Researchers in a variety of disciplines have found that participants take less time and generate less diversity of responses when judging stimuli towards the ends of a scale than when judging those near the center. Three types of models, connectionist, exemplar, and anchor models, can account for these inverted-U effects. Anchor models assume that stimuli near the ends of the scale are used as anchors to compare with the other stimuli, implying that anchor representations are activated for each judgment. Therefore, participants should learn the anchors better than the other stimuli. Participants were 40 students from the Department of Psychology at McGill University (5 men; M age = 20.5 yr.; SD = 1.7). The experiment involved two tasks: first participants judged facial gender and then performed a recognition task. The results showed no correlation between the position on the gender scale and recognition accuracy. Several hypotheses were offered to explain these results.