We compare contrarian to conformist advice, a contrarian expert be- ing one whose preference bias is against the decision-maker's prior optimal decision. Optimality of an expert depends on characteristics of prior information and learning. If either the expert is fully informed, or if fi ne information can be acquired cheaply, then for symmetric distributions F of the state a conformist (contrarian) is superior if F is single-peaked (bimodal). If only coarse information can be acquired then a contrarian acquires more on average, hence is superior. If information is verifi able a contrarian has less incentive to hide unfavorable evidence, and again is superior.