Scientists and philosophers have searched for centuries for a parsimonious answer to the question of what constitutes beauty. We approached this problem from both an evolutionary and information-processing rationale and predicted that faces representing the average value of the population would be consistently judged as attractive. To evaluate this hypothesis, we digitized samples of male and female faces, mathematically averaged them, and had adults judge the attractiveness of both the individual faces and the computer-generated composite images. Both male (three samples) and female (three samples) composite faces were judged as more attractive than almost all the individual faces comprising the composites. A strong linear trend also revealed that the composite faces became more attractive as more faces were entered. These data showing that attractive faces are only average are consistent with evolutionary pressures that favor characteristics close to the mean of the population and with cognitive processes that favor prototypical category members.

We reported in this journal (Langlois & Roggman, 1990) findings showing that attractive faces are those that represent the mathematical average of faces in a population These findings were intriguing because they provided a parsimonious definition of facial attractiveness and because they supported explanations of attractiveness from the point of view of both evolutionary and cognitive-prototype theory Since our 1990 report, several alternative explanations of our findings have been offered In this article, we show that none of these alternatives explains our results adequately