Literature on inductive reasoning shows that when testing hypotheses, people are biased toward the use of confirmatory strategies (P. C. Wason, 1960). In the present article, the authors presented 2 studies showing how people use confirmation and disconfirmation strategies during actual interaction in problem solving. Study I showed that participants were able to learn to use disconfirmation when confronted with a low-competence, nonthreatening partner. When the partner was high in competence (thereby threatening the participant's competence), participants used confirmation, even when the partner used disconfirmation. In Study 2, the authors aimed at generalizing the aforementioned results by exploring the hypothesis that disconfirmation stems from the possibility of diverging from norms. Participants who were confronted with the violation of a conversational norm used a high proportion of disconfirmation, whatever the source of influence. When there was no violation but there was a low-competence partner, the proportion of disconfirmation was high; when there was no violation but there was a high-competence partner, the proportion of disconfirmation was low. The authors discussed the interpersonal functions of confirmation and disconfirmation.