Diagnosing prosopagnosia: Effects of ageing, sex, and participant–stimulus ethnic match on the Cambridge Face Memory Test and Cambridge Face Perception Test

Abstract

The Cambridge Face Memory Test (CFMT) and Cambridge Face Perception Test (CFPT) have provided the first theoretically strong clinical tests for prosopagnosia based on novel rather than famous faces. Here, we assess the extent to which norms for these tasks must take into account ageing, sex, and testing country. Data were from Australians aged 18 to 88 years (N = 240 for CFMT; 128 for CFPT) and young adult Israelis (N = 49 for CFMT). Participants were unselected for face recognition ability; most were university educated. The diagnosis cut-off for prosopagnosia (2 SDs poorer than mean) was affected by age, participant–stimulus ethnic match (within Caucasians), and sex for middle-aged and older adults on the CFPT. We also report internal reliability, correlation between face memory and face perception, correlations with intelligence-related measures, correlation with self-report, distribution shape for the CFMT, and prevalence of developmental prosopagnosia.

Keywords:

• Prosopagnosia
• Face recognition
• Ageing
• Gender
• Other-race effects
• Aging

Acknowledgments

Supported by the following grants: Australian Research Council DP0450636 and DP0984558 to E.M., Economic and Social Research Council RES–061–23–0400 to B.D., Australian Research Council Special Research Centre for Cognitive Science and Cognitive Neuropsychology S0001507 for work at Macquarie Centre for Cognitive Science (MACCS), and Marie Curie IRG 046448 to G.Y. We thank Kate Crookes for assistance in manuscript preparation and Prof Evelyn Teng of the University of Southern California for providing instructions on administering and scoring the MAT test.

Notes

¹ Ethnic background of Caucasian New Zealanders is similar to that of Caucasian Australians.

² Statistical results with the MAT exclusion applied did not differ in any important ways from those reported.

³ Of course, education level in our sample may be only rather loosely related to intelligence, partly because many of the young adults were only beginning university degrees, and partly because for the older groups (particularly the over-65s) university education in Australia was less widely available than today, particularly for women.

⁴ Supporting evidence for this interpretation was obtained. For each individual target face in the CFPT, we calculated the mean error score (i.e., averaging across participants) and correlated this with the “default” error score for that face—namely, the score that would be obtained if participants made no changes at all to the initial left–right ordering of the faces on each trial. If CFPT participants' perception of facial identity is reliable, we would expect the correlation to be zero; that is, starting-point ordering should not matter. Results showed the expected independence of starting and final order for upright faces (r = –.19), but there was some evidence of a relationship for inverted faces (r = .43), suggesting that participants to some extent tended to leave the inverted faces in their original order.

⁵ A very similar correlation of r = .27 was found by Germine, Duchaine, and Nakayama (2009), in which 47,471 participants rated their face recognition relative to the average on a 5-point scale and completed an online version of the CFMT that used different faces from the original CFMT.

⁶ This can also occur with congenital perceptual deficits. Even with colour blindness, a phenomenon that is very well known in the general community (unlike prosopagnosia), affected individuals can make it to adulthood without realizing they are colour blind.

⁷ For a report on people from the high end of the distribution, see Russell, Duchaine, and Nakayama (2009).