Is coherence truth conducive

In recent papers Klein and Warfield (1994, 1996) argue that coherence per se is not truth conducive on the ground that the more coherent of two sets of beliefs is often more likely to contain a false belief. Merricks (1995) responds on behalf of the coherentist that we should not evaluate truth conduciveness on the level of sets, as Klein and Warfield do, but on the level of individual beliefs. A proper examination of these claims requires a clear understanding of what coherence per se amounts to. This paper proposes a precise definition of coherence, and uses it for untangling the epistemological knots of coherence and truth. My analysis reveals that (pace Klein and Warfield) coherence per se is truth conducive on the level of sets; but (pace Merricks and many coherentists) it is not truth conducive on the level of individual beliefs.