Much evidence outside the overconfidence literature indicates that confidence in a particular hypothesis is influenced more by evidence for and against that hypothesis than by evidence for and against the alternative. This article focuses on the impact that underweighting the alternative has on overconfidence and other related measures. Computer simulations showed that underweighting the alternative is sufficient for producing overconfidence under quite general conditions. In addition, data from two previous empirical studies were reanalyzed. In these simulated medical diagnosis studies, one group of subjects was known to take into account the alternative, and one to underweight it. The pattern of differences between the two groups was similar to the pattern found in the computer simulations where weighting of the alternative was manipulated. Furthermore, encouragement to take into account the alternative had no effect on the former group's behavior, but affected the latter group in a manner predicted by the simulations. Both the simulations and reanalyses provide direct evidence of the effect of underweighting the strength of alternatives on overconfidence.