This article investigates two factors posited to affect consumers' ability to learn a novel attribute relationship (e.g., "no pesticides → USDA organic symbol") and apply this recently acquired knowledge when making judgments in a new product category. The first factor concerns the nature of the attribute encoding process and, in particular, whether it allows for comparison of examples. The second factor focuses on the relationship between the learning and transfer domains, and examines the influence of perceptual similarity (manipulated two ways: similarity in the elements comprising the attribute relationships, and similarity between the base and target domains) on the transfer process. Study 1 revealed a significant improvement in inter-domain transfer when a relationship was learned in a multiple-domain/multiple-attribute learning environment over a single-domain/single-attribute environment. When attribute and domain diversity were examined separately (Study 2), it was found that attribute, but not domain, diversity produced the greatest improvements in transfer. Study 3 showed that the benefits of adding a perceptual similarity cue were most apparent when the transfer conditions were neither too easy (within domain and within attribute) nor too difficult (inter-domain and inter-attribute).