Church’s objections to the yoked control design are reviewed and criticized. To deal with them more effectively, a reciprocal yoked control design is presented, in which each S of a yoked pair receives training with two CSs, one of which involves an instrumental contingency while the other is in a yoked control arrangement with the other S’s instrumental CS. Thus, each S in the pair serves as a yoked control for the other, in reciprocal relation to the two CSs used. In addition to eliminating inadvertent confounding stemming from individual differences in sensitization and responsiveness, the reciprocal yoked control design also provides a single-session method for identifying Ss who are equally classically conditionable, so that instrumental-classical comparisons can be made free of bias from differences in classical conditionability. In addition, this design provides a means for testing the empirical soundness of Church’s arguments regarding differences in conditionability and biased outcomes Methods for statistical analysis of the reciprocal yoked control design are also considered.