The face-number effect: a new test of face discrimination

Abstract

Background: The recent many-to-many hypothesis claims that faces and words share for neural resources, leading to studies comparing face and visual word recognition. However, one difficulty of these studies is that the tests used to assess face and visual word processing and diagnose alexia and prosopagnosia differ in their measures. The key feature of alexia is an elevated word-length effect, while the diagnosis of prosopagnosia rests upon reduced accuracy on face recognition tests. Objective: Our goal was to develop an assessment of face discrimination that a)had response time as the outcome variable, and b)indexed the perceptual processing load. Method: We created two tests, in which subjects made a same/different judgment with an array of faces that could vary in the number of faces shown. In the first test, each trial showed a display of faces that varied slightly in expression. On half the trials all the faces were from the same individual, while on half, one of the faces was of a different person. In the second test, the expressions were either all the same or differed in one face. Accuracy and response times were measured. Results: Accuracy was similar in both healthy and prosopagnosic subjects. The time needed to respond correlated with the number of faces in the array. Thus, we defined the 'face number effect' as the slope of the linear regression between the number of faces and the response time. For identity, this was about 200 ms/face in healthy subjects, and 600 ms/face in prosopagnosic subjects. For expression, the face-number effect was about 170ms/face, and slightly increased at a mean of 375ms/face in prosopagnosic subjects. Conclusion: The face-number effect can reveal a reduced perceptual capacity for processing face identity and may provide a parallel to the word-length effect, in assessing processing as a function of perceptual load.

Meeting abstract presented at VSS 2018