Previous findings suggest that the gender-stereotypic physical appearance of a person influences how this person is regarded. In earlier studies, the physical appearance was manipulated by preselecting stimulus persons with feminine- or masculine-looking faces. In our study, we used a method of image-manipulation to produce the stimuli. Each image was analyzed by actively reconstructing it with a Computer Graphics based 3D Morphable Face Model. These images were manipulated by adding a gender-vector that was learned from exemplar human heads. All heads in the model are labeled male or female. The gender vector is the gradient vector of the regression function that describes the labeled data best. By applying this vector to an image, we synthesized two novel photo-realistic images: a more feminine- and a more masculine-looking one. The information in the model for specifying gender is not only effective, but effective enough to induce the well-known social psychological phenomenon of stereotyping of males and females. Participants were asked to judge the social skills and aptitude of leadership applicants on the basis of images and fictive job references. Thus sex, appearance and attributes were manipulated independently. The results indicate that feminine-looking applicants are judged to be more socially skilled than masculine-looking ones, independent of their sex and attributed characteristics. Furthermore, feminine-looking applicants seem to suit best to the job when characterized as dominant, masculine-looking ones, when characterized as socially skilled. Our novel method of image manipulation has the advantage to let us change one parameter of faces independent of identity information.