Responses to information were facilitated by the rapid prior presentation of evaluatively congruent material. This fundamental discovery (R. H. Fazio, D. M. Sanbonmatsu, M. C. Powell, & F. R. Kardes, 1986) marked a breakthrough in research on automatic information processing by demonstrating that evaluative meaning is grasped without conscious control. Experiments employing a word naming task provided stringent tests of the automaticity of evaluation and found support for it. More strikingly, a previously unobserved reversal of these effects (i.e., slower responses to evaluatively matched rather than mismatched items) was found when primes were evaluatively extreme. Procedural variances across 6 experiments revealed that the reverse priming effect was highly robust. This discovery is analogous to demonstrations of contrast effects in controlled judgments. It is theorized that the reverse priming effect reflects an automatic correction for the biasing influence of the prime.