Judgments of genetic relatedness of children

Our object was to estimate the information about genetic relatedness visually available from facial photographs of children. The stimuli were forty-eight pairs of photographs of children. In half of the pairs, the children were siblings; in the other half, the children were not. Forty-eight observers, in a signal-detection task, judged which pairs depicted siblings. There were three occlusion conditions: sixteen observers viewed full-face photographs of the children, sixteen viewed the same photographs with the eye regions of both children masked, and sixteen viewed the photographs with the mouth regions masked. For each occlusion condition, we converted the observers' performances into estimates of Shannon information, an alternative to signal-detection measures that require distributional assumptions. The maximum possible information conveyed per pair is 1 bit. The results for each occlusion condition were: 0.36 bit (full face), 0.47 bit (eyes masked), 0.63 bit (mouth masked). We find that observers can extract considerable information about genetic relatedness from photographs of children's faces. It is remarkable that observers were able to extract more information with parts of the face occluded than they could from full faces, as if the intact face masked the facial features relevant to the task