Rewriting the textbook on memory
How one woman’s amnesia helped shed light on the way the brain orchestrates experience and memory. Michael D. Lemonick investigates.
On a cloudless, unseasonably warm morning in October, Lonni Sue Johnson arrived at the Princeton Neuroscience Institute in New Jersey, ready for a long day of testing. Johnson, a cheerful, slightly plump, attractive woman in her mid-sixties, was curious about what was going to happen. Despite the fact that she had been Nicholas Turk-Browne’s experimental subject for several years, she couldn’t tell you the name of the scientist who would spend the morning putting her through scans in an fMRI machine.
Lonni Sue Johnson is what neuroscientists call “densely amnesic”. She can’t remember much about the past, and can’t form new memories. She developed this condition in 2007, the result of a brain infection with herpes simplex virus 1. The virus, which normally causes nothing worse than cold sores, had burned its way through her hippocampi, twin seahorse-shaped structures buried deep in her brain, one in each hemisphere, leaving nothing but a fluid-filled void behind. Neuroscientists know that the hippocampus is essential to the formation and recall of what’s termed declarative memory—the kinds of memories you can describe in words, including facts, general knowledge and autobiographical recollections about specific events in our lives. They know this thanks to a man named Henry Molaison, whose hippocampi were surgically removed in 1953, inadvertently launching the modern understanding of how memory works.
But in recent years, neuroscientists have begun to suspect that the hippocampus and surrounding tissues in the brain’s medial temporal lobes do far more than simply form and retrieve memories—crucial as that is. The hippocampus evidently serves as a mental map that allows us to navigate, not just our memories, but also our expectations about the future, our physical location in time and space, and more. And Lonni Sue Johnson is perhaps the most important subject yet in the ongoing quest to figure out exactly how the hippocampus orchestrates the rest of the brain to accomplish these feats.
Before 1953, nobody really had much of an idea of what the hippocampus was good for. The best guess was that it was involved with our sense of smell. All of that changed when a young man named Henry Molaison went under the knife in a hospital in Hartford, Connecticut, during an experimental surgery designed to cure his uncontrollable epilepsy. Molaison’s seizures were so frequent and so violent—he had one or more every day—that he’d had to drop out of high school for several years. When he finally graduated, in his early twenties, he was too disabled to take any job more sophisticated than winding copper coils on electric motors at a manufacturing plant.
Finally, in desperation, he and his parents agreed to a radical operation: William Beecher Scoville, a prominent local neurosurgeon, would drill into Henry’s skull, insert a thin metal tube, and suction out the soft, spongy tissue of his hippocampi, where his seizures seemed to originate. Because a sense of smell wasn’t essential, Scoville believed, removing them wouldn't violate his Hippocratic Oath to do no harm.
Molaison’s surgery cured his epilepsy but it soon became clear that something else was terribly wrong. A nurse would come into his room to take vital signs, and although she had been there several times already, and introduced herself every time, the patient had no idea he’d ever seen her before. Scoville himself dropped in frequently to check on Molaison’s recovery, and he, too, was greeted as a complete stranger every single time. Molaison did recognize his parents, but nothing else appeared to stick. It was clear he could no longer form new memories.
When it came to old memories, he could recall general things—that his father had been born in Louisiana, for example and that his parents would take him on family vacations driving the Mohawk Trail, a scenic highway in neighboring Massachusetts. But if you asked him to talk about a specific episode, say something that happened on one of those drives, he didn’t have a clue.
With that single, devastating medical experiment, scientists finally knew what the hippocampus did. Its job seemed to be to knit together our experiences—the sights, sounds, smells, emotions, and every other aspect of a given moment, each processed in a different part of the brain, into a single remembered incident that could be called up in the future and re-experienced, or described. That sort of recollection would come to be called “episodic memory”—the subset of declarative memory that relates to specific experiences we ourselves have had. The other subset is “semantic memory,” which is also declarative, but which describes general facts about the world or about ourselves—that we used to take vacations on the Mohawk trail, for example, or that Adolf Hitler ruled Germany during World War II. Episodic memory is thought to be a uniquely human invention, while semantic memory is what we share with other mammals. Your dog’s semantic memory will remind her to avoid the mean ginger cat around the corner. But while you remember the details of the episode where the cat attacked your dog, your dog has a general memory that tells her to avoid cats on that corner.
This very human ability to tell richly detailed stories of our own lives, to ourselves or to others, is only possible with at least one intact hippocampus. Nor is it possible to knit together new experiences to carry declarative memories of both kinds forward into the future without this organ. The inability to remember the past is known as retrograde amnesia; the impossibility of forming new memories, anterograde amnesia. Patients who have both forms of amnesia, like Henry Molaison, or like Lonni Sue Johnson, are profoundly disabled.
Over the next few decades, researchers—primarily the pioneering neuroscientists Brenda Milner, of McGill University, in Montreal and Suzanne Corkin, of MIT, in Boston, would perform dozens upon dozens of studies on Molaison, to try and understand the limits of what he could and could not do (until his death in 2008, he was known in the literature only as H.M., in order to protect his privacy). Amongst the things that remained unimpaired was his speech and his ability to read; so was his ability to walk. These largely unconscious acts obviously involved some sort of memory, since he wasn’t born knowing how to do either. But retrieving these memories (unlike episodes of his past life) was evidently not governed by the hippocampus.
Molaison was also able to form new memories ‘unconsciously’ as evidenced by the following experiment. Milner had Molaison try something he’d never done before: trace the image of a star on a piece of paper while looking, not at his hand, but at the image of his hand in a mirror. This is immensely awkward at first for anyone, but it inevitably gets easier with practice. It was true for Molaison as well: each time he tried the exercise, he was better at it, even though he had no conscious memory of having tried it before. After several sessions, he was reportedly quite surprised at how well he could trace the star, given (he thought) that he was attempting it for the very first time.
These experiments made it clear that the hippocampus isn’t absolutely necessary for acquiring new skills, and established a separate class of memory known as procedural memory, or “muscle memory,” as it’s popularly known. The classic example is riding a bike: once you learn, you do it unconsciously.
Nobody knows why the virus decides to target the brain every so often, but that’s what happened, at the end of 2007, to Lonni Sue Johnson, who was living alone on her farm in upstate New York at the time. She would almost certainly have died had her neighbor not found her in a state of confusion and rushed her to the hospital. Instead, she, like Molaison, became densely amnesic.
Unlike Molaison, however, Johnson was an immensely creative person with a rich set of talents. She was an accomplished commercial artist who drew covers for the New Yorker magazine. She was a talented amateur viola player. She was a pilot who owned and flew two small planes. She wrote a popular column for the local newspaper. She had, in short, a lifetime’s worth of diverse experiences and knowledge, so she gave neuroscientists a wealth of targets for testing.
Molaison was Patient Zero—the one who launched this entire field of research—but Johnson promised to expand its boundaries considerably.
Fortunately, she was also perfectly amenable to being tested. On that warm October morning at the Princeton Neuroscience Institute, she spent the morning in the fMRI scanner for a study of ‘adaptive learning’. That is the tendency of our brains to gradually pay less attention to familiar objects or scenes in favor of novel ones. When I’m driving, for example, I don’t pay any significant attention to the cars whizzing by on the other side of the road. But if one suddenly veers toward me, my brain goes on red alert. Without a hippocampus, it’s turning out, Johnson’s brain can still do some adaptive learning, but only in a very in a limited way.
After the fMRI session ended, she had another series of tests in another lab—this one designed to address a question of growing interest to neuroscientists. It’s been an article of faith, thanks to Milner’s and Corkin’s work on H.M., that procedural learning—like learning to ride a bike —happened somewhere other than the hippocampus.
Neuroscientist, Jordan Taylor, who was conducting the second set of experiments on Lonnie Sue, wanted to revisit that idea. The tests Taylor would be doing on Lonni Sue that afternoon involved a sort of flight simulator. She sat at a table in front of a screen that displayed an icon representing an airplane and another representing a runway. Her task was to push on a control stick to guide the plane onto the runway. There were two complications. First, on each attempt, the runway would appear in a different place—straight ahead, straight behind, or to the right or left. The second complication was that the lever was programmed to resist her movements. If she tried to move it forward, it would push to the right or to the left, and she’d have to compensate with a counterforce of her own in order to land. “It’s like a crosswind,” she said. “I wish this airport had a windsock!”
Lonni Sue did figure out how to compensate. She showed procedural learning. But she didn’t do it as well as normal controls. (It turns out that Henry Molaison’s ability to learn mirror drawing was also not quite as good as that of controls). The other salient fact about this experiment, like the ones with H.M. and also some music sight-reading experiments a different group of neuroscientists had done with Lonni Sue, in which she got better at playing an unfamiliar piece on the viola without realizing she was practicing it again and again, was that none of these truly involved a new skill. “Trace the star while looking in the mirror” is really just a mechanical task. Henry already knew how to hold a pen and follow a line; he simply had to let the link between his visual and motor systems reset itself. Lonni Sue already knew how to read music and how to translate the notes on a page into motor movements; she just had to apply those skills to a composition she hadn’t seen before.
But what if she tried to learn an entirely new musical instrument—the trumpet, say, where you produce the sound not by stroking with a bow but by buzzing your lips into a mouthpiece, tightening and loosening them to make the pitch higher and lower, and where you form the notes by pressing down three piston-like keys in different patterns? Could she do it? Taylor suspects, maybe not.
Taken together, the Princeton and Johns Hopkins experiments with Lonni Sue have helped reinforce a consensus that has been growing over the past few years: that the hippocampus is intimately involved in far more than simply creating and accessing declarative memories. It’s not surprising that neuroscientists thought otherwise at first. Lonni Sue and Henry Molaison, and many other famous amnesics lost significant chunks of their declarative memory function when their hippocampi and surrounding tissues were destroyed. But as researchers have looked closer, they’ve begun to understand that the cognitive losses are more widespread.
In many areas of science, the things you discover are the things you’re looking for. In this case, it was the researchers who were interested in memory who were paying attention to the hippocampus. But those interested in other aspects of brain function weren’t. People who studied decision making were focused on the frontal lobes. People who studied adaptation learning looked at the object selective cortex. People who were focused on motor learning looked at the cerebellum and the striatum. There was no reason to think the hippocampus was involved.
But that view has begun to change. In Lonni Sue, the loss of her hippocampus has destroyed much of her declarative memory, and her ability to form new episodic memories, but it’s also diminished her ability for unconscious learning, compromising her procedural memory and motor learning. It has also hampered her visual system’s ability to adapt to the sight of familiar objects.
So now, if you asked the question: “what does the hippocampus do?” you’d still have to answer: “mostly declarative memory”. But you’d also have to note it’s involved in varying degrees in other kinds of memory.
But saying that the function of a brain structure is to do mostly this, but also a little of that, plus a medium amount of something else, doesn’t really sound like science, and it doesn’t make a whole lot of sense from an evolutionary perspective. Memory systems arose early in mammalian evolution, and have persisted in virtually all the descendant species of those remote ancestors. That being the case, Daphna Shohamy, a neuroscientist at Columbia University, told me, “this traditional focus on declarative memories doesn’t quite fit. It almost seems to ignore the fact that animals [which apart from us, don’t have episodic declarative memory] have a hippocampus.” To Shohamy, as to many other neuroscientists, this suggests that the question needs to be reframed. “If you look at the connectivity between the hippocampus and other parts of the brain, the idea that the hippocampus is just a module for conscious awareness doesn’t seem like the most plausible explanation.”
Rather than acting like a discrete memory module, it now appears that the hippocampus is highly specialized at creating associations between objects, spaces and experiences. It helps the brain link all of these elements into networks that help us transform a chaotic jumble of sensory impressions that pour into our brains at every minute into a comprehensible whole. Neuroscientists call this linking capability “relational processing,” and while it’s crucial to our having rich memories of the past, it’s also crucial to integrating our experience into the present, and for using it to think about the future.
During my visit to the University of Illinois, Neal Cohen gave me an example. Imagine that he and I had just met, and that I’d mentioned I was writing about a patient Nick Turke-Brown was working with at Princeton. Cohen’s immediate response might have been to think of the last time he’d seen Nick, and about their conversation, and who else was there. That ability to move back and forth between related events, separated in time, is impaired in patients with hippocampal damage, Cohen told me. “That’s not a deficit someone would come into the clinic and complain about,” he said. They’d complain about not remembering what happened yesterday, or about not knowing who their spouse is, or something like that. These deficits are more subtle, but they prevent patients from relating new experiences to old ones, for predicting, based on past experience, the possible outcomes of every decision we make, every moment of the day. Relational processing, Turk-Browne told me, “is how you select what actions to perform or what to say or what to think.” It doesn’t matter whether those decisions are conscious or unconscious.
One of the first hints that relational processing might be what the hippocampus does came out of studies published by John O’Keefe, at University College London, in 1971. By measuring the neural activity of rats moving around an enclosure, O’Keefe and his colleague Jonathan Dostrovsky discovered individual cells in the hippocampus that fired as the animals moved though their environment. The neurons, which they called “place cells,” evidently formed a mental representation of space—a neural map that recorded the relationships between locations, allowing the rodents to navigate more efficiently with time.In the early 2000’s May-Britt and Edvard Moser, of the Norwegian University of Science and Technology, found another type of cells in the entorhinal cortex, a medial temporal lobe structure that feeds directly into the hippocampus, and receives information from it. Named grid cells, they keep a record of where the rodents are located within their environment at any given time. These two discoveries earned O’Keefe and the Mosers the 2014 Nobel Prize in Physiology or Medicine. Later, the Mosers and other neuroscientists went on to find cells in the medial temporal lobes that tell the rats how their heads are oriented within the environment; cells that fire off to signal the rats how far they are from the walls of their enclosures; and cells that keep track of the animals’ running speed, which they might use to gauge how far they’ve traveled from their starting points.
Working together, all of these neural cells evidently serve as a sort of natural GPS system. The hippocampus also has a set of cells that orient rats in time. Known as (what else?) time cells, they were discovered by Howard Eichenbaum, at Boston University. Since memory systems have been conserved during mammalian evolution, humans presumably have them as well.
Even before the Mosers discovered grid cells or Eichenbaum discovered time cells, John O’Keefe and his collaborator Lynn Nadel, of the University of Arizona, had begun to describe the hippocampus, not primarily as a memory structure, but as a cognitive map of the world. But to read the map, you have to be able to retrieve it from memory—“not,” Turk-Browne told me, “by consciously thinking ‘ok, I turn left here.’ Sometimes you do that, when you’re going to a new place, but most of the time, you’re just kind of efficiently moving through the world.” It’s the hippocampus that lets you do that. “You know about those classic studies of taxi drivers, right?”
He was talking about research by Eleanor Maguire, of London’s Institute of Neurology, on the brains of London cabbies. Unlike their counterparts in other major cities, cab drivers in London have to pass an exhaustive test of their geographical knowledge—known simply as “the Knowledge”—before they can get a license. Once they’re on the streets, that knowledge is further informed by the dozens of acts of navigation they perform each day, adding a visceral component to their factual understanding of how to get from here to there. When Maguire and her colleagues put London cab drivers into MRI machines and scanned their brains, they found that the cab drivers’ hippocampi were significantly enlarged compared with non-cabbies in the part of the organ where spatial information is processed.
Space and time are two things we navigate (or they’re just one thing, if, like Einstein, you think they’re joined together into “spacetime”). So are the world of objects, faces, tastes, body sensations, customs, social hierarchies, landscapes and more. What the hippocampus does, in the metaphor of Princeton’s Ken Norman, is to tie together everything we’re experiencing, moment by moment. “If you think of the pattern of brain activity during some event as just a bunch of balloons bobbing around, the hippocampus ties all the strings of those balloons together into a little knot,” Norman said. The little knot, he said, is the hippocampal code for that event. Once it’s made the knot, then all of that brain activity—sights, sounds, emotions, sensations—is linked. “So now when you tug down on that knot, all the balloons go down, or if you tug on one balloon you can access the other balloons.”
These networks of information are stored in other parts of the brain, including the visual cortex, the auditory cortex and so on. But the hippocampus is what linked them together in the first place—the host at the party, or the balloon man, whatever you prefer. It processes the constant stream of impressions flowing in through our senses and into the brain, and works with the cortex to sort out what’s new and what needs updating (“birds fly” updated with “except penguins and ostriches, even though they’re also birds). It establishes relationships between things in the world. “There’s been a slow, but sure, change in the consensus,” Shohamy told me, “and it’s an exciting thing to see.”
Lonni Sue would have been fascinated with all of this before her encephalitis; her expertise lay outside the sciences, but her intellectual curiosity was virtually limitless. There’s no way she could understand it now, given her limited attention span and her inability to absorb new information. But every time she tries to land that virtual plane in a Princeton laboratory—or slides into an fMRI scanner, or takes on one of the new memory test researchers keep devising to try and understand what she’s lost and what remains—she is further ensuring that her name will be prominent in the textbooks that will train a new generation of neuroscientists.
This is an edited extract from The Perpetual Now: A story of amnesia, memory and love by Michael D. Lemonick, published by Penguin Random House.