Two studies were conducted to test the hypothesis that stronger degrees of ambivalence attenuate the attitude-behavior correspondence. We also tested the hypothesis that the weaker attitude-behavior correspondence at higher levels of ambivalence may be due to the lower temporal stability of highly ambivalent attitudes. To ensure that these hypotheses be adequately tested, both studies had a prospective design. After assessing their attitudes toward certain behaviors and ambivalence, participants recorded how often they performed these behaviors during the subsequent 14-day period. The investigated behaviors included everyday behaviors such as consuming fast food, preparing classes, and watching entertaining programs on TV. To assess attitudinal stability, attitudes were measured again after the 14-day period. As the results show, the expected moderating effect of experienced ambivalence on the attitude-behavior correspondence was obtained, however, only for a minority of the investigated behaviors. As predicted, experienced ambivalence was related to decreased attitudinal stability in each of these cases. A meta-analytic summary of the two studies revealed that, across different behaviors, a weak but reliable moderating effect of ambivalence on attitude stability and the correspondence between attitudes and behaviors exists.