We recently reported that gender aftereffects, similar to those observed with front-view face images (e.g., Webster, Kaping, Mizokami, & Duhamel., 2004), also occur with face profile silhouettes (Davidenko, Winawer, & Witthoft, VSS 2006). Here we tested the extent to which these aftereffects are sensitive to differences in image properties between adapting and test faces. Because the silhouettes are defined by a single contour, it is important to determine whether aftereffects result from face-specific adaptation or are just due to general properties of the contours, such as tilt and curvature. In a brief implicit adaptation paradigm, subjects classified 8 male (or 8 female) face silhouettes on non-gender characteristics and then made a gender classification on a 9th, gender-neutral face silhouette. The nine faces were arranged in three rows on a single sheet of paper. Judgments were considered aftereffect-consistent if the 9th face was classified with the gender opposite the gender of the 8 previous faces. Gender aftereffects were undiminished by a reversal of left-right orientation (Study 1) or a reversal of contrast polarity (Study 2) between the 8 adapting silhouettes and the 9th test silhouette. The percent of aftereffect-consistent judgments was 84±12% (Study 1) and 83­±8% (Study 2), not significantly different from a previous study with neither transformation (79±7%; means ± 95% CIs). Invariance to these two image transformations suggests that face aftereffects occur at a relatively high level of representation. In Study 3, adapting to upside-down face silhouettes did not lead to significant aftereffects on the test silhouette (58±11% aftereffect-consistent judgments). Although all three manipulations modify the silhouette contours, only vertical inversion, which significantly reduces the face percept (Davidenko, 2007), eliminates aftereffects, suggesting that face-specific processing is necessary to induce face-aftereffects in silhouette profiles. We discuss the implications of these findings with respect to a face-space framework.