People report recognizing discourse inferences at rates that approach target acceptance. Brainerd et al. [Brainerd, C. J., Wright, R., Reyna, V. F., & Mojardin, A. H. (2001). "Conjoint recognition and phantom recollection." "Journal of Experimental Psychology: Learning, Memory, and Cognition, 27", 307-329] proposed that memory retrieval in contexts associated with very high levels of false memory involve a process of illusory recollection which complements the impact of recollection and familiarity [Jacoby, L. L. (1991). "A process dissociation framework: Separating automatic from intentional uses of memory." "Journal of Memory and Language, 30", 513-541]. Experiments were conducted to compare three multiprocess models of text retrieval: A three-process "phantom recollection" model; and two dual-process models, respectively lacking mechanisms of veridical recollection and phantom recollection. Participants read lists of brief texts and then evaluated explicit, implicit, and foil memory probes. Different participant groups were instructed to use verbatim, verbatim plus gist, or gist-only memory-criteria. Multinomial processing tree analysis indicated that both immediate and delayed testing require the involvement of phantom recollection (Experiments 1 and 2, respectively). When the participant's extraction of text meaning is impaired, a dual-process model is adequate to fit the data (Experiment 3). (Contains 1 figure and 3 tables.)