A seven-judge committee must decide whether to promote a candidate to a position requiring a young, trilingual person. Each judge estimates whether the candidate is young, and whether she is trilingual (see Table 1). Individual judgments on whether the candidate has the correct profile follow logically by conjunction. The candidate is young for a majority of judges (4/7) and is trilingual for a majority of judges (4/7), yet only two judges think she has the right profile and this group-level inconsistency arises even though each individual set of judgments is consistent. How should the committee proceed? Should the judges vote on the whole conjunction (the profile) and reject the candidate, or should they vote separately on each conjunct (the criteria) and promote the candidate? This problem, an instance of the more general doctrinal paradox, has generated a substantial literature in law, economics, political science, philosophy, and computer science (Bovens & Rabinowicz, 2006; Brennan, 2001; Chapman, 1998; Dietrich, 2006; List, 2003, 2005; List & Pettit, 2002, 2004; Pettit, 2001). However, it has never been addressed from an empirical-behavioral perspective (although see Kameda, 1991). In this article, I report a study testing three factors that contribute to judges' preferences for voting on the whole conjunction or voting separately on each conjunct. (See List & Pettit, 2002, for other escape routes from the paradox.) The three factors are as follows: