Home Research Research Highlights
Red Men, Green Women
What do we know about the color of men and women?

Michael Tarr, Brown University

Surprisingly, there are reliable color differences between the faces of men and women. For instance, Caucasian men are redder than Caucasian women. At issue is whether we unconsciously “know” this fact and use it to help determine the gender of a face. Adrian Nestor and Michael J. Tarr, from Brown University and members of the NSF-funded Temporal Dynamics Learning Center, examined whether observers make use of color as a cue in facial gender recognition, using a method called ‘superstitious perception’ (Gosselin & Schyns, 2003). This technique is based on the fact that we often ‘hallucinate’ meaningful structure in random noise, for example, seeing shapes in clouds. In this study, they showed subjects pictures containing chromatic noise on top of a single gender-neutral face but we told subjects that each image actually contained a noisy male or female face whose gender they had to identify.

The researchers found that observers systematically exploit color information when classifying the androgynous face embedded in chromatic noise. Moreover, they were able to identify the specific spatial locations and tease apart the components of color used by observers. Given the nature of the task, this consistent use of color in gender recognition must have arisen from observers’ pre-existing mental representations of the categories in question, male and female faces. Although the color patterns found with human observers did not completely mirror objective natural color differences, humans were sensitive to the actual contrast between the main features and the rest of the face. Perhaps more interestingly, the results suggest this sensitivity is not restricted to luminance, but extends to the red:green and yellow:blue color channels. Overall, the study provides evidence that observers encode and can use the local color properties of faces in tasks where the availability of other cues is reduced.


From left to right: androgynous base image, one example of an experimental stimulus (base image plus noise) and male / female prototypes reconstructed from classification results across observers.