Traditional theories of confidence and information processing suggest that people engage in greater processing activity when they feel doubtful as opposed to confident. Recent research, however, has hinted at the possibility that this effect might be malleable. The current research tests a confidence-matching hypothesis to determine when increased confidence yields increased processing and when increased confidence yields decreased processing. Based on recent advances in research on matching effects, it is proposed that the classic negative effect of confidence on information processing will reverse when messages are framed in terms of confidence. This hypothesis is tested by presenting participants with a persuasive message containing strong or weak arguments under confidence or doubt conditions. As predicted, when the message is framed in confidence terms, participants engage in greater message processing when they feel confident rather than doubtful, as indicated by greater argument quality effects on attitudes and thought favorability.