Recent evidence suggests that certain features on the human face indicate hormonal levels during growth, and that women judge the attractiveness of potential partners based on the appearance of these features. One entrenched notion is male facial features that are affected by testosterone are used as direct cues in mate preference. Testosterone may be particularly revealing as it is purported to be an honest indicator of male fitness. Increased testosterone may impose an immunocompetence handicap on the bearer and only the best males can carry this handicap. To date, tests of this theory have been indirect, and have relied on digital manipulations that represent unrealistic continuums of masculine and feminine faces. We provide a much more direct test by manipulating digitally male faces to mimic known shape variation, caused by varying levels of testosterone through puberty. We produced a continuum of faces that ranged from low to high levels of testosterone in male faces and asked women to choose the points on the continuum that appeared most attractive and most physically dominant. Our data indicate that high testosterone faces reveal dominance. However, there is no evidence of directional selection for increased (or decreased) testosterone in terms of attractiveness to the opposite sex. We discuss the relevance and applicability of evolutionary interpretations of our data and, contrary to predictions, provide evidence of stabilizing selection acting on testosterone through mate preferences.