In Search of Memory: The Emergence of a New Science of Mind (37 page)

I found this work enthralling. As a scientific contribution to brain science, it stands as the most fundamental advance in our understanding of the organization of the cerebral cortex since the work of Cajal at the turn of the last century. Cajal revealed the precision of the interconnections between populations of individual nerve cells. Mountcastle, Hubel, and Wiesel revealed the functional significance of those patterns of interconnections. They showed that the connections filter and transform sensory information on the way to and within the cortex, and that the cortex is organized into functional compartments, or modules.

As a result of the work of Mountcastle, Hubel, and Wiesel, we can begin to discern the principles of cognitive psychology on the cellular level. These scientists confirmed the inferences of the Gestalt psychologists by showing us that the belief that our perceptions are precise and direct is an illusion—a perceptual illusion. The brain does not simply take the raw data that it receives through the senses and reproduce it faithfully. Instead, each sensory system first analyzes and deconstructs, then restructures the raw, incoming information according to its own built-in connections and rules—shades of Immanuel Kant!

The sensory systems are hypothesis generators. We confront the world neither directly nor precisely, but as Mountcastle pointed out:

SUBSEQUENT WORK ON THE VISUAL SYSTEM SHOWED THAT IN
addition to dissecting objects into linear segments, other aspects of visual perception—motion, depth, form, and color—are segregated from one another and conveyed in separate pathways to the brain, where they are brought together and coordinated into a unified perception. An important part of this segregation occurs in the primary visual area of the cortex, which gives rise to two parallel pathways. One pathway, the “what” pathway, carries information about the form of an object: what the object looks like. The other, the “where” pathway, carries information about the movement of the object in space: where the object is located. These two neural pathways end in higher regions of the cortex that are concerned with more complex processing.

The idea that different aspects of visual perception might be handled in separate areas of the brain was predicted by Freud at the end of the nineteenth century, when he proposed that the inability of certain patients to recognize specific features of the visual world was due not to a sensory deficit (resulting from damage to the retina or the optic nerve), but to a cortical defect that affected their ability to combine aspects of vision into a meaningful pattern. These defects, which Freud called
agnosias
(loss of knowledge), can be quite specific. For example, there are specific defects caused by lesions in either the “where” or the “what” pathway. A person with depth agnosia due to a defect in the “where” system is unable to perceive depth but has otherwise intact vision. One such person was unable “to appreciate depth or thickness of objects seen…. The most corpulent individual might be a moving cardboard figure; everything is perfectly flat.” Similarly, persons with motion agnosia are unable to perceive motion, but all other perceptual abilities are normal.

Striking evidence indicates that a discrete region of the “what” pathway is specialized for face recognition. Following a stroke, some people can recognize a face as a face, the parts of the face, and even specific emotions expressed on the face but are unable to identify the face as belonging to a particular person. People with this disability (prosopagnosia) often cannot recognize close relatives or even their own face in the mirror. They have not lost the ability to recognize a person’s identity, they have lost the connection between a face and an identity. To recognize a close friend or relative, these patients must rely on the person’s voice or other nonvisual clues. In his classic essay “The Man Who Mistook His Wife for a Hat,” the gifted neurologist-neuropsychologist Oliver Sacks describes a patient with prosopagnosia who failed to recognize his wife sitting next to him and, thinking she was his hat, tried to pick her up and put her on his head as he was about to leave Sacks’s office.

How is information about motion, depth, color, and form, which is carried by separate neural pathways, organized into a cohesive perception? This problem, called the binding problem, is related to the unity of conscious experience: that is, to how we see a boy riding a bicycle not by seeing movement without an image or an image that is stationary, but by seeing in full color a coherent, three-dimensional, moving version of the boy. The binding problem is thought to be resolved by bringing into association temporarily several independent neural pathways with discrete functions. How and where does this binding occur? Semir Zeki, a leading student of visual perception at University College, London, put the issue succinctly:

WHEN A COGNITIVE NEUROSCIENTIST PEERS DOWN AT THE
brain of an experimental animal, the scientist can see which cells are firing and can read out and understand what the brain is perceiving. But what strategy does the brain use to read itself out? This question, which is central to the unitary nature of conscious experience, remains one of the many unresolved mysteries of the new science of mind.

An initial approach was developed by Ed Evarts, Robert Wurtz, and Michael Goldberg at NIH. They pioneered methods for recording the activity of single nerve cells in the brains of intact, behaving monkeys focusing on cognitive tasks that require action and attention. Their new research techniques enabled investigators such as Anthony Movshon at NYU and William Newsome at Stanford to correlate the action of individual brain cells with complex behavior—that is, with perception and action—and to see the effects on perception and action of stimulating or reducing activity in small groups of cells.

These studies also made it possible to examine how the firing of individual nerve cells involved in perceptual and motor processing is modified by attention and decision making. Thus unlike behaviorism, which focused only on the behavior stemming from an animal’s response to a stimulus, or cognitive psychology, which focused on the abstract notion of an internal representation, the merger of cognitive psychology and cellular neurobiology revealed an actual physical representation—an information-processing capability in the brain—that leads to a behavior. This work demonstrated that the unconscious inference described by Helmholtz in 1860, the unconscious information processing that intervenes between a stimulus and a response, could also be studied on the cellular level.

The cellular studies of internal representation in the cerebral cortex of the sensory and motor world were extended in the 1980s with the introduction of brain imaging. These techniques, such as positron emission tomography (PET) and functional magnetic resonance imaging (fMRI), carried the work of Paul Broca, Carl Wernicke, Sigmund Freud, the British neurologist John Hughlings Jackson, and Oliver Sacks a giant step forward by revealing the locale in the brain of a variety of complex behavioral functions. With these new technologies, investigators could look into the brain and see not simply single cells, but also neural circuits in action.

I HAD BECOME CONVINCED THAT THE KEY TO UNDERSTANDING
the molecular mechanisms of spatial memory was understanding how space is represented in the hippocampus. As one might expect because of its importance in explicit memory, the spatial memory of environments has a prominent internal representation in the hippocampus. This is evident even anatomically. Birds in which spatial memory is particularly important—those that store food at a large number of sites, for example—have a larger hippocampus than other birds.

London taxi drivers are another case in point. Unlike cabbies elsewhere, those in London must pass a rigorous examination to obtain their license. In this test, they must demonstrate that they know every street name in London and the most efficient routes for traveling between two points. Functional magnetic resonance imaging revealed that after two years of this rigorous orientation to the streets of the city, London taxi drivers have a larger hippocampus than other persons the same age. Indeed, the size of their hippocampus continues to increase with time on the job. Moreover, brain-imaging studies show that the hippocampus is activated during imagined travel, when a taxi driver is asked to recall how to get to a particular destination. How, then, is space represented on the cellular level within the hippocampus?

To address these questions, I brought the tools and insights of molecular biology to bear on existing studies of the internal representation of space in mice. We had begun by using genetically modified mice to study the effect of specific genes on long-term potentiation in the hippocampus and on explicit memory of space. We were now ready to ask how long-term potentiation helps stabilize the internal representation of space and how attention, a defining feature of explicit memory storage, modulates the representation of space. This compound approach—extending from molecules to mind—opened up the possibility of a molecular biology of cognition and attention and completed the outlines of a synthesis that led to a new science of mind.

I
n all living creatures, from snails to people, knowledge of space is central to behavior. As John O’Keefe notes, “Space plays a role in all our behaviour. We live in it, move through it, explore it, defend it.” Space is not only a critical sense but a fascinating one because unlike other senses, space is not analyzed by a specialized sensory organ. How, then, is space represented in the brain?

Kant, one of the forefathers of cognitive psychology, argued that the ability to represent space is built into our minds. He pictured people as being born with principles of ordering space and time, so that when other sensations are elicited—be they objects, melodies, or tactile experiences—they are automatically interwoven in specific ways with space and time. O’Keefe applied this Kantian logic about space to explicit memory. He argued that many forms of explicit memory (for example, memory for people and objects) use spatial coordinates—that is, we typically remember people and events in a spatial context. This is not a new idea. In 55
B.C.
, Cicero, the great Roman poet and orator, described the Greek technique (used to this day by some actors) of remembering words by picturing the rooms of a house in sequence, associating words with each room, and then mentally walking through the rooms in the right order.

Because we do not have a sensory organ dedicated to space, the representation of space is a quintessentially cognitive sensibility: it is the binding problem writ large. The brain must combine inputs from several different sensory modalities and then generate a complete internal representation that does not depend exclusively on any one input. The brain commonly represents information about space in many areas and many different ways, and the properties of each representation vary according to its purpose. For example, for some representations of space the brain typically uses
egocentric
coordinates (centered on the receiver), encoding, for example, where a light is relative to the fovea or where an odor or touch comes from with respect to the body. Egocentric representation is also used by people or monkeys for orienting to a sudden noise by making an eye movement to a particular location, by
Drosophila
in avoiding of an odor with unpleasant associations, or by
Aplysia
in generating its gill-withdrawal reflex. For other behaviors, like memory for space in the mouse or in people, it is necessary to encode the organism’s position relative to the outside world and the relationship of external objects to one another. For these purposes the brain uses
allocentric
coordinates (centered on the world).

Studies of the simpler sensory maps for touch and vision in the brain, which are based on egocentric coordinates, provided a springboard for studies of the more complex representation of allocentric space. But the spatial map discovered in 1971 by O’Keefe differs radically from the egocentric sensory maps for touch and vision discovered by Wade Marshall, Vernon Mountcastle, David Hubel, and Torsten Wiesel because it is not dependent on any given sensory modality. Indeed, in 1959, when Alden Spencer and I tried to decipher how sensory information comes into the hippocampus, we recorded from individual nerve cells while we stimulated different individual senses, but we failed to obtain a brisk response. We did not realize that the hippocampus is concerned with perception of the environment and therefore represents multisensory experience.

John O’Keefe was the first to realize that the hippocampus of rats contains a multisensory representation of extrapersonal space. O’Keefe found that as an animal walks around an enclosure, some place cells fire action potentials only when that animal moves into a particular location, while others fire when the animal moves to another place. The brain breaks down its surroundings into many small, overlapping areas, similar to a mosaic, each represented by activity in specific cells in the hippocampus. This internal map of space develops within minutes of the rat’s entrance into a new environment.

I BEGAN TO THINK ABOUT THE SPATIAL MAP IN
1992,
WONDERING
how it is formed, how it is maintained, and how attention directs its formation and maintenance. I was struck by the fact that O’Keefe and others had found that the spatial map of even a simple locale does not form instantaneously but over ten to fifteen minutes of the rat’s entrance into the new environment. This suggests that the formation of the map is a learning process; practice makes perfect also for space. Under optimal circumstances this map remains stable for weeks or even months, much like a memory process.

Unlike vision, touch, or smell, which are prewired and based on Kantian
a priori
knowledge, the spatial map presents us with a new type of representation, one based on a combination of
a priori
knowledge and learning. The
general
capability for forming spatial maps is built into mind, but the
particular
map is not. Unlike neurons in a sensory system, place cells are not switched on by sensory stimulation. Their collective activity represents the location where the animal
thinks
it is.

I now wanted to know whether the same molecular pathways needed to induce long-term potentiation and spatial memory in our experiments on the hippocampus also form and maintain the spatial map. Although O’Keefe had discovered place cells in 1971, and Bliss and Lømo had discovered long-term potentiation in the hippocampus in 1973, no attempt had been made to connect the two findings. When we began studying spatial maps in 1992, nothing was known about the molecular steps whereby one is formed. This situation illustrates once again why working at the borders between disciplines—in this case, between the biology of place cells and the molecular biology of intracellular signaling—can be highly informative. What a scientist explores in an experiment is in good part determined by the intellectual context in which that scientist functions. Few things are more exhilarating than bringing a new way of thinking to another discipline. This cross-fertilization of disciplines is what Jimmy Schwartz, Alden Spencer, and I had in mind back in 1965 when we called our new division at NYU “neurobiology
and
behavior.”

In collaboration with Robert Muller, one of the pioneers in studying place cells, we found that some of the same molecular actions responsible for long-term potentiation are indeed necessary for preserving a spatial map over a long period. We knew that protein kinase A turns on the genes and thus initiates the protein synthesis necessary for the late phase of long-term potentiation. Similarly, we found that although neither protein kinase A nor protein synthesis is needed for the initial formation of a map, they both are essential for the map to become “fixed” over the long term, so that the mouse recalls the same map every time it enters the same space.

Finding that protein kinase A and protein synthesis are required for the stabilization of the spatial map raised a further question: Does the spatial map we record in the hippocampus enable animals to have explicit memory—that is, to act as if they are familiar with a given environment? Are these maps the actual internal representation, the neural correlates of explicit memory of space? In his initial formulation O’Keefe considered the cognitive map an internal representation of space that the animal uses for navigation. He therefore saw the map as more a navigational representation, much like a compass, than a representation of the memory itself. We explored this question and found that indeed, when we blocked protein kinase A or inhibited protein synthesis, we interfered not only with the long-term stability of the spatial map but also with the ability to retain long-term spatial memories. Thus we had direct genetic evidence that the map correlates with spatial memory. Moreover, we found that in spatial memory, as in the simple implicit memory underlying the gill-withdrawal reflex in
Aplysia
, there is a distinction between the processes involved in acquiring the map (and holding onto it for a few hours) and maintaining the map in stable form for the long term.

DESPITE CERTAIN SIMILARITIES, THE EXPLICIT MEMORY OF SPACE
in people differs from implicit memory in profound ways. In particular, explicit memory requires selective attention for encoding and for recall. Therefore, to examine the relation between neural activity and explicit memory, we now needed to address the issue of attention.

Selective attention is widely recognized as a powerful factor in perception, action, and memory—in the unity of conscious experience. At any given moment, animals are inundated with a vast number of sensory stimuli, yet they pay attention to only one or a very small number of them, ignoring or suppressing the rest. The brain’s capacity for processing sensory information is more limited than its receptors’ capacity for measuring the environment. Attention therefore acts as a filter, selecting some objects for further processing. It is in large part because of selective attention that internal representations do not replicate every detail of the external world and sensory stimuli alone do not predict every motor action. In our moment-to-moment experience, we focus on specific sensory information and exclude the rest (more or less). If you raise your eyes from this book to look at a person entering the room, you are no longer paying attention to the words on the page. At the same time, you are not attending to the decor of the room or other people in the room. If asked to report your experience later, you are more likely to remember that a person had entered the room than, say, that there was a small scratch on the wall. This focusing of the sensory apparatus is an essential feature of all perception, as William James noted in his seminal book,
The Principles of Psychology
, in 1890:

Attention also allows us to bind the various components of a spatial image into a unified whole. Cliff Kentros, a postdoctoral fellow, and I chose to address the link between attention and spatial memory by asking whether attention is required for the spatial map. If so, does attention alter the formation or the stability of the map? To test these ideas, we exposed mice to four conditions that require increasing degrees of attention. The first, basal or ambient attention is the attention that is present even in the absence of further stimulation. Here, animals walked around in an enclosure without distracting stimuli. Second, we required animals to forage for food, a task that necessitates a bit more attention; third, we asked the animals to discriminate between two environments; and finally, we demanded that the animals actually learn a spatial task. We engineered things so that as the mouse walked around in its enclosure, lights and sounds, which the mouse hates, would periodically come on. The only way the mouse could turn them off was to find a small, unmarked goal region and sit there for a moment. Mice learn this task very well.

We found that even ambient attention is sufficient to allow a spatial map to form and become stable for a few hours, but such a map becomes unstable after three to six hours. Long-term stability correlates strongly and systematically with the degree to which an animal is required to pay specific attention to its environment. Thus, when a mouse is forced to pay a lot of attention to a new environment, by having to learn a spatial task at the same time that it is exploring the new space, the spatial map remains stable for days and the animal readily remembers a task based on knowledge of that environment.

What is this attentional mechanism in the brain? How does it contribute to the strong encoding of information about space and the ready recall of that information after long periods of time? I already knew that attention was not simply a mysterious force in the brain but a modulatory process. Michael Goldberg and Robert Wurtz at NIH had found that in the visual system, attention enhances the response of neurons to stimuli. A modulatory pathway that had been strongly implicated in attention-related phenomena was the one mediated by dopamine. The cells that make dopamine are clustered in the midbrain, and their axons project to the hippocampus. Indeed, we found that blocking the action of dopamine in the hippocampus blocked the stabilization of the spatial map in an animal that was paying attention. Conversely, activating dopamine receptors in the hippocampus stabilized the spatial map of an animal that was
not
paying attention. The axons of the dopamine-producing neurons in the midbrain send signals to a number of sites, including the hippocampus and the prefrontal cortex. The prefrontal cortex, which is recruited for voluntary action, signals back to the midbrain, adjusting the firing of these neurons. Our finding that some of the same regions of the brain that are recruited for voluntary behaviors are also recruited for attentional processes reinforced the idea that selective attention is critical to the unitary nature of consciousness.

In
The Principles of Psychology
William James pointed out that there is more than one form of attention. There are at least two types: involuntary and voluntary. Involuntary attention is supported by automatic neural processes, and it is particularly evident in implicit memory. In classical conditioning, for example, animals learn to associate two stimuli if, and only if, the conditioned stimulus is salient or surprising. Involuntary attention is activated by a property of the external world—of the stimulus—and it is captured, according to James, by “big things, bright things, moving things, or blood.” Voluntary attention, on the other hand, such as paying attention to the road and traffic while driving, is a specific feature of explicit memory and arises from the internal need to process stimuli that are not automatically salient.