Short title: Neuropsychology of Autobiographical Memory
Daniel L. Greenberg and David C. Rubin
Department of Psychological and Brain Sciences
This special issue of Cortex focuses on the relative contribution of different neural networks to memory and the interaction of ‘core’ memory processes with other cognitive processes. In this article, we examine both. Specifically, we identify cognitive processes other than encoding and retrieval that are thought to be involved in memory; we then examine the consequences of damage to brain regions that support these processes. This approach forces a consideration of the roles of brain regions outside of the frontal, medial-temporal, and diencephalic regions that form a central part of neurobiological theories of memory. Certain kinds of damage to visual cortex or lateral temporal cortex produced impairments of visual imagery or semantic memory; these patterns of impairment are associated with a unique pattern of amnesia that was distinctly different from the pattern associated with medial-temporal trauma. On the other hand, damage to language regions, auditory cortex, or parietal cortex produced impairments of language, auditory imagery, or spatial imagery; however, these impairments were not associated with amnesia. Therefore, a full model of autobiographical memory must consider cognitive processes that are not generally considered ‘core processes,’ as well as the brain regions upon which these processes depend.
Daniel L. Greenberg
Department of Psychological and Brain Sciences
Duke University, PO Box 90086
Durham, NC 27708
We would like to thank Martin Conway and two anonymous reviewers for helpful comments on the manuscript. This work was supported in part by NIA grant RO1 AG16340 and a McDonnell-Pew Cognitive Neuroscience Program Individual Grant, both awarded to DCR.
Over the last few years, we have been developing a multi-system model of memory by combining our basic understanding of neuropsychology and neuroanatomy with behavioral studies. This development began with a study of oral traditions (Rubin, 1995), and was then extended to autobiographical memory (Rubin, 1998; Rubin & Greenberg, 1998, in press). Researchers disagree about the precise meaning of “autobiographical memory”; for example, some view it as a form of episodic memory (Kopelman and N. Kapur, 2001; Rubin, 1998), while others use different definitions of these terms and thereby arrive at the opposite view (Conway, 2001). Still others prefer the term “recollective memories” (Brewer, 1995). When we use the term “autobiographical memory,” we refer to memories that have several properties. First, like episodic memory, autobiographical memory “receives and stores information about temporally dated episodes or events and temporal-spatial relations among them” (Tulving, 1983, p. 21). Second, such memories involve something more than the mere retrieval of stored data: the person remembering the memory must be conscious of the prior conscious experience, a self-reflective mental state that Tulving terms autonoetic consciousness (Tulving, 1985). As with many of the other cognitive processes we will discuss, autonoetic consciousness is not sufficient for autobiographical memory (it plays a role in other processes, such as prospective memory (Wheeler, Stuss, and Tulving, 1997)) but it is a necessary feature: many philosophical accounts (Brewer, 1995) and the reports of some amnesics (e. g. Crovitz, 1986; O'Connor et al., 1992) suggest that an autobiographical memory should be accompanied by a sense of reliving as well as the belief that the remembered event actually occurred. We define an autobiographical memory as a memory of a personally experienced event that comes with a sense of recollection or reliving. Autobiographical facts (Brewer, 1995), on the other hand, are bits of personally relevant information that are retrieved without this sense of reliving.
Most theories of memory claim that memory requires an interaction between medial temporal lobes, frontal lobes, and the rest of the cortex (Conway and Pleydell-Pearce, 2000; Damasio, 1989; Fuster, 1995; Kopelman, 2000; Kopelman and N. Kapur, 2001; Mayes and Roberts, 2001; Markowitsch, 2000; McClelland et al., 1995; McDonald et al., 1999; Murre, 1999; Murre et al., 2001; Shastri, 2002; Squire, 1992). The implications of this claim have rarely been investigated, however. In particular, most major theories do not dissect memory storage into modality-specific components. Conway and Pleydell-Pearce (2000), for example, state that memories are stored in an “undifferentiated pool” called “event-specific knowledge” (Conway and Pleydell-Pearce, 2000, Figure 1, p. 265; also see Conway, 1992, 1995a and Conway and Rubin, 1993 for further discussion of this idea). Kopelman’s model provides a detailed analysis of executive and emotional systems that play a role in memory, but devotes only a single module to storage (Kopelman, 2000, Figure 6, p. 608; Kopelman and N. Kapur, 2001, p. 1417). Along the same lines, McClelland and colleagues (1995, Figure 14, p. 444), Squire (1992), Murre (1999, Figure 1, p. 269), and Markowitsch (2000) all focus on medial temporal regions and devote little attention to posterior neocortical storage sites. McDonald and colleagues (1999) mention the hippocampus, the amygdala, the frontal lobes, the thalamus, and the basal ganglia—in fact, almost every region besides the posterior neocortex. Even theories that do mention memory in the neocortex do not address the relative contributions of several different cortical areas to memory in general (with the exception of studies addressing anterior and lateral temporal regions, which we review later in the paper). We are not claiming that we know little about the roles of these neocortical regions—in fact, the contrary is generally true—but rather that any such knowledge has rarely found its way into neurobiological theories of memory and autobiographical memory specifically. Nor are we embarking on a rehash of current theories or a simple assignment of a function to a region; instead, we are attempting to ask more detailed questions: given what we know about autobiographical memory on the behavioral level, which brain regions should be involved? Most importantly, what are the particular contributions of each of these regions to autobiographical memory, and what happens when they are damaged or destroyed?
We attempt to answer these questions by reviewing relevant neuropsychological case studies. We organize the paper around five interacting cognitive processes that have been identified in the behavioral data as important components of autobiographical memory: explicit memory, imagery (in visual, auditory, and other modalities as well as multi-modal spatial imagery), language, narrative, and emotion. After reviewing the importance of these behavioral processes, we identify the brain regions on which they depend; then, we search the neuropsychological literature to discover how autobiographical memory changes after impairments of these processes and damage to these regions. Some of these components will turn out to be relatively unimportant; in other cases, neuropsychological evidence will force the addition of a component to the overall model.
We selected these processes for several reasons. First, neuropsychological studies have shown double dissociations between pairs of these processes. Second, these neuropsychological data, when coupled with evidence from functional imaging and neuroanatomical studies, demonstrate that these processes have distinct neurological substrates. Third, the literature on individual differences demonstrates that linguistic and imagery abilities vary along similar lines (Carroll, 1993). Fourth, perhaps for these reasons, they are also treated as distinct cognitive processes in the behavioral literature (and are even given separate sections in many textbooks and separate names in common language). While these processes can certainly be divided further, evidence on all levels indicates that they can be treated as separate and discrete behaviorally and neuropsychologically.
These processes and brain regions might seem excessively broad; however, several factors compel a broad analysis. First, behavioral research has largely dealt with broadly defined cognitive processes; therefore, if this investigation is to have clear relevance to the behavioral data, it must operate at a similar level. Second, the neuropsychological literature does not necessarily report subtle disorders, especially if they occur in a patient with another interesting disorder; therefore, relevant data are often absent. Third, this investigation is in its early stages, and an examination of broadly defined processes and brain regions is most likely to detect interesting results. We will see if the most severe impairments will have any effect on autobiographical memory; if they do, we can try to uncover the roles of the subcomponents of these cognitive processes by studying the effects of circumscribed impairments.
Throughout this paper, we make claims about where parts of memories are stored—a point that requires some explanation. Psychology uses two opposing metaphors of memory. The first claims that memories are stored at encoding and later retrieved. The second maintains that the mind/brain changes with experience, so that it will respond differently when exposed to stimuli in the future. Under the latter metaphor, a memory is not stored; rather, the system that perceives and acts is changed. Although the latter metaphor fits better with our general approach, we tried to write the paper so it could be understood in terms of either metaphor, because we have found that readers who think in terms of one metaphor often find the other extremely difficult to understand. When we say, for example, that visual aspects of an autobiographical memory are stored in the visual system and not the medial temporal or frontal lobes, under the first metaphor we really mean the following longer and more awkward statement: “It is possible to damage parts of the visual system and remove visual information that is not available outside the visual system.” Under the second metaphor, we really mean “The visual system processes visual information, not all of which is shared with non-visual areas. Thus, when the visual system is damaged, the system that would respond to a cue with more detailed visual information and analysis is no longer available.” Our claim is that most visual information is stored and/or processed solely within the visual system, and so can be lost with damage to just that system. We will make similar claims for other sensory modalities, emotion, language, and narrative. In the last two cases, terms like “analysis” or “processing” may seem better to the reader than “storage,” but it must be processing that is particular to a specific autobiographical memory that reflects past experience. We note, though, that while the brain is largely modular (Fodor, 1983), some visual information, or binding or indexing of that information, is stored in the medial temporal or frontal lobes. There must be such information outside the visual system if there is to be both modularity and integrated autobiographical memories.
We base our approach on four well-supported observations:
1. Almost all (and perhaps all) cognition is affected by past experiences. Searching for a single neural location of memory is a fool’s errand. Memory is stored everywhere, and at every level of analysis (Fuster, 1995; Toth and Hunt, 1999). Autobiographical memories—which consist of multi-modal stimuli, extend over time, and are organized along narrative and emotional dimensions—are probably distributed throughout the brain. Neuropsychologists have long maintained that complex cognitive processes are localized not in specific brain nuclei, but in the coactivation of diffuse neuroanatomical substrates (Lashley, 1950/1960; Luria, 1966; Penfield and Perot, 1963). Thus, we are not proposing a new idea; instead, we are gathering the evidence needed to apply this idea to autobiographical memory.
2. Much of the brain is organized by sensory modality; we know at least as much about these areas as any other areas—and more about the cognition they subserve. These sensory areas qualify as modules: their processing is quick, obligatory, and cognitively impenetrable, and their output is shallow. Moreover, they are domain-specific—damage to one module does not affect cognition in another module (see Fodor, 1983 for a discussion of modularity and Moscovitch, 1992 for an application of modularity theory to the neuropsychology of memory).
3. Medial temporal and frontal structures are required for the recall of specific events, but the rest of the brain—especially the posterior cortices that process and store the sensory components of autobiographical memory—contributes something.
4. Because consciousness is such a difficult topic, and is defined in so many different ways by so many different people, we restrict our consideration of it to the phenomenological report that people often recollect—that is, they relive their past states of consciousness when they recall an event. We include this observation because many psychologists and many philosophers of mind consider it a defining feature of autobiographical memory (Brewer, 1986, 1995; Rubin and Greenberg, 1998; Wheeler et al., 1997).