This chapter puts a specific priming effect in memory under the microscope, from a probabilistic point of view. Priming effects in word recognition and memory have typically been viewed as side-effects of the mechanisms of recognition — e.g. as arising from associations between lexical items, which operate automatically. It suggests, instead, that many priming phenomena may arise from the structure of the probabilistic reasoning problem that the perceiver faces. The perceiver has a range of pieces of evidence, but has to infer their likely source. When a piece of evidence from one source is attributed to another, priming may be observed. Huber demonstrates that experimentally observed priming is sensitive to the rational structure of the ‘attribution’ problem, rather than to the operation of mere associative mechanisms.