A Conversation with Emery Brown
CHANCE asked former Editor Sam Behseta and colleague Rob Kass to interview their friend and colleague Emery N. Brown to learn about how his unique career has created intersections between statistics, neuroscience, anesthesiology, mathematics, and much more.
Sam Behseta: Welcome, everyone. My name is Sam Behseta, former editor of CHANCE magazine. I’m glad to have Rob Kass here today for a special conversation with our guest of honor, Professor Emery N. Brown.
Emery Brown is the Edward Hood Taplin Professor of Medical Engineering and Computational Neuroscience at the Massachusetts Institute of Technology (MIT). He is also the Warren M. Zapol Professor of Anesthesia at Harvard Medical School, and an anesthesiologist at Massachusetts General Hospital.
Brown received his BA in applied mathematics from Harvard College, his MA and PhD in statistics from Harvard University, and his MD from Harvard Medical School.
Brown is also an anesthesiologist statistician who is recognized for developing signal processing algorithms for neuroscience data analysis and for characterizing the neurophysiological mechanisms of general anesthesia. He was a member of the NIH BRAIN Initiative working group and has received an NIH Director’s Pioneer Award, National Institute of Statistical Sciences (NISS) Sacks Award, American Society for Anesthesiologist Excellence in Research Award, the 2018 Dickson Prize in Science, and the 2020 Swartz Prize for Theoretical and Computational Neuroscience. He is a member of the National Academy of Medicine, National Academy of Science, and National Academy of Engineering.
Kass is Maurice Falk Professor of Statistics and Computational Neuroscience at Carnegie Mellon University (CMU); and an old friend and collaborator of Dr. Brown; and my PhD advisor, so I’m glad to have both of you, and especially thank you, Emery, for agreeing to be interviewed by CHANCE.
Kass: Emery, when I spoke at your 60th birthday celebration symposium dinner at MIT and when I introduced you for the Dickson Prize here at Carnegie Mellon University (CMU), I commented on a couple of things that have always seemed crucial to your development and success—first, your undergraduate background in applied math, which included an exposure to both dynamical systems and statistics.
You saw the enormous potential for exploiting the connection between dynamical system models—represented by differential equations in continuous time—and certain time series models, namely state-space models, which can be considered discrete analogs, where differential equations are replaced by difference equations. This has informed a good deal of your technical work throughout your career.
At the highest level, I’ve taken the message to be that if you want to understand dynamic neural phenomena, you should use dynamic methods, which include both mathematical models to aid understanding and statistical models to guide data analysis. Do you think that’s accurate?
Brown: Yes, I think that’s right. The applied mathematics did give me insights into dynamical systems. At the same time, I was studying statistics and my interest in statistics was in time-series analysis—the stochastic processes that are dynamic. I think the first time I really saw the juxtaposition of a statistical model, representing dynamics, and a mathematical model representing dynamics, was when I was studying circadian rhythms for my PhD.
There was this one question that bothered me that I had worked on and it had me appreciate that you could have a dynamical systems representation, but when you looked at the real data, you should have a statistical representation to capture that dynamical systems formulation as closely as possible.
Kass: I’m curious: Could you say a bit more about your undergraduate experience? Was it Fred Mosteller you were exposed to? How did you get drawn into statistics in the first place?
Brown: The truth is, I was drawn into statistics because of my roommates. I started off in college at Harvard and my thing was going to be romance languages: I figured I was going to be a physician who worked for an organization like Médecins Sans Frontières or the World Health Organization, because I was interested in languages and I wanted to be a doctor. My roommates were majoring in economics and they kept talking about economics and statistics and that seemed really cool to me, so I decided, my sophomore year, to switch my major. Then Ken Wachter, who was, at the time, the senior tutor in the statistics department at Harvard (now at Berkeley), said, “Well, if you’re going to go to medical school, you should probably write a thesis on a medical problem. Why don’t you go talk to Professor Mosteller?”
Fred ran this very large surgical research group that was trying to understand how to measure the benefits of surgery—which is something that is not usually done with randomized controlled clinical trials, but this is the sort of thing that Fred was very good at. He would take on these messy problems, create structure from them, and answer these big questions. Everyone wanted to work with Fred.
I remember when I went to see him the first time; it was lunchtime. He said, “Pardon me, I hope you don’t mind if I eat my lunch while I talk to you.” I said, “No, go right ahead.” And I said, “Professor Wachter told me I should come and talk to you, because he thought I could do my thesis project with your surgery research group.” And he said, “Easy as pie. We meet twice a month, on Tuesday evenings, over at the Harvard School of Public Health. We start with dinner. Then we have a meeting after that. You’re welcome to join us at the next meeting.” Then there was the silence. I said, “Thank you,” and he said, “I don’t think we have anything else to talk about, do we?”
Kass: That’s Fred. My experience with him was so similar.
Brown: That was probably one of the most amazing experiences—just watching Fred work as a scientist: watching him get people to work together, watching him think; he always led from the back of the room. He would make suggestions and he got the group to think about them.
Kass: So you were planning to go to medical school. What made you decide, in addition, to do the PhD in statistics?
Brown: Statistics seemed to explain everything. I pretty much decided after my junior year—after the first couple of courses I had taken. Actually, my intro statistics course wasn’t in statistics. I took the introductory econometrics course for graduate students in economics. That turned out to be a great influential experience, because it was taught by Gary Chamberlain and he taught it with DeGroot’s book Probability and Statistics.
Kass: Oh, wow.
Brown: He taught classical statistics and Bayesian statistics on par with each other.
Kass: It was not the norm back then.
Brown: It was not the norm at all. I am so grateful for that because I didn’t grow up with this thing called Bayesian statistics, but it seemed like it was totally feasible, and I find now that I teach my students the same way. It was that junior year experience and working with Fred that made it pretty clear: I knew I wanted to do a PhD in statistics.
Behseta: I’m curious, Emery, when I look at your CV—your PhD in statistics and your medical degree are both almost in the same year: ’87, ’88, if I’m not mistaken.
Brown: Yeah—I was accepted at Harvard Medical School, then I spent a year in Grenoble, France, studying math, then I came back to join the first-year class at Harvard and the requirement, at that time, was that you had to apply to the MD PhD program once you were in medical school. I did the first two years of medical school and took off four years and did my PhD in statistics. Then, I finished the last two years of medical school. I didn’t submit my PhD in statistics until I finished medical school. That’s why I effectively got my PhD and my medical degree at the same time.
Kass: I was trying to think back to the times we first met—I think we had known each other for a few years, but then bumped into each other at the fifth international meeting on Bayesian statistics in Valencia in 1994. I remember walking and talking with you for quite a while and you told me, excitedly, about your interest in neuroscience. You must have started working with Matt Wilson—an MIT neuroscientist—already, or maybe if not then, shortly after that? Because your paper on recovering a rat’s foraging path from neural recordings in the hippocampus was submitted in 1997 and published in 1998.
Brown: That’s right.
Kass: That seemed like something of a triumph for you, in the sense that it demonstrated both the predictive power and the explanatory merit of state-space modeling for understanding neural activity. Did you see it that way? Did you anticipate how influential it would be?
Brown: No, I don’t know if I can honestly say that, and I wasn’t expecting it to be as influential as it turned out to be. I really enjoyed the problem, but that collaboration was serendipity.
I had made a conscious decision to start studying neuroscience. I’d been on the faculty in the anesthesia department (at Massachusetts General Hospital) for two years—circadian rhythms is what I had worked on for my PhD—but I wanted to branch out and have something with more depth to it. I did a systematic search across the major areas of the National Institutes of Health, which was funding at that time. Women’s health was just coming into being; genetics and genomics were starting to blossom. There was the environment. There was cancer.
Then there was also the brain and neuroscience, and my friend Steve Massaquoi—a neurologist at Massachusetts General Hospital—had always been telling me, “Emery, you should really look at the brain. I mean, that’s going to be really, really cool.” Matt Wilson had two papers that came out in Science, one in 1993 and one in 1994, showing that the place cells—cells in the hippocampus—actually represented an animal’s location in space (Wilson and McNaughton. 1993 and 1994). He had come up with a technique for decoding them, going from the firing patterns of the neurons to saying where the animal was in space.
I’d read that paper and I thought, “When I know enough, I will go and talk to Matt. He’s right here in Cambridge.”
At the time, I was teaching a course called “Statistics in Physiology” with Art Dempster from the Harvard stat department and we decided to devote the second semester of the course specifically to neuroscience—not broadly, but specific topics in neuroscience—and I was learning things one lecture ahead of the students and presenting it to them.
Then, we got a call from one of Matt’s graduate students, a postdoc—Ken Long, who’s now at Harvard, running one of the neuroscience programs—and Loren Frank, who was a graduate student with Matt. It turned out MIT didn’t have a statistics department, and they had questions about their data analysis. The secretary in the department said, “Art and Emery are teaching this seminar. Why don’t you guys come over and give a presentation?”
Kass: That was just coincidental?
Brown: That was just coincidental. Ken and Loren came over and Matt finally joined in, and that’s how I got started. I said, “You know, I had planned to talk to you, but I didn’t think I knew enough to talk to you yet.” That summer, I took the month-long Woods Hole course on computational neuroscience and started working on a hippocampus decoding problem for my project. That was the summer of 1995. Loren and I made an agreement at that time: I said, “You teach me neuroscience, I’ll teach you statistics,” and that’s how we got started.
Kass: You ended up on his thesis committee.
Brown: I was on his thesis committee and then he did two years of a postdoc with me, then he flew off to the University of California, San Francisco (UCSF) to take a faculty position.
Kass: He’s now a very successful neuroscientist himself. He’s got a great lab. He’s had a wonderful career.
Brown: He’s even a Howard Hughes principal investigator—he’s been extremely successful. But, again, our working together was serendipity.
Kass: The next thing I remember was in 1998: I asked you to give a talk at an ENAR meeting. I asked you and a couple of other people to give talks but they were not well-attended because I think statisticians didn’t appreciate that this was actually super-interesting. It was there—over lunch, I believe—that we decided we ought to write a review paper about statistics and neurophysiology.
Once we got going on the paper, it turned out that all the things we thought we should be reviewing actually didn’t exist. At least, that’s the way I remember it. We kind of had to create the things ourselves and we each mostly went our own way, although we collaborated a little bit. It took several years before we finally published this paper with Valerie Ventura that was in the Journal of Neurophysiology (Kass, Ventura, Brown. 2005).
Brown: The work that I did on the decoding problem with Matt and trying to go from neural firing activity to the rat’s position, by formulating it as a state-space problem—you’re trying to figure out the state that was the animal’s position, and you’re getting it, not from continuous measurements, but from this multivariate point process, which is the firing activity of the neurons. Basically, a one-sentence summary is you had to figure out how to build a Kalman filter for point processes. That’s the simplest way to say it.
The main thing—and this is where I felt I had a strong comparative advantage with the work I had done as a PhD student—was that people were using static methods to track dynamical processes. All I did was come along and say, “Well, let’s just use dynamic methods—time series techniques with state-space methods—to capture these phenomena.” Using static techniques was not as good as having techniques that were indeed dynamic.
Kass: Let me add one thing I remember really well. I’m sure you will, too—the first time I visited at Woods Hole was during the Neural Data Analysis course that you ended up running with David Kleinfeld and Partha Mitra. Andy Schwartz was there, and he had demonstrated the potential for using brain signals to control a robotic device. The three of us were talking and I had already—from hearing you talk about your methods with the hippocampus—thought about the possibility of doing a brain-computer interface with this approach; it seemed natural. When we talked with Andy and further developed that idea, I continued along those lines, and now it’s a standard approach to brain computer interfaces. That’s how people do it. It’s for exactly the reasons, I think, that you saw initially.
Brown: Right. I think that’s the case.
You mentioned the Woods Hole course. Kleinfeld and Mitra had the amazing foresight to set that up. They called it “Neuroinformatics.” I worked with them for 10 years, teaching, and then eventually running that course because it was much-needed. That allowed me to get a very broad sense of what the data analysis problems in neuroscience were and also to teach some of these ideas to a number of young neuroscientists, graduate students, and postdocs who were coming along—putting forward this idea of using dynamic methods to analyze dynamic data. And you came and helped with that, and Satish Iyengar.
Kass: Yes, and Valerie Ventura.
Brown: Yes, Ventura was a very loyal participant in that effort, and I think it had a tremendous impact on neural data analysis or computational neuroscience.
Kass: It’s interesting to think about the SAND meetings [Statistical Analysis of Neural Data], too, because I remember when the very first one was held and I’m sure you remember, too. We were basically trying to figure out what were the problems.
I think the year was 2001 or 2002, and we still didn’t really understand what the statistical research priorities should be. You had already identified one very tractable problem, but there were others out there and we were trying to figure it out, and trying to penetrate some of the jargon. In fact, the neural coding jargon the theoretical neuroscientists were so fond of using (based on information theory)—you started seeing through it right away. It took awhile to have the field move away from what it was doing and toward something that was way more productive.
Brown: Well, the barrier to entry—we’ve talked about this—for statisticians to get into this field and make a contribution is high; you really have to study the neuroscience.
It’s extremely difficult to be a dilettante in this field and come in and throw a technique on something, then claim victory and march out. It won’t be effective. There are a lot of talented quantitative people in the field, and to convince them that what you’re doing is the right approach, you have to argue, not only from the methodology standpoint, but also from the standpoint of understanding the impact it’s going to have on neuroscience. You’re going to have to care about both. That’s really the key.
Behseta: This prompts me to ask a question about the current state of statistics in neuroscience. Can you elaborate a bit more about the role that statisticians getting involved with neuroscience can play in all of this? I did some neuroscience, mainly because Rob was my advisor and I became fascinated by it, and he sent me to the University of Pittsburgh to sit through some basic neuroscience courses. But…there wasn’t a degree called “statistics in neuroscience.”
Brown: Right: Rob and I have had this discussion, many times. And again, my personal bias is that for this particular field (it’s probably true for other areas), which I’ve worked in for a number of years, the best way to have an impact is to get into the field and understand the problem or the problems that you’re working on deeply from a neuroscience perspective. Then, build the methods from that perspective.
For one thing, what I found is that it puts constraints on the problems that actually simplify the methodology development and, as I was saying before, being able to make a credible statement about why the method should be accepted or why the approach is to be adopted. Because, as you know, a high fraction of the people you’re talking to don’t share our perspective on statistics. They can find very plausible arguments, looking at problems from other perspectives—in dynamical systems or physics—for why a certain approach ought to be taken. I think if we’re going to argue from a statistical standpoint, we have to know both the statistics and the neuroscience to do that.
Going back to the training, what does that mean? It means we have to train statisticians—our postdocs, our graduate students—in neuroscience at the same time as we’re training them in statistics. That’s something that Rob at Carnegie Mellon has put a good deal of emphasis on. I see it all the time. I see my bio-engineering students and my physics students having no fear about acquiring in-depth knowledge of neuroscience, and I think we have to promulgate that culture among our statistics students as well.
Behseta: Can you take us through your weekly schedule? What days are you actively a neuroscientist or a statistician?
Kass: He is always a statistician.
Behseta: Right, but I do only one of those, so I’m curious about what sets of experiences, knowledge, skills you take from Job A to Job B?
Brown: My typical week for this semester is teaching my undergrad statistics course on Mondays and Wednesdays, and then the balance of the day, I usually have meetings—they could be group meetings or research meetings or various committees; most of them have recently been a lot of committees. Every Tuesday, I’m in the operating room. In pre-COVID-19 times, I tried not to schedule anything like trips for Tuesdays, because I don’t like missing my operating room days. Thursdays and Fridays, I’m again in meetings.
That’s my typical week…I’ve been doing that for 20 years or so.
Kass: As Sam mentioned, you were on the NIH BRAIN Initiative steering committee, and there, you made sure that advanced methods for data analysis became one of the priorities for research, which has been, I think, really important. You were selected as the 2020 winner of the Swartz Prize for Theoretical and Computational Neuroscience of the Society for Neuroscience, which is a major prize—it’s not only a recognition of you, but also of the importance of statistics in neuroscience.
It seems like we’ve come a pretty long way since that beginning point. How do you see the current status of statistics in neuroscience? We can claim a certain number of victories, but obviously there’s work to do.
Brown: I think the opportunities are boundless. For one thing, data generation in the neurosciences is exploding just like data generation everywhere, across a number of fields. It’s not only Big Data, but it’s big dynamic data; it’s big dynamic multiple types of data—not only point processes but continuous processes. In addition, it is being spurred very forcefully by the BRAIN Initiative. The idea behind the BRAIN Initiative was to create the tools necessary for neuroscientists to answer questions about the brain, which means building a lot of new techniques, so there are a lot of new types of data coming out of all the new recording methods.
What I think is really cool about it is these Big Data problems involve highly structured experiments. You make a decision about how many animals, where to record from, how long you record—so there’s a design. Thinking about how you improve that design is an area that is essentially untouched, so I think that the opportunities are even bigger than when we got into the field, because of the data explosion and the unique characteristics of these sorts of data. But again, that requires the neuroscience insights as well as the statistical insights.
Kass: One thing that has happened in our conversations many times is that we talk a lot about statistics and statistics training, and then the subject changes to anesthesiology. I’ve watched you develop as a statistician and I’ve heard a lot about your development as an anesthesiologist.
It seems to me that early in your career, the two didn’t have anything to do with each other. You worked as a clinical anesthesiologist while doing statistics and learning neuroscience, and then somehow, it all came together. I don’t know how long ago it was—it might have been as much as 10 years now—when you started to identify these previously mysterious mechanisms that are involved in losing and regaining consciousness. I’d like to hear you talk about these dual interests and how they’ve interacted.
Brown: Sure. I decided that I wanted to be an anesthesiologist in my third year in medical school. I did anesthesia for two weeks during my surgical rotation. I really enjoyed it. I liked the physiology, the pharmacology, the fact that we did procedures. I liked how anesthesiologists took care of patients. I knew I wanted to do anesthesiology, and I knew I wanted to do statistics. I continued working on circadian rhythms once I finished my training in anesthesiology. The two had nothing to do with each other: One was a day job and the other a night job. There was my research world centered around circadian rhythms and thinking about neuroscience and data analysis, and I took care of patients clinically during the day.
But I’ll be frank with you. I kept this thing, which I call my BS list, because there was a lot of stuff in anesthesiology that made no sense. Many people would tell you things based on these unfounded arguments—but they were the right answer to the questions on the board exam, so you learned: That’s the way you did it.
The one thing that bothered me especially was we knew that drugs bind to certain receptors: To be very concrete, propofol binds to GABA receptors or sevoflurane, or isoflurane binds to the GABA receptors. Okay, that’s fine—we know what the molecular targets are, but why do you lose consciousness? Why do you have an altered state? Knowing the receptors alone doesn’t tell you that, so I started looking at the anatomy of the brain.
I must admit, I was pushed by one thing. I would go to neuroscience meetings and we would talk about the statistical methods or analyses that I was doing and then people would say to me, “Oh, I hear you’re an anesthesiologist,” And I would say “Yeah,” and they would ask, “How does that work?” And I would say, “We don’t know.” And then, people would look at me—”You do this in the operating room, you’re putting these drugs in people, and you don’t know what they’re doing?”
After a while, you hear yourself say that to neuroscientists—people who are bent on understanding the mechanisms of how parts of the brain work or how the brain represents our information—and I realized that that didn’t sound too cool, and that the paradigms that Matt Wilson and I were working with could be used to study anesthesia. Matt’s lab put electrodes in rodents’ brains, so why didn’t we just give the rodents anesthesia and see what the brain was doing?
I came up with this idea for having not only rodent studies, but also non-human primate studies. Doing both of these kinds of experiments to try to learn as much as we could took a team effort. Once we started, I wrote out detailed discussions of how the drugs could act in the brain and the circuits to generate altered states of arousal in a couple of reviews. One appeared in the New England Journal of Medicine in 2010 and then one was in the Journal of Neuroscience in 2011.
Around that same time, we began working with Nancy Kopell, trying to build models of the circuits and the dynamics that we were seeing.
Kass: Nancy is a mathematician at Boston University and a leader in this kind of modeling.
Brown: She’s an expert in dynamical systems. I had, at that time, also been working with Patrick Purdon, starting to do human experiments to collect data on humans and looking at the human brain under anesthesia. The one thing that was apparent was that the drugs created oscillations in the brain.
Kass: This is a pretty well-known, yet somewhat mysterious, phenomenon—that neural circuits in the brain will oscillate at particular frequencies under certain conditions.
Brown: Every drug creates oscillations that are basically mechanism-dependent, so you had a different mechanism, or drugs in the same class, that showed similar oscillations and if you changed the drug class, you got different oscillations. This seemed to be an entrée into trying to get some insights on how the anesthetics were affecting the brain and how they produce unconsciousness, but talking about oscillations puts you right back in the realm of time series analysis, especially spectral analysis, which I’d worked on in my PhD.
In that sense, I was trained to start looking at these problems. I realized there was this problem that was sitting right under my nose. We’ve made a lot of progress not by inventing new methods, but by applying a lot of basic methods correctly and diligently. We’ve gained a lot of insights.
Behseta: So you contributed to the field through your statistics knowledge.
Brown: One thing that we did was we took time and characterized the dynamics of each of the oscillations for each of the drugs in great detail. In other words, we just did the statistics right—in detail—so we could benefit our modeling colleagues like Nancy Kopell. We’d say, “Look, for propofol, you have slow oscillation and alpha oscillations concentrated in the front of the head. Seemingly, some of the anatomy must favor that—what could that be? What sort of model can generate this?”
It was the data, it was doing the work—laying the foundation and the framework of the neurophysiology. Characterizing the dynamics in detail that we observed when we looked at patients or animals, and then saying, “Okay, your model is constrained, because whatever you do, you have got to reproduce this.” Then making predictions with those models and going back and doing experiments that will either verify or refute those predictions.
Statistics plays this crucial role of establishing the quantitative constraints that led us to understand what the dynamics of the brain responses are when these drugs are administered to patients or animals.
Kass: I want to switch gears now. I want to talk about a different kind of thing, involving some of your personal experiences, and I know Sam wants to follow up on this as well. The path of every successful person results from some combination of talent, persistence, and luck. In your case, what struck me the most are two different intertwined abilities.
The first is your great patience and confidence in turning away from short-term gain and, instead, investing time and energy in learning what you need to know and creating the environment necessary to achieve results you recognized as truly important. Could you comment on that first?
Brown: Well, let’s see: patience. I guess when you decide to do both an MD and a PhD, that implies a certain degree of patience.
Behseta: It’s a huge amount of patience.
Kass: But it’s also setting up the anesthesia experiments. You were putting a lot of time and energy into something where you didn’t know exactly what was going happen. That’s why I say, you had to be both patient and confident.
Brown: Right. I guess there were two things. I had this vision of having a team working on this and I had this vision of doing human experiments, non-human primate experiments, and rodent experiments. There were colleagues around who agreed to take on those various components, so it seemed feasible. That was one thing.
I think the other thing was that a lot of people were working on mechanisms of anesthesia, but nothing they told me changed in any way what I did in the operating room—it had no impact, so there was an incentive to try to do something that was going to have a real impact.
The thing that was so compelling to me was that anesthesiology could be considered as a sub-discipline in neuroscience. There’s just a fundamental flaw in the thinking in the field that persists today. I think it’s one of the things that impeded the field from making progress. We are a sub-discipline of clinical neuroscience, like neurology, neurosurgery, psychiatry, sleep medicine, psychology. We have our impact by giving pharmacologic agents, but they’re acting in the brain and the central nervous system.
Once I realized that, I thought, “This is the right way to think about it.” Despite all the other things that were going on, I felt this had to be successful on some level. I didn’t know what level yet, but it was definitely worth trying.
In fact, when I applied for, and won, the NIH Pioneer Award, back in 2007, my application was very simple: “I’m going to use neuroscience to study anesthesia.” That was kind of the breakthrough idea, if you would.
Another thing is, as I started doing little neuroscience things in the operating room, they just reinforced my intuition. For example, people under anesthesia have EEG patterns that look just like the EEG patterns you see in patients who are in a coma because of structural brain lesions. The cool thing about it is that with anesthesia, you see effects almost immediately. A patient who has an injury might take several days or weeks to recover enough that you see a change in the EEG, but with anesthesia, you can see these oscillations in the EEG appear after the drug is introduced, and disappear shortly after we stop the drug.
As I’m sitting there watching these things, I’m saying, “I don’t care what everybody says; this is the way to think about what’s going on here,” and it all started to tie together: the data analysis and statistics, the modeling that we were doing with Nancy, the clinical things that I was doing in the operating room. More recently, coming up with different strategies for dosing drugs because, as we monitor a patient’s brain under anesthesia, we can infer, very reliably, that they’re unconscious at far-lower doses of anesthesia than are typically given.
Just last Tuesday, I was in the operating room and I took care of a gentleman who was quite sick. He had to have a procedure that wasn’t going to take very long. The resident got consent from his wife because he was so sick. We were very lucky—we were able to do the procedure in about 10 minutes or so, and then we were ready to wake him up. He came to, regained consciousness very quickly, about two minutes after we finished—literally like that;and he was very clear and alert. Then the resident said to me, “Oh, I forgot to tell you his wife said he always takes a long time to wake up from anesthesia.”
It wasn’t that he takes a long time to wake up. It’s just that he’d always been overdosed before, I’m pretty sure of that. I mean it just makes sense. Right?
Kass: The other personal characteristic that I’ve commented on many times in settings where I’ve introduced you is your ability to keep moving toward your goals. This has required not only foresight and persistence, but also a kind of dexterity in sidestepping the inevitable pitfalls. You know there are pitfalls with every project, and you’ve been unusually good at dodging them. You emphasized the importance of this years ago, when we invited you to give an informal talk at Carnegie Mellon to a group of underrepresented minority students.
Could you describe some of the specific challenges you faced, the kinds of things that they would have been concerned about, and then the importance that you’ve attached to dealing with those situations productively?
Brown: One of the things that my mother used to tell me is nothing is supposed to work; you have to make it work, and that requires putting in time and energy. I remember—I think I’m quoting him correctly—Brad Efron saying one time, “You really don’t know how smart you are until you work hard.” I think that’s the case.
For me, I get interested in something like statistics and I can’t imagine myself doing anything else. When I decided I wanted to be a statistician, I thought that was the coolest thing I could be. And I can tell you now, if I were about to start over, I would do exactly the same thing: I would go after statistics because I developed a passion for it. Once I have the passion for something, nothing is going to stand in my way. Nothing is going to keep me from achieving that goal. I wanted to be an anesthesiologist and—basically the same thing—I wanted go to medical school.
There are going to be people, for various reasons, whether it’s race or because they don’t like you, who are going to oppose what you want to do. That’s a given. The world is just set up that way. But you can’t let them stop you, because if you’re thinking about solving important problems, you’re going to, ideally, make things better for humankind. Then you have a bigger goal that you have to address, and a lot of times, what people want to do is to try to take you off the mark by engaging and getting you into tussles and all these sorts of things, and you don’t have time for that. “I want to figure out how anesthesia works, so I’m going to keep working very hard on this” or “I want to get my PhD in statistics, so I’m going to keep working hard on that.”
I was very fortunate—like working with Professor Mosteller early on. In working with him, he came away with the impression I was very serious and focused and wanted to become a statistician. I think that his encouraging me made me feel that I could be successful, even with this small problem that I worked on for my undergraduate thesis.
It’s not always going to happen, but you can’t let the naysayers dissuade you from pursuing your goal. You just can’t do that.
I can honestly say there are far more people who’ve been extremely helpful to me than people who impeded me.
Kass: Is there any chance I could push you to go back to specific anecdotes, just to be more concrete? Because examples are nice. I was talking to my wife, and she immediately said she remembered that time you talked about your personal experiences at a dinner. You started in segregated schools, up until age 9 or 10?
Brown: Up to sixth grade. Well, I can tell you about the language prize. There was a summer program that my city (Ocala, Florida) had with its sister city, which was Sincelejo, Colombia. Every summer, students in Sincelejo would come to Ocala and students from Ocala would come to Sincelejo. It was a big deal. It involved the students who were in the second- and third-year Spanish—
Kass: In high school?
Brown: In junior high. This was during seventh grade. I went to school with white kids for the first time. I had never gone to school with white kids before that. And I went to school on Ocala’s Freedom of Choice. That was the way they integrated the schools. What “Freedom of Choice” meant was a few black students elected to go to the white schools, but no white students elected to go to black schools.
I elected to go to Ocala junior high school. I did pretty well, grade-wise and everything. I didn’t have any problems—it didn’t change what I was doing, and the students realized that I was quite capable in languages. I really wanted to study languages, and that was my first chance to study Spanish. Then, the next year, I could apply to be one of the students who would go to Sincelejo, and I hadn’t been out of the country at the time, so this was a big deal.
I wrote my application and everybody knew I was the top student as far as languages were concerned. Then, my mother went to a meeting—she was part of the League of Women Voters, and they were discussing the whole policy of sending students to Sincelejo. When she got home, she said, “I don’t want you to be disappointed, but they’re not going to pick you. They’ve decided to draw a line, saying that only students who are actually in Ocala, and not in the county, are going to go to Sincelejo.”
Kass: In other words, white students.
Brown: Yes. I was bummed. I was totally bummed because this was a big thing, but I still knew I was the best student; they couldn’t take that away, and they knew it, too. I’ve had to remind myself of that kind of thing on several occasions, I’ve had to rely on that sort of internal conviction.
The next summer, I took French. I was going to an integrated school in the northern part of the county—an integrated school in the ninth grade—and they didn’t have a Spanish teacher, so I had to stop taking Spanish, and I started taking French. I got into it, and my French teacher told me there was a summer program for students who wanted to go to France. I didn’t go to Colombia, but I went to France.
The point is, they weren’t going to stop me just because I couldn’t go to Colombia.
You know, those sorts of things play themselves out over and over again. Recently, they have played themselves out overtly because of the social climate we are in now. More commonly, particularly in professional settings, they play themselves out with micro-aggressions. It’s very easy to let those things take you off your game. They can be very simple things, like someone comes into the room—there are three people in the room and you’re one of the three people. They say “Good morning” to two people and don’t say good morning to you. It’s not like they were racist, right? But again, you see what the dynamic is like.
Fine. I was having a good morning before you came.
I try not to let it bother me, but I think that doing some planning for situations—thinking about these situations preemptively—is very important, because there are certain battles you do have to win to be successful.
Kass: There is one thing this is reminding me of and I’m a little embarrassed about it. I don’t know whether you remember this, but you had taken on a new position and not long after you started to settle in there, you told me, speaking of your colleagues, that they didn’t really know what to make of you. And I looked at you and I said, “You mean because you’re so interdisciplinary?”
I don’t remember what you said, but Loreta, my mixed-race wife, used to call me naïve; now she calls me privileged [laughs]. Again, it was something you had to deal with.
Brown: I know exactly what you were referring to. Let’s just say African Americans in the sciences and academia are not a dime a dozen. For a lot of white people, the African Americans whom they know who are successful are people in entertainment, in sports, or on the staff at the university, but are not the professors. They don’t know whether they should high-five me or say something like, “Hey, man, what’s going on?” or have a normal conversation—”Do you play golf?” That’s very real.
I’ve been around hundreds of people in those situations, whereas looking at it from the other way around, for them, it’s a rare encounter. It’s just part of the reality. I remember one of the first times I was going up the elevator in the neuroscience department at MIT and this guy says to me, “Whose lab are you in?”
Kass: Did you say “Mine”?
Brown: I did. I said, “My own. That’s my lab right over there.” “Oh, I see, oh,” he said.
Here is another example. I came to the operating room one day. They told me the first case had been changed. There’s a guy who…is going to have some sort of abdominal surgery and the resident was seeing him. I went up to see him and I asked the resident if there was anything I should know. The resident gives me a little report, and the nurse comes in, rather gleefully, and she says, “Oh, are you here from Transport?” I said, “No.” The resident said to her, “You know, that’s Dr. Brown; you should Google him.”
This wasn’t that long ago—this wasn’t 20 years ago or something. There’s this reality out there you can’t ignore, but the thing is, you can’t let it take you down. Getting in some ridiculous argument with her is exactly the sort of thing you don’t want to do.
Kass: Did she ever apologize?
Brown: Oh, no. She was happy about her mistake: “Oh, I thought you were from Transport.”
Kass: I guess what you’re saying is your trick here is just don’t let it change your mindset.
Brown: Don’t let it change your mind and don’t let it get under your skin.
Behseta: Do you have any suggestions, any practical solutions? We are doing this for the American Statistical Association. Committees are being formed on promoting statistics for underrepresented individuals, but what practical idea or solutions come to mind to promote statistics and sciences to minorities?
Brown: Just do the obvious things and follow through on them. While it’s great to have more African American mentors, and maybe mentors who are of Latinx descent, those don’t have to be the mentors for these things to be successful. For example, the program to integrate Harvard Medical School was actually formulated and driven by four white guys. I think a large part of is just having the heads and the hearts in the right place, doing the right thing.
The pipeline issues are very real, but they can be fixed. The summer programs work. The postdoc programs work to get students into statistics. In the African American community, people think about going into medicine because they really want to help people. When I was growing up, one of the most-respected in the community was the doctor. In my community, there were two African American doctors.
You can do important things, and have the same impact, as a statistician. Think of the drug trials that just finished. Somebody has to sign off and say that they were properly done, the inferences are correct. That’s not the infectious disease specialist; that’s the statistician.
These jobs can have just as much of an impact on society as physicians. I think it’s something we need to emphasize.
There’s no substitute for doing the hard work. There’s nothing magical. We have to create the environment that makes it possible for underrepresented minorities to see that it is totally doable.
There are some interesting case studies where this has been made clear. The University of Maryland, Baltimore County, has had an amazing success rate at getting African Americans into the sciences and mathematics. I’ve had at least two of those students in graduate school here at MIT. There’s nothing about the soil in Baltimore County; it’s no different than it is anywhere else. It’s not like Freeman Hrabowski [president of the University of Maryland system] just decided to make these guys successful. They come in with this attitude of “We can do this” and they’re part of the community, and they’ve been successful. Those sorts of efforts can be reproduced and propagated, and they will yield fruit.
Behseta: Thank you, Emery, for this fantastic interview.
Brown: Thank you, Sam, for setting up and arranging the interview and again, Rob, for conducting the interview. It’s a real privilege and a real honor for me to do this with both of you.
Further Reading
Brown, E.N., Lydic, R., and Schiff, N.D. 2010. General anesthesia, sleep, and coma. New England Journal of Medicine 363:2638–50.
Brown, E.N., Purdon, P.L., and Van Dort, C.J. 2011. General anesthesia and altered states of arousal: a systems neuroscience analysis. Annual Review of Neuroscience 34:601–628.
Ching, S., Cimenser, A., Purdon, P.L., Brown, E.N., and Kopell, N.J. 2010. Thalamocortical model for a propofol-induced alpha rhythm associated with loss of consciousness. Proceedings of the National Academy of Sciences 107: 22665–70.
Cimenser, A., Pierce, E.T., Salazar, Gomez A.F., Walsh, J., Harrell, P.G., Tavares-Stoeckel, C.L., Habeeb, K., Purdon, P.L., and Brown, E.N. 2011. Tracking brain states under general anesthesia by using global coherence. Proceedings of the National Academy of Sciences 108:8832-7.
Kass, Robert E., Ventura, Valerie, and Brown, Emery N. 2005. Statistical issues in the analysis of neuronal data. Journal of Neurophysiology 94(1): 8–25. Dor: 10.1152/jn.00648.2004.