Two studies explored the application of feature matching and cancellation models to self-other comparisons. In both studies, college participants completed a questionnaire about their own religious behaviors and received another copy of the questionnaire supposedly completed by another student. Participants appeared to "cancel out" behaviors shared by the self and other when rating the religiousness of the other student (i.e., they gave lower ratings when there was more overlap), but not when rating themselves. Participants in the first study (N = 114) who were explicitly provided direction of comparison instructions showed a direction of comparison effect, rating the person whose description they saw last as more religious. Participants in the second study (N= 103), who were not given explicit direction of comparison instructions and who viewed the descriptions in close succession, did not. Measures of self-esteem and self-clarity did not moderate effects, but there was some evidence that the self-other asymmetry was more pronounced in low schematics for religiousness.