The effect of self-affirmation on reasoning biases was explored. After participants wrote about a value that was important to them (self-affirmation) or a value that was not important to them (no affirmation), they tested a hypothesis using a task commonly used to study the confirmation bias (Study 1) and assessed correlation from data presented in a 2 x 2 frequency table (Study 2). In both tasks, participants assessed the validity of a hypothesis that had either threatening or nonthreatening implications for their self-concepts. Nonaffirmed participants who tested threatening hypotheses exhibited the confirmation bias less frequently (Study 1) and assessed correlation more accurately (Study 2) than self-affirmed participants or participants who tested nonthreatening hypotheses. Results support models of motivated reasoning that propose that information processing is altered in response to threatening information. By ameliorating the threat, self-affirmations can elicit less effective reasoning strategies.