This paper explores bias in the use of evidence in policy. It argues that existing models of the evidence–policy relationship neglect the tendency for attention to be paid only to that evidence helpful to the interests of powerful social groups. An evolutionary analogy is used to explain how this bias arises, without the need for irrationality or conspiracy on the part of policy makers. Examples are given in the fields of drug, asylum and other policies, and the possible responses by researchers to the biased use of research evidence are discussed.