Cognitive Neuroscience Talks
Using psychophysics and facial statistics to understand the information underlying face identification
It is more difficult to identify a face when it is viewed upside-down, or “inverted”, compared to when it is viewed upright. This result, known as the face inversion effect, has led many researchers to believe that qualitatively different kinds of information underlie upright and inverted face identification. This belief is so widespread that the face inversion effect is often used as a tool for revealing those facial cues that are most critical for normal face recognition. However, the results from the first part of my PhD thesis demonstrate that upright and inverted faces are identified using very similar kinds of local information around the eyes and eyebrows. In addition, upright and inverted face recognition rely on the same narrow band of spatial frequencies around 9 cycles per face. Rather than using the face inversion effect to make inferences about the information supporting face identification, we decided to take a new approach by considering the statistical variation of feature-spacing in a real population of faces, and examining individual differences in a realistic face identification task. Our findings show that face identification accuracy is correlated with thresholds for the discrimination feature-spacing. However, people’s sensitivity to these cues is not sufficient to resolve the variation in feature-spacing that exists in a large population of faces. Overall, the results of my PhD thesis suggest that, contrary to popular notions, local structure, especially around the eyes and eyebrows, may be very important for face processing. Surprisingly, we find that this conclusion is actually consistent with the Thatcher Illusion. The final part of my talk discusses how this illusion, the psychophysics of face identifiation, and the statistics of facial variations, all converge on the importance of a single facial area - the eyes and eyebrows.