491

Findings from laboratory studies, largely on rats and monkeys, have reproduced the symptoms of patients such as H. M. and J. K. by injuring the animals’ medial temporal regions and basal ganglia, respectively. Other structures, most notably in the frontal and temporal lobes, also participate in certain types of explicit memory. We now consider the systems for explicit and implicit memory separately.

The dramatic amnesic syndrome discovered in H. M. in the 1950s led investigators to focus on the hippocampus, at the time regarded as a large brain structure in search of a function. But H. M. had other damaged structures, too, and the initial focus on the hippocampus as the location of explicit memory processing turned out to be misguided.

After decades of anatomical and behavioral studies sorted out the complexities, consensus on the anatomy of explicit memory had coalesced by the mid-1990s. The primary structures for explicit memory include the medial temporal region, the frontal cortex, and structures closely related to them.

Before considering the model, we first revisit medial temporal anatomy. Review Figure 14-5’s summary of findings, which show that memories for objects’ colors and motion characteristics reside in separate temporal lobe locations. Thus the medial temporal region receives multiple sensory inputs.

The macaque monkey’s medial temporal region shares many anatomical similarities with the region in the human brain. In addition to the hippocampus and amygdala, three medial temporal areas illustrated in Figure 14-8 take part in explicit memory: lying adjacentto the hippocampus are the entorhinal cortex, the parahippocampal cortex, and the perirhinal cortex. A sequential arrangement of two-way connections, charted in Figure 14-9, projects from the major cortical regions into the perirhinal and parahippocampal cortices, which in turn project to the entorhinal cortex, then to the hippocampus.

The prominent cortical input to the perirhinal region is from the visual ventral stream coursing through the temporal lobe. The perirhinal region is thus a prime candidate location for visual object memory. Similarly, the parahippocampal cortex receives strong input from parietal regions believed to take part in visuospatial processing and likely participates in visuospatial memory—using visual information to recall an object’s location in space.

492

Because both the perirhinal and the parahippocampal regions project to the entorhinal cortex, this region probably participates in more integrative memory functions. The entorhinal cortex is in fact the first area to show cell death in Alzheimer disease, a dementia characterized by severe deficits in explicit memory (see Clinical Focus 14-3, Alzheimer Disease).

The Hippocampus and Spatial Memory

We are left with a conundrum. If the hippocampus is not the key structure in explicit memory even as it receives the entorhinal connections, what does it do? John O’Keefe and Lynn Nadel were the first to advance the idea, in 1978, that the hippocampus is probably engaged in visuospatial memory processes required for places, such as recalling an object’s location.

493

Certainly both laboratory animals and human patients with selective hippocampal injury have severe deficits in various forms of spatial memory. Similarly, monkeys with hippocampal lesions have difficulty learning the location of objects (visuospatial learning), as can be demonstrated in tasks such as the ones illustrated in Figure 14-10.

Monkeys are trained to displace objects to obtain a food reward (Figure 14-10A), then given one of two tasks. In the visual recognition task (Figure 14-10B), the animal displaces a sample object for the food reward. After a short delay, the animal is presented with two objects. One is novel. The task is to learn to displace the novel object for the food reward. This task tests explicit visual object memory. Monkeys with perirhinal lesions are impaired at the task.

In the object position task in Figure 14-10C, the monkey is shown one object to be displaced for a food reward. Then the monkey is shown the same object along with a second identical one. The task is to learn to displace the object that is in the same position as it was in the initial presentation. Monkeys with hippocampal lesions are selectively impaired at this task.

From these results, we would predict that a species with an especially good spatial memory should have bigger hippocampi than do species with a poorer spatial memory. David Sherry and his colleagues (1992) tested this hypothesis in birds.

Many birds are cachers: they harvest sunflower seeds and other favored foods and hide (cache) them to eat later. Some birds can find hundreds of items they have cached. To evaluate whether the hippocampus plays a role in this activity, Sherry and his coworkers measured hippocampal size in closely related bird species, only one of which is a food cacher. As shown in Figure 14-11, the hippocampal formation is larger in birds that cache food than in birds that do not. In fact, the hippocampi of food-storing birds are more than twice as large as expected for birds of their brain size and body weight.

Sherry found a similar relation when he compared different species of food-storing rodents. Merriam’s kangaroo rats, rodents that store food throughout their territory, have larger hippocampi than bannertail kangaroo rats, which store food only in their burrow. Hippocampal size, both in birds and in mammals, appears to be related to the cognitive demands of two highly spatial activities, foraging for and storing food.

494

One prediction we might make based on the Sherry experiments is that people who have a job with high spatial demands have large hippocampi. Taxi drivers in London fit this category. Successful candidates for a cab driver’s license in London must demonstrate that they know the location of every street in that huge and ancient city. Using MRI, Eleanor Maguire and her colleagues (2000) found the posterior region of the hippocampus in London taxi drivers to be significantly larger than the same region in the control participants. This finding presumably explains why a select few pass a spatial memory test that most of us would fail miserably.

Spatial Cells in the Hippocampal Formation

Given the role of the hippocampus in spatial behavior, we might predict that individual cells would code spatial information. They do. Three classes of spatially related cells have been identified in the rat and mouse hippocampus (Figure 14-12). Together they form an internal GPS, a neural global positioning system. That is, neurons vigorously fire when an animal is in a specific place in the environment. In 2014, John O’Keefe, Edvard Moser, and May-Britt Moser were awarded the Nobel Prize in Physiology or Medicine for this discovery.

Place cells, shown in Figure 14-12A and B, discharge when rats are in a spatial location, irrespective of its orientation. Head direction cells (Figure 14-12C) discharge whenever a rat’s head points in a particular direction. Grid cells (Figure 14-12D) discharge at many locations, forming a virtual grid invariant to changes in the rat’s direction, movement, or speed.

These cells are not localized to the hippocampus but are found within a network that includes the hippocampus as its central structure. Place cells and head direction cells are located in the hippocampus and closely related structures. Grid cells are found in the entorhinal cortex, a major afferent route into the hippocampus (see Figure 14-9). Taken together, we can envision the place cell system as indicating where things are in the world, the grid system indicating how big our present navigating environment is, and the head direction and grid systems telling us where we ourselves are in the environment.

Reciprocal Connections for Explicit Memory

Explicit memory’s temporal pathway is reciprocal: connections from the neocortex run to the entorhinal cortex, then back to the neocortex (see Figure 14-9). Reciprocal connections have two benefits:

495

Although we have focused on the medial temporal regions, other structures also are important in explicit memory. People with frontal lobe injuries are not amnesic like H. M. or J. K., but they do have difficulties with memory for the temporal (time) order of events. Imagine being shown a series of photographs and asked to remember them. A few minutes later, you are asked whether you recognize two photographs and if so, to indicate which one you saw first.

H. M. would not remember the photographs. People with frontal lobe injuries would recall seeing the photographs but would have difficulty recalling which one they had seen more recently. The frontal lobe’s role in explicit memory clearly is subtler than that of the medial temporal lobe.

THE FRONTAL LOBE AND SHORT-TERM MEMORY All sensory systems in the brain send information to the frontal lobe, as do the medial temporal regions. This information is not used for direct sensory analysis, so it must have another purpose. In general, the frontal lobe appears to participate in many forms of short-term memory.

Joaquin Fuster (e.g., Fuster, Bodner, & Kroger, 2000) studied single-cell activity in the frontal lobe during short-term memory tasks. For example, if monkeys are shown an object that they must remember for a short time before being allowed to make a response, neurons in the prefrontal cortex show sustained firing during the delay. Consider the tests illustrated in Figure 14-13:

Fuster found that in each task certain cells in the prefrontal cortex fire throughout the delay. Animals that have not learned the task show no such cell activity. Curiously, if a trained animal makes an error, its cellular activity corresponds: the cells stop responding before the error occurs. They have “forgotten” the cue.

TRACING THE EXPLICIT MEMORY CIRCUIT People who have chronically abused alcohol can develop an explicit memory disturbance known as Korsakoff syndrome. In some cases, severe deficits in explicit memory extend to implicit memory as well. Korsakoff syndrome is caused by a thiamine (vitamin B₁) deficiency that kills cells in the medial part of the diencephalon—the between brain at the top of the brainstem—including the medial thalamus and mammillary bodies in the hypothalamus. In 80 percent of Korsakoff patients, the frontal lobes show atrophy (loss of cells). The memory disturbance is probably so severe because the damage includes not only forebrain but also brainstem structures (see Clinical Focus 14-4, Korsakoff Syndrome).

496

Mortimer Mishkin and his colleagues (Mishkin, 1982; Murray, 2000) proposed a neural circuit for explicit memory that incorporates the evidence from both humans and laboratory animals with injuries to the temporal and frontal lobes. Figure 14-14 presents a modified version of the Mishkin model. Anatomically (Figure 14-14A), it includes not only the frontal and temporal lobes but also the medial thalamus, implicated in Korsakoff syndrome, and the basal forebrain–activating systems implicated in Alzheimer disease. Figure 14-14B charts the information flow:

497

Amnesia often appears to be time-dependent. H. M. could not form new explicit memories but appeared to have good recall of facts and events from periods long before his surgery, including his childhood. Such findings led to the idea that the medial temporal region could not be the ultimate storage site for long-term memories; more likely that site was the neocortex (for a review, see Squire et al., 2015).

This idea led to the hypothesis that the hippocampus consolidates new memories, a process that makes them permanent. In consolidation, or stabilizing a memory trace after learning, memories move from the hippocampus to diffuse neocortical regions. Once they move, hippocampal involvement is no longer needed.

It is not clear how memories are moved, however, or how long it takes. Robert Sutherland and his colleagues (2010) propose a model of consolidation, the distributed reinstatement theory. In their model, a learning episode rapidly produces a stored memory representation that is strong in the hippocampus but weak elsewhere. The memory is replayed on the time scale of hours or days after the learning, leading to enhanced representations outside the hippocampus. Each repetition of the learning—that is, each practice—progressively enhances the nonhippocampal memory representation. Then if the hippocampus is extensively damaged, the memory remains.

Memory is not constant over time. When people get misleading information about events they have experienced, for example, their later recall of those events often is modified. Indeed, this contributes to the notorious unreliability of eyewitness testimony. (Do you remember the “smashing” and “bumping” car collisions from Section 14-1?) The fact that memories appear changeable seems to fly in the face of the consolidation concept, which presumes that once consolidated, memories are fixed.

One solution to this conundrum suggests that whenever a memory is replayed in the mind, it is open to further consolidation, or reconsolidation, the process of restabilizing a memory trace after the memory is revisited. Interest in this idea has been intense, and although the final story is yet to be written, it appears that reconsolidation does occur, at least for some types of memories.

One way to think of the process is to see memory consolidation as never-ending: new information is constantly being integrated into existing memory networks (for a review, see McKenzie and Eichenbaum, 2011). After all, we frequently recall memories, rehash them, and integrate them with new events. Our memories are not laid on a tabula rasa but must be interwoven into a lifetime of memories.

498

One implication of reconsolidation is that it ought to be possible to erase negative memories by using amnesic agents when the memory is revisited. This idea has important implications for reducing or eliminating the effects of strong emotional experiences, such as those seen in posttraumatic stress disorder (PTSD). One promising agent is the inert gas xenon, which has been used in humans as a fast-acting anesthetic and has been shown to erase memories during reconsolidation in a rat model of PTSD (Meloni et al., 2014).

Hypothesizing that the basal ganglia are central to implicit memory, Mishkin and his colleagues proposed a neural circuit for implicit memories (Mishkin, 1982; Mishkin et al., 1997). As Figure 14-15 shows, the basal ganglia receive input from the entire neocortex and send projections first to the ventral thalamus and then to the premotor cortex. The basal ganglia also receive widely and densely distributed projections from dopamine-producing cells in the substantia nigra. Dopamine appears necessary for basal ganglia circuits to function and may indirectly participate in implicit memory formation.

The connection from the cortex to the basal ganglia in the implicit memory system flows only in one direction. Most of the neocortex receives no direct information regarding activities in the basal ganglia. Mishkin believes that this unidirectional flow accounts for the unconscious nature of implicit memories. For memories to be conscious, the neocortical regions involved must receive feedback, as they do in the explicit memory system Mishkin and colleagues proposed (see Figure 14-14).

Mishkin’s models show why people with basal ganglia dysfunction, as occurs in Parkinson disease, have implicit memory deficits. People with frontal or temporal lobe injuries, by contrast, have relatively good implicit memories even as they may have profound explicit memory disturbances. Some people with Alzheimer disease can play games expertly, even with no recollection of having played them before.

Daniel Schacter (1983) wrote of a golfer with Alzheimer disease whose medial temporal system was severely compromised by the disease, but his basal ganglia were unaffected. Despite his explicit knowledge impairment, as indexed by his inability to find balls he had shot or to remember how many strokes he made on each hole, the man retained his ability to play the game.

Whether emotional memory for the affective properties of stimuli or events is implicit or explicit is not altogether clear. It could be both. Certainly people can react with fear to specific stimuli they can identify, and we have seen that they can also fear situations for which they do not seem to have specific memories. Panic disorder is a common pathology of emotional memory. People show marked anxiety but cannot identify a specific cause. Emotional memory has a unique anatomical component—the amygdala, which mediates fear conditioning (see Experiment 14-1) and seems to evoke our feelings of anxiety toward stimuli that by themselves would not typically produce fear.

Emotional memory has been studied most thoroughly in fear conditioning by pairing unpleasant stimuli, such as foot shock, with a tone. Michael Davis (1992) and Joseph LeDoux (1995) used fear conditioning to demonstrate that the amygdala is critical to emotional memory. Damage to the amygdala abolishes emotional memory but has little effect on implicit or explicit memory.

The amygdala has close connections with medial temporal cortical structures as well as with the rest of the cortex. It also sends projections to brainstem structures that control autonomic responses such as blood pressure and heart rate; to the hypothalamus, which controls hormonal systems; to the periaqueductal gray matter (PAG), which affects pain perception; and to the enteric nervous system. The amygdala hooks in to the implicit memory system through its connections with the basal ganglia (Figure 14-16).

499

Fear is not the only aspect of emotional memory the amygdala codes, as a study of severely demented patients by Bob Sainsbury and Marjorie Coristine (1986) nicely illustrates. The patients were believed to have severe cortical abnormalities but intact amygdalar functioning. The researchers first established that the patients’ ability to recognize photographs of close relatives was severely impaired.

The patients were then shown four photographs, one depicting a relative (either a sibling or a child) who had visited in the past 2 weeks. The task was to identify the person whom they liked better than the other three. Although the subjects were unaware they knew anyone depicted in the photographs, they consistently preferred pictures of their relatives. This result suggests that although the explicit, and probably the implicit, memory of the relative was gone, each patient‘s emotional memory guided his or her preference.

We tend to remember emotionally arousing experiences vividly, a fact confirmed by findings from both animal and human studies. James McGaugh (2004) concluded that emotionally significant experiences, pleasant and unpleasant, must activate hormonal and brain systems that act to stamp in these vivid memories.

McGaugh noted that many neural systems probably take part, but the basolateral part of the amygdala is critical. The general idea is that emotionally driven hormonal and neurochemical activating systems (probably cholinergic and noradrenergic) stimulate the amygdala. The amygdala in turn modulates the laying down of emotional memory circuits in the rest of the brain, especially in the medial temporal and prefrontal regions and in the basal ganglia. We would not expect people with amygdala damage to have enhanced memory for emotion-laden events, and they do not (Cahill et al., 1995).

Neural Systems Underlying Explicit and Implicit Memories

Before you continue, check your understanding.

Question 1

Question 2

Question 3

Question 4

Question 5

Answers appear in the Self Test section of the book.