In Search of Memory: The Emergence of a New Science of Mind (36 page)

Repeated trains of electrical stimuli produce a late phase of long-term potentiation that lasts for more than a day. We found the properties of this phase, which previously had not been extensively explored, to be very similar to long-term facilitation of synaptic strength in
Aplysia
. In both
Aplysia
and mice, the late phase of long-term potentiation is strongly affected by modulatory interneurons, which in mice are recruited to switch a short-term, homosynaptic into long-term, heterosynaptic change. In mice those neurons release dopamine, a neurotransmitter commonly recruited in the mammalian brain for attention and reinforcement. Like serotonin in
Aplysia
, dopamine prompts a receptor in the hippocampus to activate an enzyme that increases the amount of cyclic AMP. However, an important part of the increase in cyclic AMP in the mouse hippocampus occurs in the postsynaptic cell, whereas in
Aplysia
the increase occurs in the presynaptic sensory neuron. In each case, the cyclic AMP recruits protein kinase A and other protein kinases, which leads to the activation of CREB and the turning on of effector genes.

One of the striking things we had found in studying memory in
Aplysia
was the existence of the memory-suppressor gene that produces the CREB-2 protein. Blocking the expression of that gene in
Aplysia
enhances both the strengthening and the increase in number of synapses associated with long-term facilitation. In the mouse, we found that blocking this and similar memory-suppressor genes enhances both long-term potentiation in the hippocampus and spatial memory.

In the course of these studies, I found myself once again in an enjoyable collaboration with Steven Siegelbaum. We were interested in a particular ion channel that inhibits synaptic strengthening, especially in certain dendrites. Alden Spencer and I had studied those dendrites in 1959 and inferred that they produce action potentials in response to activity in the perforant pathway, which leads from the entorrhinal cortex to the hippocampus. Steve and I bred mice in which the gene for this particular ion channel was lacking. We found that long-term potentiation in response to stimulation of the perforant pathway was greatly enhanced in those mice, in part by dendritic action potentials. As a result, these mice were brilliant; they had a much stronger spatial memory than normal mice!

My colleagues and I also discovered that explicit memory in the mammalian brain, unlike implicit memory in
Aplysia
or
Drosophila
, requires several gene regulators in addition to CREB. Although the evidence is less complete, it appears that in mice, expression of genes also gives rise to anatomical changes—specifically, to the growth of new synaptic connections.

Despite the significant behavioral differences between implicit and explicit memory, some aspects of implicit memory storage in invertebrates have been conserved over millions of years of evolutionary time in the mechanisms by which explicit memory is stored in vertebrates. Although the great neurophysiologist John Eccles had urged me early in my career not to abandon research on the splendid mammalian brain for work on a slimy, brainless sea snail, it is now clear that several key molecular mechanisms of memory are shared by all animals.

T
he study of the explicit memory for space in the mouse drew me ineluctably to the larger questions that had attracted me to psychoanalysis at the beginning of my career. I was starting to think about the nature of attention and consciousness, mental states not associated with simple reflex actions but with complex psychological processes. I wanted to focus on how space—the internal environment in which the mouse navigates—is represented in the brain and how this representation is modified by attention. In doing so, I was moving from a system in
Aplysia
that was reasonably well understood to systems in the mammalian brain that had yielded (and to some degree still yield) only a few fascinating results and many unresolved questions. Nevertheless, the time had come to try to help move the molecular biology of cognition a step forward.

In examining implicit memory in
Aplysia
, I had built a neurobiological and molecular approach to elementary mental processes on a foundation laid by Pavlov and the behaviorists. Their methods were rigorous, but they reflected a narrow and limited definition of behavior, one that focused on motor acts. In contrast, our research on the explicit memory and the hippocampus posed enormous new intellectual challenges, in no small part because the encoding and recall of spatial memory requires conscious attention.

As a first step in thinking about complex memory of space and its internal representation in the hippocampus, I turned from the behaviorists to the cognitive psychologists, the scientific successors to the psychoanalysts and the first group of scientists to think systematically about how the outside world is re-created and represented in our brain.

COGNITIVE PSYCHOLOGY EMERGED IN THE EARLY 1960S IN
response to the limitations of behaviorism. While attempting to retain the experimental rigor of behaviorism, cognitive psychologists focused on mental processes that were more complex and closer to the domain of psychoanalysis. Much like the psychoanalysts before them, the new cognitive psychologists were not satisfied with simply describing motor responses elicited by sensory stimuli. Rather, they were interested in investigating the mechanisms in the brain that intervene between a stimulus and a response—the mechanisms that convert a sensory stimulus into an action. The cognitive psychologists set up behavioral experiments that allowed them to infer how sensory information from the eyes and the ears is transformed in the brain into images, words, or actions.

The thinking of cognitive psychologists was driven by two underlying assumptions. The first was the Kantian notion that the brain is born with
a priori
knowledge, “knowledge that is…independent of experience.” That idea was later advanced by the European school of Gestalt psychologists, the forerunners, together with psychoanalysis, of modern cognitive psychology. The Gestalt psychologists argued that our coherent perceptions are the end result of the brain’s built-in ability to derive meaning from the properties of the world, only limited features of which can be detected by the peripheral sensory organs. The reason that the brain can derive meaning from, say, a limited analysis of a visual scene is that the visual system does not simply record a scene passively, as a camera does. Rather, perception is creative: the visual system transforms the two-dimensional patterns of light on the retina of the eye into a logically coherent and stable interpretation of a three-dimensional sensory world. Built into neural pathways of the brain are complex rules of guessing; those rules allow the brain to extract information from relatively impoverished patterns of incoming neural signals and turn it into a meaningful image. The brain is thus the ambiguity-resolving machine par excellence!

Cognitive psychologists demonstrated this ability with studies of illusions, that is, misreadings of visual information by the brain. For example, an image that does not contain the complete outline of a triangle is nevertheless seen as a triangle because the brain expects to form certain images (figure 22–1). The brain’s expectations are built into the anatomical and functional organization of the visual pathways; they are derived in part from experience but in large part from the innate neural wiring for vision.

To appreciate these evolved perceptual skills, it is useful to compare the computational abilities of the brain with those of artificial computational or information-processing devices. When you sit at a sidewalk café and watch people go by, you can, with minimal clues, readily distinguish men from women, friends from strangers. Perceiving and recognizing objects and people seem effortless. However, computer scientists have learned from constructing intelligent machines that these perceptual discriminations require computations that no computer can begin to approach. Merely recognizing a person is an amazing computational achievement. All of our perceptions—seeing, hearing, smelling, and touching—are analytical triumphs.

22–1 The brain’s reconstruction of sensory information.
The brain resolves ambiguities by creating shapes from incomplete data—for example, filling in the missing lines of these triangles. If you hide parts of these pictures, your brain is deprived of some clues it uses to form conclusions and the triangles vanish.

The second assumption developed by cognitive psychologists was that the brain achieves these analytic triumphs by developing an internal representation of the external world—a cognitive map—and then using it to generate a meaningful image of what is out there to see and to hear. The cognitive map is then combined with information about past events and is modulated by attention. Finally, the sensory representations are used to organize and orchestrate purposeful action.

The idea of a cognitive map proved an important advance in the study of behavior and brought cognitive psychology and psychoanalysis closer together. It also provided a view of mind that was much broader and more interesting than that of the behaviorists. But the concept was not without problems. The biggest problem was the fact that the internal representations inferred by cognitive psychologists were only sophisticated guesses; they could not be examined directly and thus were not readily accessible to objective analysis. To see the internal representations—to peer into the black box of mind—cognitive psychologists had to join forces with biologists.

FORTUNATELY, AT THE SAME TIME THAT COGNITIVE PSYCHOLOGY
was emerging in the 1960s, the biology of higher brain function was maturing. During the 1970s and 1980s, behaviorists and cognitive psychologists began collaborating with brain scientists. As a result, neural science, the biological science concerned with brain processes, began to merge with behaviorist and cognitive psychology, the sciences concerned with mental processes. The synthesis that emerged from these interactions gave rise to the field of cognitive neural science, which focused on the biology of internal representations and drew heavily on two lines of inquiry: the electrophysiological study of how sensory information is represented in the brains of animals, and the imaging of sensory and other complex internal representations in the brains of intact, behaving human beings.

Both of these approaches were used to examine the internal representation of space, which I wanted to study, and they revealed that space is indeed the most complex of sensory representations. To make any sense of it, I needed first to take stock of what had already been learned from the study of simpler representations. Fortunately for me, the major contributors to this field were Wade Marshall, Vernon Mountcastle, David Hubel, and Torsten Wiesel, four people I knew very well, and whose work I was intimately familiar with.

THE ELECTROPHYSIOLOGICAL STUDY OF SENSORY REPRESENTATION
was initiated by my mentor, Wade Marshall, the first person to study how touch, vision, and hearing were represented in the cerebral cortex. Marshall began by studying the representation of touch. In 1936 he discovered that the somatosensory cortex of the cat contains a map of the body surface. He then collaborated with Philip Bard and Clinton Woolsey to map in great detail the representation of the entire body surface in the brain of monkeys. A few years later Wilder Penfield mapped the human somatosensory cortex.

These physiological studies revealed two principles regarding sensory maps. First, in both people and monkeys, each part of the body is represented in a systematic way in the cortex. Second, sensory maps are not simply a direct replica in the brain of the topography of the body surface. Rather, they are a dramatic distortion of the body form. Each part of the body is represented in proportion to its importance in sensory perception, not to its size. Thus the fingertips and the mouth, which are extremely sensitive regions for touch perception, have a disproportionately larger representation than does the skin of the back, which although more extensive is less sensitive to touch. This distortion reflects the density of sensory innervation in different areas of the body. Woolsey later found similar distortions in other experimental animals; in rabbits, for example, the face and nose have the largest representation in the brain because they are the primary means through which the animal explores its environment. As we have seen, these maps can be modified by experience.

In the early 1950s Vernon Mountcastle at Johns Hopkins extended the analysis of sensory representation by recording from single cells. Mountcastle found that individual neurons in the somatosensory cortex respond to signals from only a limited area of the skin, an area he called the receptive field of the neuron. For example, a cell in the hand region of the somatosensory cortex of the left hemisphere might respond only to stimulation of the tip of the middle finger of the right hand and to nothing else.

Mountcastle also discovered that tactile sensation is made up of several distinct submodalities; for example, touch includes the sensation produced by hard pressure on the skin as well as that produced by a light brush against it. He found that each distinct submodality has its own private pathway within the brain and that this segregation is maintained at each relay in the brain stem and in the thalamus. The most fascinating example of this segregation is evident in the somatosensory cortex, which is organized into columns of nerve cells extending from its upper to its lower surface. Each column is dedicated to one submodality and one area of skin. Thus all the cells in one column might receive information on superficial touch from the end of the index finger. Cells in another column might receive information on deep pressure from the index finger. Mountcastle’s work revealed the extent to which the sensory message about touch is deconstructed; each submodality is analyzed separately and reconstructed and combined only in later stages of information processing. Mountcastle also proposed the now generally accepted idea that these columns form the basic information-processing modules of the cortex.

OTHER SENSORY MODALITIES ARE ORGANIZED SIMILARLY. THE
analysis of perception is more advanced in vision than in any other sense. Here we see that visual information, relayed from one point to another along the pathway from the retina to the cerebral cortex, is also transformed in precise ways, first being deconstructed and then reconstructed—all without our being in any way aware of it.

In the early 1950s, Stephen Kuffler recorded from single cells in the retina and made the surprising discovery that those cells do not signal absolute levels of light; rather, they signal the contrast between light and dark. He found that the most effective stimulus for exciting retinal cells is not diffuse light but small spots of light. David Hubel and Torsten Wiesel found a similar principle operating in the next relay stage, located in the thalamus. However, they made the astonishing discovery that once the signal reaches the cortex, it is transformed. Most cells in the cortex do not respond vigorously to small spots of light. Instead, they respond to linear contours, to elongated edges between lighter and darker areas, such as those that delineate objects in our environment.

Most amazingly, each cell in the primary visual cortex responds only to a specific orientation of such light-dark contours. Thus if a square block is rotated slowly before our eyes, slowly changing the angle of each edge, different cells will fire in response to these different angles. Some cells respond best when the linear edge is oriented vertically, others when the edge is horizontal, and still other cells when the axis is at an oblique angle. Deconstructing visual objects into line segments of different orientation appears to be the initial step in encoding the forms of objects in our environment. Hubel and Wiesel next found that in the visual system, as in the somatosensory system, cells with similar properties (in this case, cells with similar axes of orientation) are grouped together in columns.