Proc. Natl. Acad. Sci. USA Vol. 95, pp. 891–898, February 1998 Colloquium Paper This paper was presented at a colloquium entitled “Neuroimaging of Human Brain Function,” organized by Michael Posner and Marcus E. Raichle, held May 29–31, 1997, sponsored by the National Academy of Sciences at the Arnold and Mabel Beckman Center in Irvine, CA. Functional neuroimaging studies of encoding, priming, and explicit memory retrieval (implicit memoryyprefrontalyvisualyfunctional MRIypositron-emission tomography) RANDY L. BUCKNER*†‡§ AND WILMA KOUTSTAAL¶ *Departments of Psychology, Anatomy and Neurobiology, and Radiology, Washington University, St. Louis, MO 63130; †Nuclear Magnetic Resonance Center, Massachusetts General Hospital, Charlestown, MA 02129; ‡Department of Radiology, Harvard University Medical School, Boston MA 02115; and ¶Department of Psychology, Harvard University, Cambridge, MA 02138 ABSTRACT Human functional neuroimaging techniques through which to view the neurobiological basis of human provide a powerful means of linking neural level descriptions cognition (1). These techniques are extremely powerful in that of brain function and cognition. The exploration of the they allow researchers to identify specific brain areas and functional anatomy underlying human memory comprises a pathways that are recruited and differentially activated during prime example. Three highly reliable findings linking mem- the performance of various cognitive tasks—including mem- ory-related cognitive processes to brain activity are discussed. ory tasks—and to do so noninvasively, in normal, awake First, priming is accompanied by reductions in the amount of humans. These techniques are limited, however, in that the neural activation relative to naive or unprimed task perfor- spatial scale within which different activations can be resolved mance. These reductions can be shown to be both anatomically is presently a few millimeters and the temporal resolution is '1 and functionally specific and are found for both perceptual s. The topic we address in this paper is how these techniques and conceptual task components. Second, verbal encoding, can be applied to the study of human memory and findings that allowing subsequent conscious retrieval, is associated with have emerged so far. activation of higher order brain regions including areas within Memory refers to a function that unfolds over time. Within the left inferior and dorsal prefrontal cortex. These areas also the context of human long term memory, the phenomenon to are activated by working memory and effortful word genera- be examined concerns how information processing at one point tion tasks, suggesting that these tasks, often discussed as in time influences processing at another—later—point in time. separable, might rely on interdependent processes. Finally, Such a notion has inherent in it the idea that memory can be explicit (intentional) retrieval shares much of the same func- differentiated into two kinds of processes: (i) active informa- tional anatomy as the encoding and word generation tasks but tion processing that is isolated in time and (ii) processes or is associated with the recruitment of additional brain areas, including the anterior prefrontal cortex (right > left). These mechanisms that maintain and consolidate information over findings illustrate how neuroimaging techniques can be used extended periods of time. Classic models of memory capture to study memory processes and can both complement and this idea and conceptualize memory processes as involving extend data derived through other means. More recently three separate stages: encoding, storage, and retrieval (or developed methods, such as event-related functional MRI, will more elaborate expansions of these three stages; see, e.g., ref. continue this progress and may provide additional new direc- 2). Encoding and retrieval are active processes that occur at tions for research. relatively specific points in time; encoding refers to the initial processing of information that potentially instantiates a mem- ory trace, and retrieval refers to newly evolved processing that Human memory is a remarkably complex cognitive function. It is inherently intertwined with all other aspects of cognition, results from, and often requires access to, prior encoding shaping (and being shaped by) our thoughts and behaviors as episodes. Somewhere between these two sets of active pro- we interact with the world. The study of the neurobiological cesses occur the more temporally distributed processes in- basis of memory is therefore, not surprisingly, an enormous volved in storage and consolidation, the mechanisms that undertaking, requiring understanding at many levels ranging convert the otherwise transient encoding event into a more from the molecular to neural systems to cognitive approaches. enduring form. (For review and discussion see, for example, Unfortunately, our experimental methods to bridge the gaps refs. 3 and 4.) between these levels are limited. Linking neural systems and The question relevant here is: Which, if any, of these cognitive approaches requires observations or inferences processes can we observe with neuroimaging techniques? At about neuronal function in relation to human memory pro- present, it seems likely that neuroimaging methods allow us to cesses. Key methods that long have been available include the observe the subset of memory processes described above as use of primate and other animal models (allowing both active memory processes that can be isolated in time. The two single-unit physiological recordings and ablation studies) and most commonly used neuroimaging techniques [PET blood the study of human clinical populations (examination of pa- flow and blood oxygenation level-dependent (BOLD) func- tients with lesions and intraoperative recordings). Human functional neuroimaging techniques recently have emerged as Abbreviations: PET, positron-emission tomography; fMRI, functional promising alternative methods, providing a new window MRI. §To whom reprint requests should be addressed at: Nuclear Magnetic Resonance Center, Massachusetts General Hospital, Building 149, © 1998 by The National Academy of Sciences 0027-8424y98y95891-8$2.00y0 13th Street, Charlestown, MA 02129. e-mail: [email protected]. PNAS is available online at http:yywww.pnas.org. harvard.edu. 891 Downloaded by guest on September 26, 2021 892 Colloquium Paper: Buckner and Koutstaal Proc. Natl. Acad. Sci. USA 95 (1998) tional MRI (fMRI)] appear to be sensitive to physiologic were either living or nonliving. This meaning-based word events that correlate with acute changes in neuronal activity on processing task led to 75% correct recognition of the words. the order of seconds (5–8). This means that we are capable of Imaging data contrasting this deep encoding task and a shallow asking a range of questions: Which brain areas are active encoding task (decide whether the word contains the letter during tasks known to promote long term memory storage ‘‘a,’’ 57% correct recognition) demonstrated robust left pre- (e.g., deep encoding tasks)? Which brain areas are active when frontal activation overlapping with the regions previously information is intentionally retrieved from memory (e.g., activated by the word generation tasks. This basic pattern of explicit recall and recognition)? How do the brain areas that findings also has been demonstrated with fMRI. For example, are initially activated by a task change after repeated perfor- in a series of fMRI studies, Gabrieli and colleagues have mance of the task or prior exposure to information involved in explored the functional–anatomic correlates of another deep the task (e.g., skill learning and priming)? However, it seems encoding task, in which participants view words and then unlikely that we can observe the temporally distributed pro- decide whether they fall into the category of abstract (e.g., cesses related to storage and consolidation directly. hope) or concrete (e.g., tree) words. They found significantly A consequence of this notion is that, when we are observing greater left prefrontal activation during this deep encoding active memory processes with neuroimaging alone, we are, to task than during a shallow encoding task in which subjects a greater or lesser degree, simply observing what might be simply decided whether words were presented in uppercase or thought of as ‘‘specialized’’ instances of typical information lowercase letters (23, 27). Our laboratory, by using fMRI, has processing—tasks that are known to create, or benefit from, followed up on some of these findings, and representative data storage- and consolidation-related processes that bridge gaps are shown in Fig. 1. over time. Often these memory tasks are the same tasks that Fletcher et al. (28) approached the issue in a different are elsewhere described as language tasks or attention tasks (9, manner and compared later recall performance when partic- 10). The fundamental basis for this interdependency might be ipants first engaged in word generation concurrently with an that many processes subsumed under the concept of memory easy distractor task (yielding high levels of recall, 83%) to a are a by-product of normal information processing (11–13). dual task situation in which word generation was paired with Where information processing ends and ‘‘memory encoding’’ a difficult distractor task (yielding moderate levels of recall, begins, for example, is a blurry distinction at best. After all, in 69%). Word generation paired