Mental Rotation of Naturalistic Human Faces Jennifer Zhu Stephen P. Joy, Ph.D. Amity Regional High School Albertus Magnus College Results Abstract As shown in Figure 2, accuracy was greatest when the target face was shown in full-face (the original orientation). Accuracy declined in a linear fashion the greater the angle of rotation. When the target appeared in profile (90o rotated), accuracy was reduced to an approximately chance level. As shown in Figure 3, reaction time was most rapid when the target was presented in full face and increased linearly with the angle of rotation – with the exception that at 90o of rotation, when facial recognition was at chance levels, participants responded more rapidly than at 60o of rotation. With only three males in the sample to date, we cannot test meaningfully for gender effects. However, the overall accuracy of the males (mean = .51, SD = .11) did not differ appreciably from that of the females (mean = .49, SD = .08). Nor did reaction times appear to differ by gender. We studied mental rotation of human faces to see whether performance conformed to established findings using abstract shapes. Sets of faces (1 target, 2 foils) were presented in varying rotations under each target; 350 trials were completed per participant. Accuracy diminished, and reaction time increased, with increasing rotation. Introduction Mental rotation tasks involve identifying objects presented from different angles. Previous research has used abstract shapes (constructed of blocks) as stimuli and consistently found that greater degrees of rotation produce lower accuracy and longer reaction times; people actually rotate the mental image until they find (or fail to find) a match. Another consistent finding has been a male advantage on this task. Maeda & Yoon (2013) report a mean effect size of .57 (Hedges’ g). This is a typical stimulus set:   A few studies have used other types of stimuli, but no previous study has examined whether these principles hold for human faces. The closest approach involves rotating faces away from the vertical (Valentine & Bruce, 1988) or studies of the facial inversion effect, in which some faces are presented upside-down (Rakover & Teucher, 1997). Recognizing people from different angles is an essential social skill. However, since identification of human faces involves specific neural pathways (distinct from those used for analyzing spatial figures), it cannot be assumed that both tasks are performed similarly. Furthermore, the gender effect found with such shapes may not generalize to faces. Any modest male superiority on spatial tasks could well be offset by a modest female superiority in perceiving socially important cues. Method We created realistic human faces using FaceGenTM. Five male and 5 female target faces were generated. For each target, 2 foils were created; these were 30% randomized. This degree of relationship is comparable to that between siblings; the intent was for the foils to resemble, but be distinguishable from, the targets. In each trial, one target was presented in the upper portion of the screen, with the target and two foils in the lower portion. The three faces in the lower half could be presented in any combination of 7 orientations: full-face or 30, 60, or 90 degrees rotated to left or right. These rotations were randomized, as were the positions of the correct options. Trials were presented in blocks of 5, male and female blocks alternating. Altogether, 350 slides (35 male and 35 female blocks) were created. SuperLabTM software was used to present the stimuli and measure accuracy and reaction times.   Participants (N = 21 to date, including 18 females) were college undergraduates who received extra credit. We are aware of the need to add more participants, but our results (with the exception of gender analyses) are clear enough to present at this time. Figure 1 shows two of the slides: one male set and one female set. In the first case, the correct response is the face on the right, seen in full-face. In the second case, the correct response is the middle face, see in profile. Figure 1 Sample Stimuli STICIM Min Max M SD Interdependence 2.29 6.00 4.57 0.93 Sexual Intimacy 3.40 4.93 0.95 Comfort Intimacy 1.43 5.86 4.16 1.23 Trust 3.25 4.53 0.92 Warmth Motivation 2.50 1.00 Sexual Desire 3.20 5.21 0.91 Trust Motivation 3.75 5.18 0.64 IC Overall 3.08 5.84 4.55 0.77 IM Overall 3.70 4.99 0.70 Graphic Indicators M SD Colors 4.70 2.34 Warm Colors 2.45 1.59 DYR Overall 19.33 8.49 Discussion Not surprisingly, these findings suggest that people use a similar mental rotation mechanism when processing human faces to that used when processing abstract shapes. It also seems likely that the male advantage on mental rotation may be absent when human faces are used as stimuli, but this will require additional data. Such a finding would make evolutionary sense and would call in question the idea that males always display superiority on mental rotation tasks. The consistent male advantage identified to date may be an artifact of the type of stimuli used. Aside from the currently inadequate sample size, the major limitation of the present study was the fact that the faces turned out to be more difficult to distinguish than originally intended – so much so that accuracy diminished to chance levels beyond a certain degree of rotation. Future variations on this experiment should use faces with varying degrees of relatedness. It will also be interesting to test the effects of rotation along other axes and the possibility of interaction effects when more than one axis is involved. References Maeda, Y., & Yoon, S.Y. (2013). A meta-analysis on gender differences in mental rotation ability measured by the Purdue Spatial Visualization Test: Visualization of Rotations. Educational Psychology Review, 25, 69-94. Rakover, S.S., & Teucher, B. (1997). Facial inversion effects: Parts and whole relationship. Perception & Psychophysics, 59, 752-761. Shepard, R.N., & Metzler, J. (1971). Mental rotation of three-dimensional objects. Science, 171, 701-703.  Valentine, T., & Bruce, V. (1988). Mental rotation of faces. Memory & Cognition, 16, 556-566. For further information contact: sjoy@albertus.edu