The authors investigated depth-of-processing effects on conceptual priming by comparing incidental (implicit) and intentional (explicit) tests of word association. In Experiment 1, depth of processing at study influenced priming of weak and medium associates but not of strong associates. In Experiment 2, depth of processing influenced priming of weak associates but not of compound phrases (e.g., coathanger), whose preexperimental association strength matched that of weak associates. In Experiment 3, the same pattern persisted when study was auditory and test was visual, ensuring that priming was conceptual and not perceptual. In all experiments, in matched intentional tests, depth-of-processing effects occurred for all association strengths and for both phrases and associates, suggesting that the incidental tests were uncontaminated by voluntary retrieval, because they showed depth-of-processing effects only for some materials and not others, within the same participants and tests. Because depth-of-processing effects on involuntary free-association priming depend on the presence versus absence of a cohesive preexperimental representation, the memory-systems and conceptual/perceptual processing approaches to memory-test dissociations require modification to account for component processes of conceptual priming.