Cross-correlation in face discrimination

An extensive body of literature suggests that face perception depends critically upon specialised face processing mechanisms. Although it seems clear that specialised face processing is required to explain face recognition, face discrimination is a simpler task that could possibly be solved with a general pattern discrimination mechanism. Observers were presented with face images that were either identical or not and judged whether they were the same or different. Face discrimination performance was well described by the point-by-point cross-correlation between the face images, which is a simple mechanism of the type used for discriminating patterns such as gratings. This result held for male and female faces viewed frontally or in profile. Results for inverted and contrast-reversed faces were also well described by cross-correlation, with observers having lowered efficiency relative to normal faces.

- Discrimination of simple 2D patterns is governed by their cross-correlation.
- We find that the same is true for male and female faces in frontal and profile view.
- Less efficient cross-correlation explains inverted and contrast-reversed results.