A recent series of replications of a list-learning paradigm first published by Deese (1959) has tested conditions that elicit high levels of false recall and false recognition. After exposure to words associated with an unpresented word, participants often falsely recall and recognize the unpresented word. In the present study, the nature of imaging tasks at encoding was varied to test the hypothesis that, whereas a focus on conceptual processing promotes false memories for unpresented words, a focus on perceptual processing inhibits false memories. Participants were classified as good or poor imagers based on the results of an imagery test. The results showed that forming images of the words' referents led to high levels of both correct and false recognition, especially for good imagers; forming images of words in specific type fonts, however, did not inhibit false memories. The results are contrasted with those of other studies using the Deese paradigm.