Anthony DeCostanzo Home

Deconstructing The Mind

Adapted from Original Publication in Columbia Science Review, Volume 2, Issue 2. Fall 2005

“I remember coming to the realization at some point that every thought, and every perception, were physical changes in my brain.” Kenneth Miller is explaining to me how he came to study the interface between mind and universe. His Einsteinian hair seems perpetually windblown, and his facial expressions rapidly shift between deep concern and amusement, while remaining an amalgam of simultaneous surprise and curiosity. “I see the mind as the last frontier scientifically. We seem to have a general understanding of matter, the creation of the universe, evolution of life…[but] we can’t really say much about what the mind actually is.” Ken studies how the external world enters our mind and becomes a physical reality in our brain that we can store for perhaps our entire lives. This miraculous process is divided into sensory transduction, wherein energy from our environment is converted into electrical impulses in the brain, and memory, wherein these impulses are stored. These two processes interact to produce a rich and seamless yet subjective representation of the external world within our minds. We often take this internal facsimile of reality for granted, unconcerned with the cellular and molecular events underlying it. But a group of researchers at Columbia, Ken Miller included, are immersed in the physical construction and deconstruction of the human mind.

Our memories define us. We can often recall experiences or feelings from infancy, before we could even speak. When we think of who we are, we search our memories, and we identify with what we find. If there is any doubt about how strongly our memories define us, we can simply try to imagine that suddenly they were all taken away from us….

What, then, are memories? We hear about memories that are short-term or long-term, or perhaps declarative or procedural. We may even be familiar with some of the brain structures that process memories, such as the hippocampus and neocortex. But how, on the level of the cell or molecule, do our experiences of external physical phenomena become psychological phenomena that we may store for our entire lives?

Our memories are the stored constructions of experiences that enter us through the five senses. In order to observe the world around us, we transduce energy from the physical world into electrical impulses in our nervous system. Light energy enters our eyes and strikes photoreceptors in our rods and cones, which then transduce it from photons into an electrical impulse that travels through neurons and into our brain (the visual cortex in the occipital lobe). So the physical world never ceases to be physical while it enters us and we perceive it. From this point of view, we are merely energy transducers, sampling the energy around us (e.g. light) and converting it into another form (nerve impulses in the visual cortex) that enters our brain and ultimately encodes our memories.

Perhaps no other sense impresses more upon our memories than vision—not surprising considering that it involves as much as 50% of the human brain. Vision is processed in a region of the occipital lobe (the back of the brain) known as the visual cortex. Ken Miller came to Columbia last year to establish the Center for Theoretical Neuroscience with Larry Abbott, which so far includes six laboratories. Ken’s work mostly concerns a part of the visual cortex called V1, the junction in which nerve signals originating in the eye are translated into meaningful representations in the brain. V1, also called the primary visual cortex, is noted for containing neurons that respond to edges of a specific orientation. There is a set of neurons in V1, for example, that only activate strongly when the eye is shown a vertical line. If you show the eye a horizontal line, those neurons will not activate nearly as strongly. Processing in V1 could be thought of as analogous to an artist making a drawing. The artist consciously breaks down a subject into a series of lines or edges with different orientations. Each neuron in V1 has a preferred orientation, such that it is most responsive when the retina of the eye sees an edge that is either vertical, horizontal, or anything in between. This is called “orientation selectivity,” a property of most cells in V1. This phenomenon has been extensively documented since its discovery by Nobel laureates David Hubel and Torsten Wiesel in 1959, with recordings in the cat visual cortex (3). Orientation selectivity combines with other selectivities (such as color selectivity and motion selectivity) in different regions of the visual cortex to form our brain’s representation of images.

“I’m more of a cortex person,” Ken says, to contrast his work with that of other theoretical  neuroscientists who also study visual processing. “I see V1 as a piece of primary sensory cortex that is convenient to work with since it has a clearly defined set of inputs.” The inputs Ken is referring to come from the thalamus, a central relay system for signals from all five senses. The part of the thalamus that relays visual information from the retina is called the Lateral Geniculate Nucleus (LGN). LGN activity during development is important for constructing the mature properties of V1. Ken has recently created a model that accurately predicts the patterns of LGN activity required to achieve the remarkable property of orientation selectivity in V1 (5).  The work of Ken and other researchers exemplifies how a network of cells can develop emergent properties such as selectivity for elementary feature characteristics such as orientation, color, motion and depth. These emergent properties are the first step in our subjective reconstruction of the physical world.

Aniruddha Das is determinedly rifling through a colossal filing cabinet that contains a fair sampling of all human knowledge about vision, discretely organized into thousands of folders. Aniruddha also studies the interface between mind and universe (1). “Can you free-fuse?” he asks, as he darts back to the table while staring at the article he has just found. After realizing what free fusing is, I find myself staring at an imaginary point in distant space trying to get my brain to see two separate images as one. Aniruddha speaks quickly and precisely: “Vision was regarded as bottom–up, historically, while other forms of sensory experience seemed very interactive. Now it’s recognized that low-level visual processing is influenced by our interpretation of surfaces.”

Since completing a PhD in physics in the laboratory of Nobel laureate Charles Townes at the University of California, Berkeley, Aniruddha has turned his attention to studying vision, first as a postdoctoral fellow in the laboratory of Charles Gilbert at Rockefeller, and now as an Assistant Professor at Columbia University’s Center for Neurobiology and Behavior. He is standing in front of me explaining the kinds of questions he seeks to answer. He has handed me a page that contains two stereograms organized as three separate images (Figure 1). His smile indicates that he’s sure I will understand what he’s trying to show me. For the time being, he’s wrong. I continue to strain my eyes and tilt my head, wondering what he’s getting at.

Figure 1: Image from Nakayama K, et al. (6).
What Aniruddha’s not telling me is that there has recently been a revolution in our understanding of visual processing. For some time it was thought that orientation, color and motion are sensed as discrete features of the world that are subsequently assembled into a whole. For instance, when we are shown a drawing of a green square, different sets of neurons in our visual cortex that respond to horizontal lines, vertical lines and the color green all activate in a specific pattern that causes our brain to “see” a green square. However, in recent years it has become clear that this is not entirely accurate. Even intuitively, we’ve all experienced some form of optical illusion that makes us “see” things that are not really there. This notion seems to extend to the very basics of visual processing: as much as a series of lines and colors can form a surface to our brain, the surface we subjectively see can influence our perception of lines and colors.

After a few minutes of straining my eyes and moving the page forward and back, I finally see what he’s been explaining. Fusing the image on the left with the one in the middle yields a three-dimensional stereo image of a black screen with four holes on top of a green square. If the middle image is fused with the one on the right, however, it appears that the green square is transparent and hovering above the black screen. This simple optical illusion, Aniruddha explains, is actually both optical and cortical. The optical illusion is in seeing the four green corners as closer to the viewer; this causes a cortical illusion wherein we “see” the green spill over into the black. Our brain falsely adds green to account for its own expectations (6). This demonstrates that our brain interactively constructs an image rather than simply building it from basic information such as lines and color; our brain adds these basic elements to our “vision” where it sees fit.

All of these processes happen simultaneously, without any conscious effort on our part, to give us a continuous visual representation of the external world. Our brain acquires information about lines, colors and depth and arranges this information into surfaces and forms in higher cortical areas, and still further into objects and scenes in yet higher areas. Once arranged, the higher areas that represent, for example, surfaces can then feed back into the lower areas to influence the perception of lines and colors, which then feed again into the higher areas in a rapid cycle of reconstruction, to ultimately give us a seemingly unadulterated image. But our perception is in fact interactive at the most basic levels of processing. What our brain sees is not necessarily what our eyes see. These illusions seem to be a normal part of the cortical representation of vision, implying that some of our most cherished possessions are something of a compromise between objective reality and practical biological constraints.

Although these curiously adjusted facsimiles are somewhat suspect, the fact that our sensory experience is tweaked by our brains is not going to stop us from treasuring our memories. This leads us to ask how sensory information gets stored. Indeed, our sensory experiences would simply enter and exit our brain without a trace if it weren’t for our memory. This is the juncture wherein we become ourselves, and cease to be merely sensory processing machines. This is the principle that has motivated generations of neuroscientists to attempt to resolve the elemental question: How do we form memories?

In 1953 William Scoville, a surgeon at Hartford Hospital, removed the entire medial temporal lobe of one of his patients (7). The patient, referred to by his initials HM, has himself become a chapter in the history of neuroscience. The surgery was a treatment for an epilepsy from which HM had suffered for more than ten years and which had continued to grow more acute. After this radical surgery, HM suffered profound anterograde amnesia; that is, although his short-term memory remained intact, he could no longer consolidate information into his long-term memory. This fact marked the beginning of a revolution in neuroscience wherein the hippocampus, which resides in the medial temporal lobe, came to be understood as the site for memory formation.

This leads us to the work of Steven Siegelbaum in Columbia’s Pharmacology Department and the Center for Neurobiology and Behavior, whose laboratory is among the few sponsored by the Howard Hughes Medical Institute. He places his research in context: “The mind is a complex circuit phenomenon, and by understanding circuits we can begin to gain insight into this higher order phenomenon.” Steve studies hippocampal synaptic plasticity, the cellular and molecular circuit representation of learning and memory. When one looks at someone’s face, for example, information throughout the visual cortex is processed to give an image of the face; then this information is passed along through other cortical regions into the hippocampus. Then, over a period of minutes to days or weeks, the hippocampus facilitates the storage of information about the face through an interactive process of communication with those areas of the visual cortex that were originally activated. This process often takes place without any conscious effort—we simply remember the face of someone we’ve just met. Without the hippocampus, information enters the visual cortex and is then passed along to the prefrontal cortex for short-term storage for up to a few minutes, but it is then lost. Indeed, although HM’s recollections of early childhood remained perfectly intact after the surgery, he never recognized the faces of his own doctors despite being visited by them several times a day. He had lost the ability to construct new long-term memories.

Steve Siegelbaum’s work focuses on the cellular and molecular events in the hippocampus that allow it to perform the task of memory construction (9). When we think of a memory of someone’s face, we are experiencing an emergent property of a network of cells. This memory does not reside in any single cell, so how can we conceive of the cellular component of memory? In the most abstract sense, a memory can be thought of as any trace; something that once formed will, by definition, remain for some period of time. So by this definition, even a letter drawn in the sand qualifies as a memory; the letter once drawn has left a trace—the sand “remembers” the letter. By extension, that same letter carved into a granite slab is also a memory, and we can accordingly contrast short and long–term memories: while our sand-drawn letter will soon disappear as a result of the elements, our granite-carved letter will likely survive for our entire lives. This most sparse definition is intended to emphasize the physicality of memory. This principle is no different when it comes to the human brain. Through modulation of neuronal activity, physical changes occur within our brain. These physical changes are our memories.

The brain is said to be plastic in its ability to change itself to store information. The physical changes associated with memory formation happen at synapses between neurons, and therefore fall under the rubric long term synaptic plasticity. This type of plasticity includes both long-term potentiation (LTP) and its physiological opposite, long-term depression (LTD), which are respectively a long-term strengthening and a long-term weakening of synaptic strength. Since the discovery of LTP in 1966 by Terje Lomo, this field of cellular neuroscience has rapidly expanded and was the focus of the research  hat recently garnered Eric Kandel, another professor in the Columbia Center for Neurobiology and Behavior, the 2000 Nobel Prize in Physiology or Medicine. It is now thought that these changes in synaptic strength (which occur at a cellular level) underlie the emergent, network-level formation of

It is impossible to discuss the activities of neurons without describing the action potential, the  electrophysiological basis of neuronal signaling. An action potential is an impulse that propagates very rapidly down the length of an axon. It is sometimes referred to, as in the above discussion of vision, as neuronal activation, or sometimes nerve conduction, nerve signaling, or neuronal firing. Steve Siegelbaum tells of his early inspirations in neuroscience: “In high school I read a description of the action potential written by Isaac Asimov, which described the discoveries of Hodgkin and Huxley. I was fascinated by the fact that we could understand nerve signals in the brain as well-behaved physical phenomena.” Indeed, the Hodgkin and Huxley equations, formulated from experiments done in the 1930s and ’40s, are famous for the prescient exactness with which they model the action potential (2) For their work, Sir Alan Hodgkin and Sir Andrew Huxley received the Nobel Prize in 1963.

Steve now focuses much of his energy on elucidating LTP and LTD in the hippocampus, the site of memory construction (9). These changes in synaptic strength take place according to defined rules that are based on the firing rates of the neurons involved. Slow firing rates tend to result in LTD, and the synapse is therefore “weakened,” while fast firing rates tend to cause LTP, resulting in a “strengthened” synapse. Since the discovery of LTP and LTD, it has become increasingly clear that there are in fact many ways to achieve the overall weakening or strengthening of a synapse. Each mechanism involves the activity of specific ion channels, receptors, enzymes, and other molecules. Furthermore, there is a genetic basis for the activity of each of these molecules. “I am now more concerned with assessing the role of ion channels and other molecules in plasticity using transgenic mice,” Steve says, referring to his lab’s recent use of transgenic and knockout mouse models, wherein certain genes are either artificially added to or deleted from the mouse genome. These genetically-altered mice differ from “wild-type” mice in their ability to form memories and, accordingly, have differences in their hippocampal synaptic plasticity. These are the kinds of experiments currently being done in Steve’s lab and other labs in the Center for Neurobiology and Behavior to study the role of specific genes in memory construction.9

So, we have transduction of light energy into neuronal firing (action potentials) in the visual cortex, into coordinated firing in the hippocampus, into alterations of gene expression and…voila… a memory for life. Indeed when we are at a loss for the material comforts of this world, we often comfort ourselves with the notion that our experiences and memories are possessions that cannot be taken away from us. This, paradoxically, is grotesquely and brutally untrue.

When asked how he came to study Alzheimer’s disease, Dr. Scott Small is likely to respond, “I got tired of lying to people.” The people he is referring to are his patients. Scott Small is a neurologist at Columbia University Medical Center’s Taub Institute for Research on Alzheimer’s Disease and the Aging Brain. After completing a fellowship in neurobehavior at Columbia in 1998, Scott joined the faculty of Columbia as Herbert Irving Assistant Professor of Neurology. Before subspecializing in Alzheimer’s Disease, Scott intended to study the normal physiological function of the hippocampus using a brain-scanning  technology called functional magnetic resonance imaging (fMRI). fMRI yields information about oxygen consumption in the brain, which in turn is correlated with neuronal activity. He is describing to me his reaction to treating Alzheimer’s patients. “As a neurologist, most of your patients have Alzheimer’s. Invariably you find yourself struggling to find the words to give them the bad news, and it is almost always bad news.” Scott was faced with a fundamental flaw in the human condition. His patients who may initially have only mild forgetfulness soon suffer a loss of that which they hold most dear. Along with not being able to perform routine tasks, the individual is stripped of the ability to recognize the faces of people who have accompanied him for his entire life. A lifetime of mental construction is deconstructed as the individual is plunged into confusion and disorientation.

“You find yourself wanting to give their family a glimmer of hope where there is none. I had to involve myself somehow in addressing this.” Using his skills in fMRI, Scott has been studying the brain dysfunction that causes Alzheimer’s. The early stages of the disease seem to affect the hippocampus and associated structures of the brain. Involvement of the hippocampus had been suspected for some time, since the early symptoms are essentially anterograde amnesia, similar to that experienced by HM.  Although the hippocampus and entorhinal cortex (the input to the hippocampus) are now known to be affected, Scott’s goal is to determine with high resolution fMRI the specific regions of the hippocampus and associated structures that are impaired in Alzheimer’s disease. As he puts it, “The hippocampus is a circuit, so the best way to evaluate hippocampal function is to evaluate each node in the circuit.” The nodes include the entorhinal cortex and each of the subregions of the hippocampus—Dentate Gyrus, CA1 and CA3. By comparing the brains of Alzheimer’s patients to those of healthy elderly patients, Scott has found that Alzheimer’s patients seem to have altered patterns of activity in various subregions of the hippocampus (8). Scott believes that “the spatial pattern of dysfunction among hippocampal subregions will be the unique signature of Alzheimer’s dysfunction.”

“Alzheimer’s disease appears to be a synaptic failure,” Scott says, referring to a mounting body of literature that shows a profound effect on long-term synaptic plasticity in animal models of Alzheimer’s (10). The most probable cause of impaired synaptic plasticity appears to be the accumulation of a small peptide, amyloid-β. This peptide has been shown to impair LTP when it is artificially administered to the hippocampus. Furthermore, transgenic mice that overexpress the amyloid-β precursor have impaired LTP as well as memory impairments. The normal, nonpathological function of amyloid-β is not known, but its role in diminishing LTP suggests that it may modulate normal longterm synaptic plasticity. Whatever its role in normal physiology, pathologically high levels of amyloid-β seem to effectively deconstruct what normal hippocampal physiology is designed to construct: memories.

At every level of organization, the human mind is constructive, often representing our world with a paradoxically subjective, meticulous façade. Our senses transduce energy from our surroundings into nerve signals, which are assembled into increasingly more complex and “meaningful” representations in our cortex. As exemplified by the simple illusion presented earlier, these representations are interactively constructed in the cortex between lower areas that process basic features (such as lines or colors) and higher areas that assemble these basic features and modify them to best fit expectations. These representations remain temporary until they are processed in the hippocampus, wherein they are constructed into long-term memories by interactive firing between the hippocampus and the same high-level cortical areas in which they were processed earlier. Once so constructed, these memories are maintained in the areas of the cortex that originally participated in processing; images are stored the visual cortex and sounds in the auditory cortex. However, as with anything that can be constructed, memories can be deconstructed, through mechanisms that appear to abuse the cellular machinery for construction. This begs for an immense appreciation of the delicate balance between our mind’s constructive and deconstructive tendencies.

1. Das, A., and C.D. Gilbert. “Topography of Contextual Modulations Mediated by Short-Range Interactions in Primary Visual Cortex.” Nature 399 (1999): 655-661.
2. Hodgkin, A.L., and A.F. Huxley. “A Quantitative Description of Membrane Current and its Application to Conduction and Excitation in Nerve.” Journal of Physiology (London) 117 (1952): 500-544.
3. Hubel, D.H., and T.N. Wiesel. “Receptive Fields of Single Neurons in the Cat’s Striate Cortex.” Journal of Physiology 148 (1959): 574-591.
4. Kandel, E.R., J.H. Schwartz, and T.M. Jessell. Principles of Neural Science. 4th ed. New York: McGraw–Hill, 2000.
5. Kayser, A.S., and K.D. Miller. “Opponent Inhibition: A Developmental Model of Layer 4 of the Neocortical Circuit.” Neuron 33 (2002): 131-142.
6. Nakayama, K., S. Shimojo, and V.S. Ramachandran. “Transparency: Relation to Depth, Subjective Contours, Luminance, and Neon Color Spreading.” Perception 19 (1990): 497-513.
7. Scoville, W.B., and B. Milner. “Loss of Recent Memory After Bilateral Hippocampal Lesions.” The Journal of Neurology, Neurosurgery and Psychiatry 20 (1957): 11–21.
8. Small, S.A. “Measuring Correlates of Brain Metabolism with High- Resolution MRI: A Promising Approach for Diagnosing Alzheimer Disease and Mapping its Course.” Alzheimer Disease and Associated Disorders 17 (2003): 154-161.
9. Unni V.K., S.S. Zakharenko, L. Zablow, A.J. DeCostanzo, and S.A. Siegelbaum. “Calcium Release from Presynaptic Ryanodine-Sensitive Storesis Required for Long-Term Depression at Hippocampal CA3–CA3 Pyramidal Neuron Synapses.” Journal of Neuroscience 24 (2004): 9612-9622.
10. Walsh D.M., and D.J. Selkoe. “Deciphering the Molecular Basis of Memory Failure in Alzheimer’s Disease.” Neuron 44 (2004): 181-193.