c h a p t e r 50nwkpsych.rutgers.edu/~jose/courses/578_mem_learn/...vii. behavioral and cognitive...

C H A P T E R

50

Learning and Memory: Brain Systems

INTRODUCTION

The previous chapter summarized the cellular and molecular mechanisms of memory and explained that individual neurons contain complex machinery capable of altering membrane excitability and synaptic strength. These neuronal building blocks are the fundamental basis of plasticity, and this plasticity supports behavior through memory systems, complex circuits involving a large number of interconnected neurons. The present chapter focuses on the major memory systems of the brain and considers how each system supports learning. The chapter begins with a brief history of the concept of multiple forms of memory, followed by a summary of each of the major memory systems of the mammalian brain (Fig. 50.1). The final section consid-ers how these systems might collaborate, compete, or operate in parallel to support learned behaviors.

HISTORY OF MEMORY SYSTEMS

The history of memory systems can be usefully framed by the early question of whether a specifi c memory could be localized or whether it would be distributed throughout the brain. Even early on there was evidence that the answer might lie somewhere between the extremes. In the 1920s, the psychologist Karl Lashley conducted a series of experiments in which he carefully damaged various areas in the cere-bral cortex of rats who had learned a route through a simple maze. He found that the specific location of the brain damage did not relate to memory perfor-mance as well as the total amount of cortex damaged and concluded that memory was distributed through-

out the brain. Yet a short time later, the neurosurgeon Wilder Penfield found that when he stimulated various areas of the brains of awake epileptic patients (in order to identify functional areas deemed too important to be removed during surgery to treat the epilepsy), stimulation in some brain regions, particu-larly in the temporal lobe, led the patient to experi-ence specific memories. He reasoned that, if focal stimulations bring specific memories to mind, then individual brain regions might contain individual memories. The psychologist Donald Hebb offered a reconciliation for these diverse findings by suggesting that thoughts and memories were supported by “cell assemblies,” networks of neurons, and that learning experiences resulted in changes in the connections between cells (Hebb, 1949). Thus, to the extent that a change at an individual juncture between cells (i.e., a synapse) recorded experience, memory was localized. Yet to the extent that full memories resulted from a reactivation of an entire cell assembly, memory was distributed.

Although it was not appreciated at this time that different networks, or “cell assemblies,” might support different forms of memory, several philosophers and psychologists had already suggested that an impor-tant difference existed between the type of everyday memory that one brings to mind in the form of words, pictures, or sounds and the type of memory that accrues with our actions as habits and dispositions. For example, in the early nineteenth century the philoso-pher Maine de Biran distinguished between representa-tive memory, which he imagined as the recollection of ideas and events, mechanical memory, which he described as the acquisition of habits and skills, and sensitive memory, which he described as the acquisition of affective values for otherwise neutral stimuli. In the

Fundamental Neuroscience, Third Edition 1153 © 2008, 2003, 1999 Elsevier Inc.

1154 50. LEARNING AND MEMORY: BRAIN SYSTEMS

VII. BEHAVIORAL AND COGNITIVE NEUROSCIENCE

late nineteenth century, the philosopher and psy-chologist William James similarly envisioned walking, writing, fencing, singing, and other habitual routines as being fundamentally different from memories pro-duced by conscious recollection.

Thus, by the mid-twentieth century, the work of philosophers, psychologists, and neurosurgeons would have seemed to create an intellectual readiness for the idea of multiple memory systems in the brain. However, at this time the study of memory was guided not by the pursuit of brain systems but by behavioral psychology, a field based on the principle that the opaque details of brain and consciousness were best set aside as more productive research was accom-plished by focusing on observable behavior. Despite this principled disregard of the brain, a distinction emerged between a learning based on stimulus-response habit-like learning, such as that described by Clark Hull, and a more cognitive stimulus-stimulus form of learning, such as that described by Edward Tolman. Nevertheless, both ideas were strongly entrenched in a behaviorist tradition and were based on the idea that memory was a single capacity—they disagreed on only how to best depict that capacity. Although behavioral psychology did not actually argue against the idea that particular brain areas might be specialized for particular kinds of memory, its steadfast insistence that the details of the brain be ignored created a vacuum of relevant data. Thus, despite the contributions of earlier work, there was no

clear evidence that brain areas are specialized to support different forms of memory.

Incontrovertible evidence that at least one brain region was specialized to support everyday memory came from a man who underwent brain surgery to treat severe epilepsy and who became perhaps the most famous patient in neuroscience (Scoville and Milner, 1957). This man, known as H.M., began expe-riencing seizures at the age of 10. It was unclear what caused the seizures, but it was very clear that some-thing drastic had to be done. Heavy doses of anticon-vulsant medications did not stop the seizures from worsening to a point at which they were debilitating. Acting on the idea that dysfunctional brain tissue might be causing the seizures, on September 1, 1953 the neurosurgeon William Scoville removed substan-tial tissue from the medial aspect of H.M.’s temporal lobes (Fig. 50.2).

In one sense, the surgery was a success. The fre-quency and severity of the seizures were reduced. In another sense, the outcome was tragic. On April 26, 1955, Brenda Milner, a colleague of both Penfi eld and Hebb, conducted a neuropsychological examination of patient H.M. The profound memory impairment was immediately obvious. H.M. gave the date as March, 1953 and his age as 27 (2 years younger than his actual age). He performed poorly on memory for short stories, word lists, pictures, and a wide range of other materi-als. Remarkably, it was unclear that he even remem-bered that he had undergone brain surgery.

Striatum

Striatum

Cerebellum

CerebellumHippocampus

Hippocampus

Amygdala

Cerebral Cortex

Amygdala

FIGURE 50.1 Drawing of the human brain showing components of the major memory systems.


The severity of H.M.’s amnesia was shocking—he showed almost no capacity for new learning. Also, the fact that the brain damage was known to be confi ned to a particular region (the medial temporal lobes) added to the intrigue. Yet the observation that a major neurological deficit followed resection of a substantial amount of brain tissue was perhaps not surprising. Indeed, earlier reports already had described brain damaged individuals with memory impairments. Instead, the reason that H.M.’s case ushered in the modern era of research on memory systems came from the four aspects of his mental capacity that remained intact.

First, formal testing identified that cognitive abili-ties other than memory were intact. H.M.’s IQ score generally was unaffected by the surgery, and a battery of tests found no deficits in perception, abstract think-ing, or reasoning ability. These intact abilities indicted that memory could be separated from perception and intelligence. Second, H.M. could hold on to small amounts of information as long as he was actively rehearsing the information. This fi nding suggested that the ability to maintain information online (now usually referred to as working memory) was distinct from the ability to make a lasting record in the brain (Chapter 51). Third, H.M.’s childhood memories were relatively intact. This finding suggested that, although the medial temporal lobes might be important for

forming new memories, this region was unlikely to be the final storage site for memory. Fourth, H.M. had an intact ability to acquire new motor skill learning (Milner, 1962). Over several days of practice, H.M. gradually improved at tracing the outline of a star when viewing the paper only through a mirror (a task that initially is challenging even for healthy individu-als) despite never forming a memory for the testing experience. Thus, the contrast between H.M.’s pro-found memory loss and intact cognitive abilities outside the ability to form new memory set the stage for the work that has led to our current understanding of memory systems in the brain. H.M.’s case showed that memory is a dissociable cognitive capacity and that day to day memory was supported by brain structures that differed from those that supported the acquisition of motor skills.

H.M. has participated in numerous studies in the decades since his surgery and has provided invaluable insights into the organization of memory systems in the brain. However, studies of other amnesic patients have also been crucial to the emergence of the current understanding of memory and the brain. For example, a later study by Neal Cohen and Larry Squire (1980) suggested that the domain of preserved memory in amnesia was much larger than just motor skill learn-ing. They asked a group of amnesic patients to practice reading words that had been reversed to appear as if

FIGURE 50.2 Magnetic resonance images showing the brains of amnesic patients H.M. and E.P. The images show axial sections through the medial temporal lobes and reveal damaged tissue as a bright signal. H.M.’s damage resulted from surgery, and E.P.’s damage was caused by viral encephalitis. Nevertheless, the resulting lesion was similar for the two patients. Both patients sustained extensive damage to the medial temporal lobes and are profoundly amnesic. From Stefanacci et al. (2000).

HISTORY OF MEMORY SYSTEMS 1155



they were being viewed in a mirror. Similar to H.M.’s improvement in mirror tracing, the group of amnesic patients steadily improved in the speed with which they could read the mirror-reversed words despite having great trouble remembering the practice ses-sions. The study helped to show that a great deal of learning can occur outside the scope of conscious rec-ollection and that this nonconscious learning is not limited to motor skill learning. Cohen and Squire bor-rowed terminology from work in artifi cial intelligence (Winograd, 1975) and suggested that amnesia was characterized by an impairment in declarative memoryand that procedural memory was left intact.

Another amnesic patient, patient R.B., helped to show that damage limited to the hippocampus could result in substantial memory impairment (Zola-Morgan et al., 1986). Like H.M., R.B. performed poorly on a range of declarative memory tests, although his deficit was not nearly as severe as H.M.’s. Detailed histological analysis of R.B.’s brain indicated that his memory impairment resulted from damage to a spe-cific subregion (region CA1) in his hippocampus. Thus, patient R.B.’s case showed that damage limited to the hippocampus was suffi cient to produce amnesia.

Soon after the report on patient H.M., researchers began work in experimental animals, such as rats and monkeys, in an attempt to create a model of human amnesia. Initially, the separate lines of investigation on humans and animals were divergent and suggested to some that the functions of the hippocampus might differ between species. However, subsequent fi ndings from these two areas converged, both in their charac-terizations of the kind of memory that is dependent on the hippocampal region and in their identifi cation of functional domains and anatomical pathways associ-ated with other types of memory. In early studies, this line of work focused particularly on the hippocampus. In one proposal, O’Keefe and Nadel (1978) summa-rized a large body of literature on the effects of hip-pocampal damage on different behavioral tasks and concluded that animals with damage to the hippocam-pus are severely impaired at many forms of spatial learning. It is now generally believed that the hippo-campus is important for nonspatial as well as spatial memory, yet this early proposal added to ongoing work in humans and monkeys by identifying several key properties of hippocampus-dependent memory in experimental animals. In particular, they characterized this kind of memory as rapidly acquired and driven by curiosity rather than by rewards and punishments, properties that are consistent with characterizations of hippocampus-dependent memory in humans.

Our understanding of memory systems has advanced greatly since H.M.’s surgery and it is still

unfolding. The study of amnesia has provided insights not only about the role of the hippocampal memory system in declarative memory but also about the ability of other memory systems to support learning in the absence of conscious recollection. A large body of work, including research on humans and on experi-mental animals, has allowed researchers to understand better the unique contributions of the striatum, cere-bellum, amygdala, and cerebral cortex to different forms of memory. Thus, we now know that the hip-pocampus anchors only one of several memory systems in the brain and that each of these brain structures are components of other memory systems that contribute to our ability to benefit from experience. The following sections provide an overview of these systems.

MAJOR MEMORY SYSTEMS OF THE MAMMALIAN BRAIN

Overview

The following sections consider the best understood memory systems of the brain. The discussion focuses on individual structures that play central roles in each of the different systems and presents separate sections on the hippocampus, striatum, cerebellum, amygdala, and cerebral cortex (Fig. 50.1). However, each of these regions anchors a much larger network of distributed brain areas that work in concert to support particular forms of memory. For example, the hippocampal memory system includes not only the hippocampus but also areas in the adjacent parahippocampal region as well as many distributed unimodal and polymodal neocortical areas. These areas normally work together to support conscious recollection of facts and events, a capacity termed declarative memory. Even though declarative memory depends on very specialized information processing and plasticity within the hip-pocampus, a complicated and distributed circuitry outside the hippocampus is also required for these alterations to emerge as the experience of remember-ing. Most importantly, the cerebral cortex plays a role in many types of memory, and the present chapter will discuss both its specific role as a part of multiple memory systems and the memory functions it sup-ports outside of those systems.

A memory system is most usefully defined in psy-chological terms as well as anatomical terms. That is, a memory system is most clearly identified by both a distinct circuitry and a unique set of operating charac-teristics. Memory systems are not separated according to stimulus modality (e.g., auditory vs. visual modal-ity) or between response modalities (e.g., manual vs.


verbal responses). Instead, the critical distinctions involve how each system’s anatomy and physiology support a particular type of memory representation in terms of its organization and psychological character-istics. Finally, memory systems collaborate, compete, or operate in parallel to support behavior in the course of one’s day-to-day activities, and this issue will be considered in the fi nal section of the chapter.

Hippocampus

Anatomy

The hippocampal memory system includes the hippocampus (defined here as the CA fi elds, dentate gyrus, and subiculum) and the entorhinal, perirhinal, and postrhinal cortices in the adjacent parahippocam-pal region (the postrhinal cortex is referred to as the parahippocampal cortex in primates; Fig. 50.3; Burwell et al., 1995; Suzuki, 1996). The anatomy and circuitry of these regions, especially the hippocampus, are largely conserved across mammalian species (Manns and Eichenbaum, 2007). The hippocampal memory

system also depends on diverse and widespread higher order regions in the cortex that are both the source of information to the parahippocampal region and hippocampus and the targets of projections originat-ing from these regions. The parahippocampal region serves as a convergence site for input from these corti-cal association areas and mediates the distribution of cortical afferents to the hippocampus. These parahip-pocampal cortical areas are interconnected and send major efferents to multiple subdivisions of the hippo-campus. Within the hippocampus, an intricate pattern of connectivity mediates a large network of associa-tions (Amaral and Witter, 2004), and these connections support forms of long term potentiation that could participate in the rapid coding of novel conjunctions of information (Chapter 49). The outcomes of hippo-campal processing are directed back to the adjacent cortical areas in the parahippocampal region, and the outputs of that region are directed in turn back to the same areas of the cerebral cortex that were the source of its inputs. Additional structures also have been included as components of this system, including midline diencephalic structures.

tnedoRetamirP

Neocorticalassociation

areas

Parahippocampalregion

Hippocampus

Copyright © 2002, Elsevier Science (USA). All rights reserved.

FIGURE 50.3 The anatomy of the hippocampal memory system in monkeys and rats. The hippocampal memory system includes the hip-pocampus (CA fields, dentate gyrus, and subiculum) and the parahippocampal region, which includes the entorhinal cortex, the perirhinal cortex, and the parahippocampal cortex. The hippocampal memory system also includes midline diencephalic nuclei. Multiple association areas in the cerebral cortex send outputs that converge on cortical areas in the parahippocampal region, which in turn sends its outputs to the hippocampus. The output path involves return projections from the hippocampus to the surrounding parahippocampal region, which in turn projects back to the same cortical association areas. From Eichenbaum (2001).

MAJOR MEMORY SYSTEMS OF THE MAMMALIAN BRAIN 1157



Amnesia and Declarative Memory

Damage to the hippocampus and parahippocampal region produces anterograde amnesia, a memory deficit characterized by an inability to make lasting memories of one’s daily experiences. That is, the hip-pocampal memory system ordinarily supports remembering of new facts and events and makes this information subsequently available for conscious rec-ollection. This capacity is called declarative memory. In most amnesic patients, the deficit in declarative memory extends to both remembering specifi c per-sonal events (episodic memory) and learning new facts (semantic memory). Their deficit encompasses all stimu-lus modalities and impacts nonverbal expression as well as verbal report. Moreover, amnesic patients are impaired whenever the memory task requires the explicit expression of memory, as in tests of free recall or recognition memory. Further, damage to the hip-pocampus and parahippocampal region results in loss of memory acquired for some period before the damage. This memory impairment is referred to as retrograde amnesia.

Amnesia resulting from damage restricted to the hippocampus and parahippocampal region is highly selective in four important ways, the same four ways discussed earlier with regard to the intact abilities of patient H.M. First, perceptual, motor, and intellectual functions are intact. Across a broad range of neuro-psychological tests, amnesic patients perform well on assessments of sensory capacities, motor coordination, intelligence, and language performance. Second, memory acquired long before the onset of amnesia is typically intact. The preservation of these remotely acquired memories generally is thought to be the con-sequence of a consolidation process in which the hip-pocampus plays a critical role for a prolonged period, after which retrieval of memories can be supported by other areas, including the neocortex. Thus, the impact of retrograde amnesia is typically time limited. The process of consolidation is discussed in more detail later. Third, the capacity for immediate memory is typically intact in amnesic patients, and, just as in the case for healthy individuals, they can extend the con-tents of immediate memory in time through rehearsal or elaboration, a capacity often referred to as working memory (Chapter 52). Amnesic patients can immedi-ately reproduce a list of six or seven numbers as well as healthy individuals, but the memory defi cit becomes evident as soon as immediate memory span is exceeded or after a delay is interposed that includes some dis-traction to interrupt rehearsal.

Fourth, the various forms of memory that are sup-ported by brain systems outside the hippocampal

memory system are intact in amnesic patients. For example, amnesic patients demonstrate normal acquisi-tion of a broad variety of tasks in which memory is expressed through behavioral performance such as in the mirror-guided tracing task and the mirror-reversed word reading task described earlier in the chapter. Other examples include artificial grammar learning and improving reaction times on a repetitive sequence of finger taps. One profoundly amnesic patient, patient E.P., has a memory impairment and brain damage very similar to that of patient H.M. (Fig. 50.2). Patient E.P. has shown normal performance on a broad variety of non-conscious memory tasks, despite the fact that his nearly complete anterograde amnesia has kept him from forming conscious memories of the testing situations themselves. In some of these examples, the learning was simple and reflexive, such as in the case of standard eyeblink classical conditioning (discussed in more detail later in the chapter in the section on the cerebellum).

In other examples, performance on more compli-cated tasks came to resemble that of healthy individuals despite the fact that the learning was supported by a type of nonconscious and relatively inflexible type of learning. For example, in one study E.P. was shown eight pairs of objects one at a time, each pair containing a correct object and an incorrect object. He slowly learned to pick up the correct object in each pair and after 18 weeks was able to select the correct object almost every time (Bayley et al., 2005). When he was asked how he knew which object to pick up he pointed to his head and replied, “It’s here somehow or another and the hand goes for it.” Yet, in addition to being unavailable to awareness and being very slowly acquired, E.P.’s memory for the objects also appeared to lack the flexibility of that shown by healthy individu-als, who had learned the task after only three sessions. When the task was changed by presenting all 16 objects together and by asking E.P. to sort them into correct and incorrect piles, his performance deteriorated precipi-tously. In contrast, healthy individuals performed nearly perfectly on the sorting task. In another study, over the course of 12 weeks E.P. came to learn a list of 60 three-word nonsense sentences (e.g., “speech caused laughter”), such that he eventually was able to respond correctly some of the time when asked to fill in the missing third word (e.g., “speech caused ???”; Bayley and Squire, 2002). Yet E.P. continually expressed sur-prise when informed that he had seen the sentences previously. In addition, his memory also differed from control subjects in that it appeared to be somewhat rigid. When the middle word of each sentence was replaced by a synonym, E.P. was able to answer cor-rectly for only one sentence. His performance contrasted with that of healthy individuals, who learned the sen-


tences after only two weeks and who had no trouble adapting to the new testing format.

These studies with patient E.P. illustrate that although some examples of memory supported by structures outside the hippocampal memory system can be simple and reflexive, other examples can be more complex and can include acquisition of new verbal information and biases for one of the response choices that is readily available. These studies also highlight by contrast the characteristics of memory supported by the hippocampal memory system. Mem-ories formed via the hippocampal system typically are acquired rapidly, flexible, and available to conscious recollection.

Functional Brain Imaging and the Hippocampal Memory System

Functional brain imaging studies in humans, includ-ing functional magnetic resonance imaging (fMRI) and positron emission tomography (PET), have shown that the hippocampal memory system is engaged in a variety of declarative memory tasks. For example, in one of the earliest studies, increased activity was observed in the hippocampus and parahippocampal region when participants viewed previously unseen photographs as compared to when participants viewed photographs that they had already seen on multiple occasions (Stern et al., 1996). Subsequent studies have also found that activity in the hippocampal memory system often is related to the success of memory encod-ing or retrieval. For example, in one study, participants saw a list of words one at a time while being scanned (Wagner et al., 1998). After the scanning session, par-ticipants were given a test to assess which words were remembered and which words were forgotten. The results indicated that activity in the hippocampal memory system during encoding (in this case, in the parahippocampal region) was greater for words that were subsequently remembered as compared to words that were forgotten. Many subsequent functional brain imaging studies have also identifi ed memory-related activity in the hippocampus itself. Activity in the hip-pocampus is particularly strong in tasks that demand memory for associations that compose specifi c experi-ences (Eichenbaum et al., 2007). Moreover, functional brain imaging has become a key tool for understand-ing how the hippocampal memory system, as well as other memory systems, ordinarily operate in the intact human brain.

Models of Anterograde Amnesia in Experimental Animals

Neuroscientists have used a specifi c, carefully selected set of behavioral tests in developing a nonhu-

man primate model of human amnesia to identify the particular medial temporal lobe structures that support declarative memory. The tasks involve learning about three-dimensional objects or complex pictures. In one task, delayed nonmatching to sample, subjects are shown an object once and then, after a delay, are shown two objects (the original object and a new one). The task of the monkeys is to select the new object (Fig. 50.4A). When the delay is only a few seconds, monkeys with experimental lesions that include the same medial temporal lobe structures damaged in H.M. (including the hippocampus and adjacent cortices) performed as well as normal monkeys (Mishkin, 1978). As the delay was increased, the monkeys became progressively more impaired. Thus, monkeys with medial temporal damage, like humans with amnesia, have intact imme-diate memory but perform poorly when the memory demand increases over time. Monkeys with medial temporal lobe damage also perform poorly when they must retain rapidly acquired object discriminations for a prolonged period. In contrast to these impairments in object memory, monkeys with medial temporal lobe damage have intact capacities for skill acquisition as measured in a task that involves learning to retrieve a candy by manipulating it along a bent rod. Thus, the pattern of both preserved and impaired memory in human amnesic patients is closely modeled by the per-formance of monkeys with similar brain damage (Zola-Morgan and Squire, 1985).

Using this animal model, investigators were able to identify the structures of the medial temporal lobe critical to supporting declarative memory. In H.M. the damage included the amygdala, the hippocam-pus, and the surrounding parahippocampal region. However, studies with monkeys have shown that the amygdala is not a part of the declarative memory system. In addition, the severity of memory impair-ment depends on the extent and locus of damage within the medial temporal lobe. Damage limited to the hippocampus, or to its major connections through the fornix, produces only a modest impairment. In contrast, damage that includes the adjacent cortices produces severe amnesia. Thus, the perirhinal and parahippocampal cortical regions themselves make major contributions to memory, and the hippocampus itself is a critical component of the system. Particularly compelling evidence about the role of the hippocam-pus in recognition memory has come from studies using a test known as the visual paired-comparison task (Zola et al., 2000; Fig. 50.4B). In this task, the monkey initially is shown a pair of duplicate pictures. Then, following a variable delay, two pictures again are shown. One picture is identical to the initial sample and the other is a novel picture. Memory for the sample




picture is measured in normal monkeys (and humans) by a lesser amount of time looking at the familiar picture than the novel one. On this test, damage limited to the hippocampus results in a rapid loss of memory such that, at a brief interval after presentation of the sample picture memory is intact, but after around 10 seconds memory is severely impaired. The overall pattern of findings supports the view that both the hippocampus as well as the adjacent cortical areas support performance on these and other tests of rec-ognition memory.

Studies with rodents have also offered insights into the nature of memory representations in networks of hippocampal neurons. Many studies in rats have dem-onstrated that damage to the hippocampus results in deficits in a variety of spatial learning and memory tasks. A particularly useful example is place learning in the Morris water maze task (Morris et al., 1982; Fig. 50.5). In this task, rats are trained to find a hidden escape platform submerged just below the surface in

a pool of cloudy water. Because there is no specifi c cue at the escape site, the rat must learn the location of the platform on the basis of spatial relationships among the cues that are visible in the room. Rats with hippo-campal damage are severely impaired at this task. Additional evidence for the importance of the hippo-campus in spatial memory comes from recordings of the firing patterns of hippocampal neurons in behav-ing rats. A major early finding in studies of hippocam-pal cells was that many of these neurons fire when the rat is in a particular location in its environment. This activity was shown to reflect an encoding of the spatial relationships among physical stimuli in the environ-ment, leading experimenters to call these neurons “place cells” (O’Keefe, 1976).

Although these data indicate an essential role for the hippocampus in spatial learning, all forms of spatial learning are not dependent on the hippocam-pus. For example, rats with hippocampal damage can learn simple spatial discriminations, such as whether

+

+

Sample

Recognition

Delay Delay

BA


FIGURE 50.4 Recognition memory tasks used for studies of memory in nonhuman primates. Both tasks are easily adapted for human participants as well. (A) The delayed nonmatching-to-sample task using unique objects as stimuli. The subject initially is presented with a single novel object as the sample and must displace the object. This is followed by a variable delay during which the subject cannot see any objects. In the subsequent recognition test, two objects are presented, one of which is the same as the sample and the other of which is novel. Correct performance requires the subject to recognize and avoid the sample object and instead choose the novel one to receive a food reward. (B) Visual paired comparison task. During the sample phase the monkey looks at two identical pictures. In the test phase, one of the sample pictures is represented along with a novel picture. Memory for the repeated picture is inferred by measuring the subject’s tendency to look away from the repeated picture and toward the new picture.


to turn left rather than right in a T-shaped maze. However they are impaired if they are forced initially to visit either the left or the right arm of the maze, and then asked to remember that experience and choose to visit the opposite arm. Rats with hippocampal damage can also learn to locate the escape platform in the Morris water maze when they are trained to fi nd the platform from a single starting point. However, they

cannot learn if trials from different starting points are intermixed. These studies show that rats with hippo-campal damage can sometimes use spatial memory to guide performance, but they cannot organize spatial information gained from different episodes to express memory flexibly according to the demands on a par-ticular trial.

Other experimental evidence indicates that the hip-pocampus is also critically involved in the organiza-tion and flexible expression of nonspatial memories. For example, in one experiment intact rats and rats with damage to the hippocampus were trained to dig through sand mixed with one of a list of odors to obtain a cereal reward (Bunsey and Eichenbaum, 1996; Fig. 50.6). Initially, both healthy rats and rats with hip-pocampal lesions learned a set of simple associations between pairs of odors. For example, rats learned to dig in odor B when presented with odor A and to dig in odor C when presented with odor B. Subsequently, the rats were given probe tests to determine the extent to which learned representations supported two forms of flexible memory expression. One of these tests, a test for transitivity, measured the ability to infer an asso-ciation between two odors that shared a common associate. For example, having learned that odor A is associated with odor B and that odor B is associated with odor C, could they infer that A is associated indi-rectly with C?

The other test, a test for symmetry, measured the ability to recognize associated odors when they were presented in the reverse of their training order. For example, if B is associated with C, is C associated with B? Intact rats showed strong transitivity and success-fully inferred an association between A and C that had been learned only indirectly. In contrast, rats with selective hippocampal damage were severely impaired in that they showed no evidence of transitivity. In the symmetry test, healthy rats showed their associations were indeed symmetrical. In contrast, rats with hip-pocampal damage again were severely impaired, showing no detectable capacity for symmetry. Corre-spondingly, in humans the hippocampus is activated during retrieval of indirect associations during the same transitive inference judgment (Preston et al., 2004; Fig. 50.7). These findings show that the role of the hippocampus in rats extends beyond spatial memory and suggest that the hippocampus encodes and retrieves nonspatial as well as spatial memories in both animals and humans.

Consistent with this view, numerous recording studies in rats, monkeys, and humans have shown that hippocampal cells also fire in association with visual, auditory, and olfactory stimuli, as well as combina-tions of these stimuli and the places where they occur,

During learning

After learning

Hidden platform


FIGURE 50.5 The Morris water maze task. Early in training, rats search for the submerged platform for extended periods. After train-ing the rat swims directly to the platform.




suggesting that hippocampal neurons represent impor-tant stimuli in the context in which they are remem-bered (Ekstrom et al., 2003; Wirth et al., 2003; Wood et al., 1999). For example, in one experiment, rats per-formed a variant of the delayed nonmatching to sample task that was guided by olfactory cues that were pre-sented at several locations in an open field (Fig. 50.8A; Wood et al., 1999). The activity of some hippocampal cells correlated with spatial features of the task (i.e., the rat’s location when it encountered odors), and the activity of other cells correlated with nonspatial fea-tures of the task (e.g., the identity of odors). However, a substantial number of hippocampal cells fi red only when the animal encountered a particular combina-

tion of task features. For example, some cells fi red when the rat sniffed a particular odor at a particular place but did so only when the odor represented the correct (i.e., nonmatch) choice. Thus, the hippocampus appeared to represent not only the locations that were common across trials but also the unique combinations of nonspatial and spatial features that may have helped the rat perform the task.

Observations also suggest that hippocampal neurons can represent sequences of events and places that compose episodic memories. For example, evidence of episodic-like coding was found in a study when rats performed a spatial alternation task on a T maze (Wood et al., 2000; Fig. 50.9). Each trial began when the rat

Test for Transitivity: AC & XZ

Test for Symmetry: CB & ZY

Odor paired associates

Training set 1: AB & XY

Training set 2: BC & Y Z

Sample A X A X

Choice B+ vs Y B vs Y+

Sample

+

Choice C vs Z C vs Z

Choice C+ vs Z C vs Z+

Sample B C ZY Sample

Choice B vs Y B vs Y

A

B

Sample Choice


FIGURE 50.6 Associative transitivity and symmetry in paired associate learning. (A) On each training trial, one of two odors is presented as the sample. On the subsequent choice trial the animal must select the assigned associate, indicated by a “+”. (B) Outline of odor pairings used in training on two sequential sets of paired associates, plus stimuli used in tests for transitivity (C for A; Z for X) and symmetry (B for C; Y for Z). From Bunsey and Eichenbaum (1996).


A B B C

A C

FIGURE 50.7 fMRI activity in the human hippocampus during judgments of associative transitivity. Bilateral regions in the hippocampus (indicated by yellow areas) were more active when participants used memory of face-house pairs (A-B and B-C) to infer face-face pairings (A-C) as compared to when participants simply memorized face-face pairings. Adapted from Preston et al. (2004).

CB

A

1 2 3 4 5 6 7 8 9

64

20

64

20

1 2 3 4 5 6 7 8 9

P lace

10

68

420

1 2 3 4 5 6 7 8 9Odor

10

68

420

1 2 3 4 5 6 7 8 9P lace

Odor

Firin

g ra

te (H

z)Fi

ring

rate

(Hz)

Firin

g ra

te (H

z)Fi

ring

rate

(Hz)

Trial n lairT n lairT1+ n + 2


FIGURE 50.8 Hippocampal neuronal firing patterns. (A) A rat performing the delayed nonmatching to sample task with odorized cups as stimuli. The three panels indicate a sequence of trials that vary the position of the odor and whether the presented odor matches the odor used on the previous trial. (B) Example of a cell that fires when trials are performed at two adjacent places (2 and 3), regardless of the odor presented. (C) Example of a cell that does not fire differentially in association with trials at different locations, but fires selectively on trials when odor 5 is presented. From Wood et al. (1999).




traversed the stem of the “T” and then selected either the left- or the right-choice arm. To alternate success-fully, the rats were required to distinguish between their left-turn and right-turn experiences and to use the memory for their most recent previous experience so they could turn in the opposite direction on the next trial. Different hippocampal cells fired as the rats passed through the sequence of locations within the maze during each trial. These cells could be considered canonical place cells. However, in addition, the fi ring patterns of many of the cells depended on whether the rat was in the midst of a left- or right-turn episode, even when the rat was in the stem of the T and running similarly on both types of trials. Although the majority of the cells strongly preferred one trial type, most of the cells fired at least to some extent when the rat was at the same point in the stem on either trial type. Thus, the hippocampus encoded the left-turn and right-turn experiences using distinct representations, and these representations included information that could link them by their common features. These results suggest that the representations of event sequences, linked by codings of their common events and places, could con-stitute the substrate of a network of episodic memo-ries, consistent with the idea that the hippocampus has a broad role in declarative memory.

Retrograde Amnesia and Consolidation

Patients with amnesia due to medial temporal lobe damage suffer not only a deficit in learning new mate-rial (anterograde amnesia), but also loss of memories that were acquired before the brain damage (retro-

grade amnesia). Importantly, the retrograde defi cit is time-limited, and material acquired shortly before the damage is affected most severely, whereas items learned earlier in life are relatively spared. For example, the profoundly amnesic patient E.P. (Fig. 50.2) has almost no capacity for new declarative learning but shows good memory from his childhood. In one study, E.P. was asked to think back to his childhood neigh-borhood and describe routes he would take from one location to another (e.g., grade school to the town theater; Teng and Squire, 1999). In some instances, he was asked to imagine that a main street was blocked and to describe an alternate route. E.P. performed as well as a group of healthy individuals who had lived in the neighborhood at the same time but who had moved from the area long before the study, indicating that his memory for the spatial layout of his childhood neighborhood had escaped the damage to his medial temporal lobes. By contrast, E.P. has been unable to learn the layout of the neighborhood he moved to subsequent to the onset of amnesia.

Retrograde amnesia is another aspect of the amnesic syndrome that has been studied extensively in experi-mental animals. For example, studies have also shown that simple object discriminations that were learned by monkeys shortly before medial temporal lobe damage are poorly retained, but discriminations learned remotely are spared. This pattern of memory impair-ment has been replicated in a variety of tasks in humans, monkeys, rabbits, rats, and mice and is thought to reflect a process of memory consolidation (Squire et al., 2001).

The term consolidation has been used to character-ize two kinds of brain events that affect the stability of memory after learning. One event involves the fi xation of plasticity within synapses over a period of minutes or hours through a sequence of protein synthesis and morphological changes at synapses (Chapter 49). The other event involves a reorganization of memories, which occurs over weeks to years following new learn-ing. This prolonged consolidation occurs in the hip-pocampal memory system and is thought to involve interactions between the hippocampus, parahippo-campal region, and the cerebral cortex. Several models have been proposed to account for how the hippocam-pus might interact with other brain regions over a prolonged period in memory consolidation (Alvarez and Squire, 1995; McClelland et al., 1995). These models assume that widespread areas of the neocortex contain the details of the information that is to be remembered and that areas in the hippocampal memory system support the capacity to retrieve the memory during the period shortly after learning. Over time, areas in the hippocampus are thought to reactivate the cortical rep-


FIGURE 50.9 Selective firing during different types of memory episodes in rats performing a T maze alternation task. On left turn trials, different cells fi re as the rat runs forward through a series of locations on the maze. On right turn trials, a different set of cells fi re, even when the overt behavior and places are the same. From Eichenbaum (2000).


resentations through repetition, rehearsal, or sponta-neous activity. The reactivation is then thought to induce plasticity in cortical–cortical connections, and these connections are viewed as a possible network structure for the permanent storage and organization of the memory.

Recent studies using molecular markers of neuronal activity and synaptic change have reported evidence consistent with this idea. For example, one study of maze learning in mice observed that these markers were found in high numbers in the hippocampus one day after learning but their numbers were much lower 30 days later (Maviel et al., 2004). In contrast, the same markers were found to be at low levels in the cerebral cortex initially after learning but were observed to be high in various cortical areas 30 days after learning, areas including prefrontal, anterior cingulate, and ret-rosplenial cortices. These results suggested that plas-ticity in the hippocampus was initially important for the spatial memories but that over time plasticity in neocortical areas became more important.

There has long been interest in the role of sleep and dreaming in memory, and one idea has been that processes might occur during sleep that promotes the consolidation of hippocampus-dependent memories. Although the benefit of sleep to nonconscious exam-ples of learning is generally accepted, the status of hippocampus-dependent memory is less certain (Stickgold, 2005). Nevertheless, there are several results that indicate that processes during sleep might con-tribute to consolidation of hippocampus-dependent memory. For example, specific patterns of activity observed in the hippocampus while rats were awake were observed to “replay” during both REM sleep and slow wave sleep (Sutherland and McNaughton, 2000). These results suggest that reactivation of memories during sleep may participate in the process of consoli-dation. In another study, induction of slow-wave like oscillations in the brains of human participants while they slept led to slightly improved recall upon waking (Marshall et al., 2006). Thus, there is accumulating evi-dence regarding the role of sleep in the consolidation of hippocampus-dependent memory, but further work is needed to fully understand this process.

Summary of Hippocampus

Amnesia associated with damage to the hippocam-pal memory system in humans is characterized by an inability to retain and consciously recollect memories of facts and events. Studies with animal models of amnesia memory provide a framework for thinking about hippocampal function in terms of memories that are represented as sequences of events remembered in the context in which they occurred and that are orga-

nized according to relations among distinct experi-ences. Furthermore, hippocampus-dependent memory is accessible through a variety of routes and forms of behavioral expression and supports the capacity to make generalizations and inferences from memory.

Striatum

Anatomy

The striatum is a major component of the basal ganglia and is involved with reward and motivation (Chapter 43) as well as with motor control (Chapter 31). The focus of the present section is the role of the striatum, particularly the dorsal striatum, in memory. The striatum receives its cortical inputs from many areas of the cerebral cortex (Fig. 50.10). The projections are organized topographically such that divergent and convergent projections into modules within the stria-tum might sort and associate somatosensory and motor representations. The striatum projects to other compo-nents of the basal ganglia and to the thalamus, which project back to both the premotor and motor cortex and the prefrontal association cortex. Notably, there

Cortex

Striatum

Indirectpathway

Directpathway

ThalamusSNc/VTA

GP e

STN GP i/SNr

Brainstem

ExcitationInhibitionDopaminemodulation

Copyright © 2006 Nature P ublishing GroupNature Reviews | Neuroscience

FIGURE 50.10 Anatomy of the striatum and the rest of the basal ganglia. STN, subthalamic nucleus; GPe, external globus pallidus; GPi, internal globus pallidus; SNr, substantia nigra pars reticulata; SNc, substantia nigra pars compacta; VTA, ventral tegmental area. From Yin and Knowlton (2006).




are minimal projections from this system to the brain stem motor nuclei and none to the spinal motor appa-ratus, which suggests the system operates mainly to modify cortical motor representations rather than control behavior through direct motor outputs.

Memory for Habits, Actions, and Outcomes

The idea that the striatum is critical for many forms of stimulus-response habit learning was introduced in studies in experimental animals that dissociated this system from other memory systems, such as the hip-pocampal memory system. For example, in one study the role of the striatum in learning specifi c behavioral responses was demonstrated using a simple T maze apparatus where two possible strategies in solving the maze were compared directly (Packard and McGaugh, 1996; Fig. 50.11). In this study, rats began each trial at the base of the T maze and were rewarded with food at the end of one choice arm (e.g., left). Accordingly, the rats could have acquired the task by learning to make a specific turning response (a “response” strat-egy, left in this example). Alternatively, rats could have remembered where the reward was located relative to the surrounding stimuli in the test room, independent of any particular behavioral response required to obtain it (a “place” strategy). The critical test to distinguish these two strategies was to rotate the maze by 180° in such a way that the start point was at the opposite end of the room. After training for a week on the T maze

task, most rats used a place strategy on the probe test. However, when they were trained for another week with the maze in its original orientation and were then presented with an additional probe trial, most rats adopted a response strategy. These results suggested that under these training circumstances, initial acquisi-tion of the task was guided by memory of the reward locus, but subsequent overtraining led to development of a habitual turning response.

Packard and McGaugh (1996) also examined whether different brain systems supported these dif-ferent strategies. All animals had been implanted with indwelling cannulae that allowed injection of a local anesthetic or saline directly and locally into one of two brain structures, the hippocampus or the striatum, prior to the probe tests. The effects of the anesthetic were striking. On the first probe trial, when the stria-tum was inactivated, animals behaved just as controls. That is, they were predominantly place learners and the place strategy did not depend on the striatum. In contrast, when the hippocampus was inactivated, the animals showed no preference on the first probe trial, indicating that the hippocampus was essential for the place strategy and that this was the only strategy avail-able in the early stage of learning. On the second probe test conducted after additional training, control sub-jects had acquired the response strategy, and inactiva-tion of the hippocampus had no effect. In contrast, animals with the striatum inactivated lost the turning

tseTniarT

Goal

"P lace" "Response"

Start

S tart


FIGURE 50.11 “Place” versus “response” learning. The rat is trained initially to turn left in order to obtain a reward at a particular loca-tion. In a subsequent test, the maze is rotated and the rat is allowed to select whether it will perform the same left turning “response” or remember the “place” of the previous reward.


response and instead showed that they were relying on the place strategy. The results indicated that animals normally develop an initial place strategy that depends on the hippocampus and do so prior to acquiring a habit-like response strategy. With overtraining, a response habit mediated by the striatum is acquired, and it predominates over the hippocampal place strategy. Nevertheless, the place memory can be “uncovered” by inactivating the striatum and sup-pressing the turning response strategy. These results suggested that there are distinct types of memory that guide place and response strategies, that the response strategy tends to be acquired more slowly than place memory, and that the response strategy is supported by the striatum.

More recent research has suggested that the memory supported by the striatum is not necessarily acquired slowly. For example, one study that recorded from neurons in the striatum and prefrontal cortex in monkeys found that the striatal neurons led prefrontal neurons in tracking the reversal of action-outcome contingencies (Pasupathy and Miller, 2005). On each day of testing, monkeys were trained to move their eyes left after seeing one novel picture and to move their eyes right after seeing a second novel picture. Monkeys were rewarded with juice with each correct response. After the monkeys learned these associa-tions, the picture-direction associations were reversed (e.g., the first picture then indicated that right was the appropriate response). During the trials when the monkeys were learning to reverse their responses to the pictures, neuronal activity related to viewing the pictures in both the striatum and prefrontal cortex mirrored the monkeys’ improvement in performance. However, the striatal activity led the prefrontal activ-ity in two ways. First, on each trial the onset of activity in the striatal neurons tended to occur before the onset of the prefrontal neurons. Second, a relation-ship between the neuronal activity while viewing each picture and the subsequent correct eye movement emerged in fewer trials in the striatum as compared to the prefrontal cortex. The observation that activity in the striatum in this task reflected reversal of a learned association more quickly than the prefrontal cortex suggests that memory supported by the striatum may not necessarily be particularly slow and may be used to instruct the prefrontal cortex in mediating complex stimulus-response associations.

Research with human participants has also indi-cated that the striatal system may play a critical role in memory even for more complex forms of stimulus-response learning. For example, Knowlton and col-leagues (1996) tested patients in the early stages of Parkinson’s disease on a probabilistic classifi cation

learning task formatted as a weather prediction game (Fig. 50.12). Parkinson’s disease is associated with the degeneration of neurons in the substantia nigra and a resulting major loss of input to the striatum. Amnesic patients were also tested, and their results contrasted with that of the Parkinson’s patients. The task involved predicting on each trial one of two possible outcomes (“rain” or “sunshine”) based on cues that were pre-sented to the subject. On each trial, one, two, or three cards from a deck of four were presented. Each card was associated with the sunshine outcome indepen-dently and probabilistically, 75%, 57%, 43%, or 25% of the time, and the outcome with multiple cards was determined by the conjoint probabilities. After the cards were presented on each trial, the subject was asked to choose between rain and shine and was then given feedback (correct or incorrect). The probabilistic nature of the task made it difficult for subjects to improve by recalling specific previous trials because the same confi guration of cues could lead to different outcomes. The most useful information was the prob-ability associated with particular cues and combina-tions of cues, which could be acquired gradually across trials much as habits and skills are acquired. Over a block of 50 trials, normal subjects gradually improved from pure guessing (50% correct) to about 70% correct. However, patients with Parkinson’s disease failed to show significant learning, and the failure was particu-larly evident in those patients with more severe parkinsonian symptoms. In contrast, amnesic patients were successful in learning the task, achieving levels of accuracy similar to the controls by the end of the 50-trial block.

Subsequent to training on the weather prediction task, these subjects were given a set of multiple-choice questions about the nature of the task and the kinds of stimulus materials they had encountered. Normal sub-jects and patients with Parkinson’s disease performed very well in recalling the task events. In contrast, the amnesic patients were severely impaired. Thus this example of probabilistic classification learning was disrupted by striatal damage, whereas declarative memory for the learning events was impaired by hip-pocampal or diencephalic damage.

Summary of Striatum

The dorsal striatum supports learning that accrues over repeated trials and supports examples of stimu-lus-response, habit-like learning. However, the dorsal striatum also supports memory that goes beyond simple stimulus-response learning and includes prob-abilistic classification learning as well as examples in which subjects learn about reversals of outcomes of particular actions.




Cerebellum

Anatomy

The cerebellum consists of the cerebellar cortex, which has been divided into lobes and further into lobules, and several deep cerebellar nuclei (see Chapter 32 for more detailed anatomy). The cerebellum receives direct input from the spinal cord and brain stem as well as indirect input from a variety of sensory and motor areas in the cerebral cortex via pontine nuclei in the brain stem. The cerebellum projects directly to the spinal cord, brain stem, hypothalamus, and thalamus, and the thalamic targets in turn project to various motor and nonmotor areas in the cortex, particularly in the frontal lobes (Middleton and Strick, 1998). The internal circuitry of the cerebellum has been described as modular, with a local wiring diagram that is repeated again and again throughout the cerebellar cortex. This modularity raises the possibility that, although differ-ent regions of the cerebellar cortex receive different input, each region might perform a similar set of oper-ations on that information before passing the results back out to the rest of the nervous system (Boyden et al., 2004).

The cerebellum has long been associated with aspects of motor learning. For example, the cerebellum supports specific sensory-to-motor adaptations and adjustments of reflexes, such as changing the force that one exerts to compensate for a new load or acquiring conditioned reflexes that involve associating novel motor responses to a new stimulus. More recently, the cerebellum also has been associated with nonmotor learning and other examples of cognition and behav-ior. Based on the idea that the modular circuitry of the cerebellum offers a generalized capacity for processing of specific inputs, the present section focuses on one example of cerebellar-dependent learning as a frame-work for highlighting the role of the cerebellum in memory.

Eyeblink Classical Conditioning as a Prototype of Cerebellum-Dependent Memory

Eyeblink classical conditioning has been used fre-quently as a model of cerebellum-dependent learning, and is perhaps the best understood form of associative memory in vertebrates. In this paradigm, experimental animals (typically rabbits due to their tolerance for physical restraint) are placed in restraining chambers

In this game you are a weather forecaster.You will learn how to predict rain or shine using a deck of four cards.


FIGURE 50.12 Weather prediction task. View of the computer screen presented to subjects showing all four stimulus cards and the “sun” or “rain” response choices. From Knowlton et al. (1996).


where a well-controlled tone or light is presented as the conditioning stimulus (the CS). In classic “delay” conditioning protocol, the CS lasts 250–1000 ms and coterminates with an air puff or mild electrical shock to the eye (the unconditioned stimulus or US) that produces a reflexive, unconditioned eyeblink (the UR). After many pairings of CS and US, the animal begins to produce the eyeblink after onset of the CS and prior to presentation of the US. With further training, the conditioned response (CR) occurs earlier, and its timing becomes optimized so as to be maximal at the US onset. Amazingly enough, rabbits with no cerebral cortex, basal ganglia, limbic system, thalamus, or hypothalamus showed normal retention of the condi-tioned eyeblink response (Mauk and Thompson, 1987). Further, in humans, conditioning can proceed without knowledge of the relationships between the condition-ing stimuli (Clark et al., 2002).

The essential cerebellar and brain stem circuitry that supports eyeblink conditioning has been carefully delineated by Thompson and colleagues (Fig. 50.13;

Thompson, 2005). In their studies, permanent lesions or reversible inactivation of one particular cerebellar nucleus, the interpositus nucleus, resulted in impaired acquisition and retention of classically conditioned eyeblink reflexes without affecting refl exive eyeblinks (URs). Additional compelling data indicating an im -portant role for the interpositus nucleus in this kind of motor memory come from studies using reversible inactivations of particular brain regions during train-ing. These studies showed that drug inactivation of motor nuclei that are essential for production of the CR and UR prevented the elicitation of behavior during training. However, in trials immediately following removal of the inactivation, CRs appeared in full form, showing that the neural circuit that supports UR production is not critical for learning per se. A similar pattern of results was obtained with inactivation of the axons leaving the interpositus nucleus or their target in the red nucleus, showing that the final pathway for CR production is also not required to establish the memory trace. These results, in combination with

C erebellar cortex

C limbingfibers

M ossyfibers

ToneC S

Auditorynucle i

P ontinenucle i

A ll o ther targets o f the superiorcerebellar peduncle

Inferioro live

Trigemina lnuc leus

C orneaairpuff

U S

R ednucleus

C rania lmotornuc le i

E yeblinkU R & C R

R eticularformation

Interpos itusnucleus

c

e

b

a

d


FIGURE 50.13 A schematic diagram of principal pathways involved in classical conditioning of the eyeblink reflex. The role of structures at points a–e has been studied using reversible inactivation with a local anesthetic. Inactivation at point “c” (shaded areas) prevents learning, whereas inactivation at “a,” “b,” “d,” or “e” prevents the behavioral response during inactivation, but does not block learning. From Thompson and Kim (1996).




results from other studies that addressed candidate regions upstream from the interpositus nucleus, point to areas in the interpositus nucleus as an essential locus of plasticity in this form of motor learning. Addi-tional studies have indicated that the cerebellar cortex is important for modulating the learning, such as by influencing the timing of the conditioned response to ensure that it is adaptive. As in animals, damage to the cerebellum in humans also retards the classically con-ditioned eyeblink refl ex.

Summary of Cerebellum

Eyeblink classical conditioning is one example of learning that depends on the cerebellum. Other fre-quently studied examples include adaptations of the vestibulo-ocular reflex and coordinated motor skill learning. The circuitry essential for eyeblink condition-ing has been identified and provides an experimen-tally tractable opportunity to investigate what may be generalized learning processes supported by the cere-bellum. These processes may contribute generally to motor reflex learning or even more generally to mental processing involved in precise timing.

Amygdala

Anatomy

The amygdala is a collection of anatomically and functionally heterogeneous nuclei (Swanson and Petrovich, 1998). The two major areas of the amygdala involved in emotional memory are the basolateral amygdala complex (BLA; includes the lateral, basolat-eral, and basomedial nuclei) and the central nucleus of the amygdala (CEA; Fig. 50.14B). The BLA receives input from widespread cortical areas as well as from sensory nuclei of the thalamus and thus has access to higher-level information from association areas as well as lower-level sensory information. In addition, the BLA has reciprocal connections with other brain systems, including the hippocampal and striatal memory systems. Areas within the BLA project to the central nucleus, which is the source of outputs to sub-cortical areas controlling a broad range of fear-related behaviors, including autonomic and motor responses (e.g., changes in heart rate, blood pressure, sweating, hormone release and alterations in startle response).

The amygdala contributes to emotion in several ways, including mediating emotional infl uences on attention and perception and regulating emotional responses (Phelps and LeDoux, 2005). The topic of the present section is the two ways in which the amygdala contributes to emotional memory. First, the amygdala supports the acquisition of emotional dispositions

toward stimuli. This kind of memory includes prefer-ences and aversions that can be learned unconsciously and independent of declarative memory for the events in which the disposition was acquired. Second, the amygdala mediates the influence of emotion on the consolidation of memory in other memory systems.

Role of the Amygdala in Acquiring Emotional Dispositions to Stimuli

Much of the research directed toward understand-ing the role of the amygdala in acquiring emotional dispositions to stimuli has focused on the circuitry that supports acquisition of fearful responses to simple auditory or visual stimuli (Davis, 1992; LeDoux, 2000; Fanselow and Gale, 2003). For example, in one fre-quently used task, rats are initially habituated to a chamber and then are presented with multiple pair-ings of a tone that terminates with a brief electric shock delivered through the floor of the cage (Fig. 50.14A). Subsequently, conditioned fear elicited by the tone is assessed by measuring autonomic responses, such as changes in arterial pressure, and motor responses, such as stereotypic crouching or freezing behavior. Rats with lesions in the BLA show dramatically reduced conditioned fear responses to the tone in mea-sures of both autonomic and motor responses but still show normal unconditioned fear responses to the shock itself. Rats with lesions to the central nucleus show reduced fear responses to both conditioned and unconditioned stimuli. Thus, the BLA appears to be important for acquiring an aversion to the tone, whereas the central nucleus appears to be important for producing fearful responses in general.

This basic procedure has been adapted for use with human participants, and one study that included patients with damage to the amygdala or hippocam-pus helped to distinguish the type of memory sup-ported by these two structures (Bechara et al., 1995). In this study, participants were trained in a discrimina-tive fear-conditioning protocol in which, on some trials, a monochrome color stimulus (reinforced

FIGURE 50.14 Fear conditioning. (A) Prior to training, the tone produces a transient orienting response. During training the tone is followed by a brief foot shock. Following training, the rat is reintro-duced into the chamber and freezes when the tone is presented. (B) Anatomical pathways that mediate fear conditioning. A hierarchy of sensory inputs converges on the lateral amygdala nucleus, which projects to other amygdala nuclei and then to the central nucleus, which send outputs to several effector systems for emotional responses. BNST, bed nucleus of the stria terminalis; DMV, dorsal nucleus of the vagus; NA, nucleus ambiguus; RPC, nucleus reticu-laris pontis oralis; RVL, rostral ventral nucleus of the medulla. From LeDoux (1995).


Emotionalstimulus

Neocortex

P rimarysensory

Unimodalassociation

Polymodalassociation

Entorhinalcortex

Hippocampus

Subiculum

Sensorythalamus

Amygdala

Lateral

Central

Baso-lateral

Baso-medial

S tresshormonerelease

Para-sympathetic

controlEmotionalbehavior

Sympatheticactivation

Reflexpotentiation

Corticalarousal and

attention

Basalnucleus

BNST

Para-ventricular

Anteriorpituitary

Para-brachial

DMVNA

Centralgray RP C

Lateralhypothalamus

RVLmedulla

Sound

A

B





conditioned stimulus, CS+) was presented for 2 s and then was terminated just as an unconditioned stimulus was sounded briefly (US; a loud boat horn). Auto-nomic responses to these stimuli were measured as skin conductance changes through electrodermal recordings, a measure used often as an indicator of fear or anxiety. On other trials, different colors (unrein-forced conditioned stimulus, CS−) were presented without the boat horn. Healthy individuals typically showed strong electrodermal responses to the US and with training came to show conditioned responses to the CS+ but smaller or no responses to CS− stimuli. Thus, the subjects acquired a specifi c fearful response to the CS+. One patient with selective damage to the amygdala showed normal unconditioned responses to the US but failed to develop conditioned responses to CS+ stimuli. In contrast, a patient with selective hip-pocampal damage showed robust skin conductance changes to the US and normal conditioning to CS+stimuli and appropriately smaller responses to CS-stimuli. Control subjects and the patient with selective amygdala damage answered most of these questions correctly, but the patient with hippocampal damage was severely impaired in recollecting the training events. These findings indicated that emotional condi-tioning was disrupted by amygdala damage and that declarative memory for the learning situation was impaired by hippocampal damage.

Role of the Amygdala in the Modulation of Memory

Memories of emotionally arousing events are often more vivid, more accurate, and longer lasting than memories of more neutral events. Indeed, it is adaptive for organisms to remember important events better than trivial events. Thus, it also makes sense that the brain should have evolved mechanisms for storing information in accordance with how much the infor-mation is worth remembering. Evidence from a large number of studies in experimental animals has indi-cated that the amygdala, in particular the BLA, medi-ates this memory enhancing effect (McGaugh, 2004).

Emotionally arousing events result in the release of epinephrine and glucocorticoids by the adrenal glands, which in turn result in the release of norepinephrine in the amygdala (Fig. 50.15). The release of norepi-nephrine leads to an increase in activity in the amyg-dala, which is thought to influence consolidation of memory in the other parts of the brain, either directly, though connections of the amygdala with the striatum, hippocampus, and cortex, or indirectly, through con-nections with nucleus basalis (which innervates much of the cortex). Data in support of this model include the findings that damage to the BLA or infusions of drugs in the BLA that interfere with norepinephrine

block the memory enhancing effect of emotional arousal. In addition, damage to pathways out of the amygdala, in particular the stria terminalis, also have a similar effect. Furthermore, the modulatory infl u-ence of the BLA seems to be limited to the time of the learning event as well as a short period of time after. Inactivation of the amygdala during subsequent testing of retention does not influence memory performance. Thus, the amygdala is important in modulating storage in emotionally stressful situations, but not in the maintenance or retrieval of the memory that has been modulated.

The amygdala has been strongly implicated in the enhancement of memory for emotional material in humans as well as in experimental animals. In one study, a patient with damage to the amygdala was tested for memory of an emotional story along with a group of healthy individuals (Adolphs et al., 1997). The participants watched a series of slides and listened to an accompanying narrative that told a story about a mother and son. One portion of the story involved a traumatic accident, and all participants rated this

Amygdala

NorepinephrineAcetylcholine

GABA

Experience

Hormonalsystems

EpinephrineNorepinephrineGlucocorticoidsOpioid peptides

Memory storage

Emotionalarousal


FIGURE 50.15 A schematic representation of how hormonal systems and the amygdala complex can modulate the storage of memory for emotionally arousing events through infl uences on other brain systems. See text for details. From McGaugh et al.(1992).


portion as being strongly emotional. Compared to the healthy individuals, the patient with damage to the amygdala failed to show enhancement of memory for the emotional part of the story. However, the patient performed as well as others on neutral story material. In brain imaging studies with healthy individuals, the amygdala was activated during the viewing of emo-tional material that elicited either positive or negative emotional reactions, and this activation was related to the likelihood that participants subsequently were able to remember the emotional material. In these studies, there was no correlation between activation of the amygdala and subsequent memory performance for neutral material.

Summary of Amygdala

The amygdala is well situated anatomically to attach emotional (positive and negative) dispositions to a broad range of stimuli and to mediate the acquisition of such dispositions in the absence of conscious recol-lection of the circumstances of the emotional experi-ence. In addition, components of the amygdala also mediate the modulation of memory storage during and after emotional events. Memory-modulation mechanisms are an efficient, evolutionarily adaptive method of ensuring that the strength of a memory tends to be proportional to its importance.

Cerebral Cortex

Anatomy

As discussed in Chapter 2, the posterior half of the cerebral cortex contains many functionally distinct areas that are organized into hierarchies of serial and parallel processing for each sensory modality. Addi-tionally, there are cortical areas where information from different modalities converges (so-called “asso-ciation” areas) in the parietal and temporal lobes. The anterior cerebral cortex contains a similar hierarchy of motor areas, as well as association areas in the prefron-tal cortex involved with motor planning, higher order cognition, and working memory (Fuster, 2001; Chapter 52).

Some of the examples provided so far, such as stan-dard eyeblink classical conditioning, illustrate that some simple forms of learning can circumvent the cerebral cortex. Although simplified learning para-digms are indispensable as precise instruments to dissect local circuitry, these examples should not overshadow the fact that the global circuitry of each memory system discussed so far includes the cerebral cortex. The cerebral cortex provides access to incoming information as well as output routes to infl uence

behavior, access that is important for even the cerebel-lum. Yet plasticity in the cerebral cortex also directly supports types of memory that do not require compo-nents of the other memory systems. For each area of the cortex, the modification of its information process-ing circuitry through alterations in synaptic connectiv-ity and membrane excitability underlies its direct participation in memory. These alterations can be dra-matic, such as in examples of experience-dependent reorganization that occurs in perceptual learning, but can also be more subtle, such as in the phenomenon of repetition priming.

Perceptual Learning

Specific training experiences can result in the modi-fi cation of specifi c parts of the cortex. For example, in one study, monkeys were trained over a period of several weeks to discriminate small frequency differ-ences in tone stimuli. In subsequent recordings, increases were observed in the size of the auditory cortex representation of the task-relevant frequencies and the sharpness of tuning to these frequencies (Recanzone et al., 1993). Furthermore, changes in area of the cortical representation were correlated with the improvement in task performance. In other studies using classical conditioning in anesthetized guinea pigs, the tuning curves of single neurons in the audi-tory cortex were first characterized (Weinberger, 2007). Individual neurons typically showed preferences for a specific tone frequency. Subsequently, the animal was presented with repeated pairings of a nonpreferred tone and foot shocks. This training resulted in long-lasting shifts in the tuning curves of auditory neurons toward the frequency of the training stimulus. A similar expansion of perceptual representations for training stimuli has been observed in the somatosen-sory cortex of monkeys following acquisition of a tactile discrimination. The alterations of the sensory cortex following experience is thought to contribute to adaptations and tuning of perceptual representation systems and, in coordination with reorganization of cortical motor representations, may also contribute to procedural memories. In both cases, this contribution is derived specifically from alterations in the percep-tual and motor processing for the relevant learning tasks.

Functional imaging studies in humans have revealed that perceptual learning can occur for stimuli that are more complex than simple tones, yet similar to the studies in guinea pigs with tones, these studies have also suggested that the cortical area important for per-ceiving the stimuli is also an important area of plastic-ity. For example, in one study participants were scanned while viewing pictures of nonsense objects




both before and after performing a perceptual discrim-ination task that included half of the objects (Op de Beeck, 2006). Following the perceptual discrimination task, the pattern of brain activity in object-responsive areas in visual cortex had changed for the objects included in the discrimination task but did not differ for the objects not included in the discrimination task. Thus, it appears that many of the areas in the cerebral cortex that are specialized for specific perceptual or information processing functions—from areas involved in perceiving simple tones to areas involved in per-ceiving complex objects—are capable of supporting perceptual learning for those stimuli.

Repetition Priming

Another example of memory thought to be sup-ported directly by the cerebral cortex is the phenome-non known as repetition priming (or priming). Priming involves initial presentation of a list of words, pictures of objects, or nonverbal materials and then subsequent reexposure to fragments or very brief presentation of whole items. For example, in one version of priming, words are flashed on a screen so briefly that many of the words are unreadable. The measure of priming is the increased likelihood of identifying previously viewed words as compared to words not encountered recently, and the benefit of repetition has been found to be independent of the ability to recognize which words had been read initially. Studies with brain-damaged patients suggest that priming depends on cortical areas involved in initially processing the stimuli. In one study, the amnesic patient H.M. and a patient with a large lesion in the right visual cortex attempted to read new and repeated words fl ashed on a computer screen (Keane et al., 1995). Healthy indi-viduals and H.M. showed evidence of priming for the repeated words, but the patient with visual cortex damage did not, suggesting that the priming effect depended on perceptual areas in the visual cortex. In contrast, the patient with visual cortex damage per-formed normally on conventional memory tests. In functional brain imaging studies with healthy indi-viduals, the visual cortex showed decreased activation to recently presented materials, suggesting that less neural activity is required to identify words recently processed. These findings suggest that the neuronal basis of priming is an increase in the efficiency and bias in the direction of cortical sensory processing associ-ated with perceptual identifi cation.

Summary of Cerebral Cortex

Studies in both animals and humans show that the cerebral cortex is highly plastic in that its representa-tions can be altered after experience. Distinct cortical

areas are reorganized by specific training experiences to meet demands for specific types of perceptual rep-resentation and memory. Also, modality-relevant cortical circuits are able to support the priming of per-ceptual representations.

BEHAVIOR SUPPORTED BY MULTIPLE MEMORY SYSTEMS

The discussion of memory systems so far has focused on dissociations between systems in illustrating the unique contributions of the central structure in each system. However, behavior is a product of the entire nervous system and typically is guided by contribu-tions of more than one form of memory. For example, training in a sport normally results in declarative memory of instructions from one’s coach but also leads to motor skill learning from repetitive practice. The following section considers emerging data regarding how the memory systems discussed so far might col-laborate, compete, or operate in parallel and illustrates some progress in understanding how the distribution of memory throughout the brain ordinarily contrib-utes to our day-to-day activities.

One example of a direct collaboration between two memory systems is the influence of the amygdala on hippocampus-dependent memory for emotion-induc-ing material. The section earlier in the chapter on the amygdala discussed how emotional arousal leads to the amygdala engaging in modulation of declarative memory, such that stimuli that elicit either positive or negative emotional responses tend to be remembered better than neutral stimuli. This process is thought to result from amygdalar projections modulating synap-tic plasticity either in the hippocampus and parahip-pocampal region or in cortical areas fundamental to the hippocampal memory system. It has been sug-gested that the amygdala also modulates other memory systems, such as the striatal memory system, and that the amygdala may play a generalized role in memory modulation for emotion inducing stimuli in addition to its role in attaching positive or negative dispositions to stimuli (McGaugh, 2005).

Another example comes from an indirect collabora-tion between two very dissimilar memory systems, the hippocampal memory system and the cerebellar memory system. The section earlier in the chapter on the cerebellum detailed that the essential circuitry for standard eyeblink classical conditioning involved no forebrain structures and instead centered on deep cer-ebellar nuclei. Yet the situation changes if the param-eters for standard eyeblink classical conditioning,


referred to as delay eyeblink conditioning, are altered slightly such that the tone that normally overlaps with the airpuff is moved earlier in time such that it termi-nates as little as a half second prior to the onset of the airpuff, a version called trace conditioning (Christian and Thompson, 2003). Trace eyeblink conditioning depends on the same cerebellar and hindbrain cir-cuitry as delay eyeblink conditioning, but trace condi-tioning depends additionally on forebrain structures including the hippocampus and prefrontal cortex. In studies with healthy human participants, trace (but not delay) eyeblink conditioning was found to relate closely to participants’ knowledge about the relation-ship between the tone and the airpuff (Clark et al., 2001). A possible explanation of these findings is that the silent interval that separates the tone and airpuff may outstrip the learning abilities of the cerebellum and may cause it to require inputs from additional forebrain structures, such as the hippocampal system, to bridge that brief interval. The growing understand-ing of the relationship between hippocampus and cer-ebellum illustrates the tendency to collaborate between even the most dissimilar memory systems, and exem-plifies the progress being made toward understanding how memory systems interact in general.

In other examples, memory systems may operate in parallel or may even compete for control of behavior. For example, the previously discussed studies con-trasting the role of the striatum and hippocampus in supporting “response” learning and “place” learning, respectively, illu

c h a p t e r 50nwkpsych.rutgers.edu/~jose/courses/578_mem_learn/...vii. behavioral and cognitive...

Documents