Fig. 1. (A–C) An averaged male (A) and female (C) faces, created using morphing techniques. The consensus of these two averaged faces was then calculated to create an androgynous averaged face (B).

Fig. 2. (A–D) Examples of face stimuli used in the experiment. (A) and (C) show the original faces and (B) and (D) are the stimulus faces created using the geometry of (A) and (C) and the androgynous average texture and then masking the hairs and hairline.

Fig. 3. (A–C) Faces with the lowest (PM score near 0.1; A) and highest perceptual masculinity (PM score near 1; C) values and one face with a PM score of 0.5 (B).

Fig. 4. Distribution of PM values. Median perceptual masculinity (PM) values are visibly different for faces of male and female sex. However, 67% of the faces have a PM score within the 0.2–0.8 range and can be considered as perceptually ambiguous.

Fig. 5. Traditional ERPs analysis. Mean electric activity (electrode: T6) recorded for both male and female observers for faces with low PM values (left), with high PM values (right) and for ambiguous stimuli (center).

Fig. 6. Temporal profile of observer's ICA8 amplitude, respectively, in males (M), females (F) and both sexes (M + F). The ordinates are in ms and the abscissa is in arbitrary units.

Fig. 7. Scalp spatial distribution of ICA8. Note the right parieto-temporal location of the ICA8 over the scalp.

Fig. 8. Correlation of PM and ICA8 latency in male (solid circles) and female observers (open circles). Each point represents the latency of the response of an individual face. The solid lines represent the best linear fit; dotted curves are 95% confidence intervals for the linear fit.

Fig. 9. Spatial reconstruction of the source dipole in axial (A), coronal (B) and sagittal (C) projection. The dipole source is identified by the white arrow. The reconstruction are based on an averaged human brain.