Inversion leads to qualitative changes in face processing but not in word processing

Abstract

Face and visual word recognition are two prime examples of expert visual processing in humans. Both activate the fusiform gyrus, a key region in visual expertise, and show an inversion effect - a marked reduction in performance when stimuli are turned upside down - which is attributed to an orientation-dependent expertise because our experience with these stimuli is biased towards the upright orientation. Here, we investigated inversion effects on two invariant aspects of visual performance, efficiency and equivalent input noise. We hypothesized that visual objects like faces and words, for which we have developed expertise, would show significant gains in efficiency and/or reductions in noise when stimuli were seen upright rather than inverted, and that this effect would be greater than that for an object type (e.g., houses) for which we are not expert. Twelve subjects performed five-alternative forced-choice tasks, in which we measured identification contrast thresholds for three stimulus categories (faces, words, houses) in two orientations (upright, inverted) under two noise conditions (no-noise, white noise). Efficiency was calculated relative to an ideal observer. We found that efficiency was greater for visual words than faces, and reduced for houses compared to words and faces. Inversion profoundly reduced efficiency for faces, while it had only a modest effect on efficiency for visual words and houses. Equivalent input noise was slightly higher for faces than for visual words, but was not affected by inversion for any stimulus. These results show that even though face and word recognition are both expert processes with known inversion effects, this orientation-dependent expertise differs in its impact on efficiency. While one of the orientation-specific gains in face processing is enhanced efficiency, this is not the case for visual words, suggesting that there is a different origin for the inversion effects reported for visual word processing. Meeting abstract presented at VSS 2018