In this study, we examined whether adult subjects' beliefs regarding the empirical truth of a conclusion affected their production as well as their evaluation of a logical conclusion in a reasoning task. In addition, the relation between the ability to resolve an abstract reasoning problem correctly and the effect of belief-bias was examined. The subjects were given one of four paper-and-pencil reasoning tasks, two of them using an evaluation paradigm, and two of them using a production paradigm. Each paradigm comprised either neutral problems or belief problems. The neutral problems were constructed to be as similar as possible to the belief problems, in order to control for extraneous factors. All four tasks also included an abstract reasoning problem. The results indicate a significant belief-bias effect for both the evaluation and the production tasks. Qualitative analysis indicated that the belief-bias effect was more pervasive in the production condition. In addition, the belief-bias effect was found to exist independently of the subjects' abstract reasoning ability. The results are discussed with reference to a two-stage model, in which belief is used to resolve uncertainties in inferentially produced conclusions.