Two experiments tested the prediction based on the source monitoring framework that imagination is most likely to lead to false memories when related perceived events have occurred. Consistent with this, people were more likely to falsely remember seeing events when the events had been both imagined as seen and actually heard than when they were just heard, just visually imagined, or imagined both visually and auditorily. Furthermore, when people considered potential sources for memories or more carefully evaluated features of remembered events, source errors were reduced. On average, misattributed ("false") memories differed in phenomenal qualities from true memories. Taken together, these findings show that as different qualities of mental experience flexibly enter into source attributions, qualities derived from related perceptual events are particularly likely to lead to false claims that imagined events were seen, even when the event involves a primary modality (auditory) different from the target event (visual).