Subjects performed an oddity paradigm with both upright and inverted faces. We replicated the large inversion effects for vertical mouth position and vertical eye-and-brow position, and the small inversion effect for horizontal eye position. The critical condition, the vertical eye-alone position, showed less inversion effect than the vertical eye-and-brow position. This suggests that at least part of the effect of inversion on perceiving vertical eye position is due to the lack of short-range second-order relations for this type of shift. A failure to appreciate changes in configuration that depend upon long-range relations could be due to inefficient sampling of the whole face structure in inverted presentation, whereas preserved perception of short-range relations would be consistent with preserved local processing.