Magazine: Library of the Mind
Quick—what were you doing last Tuesday night at around, say, 7:30? Where did you park your car this morning? When’s your mother’s birthday, and did you send her a card? Did you remember to pick up bread and milk on your way home last night?
Whether or not you’re able to readily answer those questions, they all involve memory—one of the human brain’s most powerful, useful, and yet puzzling and sometimes frustrating abilities. “We form memories every moment of our lives and store them away,” notes Michael Kahana, professor in Penn’s Department of Psychology and director of the Computational Memory Lab. “They may be very hard to retrieve, but it doesn’t mean they’re not there. And sometimes when we least expect them, they pop up. Sometimes when we most want them, we can’t seem to find them. And the fact that we’re actually able to find them as often as we can is kind of a miracle.”
Studying how human memory works, the different neurological structures and pathways that make it possible, and how the brain puts it all together in the processes of learning and behavior is Michael Kahana’s passion. “What drives me when I wake up in the morning and come into the lab is that I want to understand this amazing human capacity, the ability to form and retrieve memories,” he says. In that quest, he harnesses a wide variety of techniques ranging across multiple disciplines including experimental psychology, neurobiology, computer science and applied mathematics. Such a broad approach is necessary because memory is a highly complex and elusive scientific target that manifests itself in different forms.
“My lab has mainly been focused on studying two types of memory: verbal memory and spatial memory,” Kahana explains. “Spatial memory is, imagine yourself in Disney World, and you’re walking down Main Street USA, and in front of you is Cinderella’s castle, and I tell you, point to Space Mountain. Students who have been to Disney World even just once or twice will typically point to the right. Although it is unlikely that they memorized a map of the Magic Kingdom, they have a general sense that Space Mountain is to the right. Building that representation involves linking landmarks or objects with their locations.”
Perhaps even more important is how information is remembered in its temporal context. “A more general kind of learning is linking aspects of an event to the time in which they occurred,” Kahana says. “That’s what we call episodic memory, learning about an item in its context, where context uniquely specifies when the item was encountered.”
Kahana notes that episodic memory is basically the conscious form of memory that we use all the time. “Where did I park my car, what did I eat for breakfast, who did I meet at the party, all of these things involve linking an item to its time and place,” he says. “Episodic memory and spatial memory are really two sides of the same coin: putting an event, an item, in a position either in time or space or both. In my lab, we are concerned with identifying the neural and cognitive mechanisms that support these memory functions.”
Much of Kahana’s data is collected from electroencephalograph (EEG) recordings from electrodes placed on the scalps of healthy volunteers, most of them Penn undergraduate or graduate students. Scientists have been performing EEGs for decades now, but the technique has definite limitations. Kahana explains that because the strength of electrical signals drops off rapidly with distance, “you can’t know with scalp recordings where the signals are coming from in the brain. If you’re all the way out at the scalp, all those signals inside the brain blend together. It’s sort of like listening to the orchestra from a mile away and trying to pick out a particular instrument.”
But thanks to a special collaboration with the Penn medical school and Thomas Jefferson University Hospital, Kahana is one of the few researchers in the world who is able to go further. “We record from the brains of neurosurgical patients, patients with epilepsy, patients with Parkinson’s disease,” he says. Such patients may have electrode arrays placed in their brains to monitor seizure activity or to electrically stimulate certain brain regions, and many of them volunteer to participate in Kahana’s projects. “From these recordings we can observe the responses of individual cells, neurons, in the brain, as well as the responses of the electrical field surrounding those neurons, during the learning and recall of items or objects learned in a particular context.” This allows Kahana to examine how networks of nerve cells give rise to memory.
As the name implies, not all of the Computational Memory Lab’s work involves human patients and volunteers. Kahana says, “To make sense of the behavior and the neurophysiology, you need to build mathematical or computational models. We then test these models to gain a deeper understanding of those things which we can’t observe, but must infer. The big challenge in the field is to have models that can relate measurements of neurons, of fields of neurons, of entire brain activity, all the way up to actual behavior.”
Kahana never set out to become a brain researcher. “As an undergraduate engineering and physics student I never imagined that I would become a psychologist or a neurobiologist.” A computer programming gig for a psychology professor got him interested in neural network modeling and mathematical psychology, in which he earned his Ph.D. from the University of Toronto in 1993. But even when he became an assistant professor at Brandeis University, “I had absolutely no interest in the brain, I had never taken a biology course in my life,” he says.
Then in the fall of 1996, a chance meeting with a Harvard neurosurgeon, Dr. Joseph Madsen, resulted in a dramatic change in Kahana’s research career. Madsen invited Kahana to give a “grand rounds” lecture on human memory to the neurosurgery department, and even better, an opportunity to “scrub in” and observe a neurosurgical procedure. “I was stunned to discover that neurosurgeons were routinely implanting electrodes in people’s brains, recording from these electrodes for many weeks while the patients played games, watched TV, and spoke with their family, and then, after the patient had their surgery, the data were simply deleted. They were deleting the data because they didn’t have enough storage capacity to save it.” Inspired to combine his computational modeling work with neurobiology at its most direct and fundamental level, Kahana decided to change his professional direction.
“It went better than I could have anticipated,” he says. So did his move to Penn in 2004. Kahana stresses that his work at the interface between basic science and clinical medicine would be much more difficult in most other institutions. “It is quite unusual to be at a world class university that has its scientists and physicians working in such close proximity. It makes it much easier to build the kind of collaborative efforts you need to collect these kinds of data, which are massive undertakings involving a lot of different people each with their own area of expertise.”
So does Kahana have any special insights on how the rest of us might do a better job remembering where we parked? “Being that I’m a memory researcher, what I do is walk out of the garage and forget where I parked my car, and then explain to my wife that I study memory because I am so forgetful,” he jokes. But he denies that people tend to expect a memory researcher to remember things any better than the rest of us. “People are so polite about that. My wife should give me such a hard time. I asked her about this, and she said, well, you never had a good memory, so it doesn’t bother me,” he laughs.