RIGHT-HEMISPHERE MEMORY SUPERIORITY: Studies of a Split-Brain Patient

Six experiments explored hemispheric memory differences in a patient who had undergone complete corpus callosum resection. The right hemisphere was better able than the left to reject new events similar to originally presented materials of several types, including abstract visual forms, faces, and categorized lists of words. Although the left hemisphere is capable of mental manipulation, imagination, semantic priming, and complex language production, these functions are apparently linked to memory confusions—confusions less apparent in the more literal right hemisphere. Differences between the left and right hemispheres in memory for new schematically consistent or categorically related events may provide a source of information allowing people to distinguish between what they actually witnessed and what they only inferred. People are indisputably able to interpolate, make inferences, and imagine plausible scenarios that fit with what they perceive. However, they frequently commit false alarms on new schemaconsistent or categorically related events, thinking that they had actually been presented (Bartlett, 1932; Loftus, Donders, Hoffman, & Schooler, 1989; Mandler, 1984), Many studies have been directed at how people distinguish real events from those that were only inferred or imagined. This research (Johnson, Foley, Suengas, & Raye, 1988; Schooler, Gerhard, & Loftus, 1986) indicates the importance of the clarity of the retrieved representations, the amount of sensory information or detail present, and memory for the cognitive operations. Here we suggest that people may have an additional source of information—a hemispheric difference in memory for unpresented related material. Our hypothesis is that the left hemisphere encodes and stores not only the events that it experiences in the world, but also the interpolations, extrapolations, and inferences that it adds. Thus, it will "remember" items related to the presented events as well as the presented events themselves. The right hemisphere, being less capable of generalization and inferences, tends not to store them and hence does not retrieve them. It will, therefore, exhibit more veridical memory, particularly within a certain range of relatedness to the presented events. Several studies lend credibility to our hypothesis, Phelps and Gazzaniga (1992) showed split-brain patients a sequence of slides of a schematic event—of a woman baking cookies, for example. Twenty slides of this schema were shown, free field, for 3 s each. The recognition test probes included the 20 old slides, 20 schema-consistent new slides (of the same woman baking cookies), and 20 new shdes that were unrelated to the Address correspondence to Janet Metcalfe, Department of Psychology, Dartmouth College, Hanover, NH 03755; e-mail: metcalfe® dartmouth,edu. schematic event. There were no hemispheric differences in responses to either the exact old items or the nonschematic new items. However, the left hemisphere tended to accept new schema-consistent slides, whereas the dght hemisphere tended to reject them. Although these results are suggestive, there are difficulties in interpretation. First, the patients may have treated the test as a classification task. Although the patients were asked for literal recognition—whether the slides were old or new—they may have thought they were being asked whether the slides were schematically consistent with the material they had studied. Under this assumption, saying "yes" to the new schemaconsistent slides was correct, and the left hemisphere was simply better than the right. Such results might have no bearing on memory differences. Second, the right-hemisphere superiority might have been attributable to the pictorial materials. Many studies have shown left-hemisphere specialization for language functions and right-hemisphere specialization for visual-spatial functions (Bogen & Gazzaniga, 1%5; Gazzaniga, Bogen, & Sperry, 1962, 1965; Gazzaniga & Sperry, 1967; Kosslyn, 1987; LeDoux, Wilson, & Gazzaniga, 1977; McCarthy & Warrington, 1990; Milner, 1965), Our hypothesis about differential memory for events related to the target event is orthogonal to the visualverbal differences. In the experiments that follow, we address both of these interpretive difficulties in Phelps and Gazzaniga's (1992) seminal study. Converging evidence bearing on our hypothesis comes from a study of the N400 event-related potential in split-brain patients (Kutas, Hillyard, & Gazzaniga, 1988), The N400 does not occur when two highly related words occur together, but is produced, in a graded manner, with decreasing degrees of relatedness, Kutas et al, (1988) found that provided the patient had no language production from the right hemisphere, the left hemisphere but not the right hemisphere produced N400 responses. The authors interpreted this finding as indicating that the left hemisphere but not the right hemisphere is sensitive to differences in semantic similarity. If so, then the left hemisphere should show semantic priming with related stimuli to a greater extent than should the right hemisphere, and it does (Beeman et al,, 1994, Experiment 2). If activation by similarity is triggered more strongly in the left hemisphere than the right, then the left but not the right hemisphere might tend to treat events that are similar to presented events as if they had actually occurred. Furthermore, Marsolek, Kosslyn, and Squire (1992) showed a right-hemisphere priming advantage based on exact physical matches. These results are consistent with our hypothesis that the right hemisphere may be better than the left on veridical information, whereas the left hemisphere generalizes over and (incorrectly) remembers related information. Some researchers, however, have suggested that the right hemisphere may be the seat of inferential processing, Beeman VOL, 6, NO, 3, MAY 1995 Copyright © 1995 American Psychological Society 157 PSYCHOLOGICAL SCIENCE Right-Hemisphere Memory Superiority (1993) tested reaction time to unpresented words that would have been reasonable inferences in stories. Normal subjects exhibited equal priming to inference words and presented words. Patients with right-hemisphere damage showed less priming to the inference words than to the presented words. Similarly, Brownell, Michel, Powelson, and Gardner (1983) showed that right-hemisphere-damaged patients chose inappropriate punch lines for jokes, suggesting an affective impairment and possibly a decreased ability to appreciate violations of expectations. In both of these studies, though, damage to the right hemisphere may have had its deleterious effects by distorting the input to the left hemisphere and hence altering its performance. The split-brain patient studied here has completely disconnected hemispheres and does not have such potentially distorted input. In addition, little can be concluded from experiments not including a matched ieft-hemisphere-damaged group; such a group might have shown even more severely impaired processing. Indeed, Bihrle, Brownell, Powelson, and Gardner (J986) showed that left-hemisphere-damaged patients also responded inappropriately (excessively literally) to jokes. Similarly, Chiarello and Church (1986) studied similarity judgments in both leftand right-hemisphere-damaged patients on rhyme, visual similarity, and meaning tasks. Although the righthemisphere-damaged patients showed some impairment on the semantic task, they were less impaired on all tasks, including the semantic task, than were the left-hemisphere-damaged patients, Chiareilo, Burgess, Richards, and Pollock (1990J found that closely related pairs of words, such as bee-honey or doctornurse, produced more priming in the left hemisphere than in the right. However, marginally similar word pairs, such as deerpony, tested at 500 ms revealed no priming in the left hemisphere and a small priming effect in the right hemisphere. Similarly, Beeman et al, (1994) found a small right-hemisphere priming advantage for remotely related words. The result was fragile, though, and was not replicated in their second experiment. Burgess and Simpson (1988) found that although both hemispheres revealed automatic priming to the dominant meaning of a word (with the left hemisphere showing more priming than the right), only the left hemisphere produced controlled priming indicated by first facilitation but then (by 750 ms) inhibition of the nondominant meaning. This inhibition is considered to be fundamental for the lexica! disambiguation needed for comprehension. The possibility that the inhibition process occurs oniy or primarily in the left hemisphere deserves further investigation. We are not suggesting that the right hemisphere does no priming and makes no inferences, only that it may do less than does the left hemisphere. Either because the left hemisphere stores inferences but the right tends not to, or because the left hemisphere keeps some record of primed related events but the right tends not to, we expected the left hemisphere to show less veridical memory for events than the right. In particular, we expected that the right hemisphere, to a greater extent than the left, would correctly reject lures that were highly similar to presented events,