Since I'm a white man with a white man's IQ, I was able to locate the actual study.
http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0031703#s2
Methods
The current experiment aimed to establish the relative attractiveness of individuals of three broad racial groups. These attractiveness ratings were made by people of a similar age to the individuals and of an opposite sex. The raters came from a range of different ethnicities and any differences in their ratings of different races were considered separately.
Ethics
The research was approved by Cardiff University School of Psychology research ethics committee. Informed written consent was obtained from all participants.
Participants
Forty undergraduates studying at Cardiff University took part as face raters either for course credit or for a small cash payment. Twenty were female and twenty were male and all were between the ages of 18 and 30 years. Of the male raters, 15 were White, 2 were Black and 3 were Asian. Of the female raters, 14 were White, 3 were Black and 3 were Asian.
Stimuli
Images of 600 people were used. Half were male. The images were collected from the social-networking website
Facebook.com. Images were selected from people who were members of groups associated with further and higher education bodies either in the UK (for White faces), sub-Saharan Africa (for Black faces) and East Asia (for Asian faces). These images were collected by a naive research assistant who selected images according to a set of criteria: Images had to show a clear view of a single person that was of sufficient quality such that it would be recognisable by a friend. If the face in the image had a weird expression or was possibly of a race other than the main race for the region then it was rejected. The face was also rejected if the person depicted looked to be under 18 or over 30 years old.
This method of stimulus generation was employed as being the best of those available to produce a representative set from each population. By using images that the individual freely posted on the internet to represent themselves, we avoid many of the problems associated with self selection had we used standardised photographs: that is, individuals who are particularly self conscious about their appearance will not volunteer. Selection bias is likely to be less true for facebook images as posting an image of oneself is what everyone else is doing. There would, of course, be a bias to select a photograph that presents one's most positive image but this bias would be present across all races and genders. One might assume that if these people were to use online dating websites then they would use an image similar to their facebook image in their profile. As such, this means that the stimuli employed are similar to the information presented during courtship, which is appropriate as marriage is the focus of the research.
One potential problem with this set of stimuli is the possibility that one group might be more reluctant to post images of themselves if they are less attractive than another. If this were the case then we would expect to see more facebook images that do not contain a face of the person (it might be left blank, be a scene or a cartoon character). Re-examination of the sources of the images used revealed that less than 1% of facebook accounts did not include a face of a person.
Procedure
Participants were presented with 300 opposite-sex faces, one at a time via a computer monitor. They rated each of these faces on their attractiveness. The faces were presented in a random order and the rating scale went from 1 (unattractive) to 10 (attractive)