An experiment is reported in which the attribution of gender to isolated facial features and to faces whose features have been interchanged with those of a face of the opposite gender has been examined. Sixteen male faces were averaged to create a prototype male face and sixteen female faces averaged to create a prototype female face. The prototypes were then masked to exclude ears, neck, hair, and hairline. Individual features (brows, eyes, nose, mouth, and chin) and pairs of features (brows & eyes, eyes & nose, nose & mouth, mouth & chin) from the prototypes were then presented in isolation for classification according to their perceived gender. The results showed that, for the faces used, the brows & eyes, brows alone, eyes alone, the whole jaw, the chin, the nose & mouth, the mouth alone (in descending order), ie all the features except the nose, carried some information about gender when they were seen in isolation. In the second part of the experiment different features from one prototype face were grafted into the prototype of the opposite gender and the resulting composite faces classified by their perceived gender. The results of this feature substitution showed the jaw, brows & eyes, chin, and brows (in descending order) effecting significant change in perceived gender. The difference in gender information carried by feature(s) when they were viewed in isolation from that when they were substituted for each other is attributed to the role of configuration in the perception of the gender of a face.