After failing to solve items from the Remote Associates Test (RAT), subjects showed significant priming effects when the solutions were presented in a lexical-decision task (Experiment 1). Experiments 2 and 3 found no significant priming effect when subjects were asked merely to remember the RAT elements, or for targets that were associates of only two of the three elements in incoherent RAT items. Experiment 4 showed that identifying a correct solution took longer than lexical decision, and that the probability of correct identifications for a given item was uncorrelated to the priming effect for the item. Experiment 5 yielded item-difficulty norms for 68 RAT items as well as a replication of the priming effect observed for unsolved items in Experiment 1. In Experiment 6, a significant priming was observed for targets that were solutions to hard items but not for solutions to easy items. This research provides evidence for implicit problem-solving, which is nonconscious but not automatic, and is neither a perceptual nor a purely memory-related phenomenon.