Several neuropsychological tests that ostensibly assess performance with face stimuli rely upon cognitive abilities not related to facial perception (Duchaine & Weidenfeld, 2003). Although recent efforts have provided better alternatives (Duchaine & Nakayama, 2006), we sought to develop a battery of face perception tests that could be used easily with a clinical (brain damaged) population. Our design goals included minimization of sensitivity to memory loss, and simplification of instructions and requirements for motor response. Additionally, the set of tests would include assessment of several aspects of internal face perception to provide for improved test stability. Four tests were created using computer-generated faces to assess discrimination performance for four face attributes: similarity, attractiveness, gender, and age. Each test presented between 166 and 197 trials and was completed by 62 to 117 subjects ranging between 18 and 81 years of age. The test trials were then narrowed to include the 75 trials for which between-subject agreement (across demographic populations) for each trial was greater than 80% and as a set provided average subject performance of 90%. Performance on the similarity, beauty, and age components was correlated across subjects (adjusted r2=0.45, 0.27, 0.49 respectively). The gender test performance was not correlated with the other subcomponents (adjusted r2=0.01). Performance was not significantly predicted by subject age, gender, race, or handedness. Years of education was positively related to performance on the similarity and age subtests, but not the gender or beauty components. We are now examining the test-retest reliability of scores within subject, and will assess the sensitivity and specificity of the test battery for perceptual deficits in brain damaged patients.
Burroughs-Wellcome Trust, NIMH