SUMMARY(i) To find out whether a mating preference could have initially evolved for adaptive reasons, one must determine whether the preferred trait could have provided useful information about mate quality at the time when the preference first arose.(ii) One way to do so is to determine whether the preference evolved before or after the preferred trait. If the preference evolved first, then it cannot initially have served an adaptive function in mate choice, rather it must have arisen by random drift, or as a pleiotropic consequence of selection acting on other aspects of individual perceptual abilities.(iii) A number of studies have shown that females exhibit a mating preference (e.g. for movement) in non‐sexual contexts also, which suggests that it may have evolved for reasons unconnected to mate choice. In addition, phylogenetic analyses have revealed that in several cases, females of a certain taxon exhibit a preference for a male trait that is absent in a sister taxon and in outgroup taxa, and that this preference is shared by females of the sister taxon tacking the male trait. The principle of parsimony suggests that such a preference has been inherited from a common ancestor, while the preferred trait arose only once in the lineage exhibiting the trait, i.e. that the preference predates the attractive trait.(iii) While the above evidence indicates that females may possess ‘hidden’ preferences for male traits that are not exhibited by members of their own species, and that in at least some cases males have later evolved display traits that exploit preexisting preferences of this kind, there have been too few historical studies of preference evolution to allow one to assess the frequency of such exploitation. Moreover historical studies cannot provide strong support for the adaptive origin hypothesis, because coevolution of trait and preference (as opposed to exploitation of a pre‐existing bias) is compatible with Fisherian models of preference evolution as well as with honest advertisement and the handicap principle. One can conclude only that while some mating preferences did not originally evolve for adaptive reasons, others may or may not have done so.(iv) To find out whether a mating preference is currently maintained by natural selection because the preferred trait provides useful information about mate quality, one must investigate the phenotypic and genotypic correlates of display, and the fitness consequences of mate choice.(v) A review of the published data reveals some support for the ideas of adaptive choice and honest advertisement. In a number of species, preferred display traits are correlated with putative measures of quality, and in a small proportion of these, there is evidence that reproductive success and/or offspring performance are higher for individuals mated to attractive partners. Very few studies report a failure to find any such correlates of display or any such benefits.(vi) While the above result suggests that honest advertisement does sometimes occur in extant populations (which does not necessarily imply that preferred traits originally evolved as reliable indicators of mate quality), the possibility of publication bias means that one cannot assess how widespread it is. More data is needed to remedy this problem, particularly regarding the fitness consequences of mate choice for females. Experimental rather than observational methods are the best means to gather such data. Studies that look for correlates of display, for instance, should rely on experimentally induced rather than natural variation in ‘quality’.(vii) The most common correlates of male display are age and dominance. The latter observation suggests that there may often be interactions between the processes of intersexual and intrasexual selection.(viii) There is considerably more evidence to support the idea of female choice for direct than for indirect benefits. At the same time, however, it is apparent that mating decisions are commonly influenced by more than one measure of quality, so that these two kinds of choice need not be independent. To assess this possibility will require more studies of the relationship between male attractiveness and offspring performance.(ix) Mate choice is frequently based on more than one display trait, and each trait is frequently influenced by more than one aspect of quality. ‘One quality, one trait’ views of honest advertisement are simplistic, and must be abandoned.(x) Honesty in sexual displays is sometimes maintained by cost (as in strategic handicap models) and sometimes, with approximately equal frequency, by physical necessity (as in revealing handicap models). In some cases, both mechanisms are involved in a single signalling system. To further distinguish between these possibilities will require experimental investigation of display cost, ba