One of the longest standing problems in social psychology and personahty research is the inconsistency between self-reports and behavior (Liska, 1975). A classic example of inaccurate reporting of past behavior is found in La Piere's (1934) study of the relationship between verbal and behavioral indices of prejudice among hotel-restaurant managers. On the behavioral side. La Piere found that only one out of 251 hotel-restaurant establishments actually refused to serve a Chinese couple, but in contrast, on a subsequently mailed questionnaire only a small percentage replied that they would not refuse to serve Chinese. Certainly this is a conspicuous example of the possible invalidity of self-report measures, at least self-report measures directed at prior behavior.The problem seems equally acute in the attempt to gain predictive validity. For example, in an extensive series of studies on childhood morahty, Hartshorne and May (1930) found no significant correlations between stated belief in the wrongness of cheating and actual resistance to an opportunity to cheat. Recent reviews by Mischel (1969) and Wicker (1969a) have indicated that self-report forms are generally of only slight validity, and reasonably enough. Wicker (1969aWicker ( , 1969b has emphasized the necessity of uncovering variables that would mediate the connection between self-reports and behavior. It is our purpose here to examine what may be an influential factor.A variable that motivates a person to resolve cognitive inconsistencies, including attitude-behavior inconsistencies, is sug-1.

We investigated semantic priming effects on item recognition from short (8-word) lists in one reaction-time and three interruption speed-accuracy trade-off (SAT) experiments. SAT priming conditions included modest (0.35-s) prime durations; prime as a final list member; and long (1.5s) prime durations with special instructions. Analyses tested for constant increment (bias) priming (an equivalent increase in both hits and false alarms) and enhanced discrimination priming (differential priming for targets and lures). Constant increment (bias) priming was ubiquitous, but some subjects showed enhanced early discrimination in restricted conditions. In constant increment (bias) priming, the semantic relations between prime and test either additively increase familiarity or lower criterion. Only enhanced discrimination must imply an interaction between the prime and the retrieval process, due to either interaction or use of compound cues.