A major implication of political sophistication research is that most Americans lack the knowledge necessary to make informed, rational judgments about issues, candidates, or parties they are voting for (e.g., Campbell et al ., I960; and Converse, 1964). Sniderman, Brody, and Tetlock (1991) challenge this view by arguing that voters can make accurate judgments by using attitude in place of knowledge. However, not all evidence supports their argument. To explore accuracy of policy judgments in greater depth, the following research question was asked: How do respondents' own attitudinal position and cognitive complexity as well as candidates' overall stated positions (which span a continuum of possible positions) lead to either distorted or accurate judgments about those same candidates' specific but unstated policy positions? Schema research offers conflicting answers to this question, so results were left open-ended. Two subsidiary research questions regarding recall and judgments of candidates' overall positions were also asked so that hypotheses from research on expertise and social judgment theory, respectively, could be used to interpret the openended results. The issue of abortion was selected to operationalize the study. Using an experimental design, a questionnaire was administered to 134 undergraduates. The pre-test measured abortion attitudes and cognitive complexity. The stimulus was a simulated newspaper article in which three different unknown candidates stated their positions on abortion (pro-choice, neutral, or prolife) and women's health care (the distractor). The posttest measured factual recall, judgments of overall position, and judgments of position on specific abortionrelated policies not addressed directly in the candidate's statement. For recall, higher levels of cognitive complexity, measured by knowledge, differentiation, or integration, predicted, as hypothesized, greater accuracy. Attitude and knowledge predicted accurate recall best overall, though influence from the former differed across candidate conditions. For judgment of candidates' overall positions, assimilation and contrast effects were found as hypothesized initially among pro-choice respondents only. Additional tests using the knowledge variable revealed assimilation and contrast effects among pro-life respondents and suggested that judgmental distortions occur most often among low knowledge respondents. For judgments of candidates' specific but unstated policy positions, schematic respondents were significantly more extreme than aschematic respondents at every statistical level. Schematic respondents were also found to have significantly higher levels of differentiation than aschematic respondents. Implications from these results are discussed in terms of conceptualization and measurement of accuracy.