Annu. Rev. Neurosci. 1994. 17.’519-549Copyright © 1994 by Annual Reviews Inc. All rights reservedORGANIZATION OF MEMORYTRACES IN THE MAMMALIANBRAINRichard F. Thompson and David J. KrupaProgram for Neural, Informational, and Behavioral Science, University of SouthernCalifomia, Los Angeles, Califomia 90089-2520KEY WORDS:cerebellum, learning and memory, eyeblink conditioning, memory localiza-tion, classical conditioningINTRODUCTIONPerhaps the most fundamental issue in the broad field of neuronal substratesof learning and memory concerns the physical/biological mechanisms under-lying long-term memory formation, storage, and retrieval in the mammalianbrain. Considerable progress has been made in elucidating memory storagemechanisms in simpler invertebrate systems (Abrams et al 1991, Bergold etal 1990, Clark & Schuman 1991, Crow & Forrester 1990, Nelson & Alkon1990), but similar advances have yet to be made in understanding themammalian brain. As Lashley (1929) stressed so many years ago, theoverriding problem for understanding memory mechanisms in the mammalianbrain is localization of memory storage. Mechanisms of memory storagecannot be analyzed until the memory storage sites, whether localized ordistributed, have been identified. Given the nature of the vertebrate brain,memories should, to some degree, be distributed over ensembles of neurons,but whether this is within a localized region, over several regions in a givenstructure, or over several structures, is not yet known for most forms ofmemory.Once the locus, or loci, of a given form of long-term memory has beendetermined, we should be able to analyze the mechanisms of memory storage.These will likely involve chemical/structural changes at synapses and possiblylong-lasting changes in gene expression in the relevant neurons. Determiningthese mechanisms, however, will not inform us of what the memory is; the519www.annualreviews.org/aronlineAnnual ReviewsAnnu. Rev. Neurosci. 1994.17:519-549. Downloaded from arjournals.annualreviews.orgby University of California - San Diego on 01/05/07. For personal use only.520 THOMPSON & KRUPAcontent of a neuronal memory store can only be determined by a detailedcharacterization of the neural networks that subserve the memory.We now know that different forms or aspects of memory critically involvedifferent neural systems in the brain. The hippocampus plays a key role inrelational, contextual, spatial, and olfactory learning in lower mammals(Becker et al 1980, Berger & Orr 1983, Eichenbaum et al 1986, Lynch 1986,Moyer et al 1990, O’Keefe & Nadel 1978, Ross et al 1984, Solomon et al1986b, Squire 1992). In monkeys, the hippocampus and adjacent cortex,particularly the perirhinal, parahippocampal, and entorhinal areas, are criticalfor delayed non-matching to sample and for establishing (and/or retrieving)long-term visual memories, which are often termed declarative or working,but do not appear to be the site of long-term storage (Meunier et al 1993,Squire 1992, Zola-Morgan & Squire 1990). The cerebral cortex is theproposed repository of long-term declarative memories, but conclusiveevidence supporting this belief is lacking. The amygdala plays a key role ininitial leaming of conditioned heart rate and blood pressure, conditionedpotentiation of startle, conditioned freezing to tone, and instrumental avoid-ance (Hitchcock & Davis 1986, Iwata et al 1986, Fanselow et al 1991, Lianget al 1982); however, it may not be the site of long-term storage (see Lavondet al 1993, McGaugh 1989). [Interestingly, the hippocampus is critical forinitial learned freezing to context (Kim & Fanselow 1992).] Recent evidencesuggests that the basal ganglia may play an important role in some aspects ofinstrumental learning (Packard et al 1989). A region of the cerebellum necessary for basic delay classical conditioning of discrete motor responses(see below). On the other hand, it is possible that all these memory systemsbecome engaged, to varying degrees, in multiple aspects of learning, witheach system playing its separate role (see Lavond et al 1993, Wagner Brandon 1989).With the possible exception of declarative or working memory, all theseaspects of learning exhibit similar basic parametric features (e.g. massedvs spaced practice, temporal relations of stimuli, stimulus salience, etc),which suggests that at some level common mechanisms are involved. AsRescorla stressed (1988a,b), basic associative learning, which results fromexposure to relations among events in the world, is the way organisms,including humans, learn about causal relationships in the world. For bothmodem Pavlovian and cognitive views of learning and memory, theindividual learns a representation of the causal structure of the world and,as a result of experience, then adjusts this representation to bring it in tunewith the real causal structure of the world, thus striving to reduce anydiscrepancies or errors between its intemal representation and extemal reality(see also Dudai 1989, Squire 1987).www.annualreviews.org/aronlineAnnual ReviewsAnnu. Rev. Neurosci. 1994.17:519-549. Downloaded from arjournals.annualreviews.orgby University of California - San Diego on 01/05/07. For personal use only.CEREBELLUM AND CLASSICAL CONDITIONING 521THE ISSUE OF LOCALIZATIONPermanent lesions that abolish a learned behavior (i.e. the behavioralexpression of the memory) but do not interfere with the ability of the organismto otherwise perform the behavior can serve to identify the brain structuresnecessary for a given form of memory but cannot illuminate precisely whatroles these structures play in memory storage. Recording of neuronal activitythat changes in tight correlation with learning can identify structures involvedin or influenced by formation of the memory. Such evidence, per se, cannotlocalize the site(s) of memory formation. But at the very least the locus memory storage must in some way receive information about the conditionedor signal stimulus and the unconditioned or reinforcing stimulus and/orresponse and must exhibit