97. Arne Ekstrom: Spatial navigation, memory, and invasive recordings in humans

Arne Ekstrom is a professor of psychology at the University of Arizona, where he studies spatial navigation and memory. We talk about how he got into psychology, his unusual path to getting a PhD, his work on using single-cells recordings from people, the relationship between memory and spatial navigation, why he uses multiple methods, and much more.

Support the show: https://geni.us/bjks-patreon

Timestamps
0:00:00: How Arne ended up studying psychology and neuroscience
0:06:23: Arne's route to a PhD recording single-cells in humans (via political activism in Central America)
0:20:18: The state of using VR-like tasks in the early 2000s
0:24:32: The status of spatial navigation research in the early 2000s
0:29:45: Collecting data from unusual populations
0:33:59: Why record from amygdala for a spatial navigation task?
0:41:35: Combining memory and navigation in hippocampus
1:02:04: Should I use one method or many?
1:11:29: A book or paper more people should read
1:13:51: Something Arne wishes he'd learnt sooner
1:14:51: Advice for PhD students/postdocs

Podcast links

• Website: https://geni.us/bjks-pod
• Twitter: https://geni.us/bjks-pod-twt

Arne's links

• Website: https://geni.us/ekstrom-web
• Google Scholar: https://geni.us/ekstrom-scholar

Ben's links

• Website: https://geni.us/bjks-web
• Google Scholar: https://geni.us/bjks-scholar
• Twitter: https://geni.us/bjks-twt

References & links

Episode with Lynn Nadel: https://geni.us/bjks-nadel
Episode with Nanthia Suthana: https://geni.us/bjks-suthana
Episode with Nikolai Axmacher: https://geni.us/bjks-axmacher
Episode with Nachum Ulanovsky: https://geni.us/bjks-ulanovsky

Argyropoulos ... & Butler (2019). Network-wide abnormalities explain memory variability in hippocampal amnesia. Elife.
Ekstrom, .. & Fried (2003). Cellular networks underlying human spatial navigation. Nature.
Ekstrom ... & Kahana (2005). Human hippocampal theta activity during virtual navigation. Hippocampus.
Ekstrom ... & Bookheimer (2009). Correlation between BOLD fMRI and theta-band local field potentials in the human hippocampal area. J neurophys.
Ekstrom ... & Starrett (2017). Interacting networks of brain regions underlie human spatial navigation: a review and novel synthesis of the literature. J neurophys.
Ekstrom & Ranganath (2018). Space, time, and episodic memory: The hippocampus is all over the cognitive map. Hippocampus.
Hassabis ... & Maguire (2009). Decoding neuronal ensembles in the human hippocampus. Current Biology.
Iaria & Burles (2016). Developmental topographical disorientation. TiCS.
Kunz ... & Axmacher (2015). Reduced grid-cell–like representations in adults at genetic risk for Alzheimer’s disease. Science.
Logothetis ... & Oeltermann (2001). Neurophysiological investigation of the basis of the fMRI signal. Nature.
Watrous ... & Ekstrom (2013). Frequency-specific network connectivity increases underlie accurate spatiotemporal memory retrieval. Nat Neuro.
Zhang & Ekstrom (2013). Human neural systems underlying rigid and flexible forms of allocentric spatial representation. Human brain mapping.

Transcript

[This is an automated transcript that contains many errors]

Benjamin James Kuper-Smith: [00:00:00] I thought, uh, we could start today with a slight confusion I had that, uh, I'd be grateful if you could just clarify that for me. Which is, so I've known about, uh, you and your work, well, your work more than you, but I've known your name for a while. And it was always Arne Ekström, which to me is pretty much as Swedish as a name can get, or can be.

And so, you know, I always, in preparation, look at people's CVs and that kind of stuff. Interestingly for you, I could only find a CV from 2006. On brainmapping. org But it contained the information I was interested in which was basically I thought how did this clearly swedish guy end up in the US? And then I saw your all of your education including your high school is in the US Are you american or is it it's just the name swedish or?

I

Arne Ekstrom: so my father is, uh, born and raised in Sweden. He met my mom, who's an American citizen, um, when he was [00:01:00] 35 in Switzerland and they got married in Switzerland and then they decided the professional opportunities to be better back in the U. S. Um, my dad, um, is a retired clinical psychologist. So they moved to Boston because that's where they thought to be the most opportunities for that.

Neither of them had any direct connection to Boston. Um, and, uh, so that that's how I ended up, uh, being born and raised in Boston. And so that the 1st name reflects my Swedish heritage. My brother got the opposite. His 1st name is Max and that reflects Ashkenazi heritage. So I'm, I'm half and half.

Benjamin James Kuper-Smith: Okay. Well, I mean extrem is as swedish or so

Arne Ekstrom: Completely Swedish. Yep. My, and my father is 100 percent Swedish.

Benjamin James Kuper-Smith: Yeah Okay, well at least that makes sense. Yeah, I was just really confused because then I I kept waiting, like, when is the Swedish education going to come in? But

Arne Ekstrom: uncommon for a lot of people born and raised in the U. S. from immigrants. So in Europe, your last name is much more kind of [00:02:00] deterministic of your, you know, heritage. In the U. S., it's, it's like a little more guesswork. Yep.

Benjamin James Kuper-Smith: you got to the US. You just, you were just there the entire time. Um, but yeah, just kind of to, um, to kind of set the scene of how you got to kind of do the research you do, because from what I can tell you, pretty much from the beginning, worked on spatial navigation, memory, hippocampus, that kind of stuff.

Um, yeah. Yeah, maybe, maybe it just, uh, right from the beginning you said your father was a psychologist, you studied psychology. Is that the obvious connection or is it a coincidence or? Yeah,

Arne Ekstrom: a lot about psychology growing up. I think it was a little bit of a different era back in the 80s and even the 90s where I think. Concepts of clinical psychology were, you know, to some extent, I think, more widely accepted, and the idea of doing clinical trials to prove concepts [00:03:00] about, say, therapy, is therapy effective or not, right, were not widely thought about.

I think a lot of these ideas handed down from Freud and Young and many other influential clinical psychologists were just kind of assumed. And so that, that's kind of what I grew up thinking some of, but at the same time, having this nagging sense that there's this whole discipline called science. Which, at the time, I did not feel like clinical psychology was.

And, at the same time, my father has a lifetime disability. He's narcoleptic, which means he has a sleeping disorder. He goes into REM sleep very quickly. He, you know, now takes three or four naps a day. Uh, you know, he's had to be medicated most of his life. So he had a neurological condition which could not really be answered by clinical psychology.

Without really thinking about that a lot growing up, I think it inspired me to think a lot more about what the brain contributes to behavior. But at the same time, I was just very interested in experimental work because I felt very unsatisfied by many of the dogmas that I learned about in clinical psychology.

You know, the id, the superego, [00:04:00] archetypal symbols, all that stuff was just kind of unsatisfying to me in some form. And When I started as an undergraduate at Brandeis University, I took a class in, um, Intro to Psychology with Bob Secular, he's a very famous perceptual psychologist, and I just got really fascinated by experimental psychology.

I took chemistry classes, molecular biology classes, and I thought there's this whole field where you can actually test things to figure out if they're true. What a crazy idea, right? Instead of just having to be told what, is true, which to me felt a little bit like religion. Um, here's a way of actually determining what, what is true or not through experimental testing.

Now, of course, you know, fast forward to the replication crisis, you could say, even that seems a little bit naive, right? But at the time, that really didn't matter. inspired me. And I worked in a chemistry lab and was just really very excited about the idea of what we could learn through all the different tools we had available [00:05:00] in molecular biology, organic chemistry, like NMR, like gel electrophoresis.

All these were tools that, you know, I worked with and I thought, wow, we can really figure things out in a way that's unbiased and in some form or another driven by the underlying truth. And, um, based on some of that experience, you know, I, I just kept finding myself gravitating toward the brain no matter what I did, right?

I worked in organic chemistry lab and I kept looking up molecules that would be relevant to binding with neurotransmitters. That I just, no matter what I did, I wanted to understand how this related to the brain. And so that led me to two different labs. I worked in a Drosophila lab at Brandeis University with Leslie Griffith, who's a neurogeneticist, and then Mike Gahana, who's a computational memory expert and cognitive psychologist.

And getting immersed in that, I just got a very different perspective on the brain and behavior, and I started to realize there are all these amazing tools we can use to [00:06:00] really understand How the brain and behavior link to each other. And so that was really kind of my gateway drug, so to speak, into a lot of this.

And I think that's when I really became convinced that we were at a frontier where there were amazing discoveries on the horizon. And that in some form or another, in whatever way I could be part of it, I wanted to be part of it.

Benjamin James Kuper-Smith: I remember correctly, then you, uh, I mean, you ended up doing a PhD with, with Mike Kahana, but you spent a year, well, you took a while, it seems, for you to actually end up doing a PhD in the lab you almost started with. Uh, I think you did a research assistantship first and then a master's. Uh, how did that happen and why?

Arne Ekstrom: I had a little bit of an odd trajectory, I think that's, that's fair to say. I did a senior honors thesis with Mike Gahana. I was extremely inspired both by computational modeling, I did additional computer sciences and linear algebra classes. And then, um, you know, on, on top of that, I'm just very interested in trying to do [00:07:00] cognitive psychology experiments where we could test these models.

And the idea of a model being falsifiable, right? It was not a model that was just on its face true. It was a model that we could test its validity. To me, it was just really inspiring. And at the time, Computational models were focused on, uh, the ones at least that I was working with were focused kind of on two different levels, either these parallel distributed processing networks, which later have gone on from, you know, the work that we work at the time from Jeff Hinton to become what we now know as chat GPT and AI.

These were kind of very simple units that simulated neurons. They're fired or they didn't and synapses either connected or they didn't. But they were capable of pretty complicated behaviors, like some of the mistakes that children make when they first learn to conjugate verbs and things like that. And then on the other end, there were people like Larry Abbott, Marius Usher, and others who were building much more complicated models of single neurons and the physiological dynamics of neurons.

[00:08:00] And then trying to wire these together to account for more complicated behavior. And I got very interested in the idea that by modeling the properties of individual neurons, we could start to look at emergent properties of networks without having to sort of oversimplify a neuron to be the same thing in every single brain region.

So that was kind of what I was immersed in, and I became very interested in the idea of doing single neuron recordings. Um, but I wanted to stay in Boston for another year. I was still kind of figuring out what I wanted to do for grad school. So I ended up doing a research assistantship at Harvard with Dan Schachter.

And that was mostly focused on aging and cognitive psychology. And that was interesting, um, you know, it helped pay the rent in some ways, uh, but I think in some ways it, it wasn't exactly my passion of kind of digging more into the brain, um, although I, I think I learned a lot from the experience too. And then I ended up applying to graduate school and was interested in doing single neuron recordings.

And

Benjamin James Kuper-Smith: In humans or in animals?

Arne Ekstrom: so this started with animals [00:09:00] actually, um, to kind of do the accurate trajectory here. Uh, this started with animals. Um, and, uh, I applied to University of Arizona and several other programs where I could do single neuron recordings. Um, I did apply to UCLA for graduate school to do single neuron recordings with humans.

Um, but at the time, Itzhak Fried is a neurosurgeon there, was not taking grad students. So I was not able to join his lab. And so I ended up going to University of Arizona to work with rats. And I quickly discovered that rats and me were not compatible. meant to be. I just didn't like working with rats. I didn't like killing rats.

I didn't. I just didn't work for me. And to be honest, I think it was just too much. An example where it was maybe not the right fit for me with research interests and mentorship, and, um, I ended up dropping out of grad school and not being sure what I wanted to do with my life, um, and went on an interesting detour where I traveled in Central America, a little bit of South America, got very interested in political activism in various ways, and, uh, spent a lot of time [00:10:00] thinking about, um, you know, ways we could improve the world and, you know, Which maybe is not uncommon in folks in their 20s, um, and got involved in some groups that were focused on building consensus, which is a funny term, which at the time meant everyone had to agree on a principle.

And so that meant working with 30 people where everyone had to agree on your course of action, which in practice is very difficult. But looking back on it, I think I learned something about how to work with groups of people, which in retrospect was helpful. Um, but yeah, so then I ended up spending some time trying to.

Benjamin James Kuper-Smith: So just briefly, uh, that's, uh, I guess that's not part of the, that, that wasn't on your CV. The, the years you spent in Central America.

Arne Ekstrom: I did, I did put that on the CV. Well, and it was really more of a detour, like a couple months here and there. But, I mean, it was valuable in terms of one's kind of development as a person, I

Benjamin James Kuper-Smith: Yes, it's interesting to me because I, I, I didn't do that, but I also started a PhD, um, in Europe. So I'd already done my master's, then I started a [00:11:00] PhD and stopped that after half a year or so because it just wasn't the right fit. Um, and. Yeah, I didn't, uh, my, I didn't then do something quite as cool as you did.

Uh, I just applied for new programs and got rejected a lot. But, um, it does kind of put things a bit into perspective. I think, at least to me, when you, when you, you've done it and then you go out of it and then there's a sense of like, I don't know, like before this is automatic trajectory. It seemed to me where most of the people, especially the masters I was hanging out with were doing PhDs now and that kind of stuff.

And then suddenly you're sidestepping that whole trajectory. She's like, am I, am I gonna continue doing this or not? I mean, I applied and I did, but uh, it is, it is kind of interesting experience.

Arne Ekstrom: I completely agree. I mean, I think, when I was growing up, I don't know how this was for you. Failure was really not an option. Trajectories were thought of in a very linear way. You chose a career and you succeeded at it. And the idea of failing at something was not something that we [00:12:00] really thought of as having any value at all.

And I think now there's a literature that failure can actually teach you a lot and, you know, you can, you can gain valuable knowledge. And when I ended up dropping out of graduate school, um, you know, I think it caused me to really reassess what I wanted to do and think of all the different options of what I could do and think, you know, well, what would I do differently if I were to do this again?

And I ended up going to a conference in neuroscience, and at that point I was pretty confident I was going to pursue a different career. And I ran into Mike Kahana, who'd been my advisor at Brandeis University. And we ended up talking about neuroscience, and he said, you shouldn't give up. And I had this weird kind of feeling that I would compare to falling in love again, where I said, I'll give this a shot and we'll see how it goes.

And if it's not for me, I'll do something else with my life. And at every step, I just got excited by the science again and talking about the ideas and thinking about the ways that we could understand the brain in ways that we [00:13:00] hadn't already. And Mike made me a really, uh, an offer that, you know, I could, I felt he couldn't refuse, which is if you join, join the lab and do a PhD with me at Brandeis.

You can fly out to UCLA and collect single neuron data in humans with its up freed. And I'd really wanted to do single neuron recordings in humans all along, and I saw rats as kind of like a way to learn it better, but I didn't realize that rat research, for a variety of reasons, was just not going to work for me.

And I've since come to realize I think I have a mild phobia of rats, having had to deal with them in other contexts in my house. Um, so, I started flying out to UCLA and collecting single neuron data, and

Benjamin James Kuper-Smith: So can I just briefly interrupt you there, was that already set up, that whole thing? Or like, it, it seems like a bit of a. You know, because, uh, especially, uh, Boston and Los Angeles, I'm, I'm not a U. S. expert, but they're not exactly close. Um, how was, yeah, yeah, I mean, how was that, uh, it seems like a, uh, from my perspective, a slightly random [00:14:00] suggestion that is obviously great, but like, how did, how did they even start that?

Or, or why was Itzhak Fried now interested in doing this? So, yeah. But,

Arne Ekstrom: When I'd applied to work in his lab in 1997, he really was just getting set up. And I'd had the good fortune that when I started the collaboration with, with him via Mike Kahana, and this was in approximately, um, 2000, uh, I think it was Iran. Let's see if I can get the dates right here.

It would have been about, uh, 2000, 2001. He'd already built up the infrastructure to a much greater extent and published some papers successfully. People like Gabriel Kreiman at Caltech and others. So in some ways, they'd already kind of built up what would be needed to do the single neuron recordings in humans.

And I'd been working with rat data for the last several years. And even though I hated working with rats, and me and rats just didn't get along, I loved working with the data, and I'd learned many of the techniques for programming and analyzing the data that were very helpful to human [00:15:00] neurons, because human neurons are not that different from, at the level of action potentials and things like that, from what you see in rats.

And I remember when I saw my first, uh, Human single neuron that was being recorded. I was just blown away. I mean, I just thought we're seeing the language with which the brain speaks, right? I mean, it's like understanding this hidden language that, you know, like, you, you hear someone speaking a foreign language.

You're like, what is that? That's so fascinating. And that's, that's what happened to me when I saw and heard these neurons. It was just really blown away with them and just fascinated by them and just wanted to do it all the time. And it's like became its own kind of addiction in a way. It was just.

really taken by it. And so, I've been recording place cells at University of Arizona with Bruce McNaughton, and um, that was something that then I knew the techniques for doing it, and we were able to apply it to the human data. But we discovered one of the complicated things about the human brain is that the neurons are just less selective than what you typically find in the [00:16:00] rat.

And in rats, it's worth keeping in mind that their hippocampus is proportionally a bigger part of their brain. They have a less developed neocortex. And so the hippocampus is a relatively easy place to target in rats. And it's also worth remembering with rat recordings, you drive the electrodes. So if you place the electrodes after surgery and they're not in an area that you like, Um, you can keep moving the electrodes until you get responsive neurons.

And although you shouldn't do this, it's open to the possibility that you can select the neurons that are responsive. In a task that you want, uh, I'm not saying that people do that, but it's, you know, you always want the most responsive neurons. And so when we start recording from humans, the electrodes are placed during surgery and then they cannot be moved, at least currently, because they're for, they're there for clinical reasons, they're there to treat epilepsy.

And once you start moving the neurons, I'm sorry, once you start moving the electrodes, you start damaging the brain. And you start, you know, potentially changing things in ways that [00:17:00] are not. Appropriate for treating epilepsy, right? So the, the electrodes were where they were put by the surgeon and we had to make the best of those.

And what we found was that there were neurons that responded at locations, but they were relatively weak compared to what we found in rats and we found neurons that fired that were responsive to many other variables in the task. Including goal, and then what the patient was looking at. And then most interestingly, conjunctions of these things.

So goal and place, goal and view, place and view, combinations of these things. And that was not unprecedented in the literature. Howard Eichenbaum had had a long history of reporting these types of conjunctive cells. For example, odor and place cells. Cells that fired when an odor, um, was present. The rat was sniffing an odor.

but only at a certain location when the rat was exploring. So when we found these cells, we're really excited because they both fit with the idea that they were place cells in the human brain, although they were weaker than what was [00:18:00] observed in rats. Um, however, we also found these conjunctive cells that suggested a code that was more complicated than one set up just for navigation.

And I think the biggest obstacle we faced initially was The responses didn't look like place cells in rats, and so the response we got from the community was a little mixed. One of the responses was, these don't look like place cells that we see in rats, so they can't possibly be place cells, right? And statistically, we went through a number of different steps to try to verify that these were, in fact, place responsive neurons.

We used some of the same permutation testing techniques. that is used in the rodent world for confirming that it fires selectively at a location, one location and not others. We used de novas and things like that. We used a bunch of different statistical techniques to confirm that the cells we were getting were in fact place cells and it ended up being about, um, 20 percent or so of neurons that we would say were legitimate place cells that were not confounded by other [00:19:00] variables like view.

So that, that was one piece of it that ended up being complicated. And then from the other end of things, you know, in the human world at that time, fMRI was the most widely used technique. Um, you know, single neurons and local field potential recordings were nowhere at the level that we currently have them used in the field.

And most of the techniques that were being used with fMRI were these relatively simple stimulus types of presentation paradigms where a person saw a scene. And then had to respond, did you see the scene before or not? Or is this a house or a scene, or is this a face or a scene? And so our task was much more complicated in that it involved freely navigating in virtual reality.

And so we had to go through additional steps to figure out what were people looking at, what were people doing? What was the behavioral demand that they were responding to? So I think from the human end. It was more difficult for us to so called isolate the cognitive processes that were occurring in navigation.

Although it is worth mentioning there had been [00:20:00] some other studies in virtual reality that had been done around that time as well. Sorry, long winded answer. Feel free to

Benjamin James Kuper-Smith: yeah, no, I mean, you basically hit on most of the points I wanted to address anyway, so that's great. Uh, I'll just try and see which order makes most sense. Um, maybe let's start with the last thing you mentioned, the VR thing, just because I'm, I'm curious. I mean, I've I've actually taken part in some of those studies a while back.

But that was still like, you know, 15 years after you did your study. And I remember, you know, in 2000 or whenever when I was still a little boy and playing computer games. The graphics weren't exactly amazing and those were, you know, highly produced computer games. I'm just curious kind of what was there?

How did you go about? Um, designing the tasks so it works, uh, and I don't know, also were the patients, did you have to do it like in the operating theater or could they, yeah, basically how was that whole setup and how did, how did that work?[00:21:00] 

Arne Ekstrom: That's a great question. So Itzhak Fried already had a recording set up. He was moving over to a system called Neuralinks, which is fairly widely used with humans to this day. And they had been using that, um, just starting to use that when I got there. But they'd already successfully done some recordings of people looking at faces, people looking at scenes, and people looking at houses.

That was some of Gabriel Kreiman's work. So they had the basic recording set up going. And Mike Kahana had had several programmers and, um, undergraduates helping to code in virtual reality. One of those people being Aaron Newman, who's now faculty at Ohio State. And Aaron and others, Jeremy Kaplan, who was a graduate student at the time as well, had built something called Yellow Cab.

And it was basically a virtual reality game where you could navigate around, you look for passengers, and you took those passengers to stores. And we were able to take that game, and then Jeremy had built in some of the capacity to [00:22:00] interface with these cellular recordings, uh, where you have a device that plugs into the laptop, and then that sends pulses to the recording device.

Um, so I mean, I think I was lucky in many ways that the basic virtual reality code had already been written by people in Mike Kahana's lab, and it had already been tested to some extent. Behaviorally, and it showed that people were able to learn things about the locations of the stores, and they got better.

The more they navigated, um, and then on the other side of things, it sucks technicians had already set up the recordings to do single neurons, um, and local field potential. So, that infrastructure, Eric, thank you, Tony fields, many others had already kind of had that place of working. So I was, I was lucky to kind of be able to.

Piggyback on that as well. And so what I had to do as far as technical setup was not too much. Um, the patient basically had electrodes implanted in their brain. They were recovering. from surgery. And at some point, you know, we approached them and said, Hey, would you be interested in [00:23:00] doing these cognitive tests?

And at the time it's, I had several other tests that he was also having the patient do some of which involved looking at faces and scenes. Um, you know, some of which involved, um, you know, other memory types of tests. Um, there were a variety of different ones and it turned out virtual reality was something that they Many patients enjoy doing because they had to be on the epilepsy ward for, you know, in some cases a week, a week and a half.

And so they were looking for distractions. Um, and so navigating virtual reality ended up being something that many of them enjoyed. And, uh, we were able to get a lot of data very quickly, um, from patients doing this task. So basically I was, I was lucky in that I flew out, you know, I collected about 10 patients of data and then was able to start analyzing it.

Pretty quickly.

Benjamin James Kuper-Smith: It's kind of lucky you didn't, that it's actually, it wasn't initially taking you on or anyone as a graduate student because I guess by the time you arrived then a lot of the initial technical stuff was already [00:24:00] solved.

Arne Ekstrom: I mean, I think when when one looks

Benjamin James Kuper-Smith: I mean, unless you would have really enjoyed that, I guess, then.

Arne Ekstrom: Yeah, no, I, I think when, when one looks back at one's life, you think there's periods you were lucky and, um, I think I was lucky in that the right things came together starting in about 2001, uh, 2000 to allow these recordings to happen. So, I, I feel, I feel privileged that, that things just happened the way they did and you're right, if I had arrived in 1997, maybe I would have hit my head against the wall trying to get all the recordings going and maybe I would have thought differently.

Benjamin James Kuper-Smith: Yeah. Uh, kind of a fairly general question I had was kind of, what was the status of spatial navigation at the time? Because, I mean, this was before the discovery of, of grid cells. This was, um, you, you already mentioned the, the conjunction cells in, well, this was also before, you know, most of the stuff, spatial navigation had been done in humans, before any of the, uh, fMRI studies like Costantola's paradigm or before all of that.

Uh, so I'm curious kind of what was the Yeah, in some sense you already alluded to maybe [00:25:00] with the conjunction cells and that kind of stuff But kind of I'm just curious like at the time kind of what was What was the executation of what you'd find and what did you then actually add to that literature?

Arne Ekstrom: I would say. There were two different lines of research that were really dominant. There was work in rats, and that had really been kind of capped by the discovery of the place cell in 1971 by Kiefen Dostrovsky, and then of course, Lynn Nadel and John O'Keefe's book, uh, The Hippocampus and Cognitive Map in the late 70s.

And so the place cell was really you know, a highly replicated phenomena and had been strongly linked to the idea of a rat's position. And not only their position, but their allocentric position, in other words, how they knew where they were relative to everything else in the world. And so there were, I think, very strong ideas about that based on the hippocampus as a cognitive map.

And so at that time, no one had shown place cells in humans. It was considered kind of like It was unknown whether we even had him and I think there was a strong expectation that we'd find [00:26:00] something like that because at the time people were very strongly influenced by this idea that place else allowed us to know where we are right now on the other.

And of the argument in rats, there was Howard Eichenbaum, who is very strongly arguing no, no, no, the hippocampus is a general memory system. We know this from work in humans, uh, Brenda Milner's work showing that damage to the hippocampus in, for example, the famous patient H. M., produces dense amnesia. And Howard said, no, no, no, the hippocampus is a Memory system, and part of the memory system is gluing together different associations, different elements of experience that may not necessarily belong glued together.

And, you know, that's the basis of episodic memory in many ways, right? The way that you remember something from a couple days ago is sometimes that bizarre thing that stands out that provides the piece that helps you get everything else from that experience. So Howard was really arguing, no, no, no. Place cells are just one piece of a [00:27:00] much larger picture here to a conjunctive episodic memory system.

Um, so those were two debates that were going on that we were very aware of and very interested in. And then on the other end, there was fMRI work, mostly with scenes. So Russell Epstein's work, Eleanor McGuire had also done some work with taxi drivers that we were aware of. And so there'd also been work that was suggesting in some form or another that either the hippocampal gyrus Or the hippocampus was also specialized for navigation, but in humans.

So there, there was these kind of different lines of work, which didn't connect very well, and we were really interested to try to understand, okay, let's take an unbiased look at what's going on with these cells and just see what we see. And we did find evidence for, we, we went into it saying, if we don't find evidence for place cells, we're going to publish that because that's a big deal.

But if we do find evidence for place cells, we'll definitely publish that because that, that's clearly of interest to the field. And we ended up getting this kind of. interesting story that supported [00:28:00] the idea that their landmark responsive neurons impair hippocampal gyrus. That fit with some of the scene processing evidence from fMRI, but that there was play cells and these conjunctive cells in the hippocampus, and that sort of bridged between some of the arguments John O'Keefe was making and Howard Eichenbaum had been making.

So, I mean, in an interesting way, I guess the results could have been seen as making all parties happy. Um, I don't know if that's really true, but, uh, I do think sometimes that's the reason why a paper could be published. Um, fast forwarding to much further my career, I found the more controversial a finding that I'm trying to publish.

The less, like, less likely it's to get published because, you know, we all have our own biases that we bring to the review process. So I, again, I think we were lucky in terms of. The findings, of course, we didn't know ahead of time what they would be. And so

Benjamin James Kuper-Smith: Yeah, yeah, so I mean I guess the the lucky thing was that you you You wrote a paper that made everyone happy? Um [00:29:00] Yeah.

Arne Ekstrom: with it. I think the one response that we got that suggested that people were not happy was, um, from some people that do rodent recordings that felt like the responses didn't look like rats. So they couldn't be place cells. You know, it's, it's a little bit like, looks like a duck quacks, like a duck argument.

That doesn't look like a duck. Can't be a duck, right? Um, but we had all the statistical evidence, um, and we'd recorded, um, you know, several hundred cells to say, look, statistically these things are there, you know, look at the data, even if the response doesn't look as dramatic as, you know, what, what you might be used to seeing in rats.

So I think that was, that was the pushback that we really got on the paper, but I think the theoretical advance was not strongly disputed from my perspective.

Benjamin James Kuper-Smith: Yeah, I have a kind of general question about doing science or getting data that has a natural limit to it. In the sense when you have an unusual population usually let's say some sort of clinical population Um, I mean, for example, i'm assuming well I don't know But my assumption is [00:30:00] that you were also quite limited in terms of how many people you could get I mean you had I think seven people in the published paper, uh, I don't know how many you tried to record and then had to stop because of whatever.

Um, but I'm just curious to kind of, um, I mean, I also once published a paper with, with people in prison, uh, well recently, and it's, a lot of the criticisms you get often like, It's a fair criticism, but like, what are you gonna do? Like, you just can't really do much about that. So I'm, I'm kind of curious, especially when it, you know, in this case, it looks different than the Roland stuff, but not completely different, and I don't know, how do you kind of navigate this kind of situation where it's, it's just never gonna be as clean, maybe, as it might be in, in the context you're familiar with originally.

Arne Ekstrom: Yeah. I mean, first to the small sample size point. I mean, I think at the time, many fMRI studies just included 10 to 12 subjects, and it was understood that ECOG recordings were very difficult. So that, you know, [00:31:00] six, six patients was seen as pretty good, and we did collect more than that ultimately, but some of that ended up in the dissertation and not in the paper itself.

And I think you make the best of what you have, With data, um, as far as, you know, trying to clean the noise, dropping bad data sets, like if the patient doesn't complete or you see a lot of noise in the data, and then statistics, right? So having the statistics guide the conclusions as much as possible and trying to avoid wishful thinking too with statistics when you can.

So I think, I think those are really good points. Um, I think now the standards have definitely increased for sample size in the typical intracranial study. But the number of patients at many of these larger city hospitals has also dramatically increased that are typically tested in a year. Yeah,

Benjamin James Kuper-Smith: Oh, is that just because of the techniques are there just more people trained to do it or it's it's we know how to do it or likewise, or just more people with epilepsy.

Arne Ekstrom: I mean, I think it's it's a combination [00:32:00] of many factors. I mean, I think, um, there's been a movement toward two types of recording methods. Now, the invasive techniques that we used at the time, and I think those have gone in several different directions with grid strips, stereo E. G. Um, and then classic depth electrodes.

Um, and so we were the classic depth electrodes with these micro wires added on top. Um, Now, there are many different ways of doing invasive recordings, and there's also a new technique which didn't exist back in the 90s and 2000s, which is called Neuropace. They're wireless, chronically implanted devices.

They don't allow single neuron recording. So that's something that Nantes Suthana UCLA has worked a lot with. We've worked with a little bit. And those, in many ways, were, you know, a game changer as well because you could still record from the human brain, but these were not people who were on painkillers.

Um, they were not in a hospital bed. They were free to move around. Um, the recording quality of that's lower than the wired recording still. But, um, yeah, so I think many things have changed in terms of [00:33:00] access to invasive recordings, um, that were not true back then. And I also think hospitals have invested more in epilepsy surgery in a way that was less true.

at the time. So I think the most common technique still for treating epilepsy is drugs. Um, and you know, if you didn't, you know, if they didn't think you were a great candidate for surgery, then you would just continue on anti seizure drugs, right? But that often wasn't a great outcome for the patients. So now I think there's a lot more stratification and a lot more kind of options in terms of what a patient can do and some percent of that includes these invasive recordings.

Benjamin James Kuper-Smith: Yeah, if anyone listening, you mentioned Nancy Suthano, with whom I've had an entire episode. So, anyone interested in, we talked, we talked especially about like, the current status of invasive recordings, that kind of stuff. So anyone interested in that, I'll put a description to that in the description or the show notes.

I guess that's what you can put it. Um, yeah, I mean, [00:34:00] so one, um, uh, kind of sticking with the patients a little bit here. I mean, so one, uh, you mentioned, you know, you're restricted to what is clinically sensible, uh, in terms of well, obviously who you record from, but also, uh, within the patient where you record from.

Is that why you, uh, one thing that surprised me was that I think the majority, well, the majority, but the, the area you recorded from with the most neurons was the amygdala, which is not typically where I would start if looking for spatial navigation. Was that just because you had the patients or was it supposed to be control or were you actually interested in something else also?

Why are you the amygdala?

Arne Ekstrom: you place the electrodes at, at the time in the brain region you were interested in, right? So the vast majority of recordings at the time were in the rodent hippocampus And [00:35:00] then some people had started to record like Lauren Frank in entorhinal cortex But that didn't really come into the mainstream until kind of the middle You know 2007 or so when the Mosiers really started looking at entorhinal cortex So the vast majority of recordings were in hippocampus with not a lot outside of that in rats.

Now, for seizure monitoring for these patients, the idea is that they're testing multiple ideas about where the seizure focus is. All they can see at the level of scalp EEG and maybe MEG if they do it, is that they're some source somewhere deep in the brain. And so in order to figure out where that's coming from, you put electrodes in multiple areas.

Including anterior cingulate amygdala, hippocampus, and rhomboid cortex. And it varies by patient in terms of where they think the seizures might be coming from. And they almost always implant bilaterally in healthy tissue as well, because the idea is if we want to see abnormal seizure activity, we have to have a comparison with what healthy tissue looks like.

And so that's why [00:36:00] there were recordings in so many different areas that these were clinically mandated and the electrodes were there anyway. And Mike and I felt really strongly that we should look at all the different brain areas, and not just restrict yourselves to the hippocampus, because we thought that that would produce its own bias.

And I think that's a real strength of the human recordings, is that it gives you less of a biased look at what's going on in the brain. You don't get as many choices, and therefore it removes some of the experimental bias. So with regard to the amygdala, you know, we were there to some extent because that's where the electrodes were, were put, but we were also interested to see what we would find.

And Mike Ahana later published a paper where they reported place cells in the amygdala. Um, and that kind of segues to a topic that I'll get into I assume we have time a little bit later, but we observed place cells in several different areas. Now, we weren't sure if those were above chance or not, um, but they were not just present in the hippocampus.

We saw [00:37:00] examples of place view type of cells, other types of conjunctive with place cells. in areas where we didn't think classically they would have been expected to be, uh, areas like anterior, anterior cingulate as well. And the presence of place cells outside of the hippocampus suggested that the hippocampus was not an organ that was doing something really radically different than other areas.

And the more we started to do recordings, and this is a common report in single cell recordings from humans, you see responses to face to scenes. other things like that in many different brain areas. And it suggested that brain areas were probably pluripotent. They were capable of many different coding schemes.

And we just happened to be asking a single brain region a question, revealed kind of the answers that we already thought we would get. And so with fMRI, we'd already started to get suggestions that different brain regions might were capable of [00:38:00] many different cognitive functions. And although there was a localizationist tendency to say this brain region does this function because I've, you know, used some control that doesn't activate it in this situation, I think it also became more and more clear and is clear now that most brain regions are pluripotent.

In other words, you can find place cells, and this is now, we now know this is true in rats, you can find place cells in areas like the claustrum, retrosplenial cortex, prefrontal cortex, and now with fMRI, depending on what you think of grid coding in fMRI, there's evidence for grid coding in areas outside of entorhinal cortex as well.

And so From the human recordings I did back in 2003, I'd had the early evidence that brain regions were pluripotent. There were cellular responses that shouldn't be in places like amygdala yet they were there. And so that inspired me later, and that kind of moves us forward. Slowly into the Charan Ranganath paper and some other papers to think that the hippocampus is not a unique [00:39:00] player in in memory

Benjamin James Kuper-Smith: One of the questions I wrote down was, you do have, I think, 5 percent or something of your amygdala cells also were, in this sense, place cells. Um, I guess you kind of already talked about part of that. Um, but I mean, so I mean, you mentioned that, you know, different areas can do different things. Different, the same thing and the same area can do different things.

Um, but there still does seem to be some sort of bias, right? I mean, you do still find more places cells in hippocampus than in amygdala. So like, yeah. I mean, this is a very vague question, but what does that mean? What, what

Arne Ekstrom: I I completely agree. I mean I I should mention you were correct about the the 5 percent You know, there was 5 percent or less and in the amygdala That said, I think some of that's going to be dependent on the task, um, and that we found other types of cells like place view conjunctive cells, not the so called pure place cells, but these ones that had some place component to them.

[00:40:00] but had other correlates, those were present in many different brain areas. And so your point's well taken that the so called pure place cells were primarily in the hippocampus in that study. Um, but I think now we have enough evidence from recordings in areas like retrosplenial cortex, and this would primarily be in rats and non human primates, um, areas like prefrontal cortex, that there are so called pure place cells, uh, you know, mostly responding to location that we can find in areas outside of the hippocampus.

And I think, you know, if you take it through the lens of a memory, hippocampus is a memory structure, we know that space is a powerful component of memories and, and retrieving memories. So to me, the presence of place cells there would, would be consistent with that idea, but the pluripotency of different brain areas would suggest that, for example, in the case of brain injury, There are other brain areas that have some of the neural machinery capable of compensation.

Not to the same degree, but um, you know, some of the machinery that would be needed for kind of picking up some of the slack.

Benjamin James Kuper-Smith: Yeah. [00:41:00] Um. Yeah, on that note, I did an episode with Nikolaj Axmocher and I really like kind of their, what is it, science paper about, with people who have the genetic predisposition to Alzheimer's. Um, and I really like there you have this thing where they, you know, have the same performance, like they have no deficits in that sense in their performance, but you can, uh, what is it?

You can see some sort of behavioral and some neural compensation around the people who don't have Alzheimer's yet, but are predisposed to get it. Um, so if anyone's

Arne Ekstrom: That's, that is the argument.

Benjamin James Kuper-Smith: there's an episode around that kind of theme there too, I'll add that link. Um, yeah, we want to talk in part about like all this stuff in a slightly broader context, Hippocampus, memory and spatial navigation.

Um, yeah, kind of where, where did you go kind of after that study? Yep.

Arne Ekstrom: observed in rats when they move. That had been reported by Cornelius Vanderwolf back in the late 60s. [00:42:00] And so that was another component of my dissertation, to report these Changes in low frequency oscillations during navigation as humans navigated in VR.

And we had those data from the local field potential recordings that we got on a different recording system, the clinical recording system, so that was a separate paper. But as I mentioned, we were also very interested in these fMRI findings, particularly these um, scene responsive areas like the pericampal gyrus.

And we had recorded these landmark responsive cells in the pair hippocampal gyrus. Um, so we were very interested in, you know, what does fMRI mean? And at the time there'd been sort of the debate about how can we, identify place cells with fMRI. There'd been some jokes about these things called place voxels.

Um, and Demi Hassabis before he founded Google DeepMind actually has a paper where they reported some version of place voxels. Although I should mention there's some statistical issues that have been identified by an Australian group with those findings. So I'm not sure how robust those findings are, but we were very interested in what [00:43:00] could fMRI reveal that was either similar or different.

In the brain. And it was very clear to me from doing these recordings in humans that there was a tremendous amount of information that there was no way you could see with fMRI. But at the same time, it'd been very difficult to demonstrate that experimentally. Um, and Nikos Logothetis had a really unique setup at that time in Tubingen, um, and published a paper around 2001 where he recorded simultaneously fMRI local field potentials and single neurons.

This was an anesthetized monkeys in visual cortex. We were very interested in those results because they provided what you might call Rosetta Stone, a way of translating between fMRI single neurons and local field potentials. But at the same time, We thought Logothetis's results seemed a little simplistic, right?

These were anesthetized monkeys after all. It was visual cortex, which is not the same thing as an area like hippocampus. It's a unimodal, uh, [00:44:00] sensory area rather than a multimodal sensory area. It's anatomy is really quite different. Um, so for my postdoc, I ended up working with Itzhak Fried and Susan Buchheimer, who's done a lot of work in fMRI.

She's one of the early innovators, innovators in that area. And so we ended up doing a project where we ran patients first In a version of the virtual navigation task using fMRI and then recorded single neurons and local field potentials when they did that task. And we added in some kind of more simple stimulus types of presentation stuff at the end to tax things like spatial retrieval, because that would be more akin to what you typically do in the scanner.

And that ended up being a really complicated, unsatisfying project, because basically what we found was the answer that no one really wanted, that the areas where we saw activation, um, in the patients. prior to undergoing implantation didn't match up very well with what we saw with single neurons or the local field potential.

And really what we found was weak evidence for a correlation in some of the surrounding cortical [00:45:00] areas for low frequency oscillation. power increasing with the bold signal. And that was really all we found. Um, so if anything, we would conclude that single neurons don't relate well to fMRI, at least in the medial temporal lobe, and there's a weak correlation with the local field potential.

And we submitted that paper at Nature. It got reviewed and rejected and, you know, just, you know, reviewers didn't like it because Logothetis had already solved the problem, right? We already know that the bold signal tells us mostly about gamma oscillations in the local field potential, but still a little bit about single neurons.

We already know that, so. Where are you trying to tell us something different? And even if you are, Logothetist could do it better, because he could do it simultaneously. So there just wasn't a lot of enthusiasm, and it was like the result that no one wanted, which is fMRI is really difficult to interpret, and it depends on what brain region you're in, what response you're going to get.

And one thing I've learned in cognitive neuroscience is, People don't tend to like complicated answers. The simple answers, for example, um, one you [00:46:00] mentioned, that degradation of grid cells precedes Alzheimer's disease by several decades, and that accounts for spatial disorientation being the primary deficit with early Alzheimer's disease.

That, that's a story many people like, but is very simplistic and probably in some ways wrong, right? I won't go into it, but we, we encountered a similar problem with this problem. The bold signal is not always reflected in electrophysiological activity, it's complicated, right? And I think it was one of these examples where the story was sort of, it didn't fit with the zeitgeist and we didn't exactly know what to do with it either.

And so I ended up after my postdoc taking a faculty position at UC Davis where I didn't end up doing a lot with that line of research, but I continued invasive recordings and then spent a lot more time working with fMRI and then invested a lot more time in studies related to fMRI. And. You know, is it okay just to kind of segue into that?

Benjamin James Kuper-Smith: Yep.

Arne Ekstrom: Okay. Um, [00:47:00] Yeah, so we started doing some studies where we had people navigating in, um, spatial environments and we had them learning different types of information. So we either had them learn things relative to landmarks and have to retrieve relative to landmarks, which is what we would call allocentric navigation, or they had to learn new types of contingencies that changed, which would be an example of something that's more flexible, right?

You have to remember specific instances and then change them around a little bit. And we had a strong prediction based on almost everything that had been published in fMRI and what we thought we knew from rats, that the hippocampus should only be active in allocentric conditions. And we didn't find that.

In fact, we had trouble finding any hippocampal activation related to navigation. And naively I submitted this as a paper to be published saying the hippocampus is less involved in navigation than you think. And it was just panned, rejected, and you know, I, I learned again this lesson, things that don't kind of fit with [00:48:00] hippocampus.

thinking at the time are just more difficult to publish for, for sometimes good and bad reasons. And that started us on a course of trying to figure out what does the hippocampus activate for, you know, what, what activates it with fMRI and then what other brain regions could be more relevant to spatial navigation.

And we started to get evidence from a number of different papers that the retrosplenial cortex. It was more consistently activated for a number of different spatial tasks, including the allocentric tasks, than the hippocampus, and the hippocampus was more active in a way that was consistent with a memory structure rather than a navigation organ.

Um, so that kind of led us on, on that course for several years at UC Davis, and I, I got more involved in thinking about the hippocampus as an episodic memory structure, and I was inspired by the work of people like John Ranganath and Andy Onalinas, who'd been spending a lot more time on that topic than I had before I got to Davis.

And so that's how we got into the review paper with me and Charan. I don't know if you've had a chance to have a podcast with Charan. [00:49:00] You definitely should. He's a really dynamic, fascinating personality to talk to. And he's one of those people that is just so brilliant. He can throw an idea together in one minute that will give you miles for years as far as it being interesting and worth pursuing.

And that's basically what the review paper was. Charan had this idea that we'd been sort of discussing and throwing around. The hippocampus might be involved. to some extent in navigation, but more so in memory, because both space and time are critical axes by which the hippocampus tends to encode stuff.

And if you think of many aspects of episodic memory, they typically involve a place and a time. So the idea we had in the hippocampuses all over the cognitive map was essentially a way of trying to reconcile a perspective that said, why would it be active in some instances during navigation, but why would it be more important as a memory structure?

And This paper really put forward the idea of this four dimensional axis of space and time being critical dimensions [00:50:00] of hippocampus, and that would be where you would see many patterns of activation and deficits in patients with hippocampal damage. But it also started to advance the idea that there are other brain regions that might be more important for navigation than the hippocampus.

And it started to kind of touch on these ideas of graph theory and that instead of thinking about one brain region having one function, there might be, um, this pluripotency of brain regions that could kind of pick up the slack.

Benjamin James Kuper-Smith: Yeah. I mean, in that review, I thought you had like this, uh, header that I thought was interesting, which was, um, what is the difference between spatial and non spatial? And, um, yeah, there was one part that I, I'd never quite thought about like that, which is kind of, you know, In the sense, like, how do we even know about a perceived space?

And it's kind of interesting. Maybe, uh, I'll give you a very example of, of this. Maybe you can kind of use, uh, explain the difference there. [00:51:00] Um, for example, like, where am I right now? So I know I'm in, well, by coincidence, the room I grew up in. I'm at my parents place right now. So I know, you know, there's, there's all these things around me that I can see.

And, uh, objects around me and all that kind of stuff. But I also know that I'm in Germany, uh, and in my particular case, in a very unusual place, it's the German exclave to Belgium. And if you're interested in geopolitics, you can look up exactly where that is and why it's one of the weirdest borders in Europe.

Um, but I don't, I mean, I know, for example, Belgium is about 200 meters away from where I'm right now. But like, I don't perceive, there's, there's, in this case, there's no, obviously not a wall or a fence or anything, it's just I know it's there. So like, in this sense of like, where am I right now, there are these two very different ways in terms of how I know where I am.

Um, so can you maybe pull that a little bit apart? Um, about, yeah, what is spatial, what isn't spatial?

Arne Ekstrom: Yeah, I [00:52:00] mean, we were very inspired, again, by the idea of the cognitive map and thinking about it. critically. Right? Um, so the problem is exactly what you mentioned. A place cell is active at in a rat into a lesser extent in a human and a monkey at a specific spatial location. But we have, as humans have this, Capacity to think about other locations simultaneously and still know where we are, and it produces a problem for the cognitive map, and this is something that Charn pointed out and we discussed quite in depth, that when you think about another location, how does the cognitive map resolve two place cells being simultaneously active, one for where you are and one for where you're thinking about, right?

And that, that produces a problem in terms of explanation of the cognitive map. And there was. Numerous studies that we were aware of that suggested that space could be better thought of as something that has multiple dimensions to it, right? There's the very large scale space that [00:53:00] you know about from elementary school, which is the geography of Europe and the boundaries of these different countries, which have changed over the centuries in different ways.

And you have to learn about all that stuff and learn about where that fits. But you're very unlikely to walk from where you are now, uh, to, you know, the coast of Belgium, right? That'd be quite a long walk. Yet, you know things about the geographical properties of these, you know, different countries and locations.

And so, it's pretty clear that we can think of space at many different levels, right? There's the laptop immediately in front of you, we call that peripersonal space, knowing how to reach out to, you know, keys on your keyboard. Then there's the navigable space around you, um, say your room. Or, uh, you know, the city that you're around, um, those are a little different, too, in certain ways as well.

Um, then there's geographical space. This is an old argument that was made by Dan Montello, um, back in the early 90s. And I think it's something we explored a little bit there as well, that we tend to forget about this thing called scales of space. And that [00:54:00] there are multiple different ways we can think about space.

even simultaneously, right? Um, you can even imagine yourself what it would be like in Tucson, uh, you know, based on some of the things you've seen about Arizona and, um, you know, the room, room behind you, but that doesn't mean you've actually been here, right? And so that was one of the things we were trying to explore in this paper, that the cognitive map is ultimately very limited as far as its ability to explain our ability to simultaneously imagine and know our position in multiple different places.

Benjamin James Kuper-Smith: I think this will be the last time I reference an old episode I did, but I recently did one with Nahum Ulanovsky where we talk kind of about this like large scale and spatial navigation, how the spatial code is very different for that. Uh, so I'll link that and now I'll stop linking to old episodes. Um,

Arne Ekstrom: it's great.

Benjamin James Kuper-Smith: Yeah, maybe one, uh, one question I have, uh, which I, which I like to ask occasionally, and I think no one really has an answer to, but I still like asking it, uh, is, um, for example, you mentioned here that you have this, [00:55:00] uh, this 4D space, so three spatial dimension, one temporal.

Um, one question I always have is, do we have any evidence that a, that the hippocampus, for example, can code more than four dimensions? Or kind of what's the upper limit, like how many dimensions can we, can we do at the same time?

Arne Ekstrom: so there is pretty good evidence that we can retrieve episodic memories based on emotional valence, um, and so that episodic memories could be encoded in some dimension related to you. You know, emotion, right, uh, excitement, happiness, sadness, things like that. Um, it's very clear that in depression, rumination is, is, is an associated symptom, which involves kind of replaying things that happen, um, with sort of a negative angle to it.

Um, so we, I think, briefly touched upon that idea in that review paper, that emotion could be another dimension. And there have been some papers, including some work in rats, that suggested that you get play cell like activity just based on hearing, uh, Tones organized in some ways in, in a, in a, [00:56:00] um, tonal space.

And so we touched upon that idea, but we kind of came back to the argument that space and time would be the primary and most important axes that would typically define episodic memories. Um,

Benjamin James Kuper-Smith: Yeah, yeah, I mean, yeah, so there's, my question was, I meant it slightly differently, which is, so I know that there's lots of studies about like which things can be encoded along these lines and lots of stuff also with grid cells in terms of like different variables, but it seems to me, and this, I guess my, This might be maybe more a difficulty in testing this empirically, but I'm always curious like how many dimensions you actually can encode, independent of what variables they are at the same time.

So whether there's maybe a limit to three or four, or whether that's just, it's just difficult to test more than four, in an actual setup, or whether maybe there is some sort of limitation where like we just pull the three most relevant dimensions in any, uh, given situation and ignore everything else.

Arne Ekstrom: Yeah, I think it's a really good question. I think there's a [00:57:00] new emerging literature that would focus, that focuses on low dimensional space or Roberta Buccini has been working in this area and he and I have discussed and argued, I guess, so to speak, he would view this as a cognitive map. I would view it as something different, and I would view it as more consistent with the idea of something we've kind of been discussing.

Whenever you have to do one of these experiments, you have to figure out the relevant dimensions, and in many cases it may just be one or two dimensions. But if it's more than that, you know, five, six dimensions. We can learn that not necessarily without knowing, um, but the process is something of dimension reduction, right?

We need to figure out which things do we need to focus on here? Um, you know, is it location? Is it what we see when we're navigating? Is it the goal? The thing we need to find? And weight these different dimensions accordingly to solving the task. And those are contingencies we don't necessarily know. It's what the experimenter sets up.

So I would say, yes, I mean, there's probably an upper limit on how many dimensions we [00:58:00] could reasonably encode in any experiment. But that's essentially what we ask people to do when we have them run our experiments. Figure out these dimensions or don't get paid or get your class credit, right? Um, and so the argument Char and I were trying to make is that, Um, Uh, 3D space in one, in one dimensional time would be the most relevant dimensions to many episodic memories.

And if you were able to do the principle components dimension reduction, that that would typically be one of the things that would stand out. But things like emotion and other aspects could be there as well to, depending on what subjects you're being asked to do in the experiment.

Benjamin James Kuper-Smith: Um, is there anything you definitely want to talk about or?

Arne Ekstrom: Yes. Uh, thank you. I mean, I, I figure I'll, I'll fast forward a little bit more

Benjamin James Kuper-Smith: Yeah, I mean, I would also

Arne Ekstrom: our other

Benjamin James Kuper-Smith: in general, get curious kind of what you've been doing more recently and kind of what you, what you want to do in the future, maybe.

Arne Ekstrom: Yeah. So. This idea of pluripotency of brain regions has become really driving force between, behind some of our more recent research. [00:59:00] You know, at the time when we did some of the work in the 90s, we had assumed that the patient H. M., the famous amnesia patient from the 50s, who had a bilateral medial temporal lobe lesion and dense amnesia, that he was in some ways like a pure hippocampal lesion, even though at the time we knew that wasn't completely true, that he had some thalamic damage.

that probably most of his damage was to anterior hippocampus and not posterior hippocampus. It wasn't until some of the post mortem data came out after he passed away that, um, Sue Corkin and many others were involved in to demonstrate he also had damage to his orbital frontal cortex and white matter tracts connecting the frontal lobe to the medial temporal lobe.

So he was certainly not a pure hippocampal lesion patient. If anything, he was a complicated lesion patient. Around 2017, uh, 2013, when I was at UC Davis, we'd started to get into the idea of connectivity and graph theory. I had a very talented graduate student. He's now a professor at Baylor College of Medicine who [01:00:00] applied graph theory to the local field potential recordings we were getting from patients and he showed that correctly remembering something involved interactions across many different electrodes, not just activity in the hippocampus.

So this started to move us in the direction of thinking of the brain as a network where any given brain region. could have many different functions it could do but it would be biased more toward one function than many other functions. Um, and that through interactions between brain regions you could get emergent new functions.

Um, and so that kind of ties in with this idea of dimensionality reduction that we're talking about. That there's an interaction between the brain and behavior that's dynamic and that's continual. And essentially there's a tuning of the brain with whatever behavior you're trying to figure out. And so I'd say that, that's been the new direction for the lab, which is not, not giving up on the idea of, What brain regions can do in terms of cellular responses, but the dynamics [01:01:00] of the brain in different tasks, and we focused a lot on patients with focal brain lesions and the idea that even after you have damage to a brain region, other brain regions with this pluripotent perspective are capable of some degree of compensation and picking up the slack for damaged brain regions.

Very much in like the way if you break your leg and have crutches, um, Your upper body and your non damaged leg will become stronger. It's, you know, a similar concept, but, but differently in the brain. And, yeah, so I'd say that, that's the direction of the lab now, is to look at things like aging, brain damage associated with stroke, and brain damage associated with surgical resection.

And try to understand compensatory mechanisms and how the brain might be doing things differently, not necessarily incorrectly.

Benjamin James Kuper-Smith: yeah, okay. Are you still doing singletary recordings? Or is that just something from the past?

Arne Ekstrom: We have a collaboration with Brad Legge at University of Texas [01:02:00] Southwestern where we are planning to do those, but um, have not yet started.

Benjamin James Kuper-Smith: So one thing I always wonder about is like how many methods one should do. And it sounds to me like you're more towards the size of do all of them or lots of them. Uh, is that what would you, I mean, I'm, I'm always cautious with that because I feel like it's, it's hard enough to understand one of those things.

Um, yeah, maybe why are you, why are you comfortable or maybe you're comfortable, why are you doing so many and. Uh, not saying, I'm just going to do this or that, or two, but not five.

Arne Ekstrom: It's a great question. Every method has its limitations, right? So every method has questions that can be answered and questions it's not very good for answering. So fMRI is, in my opinion, not very good for answering, um, what's going on at the cellular level, um, what's going on even at the local level, um, because it just doesn't really have the spatial precision for that.

So we use fMRI to answer [01:03:00] questions more broadly about, you know, what a brain region generally might be responding to, but in particular, how networks of brain regions may be connected. In contrast, intracranial EEG gives you these much more precise local field potentials and cellular responses. So you get a little bit more insight into what a small subset of that brain region might be doing in very high resolution time, very, very precisely in time.

And so we've been using the recordings to address questions more related to how a brain region might be active. You know, differently for one type of experiment versus another, one type of condition versus another, um, you know, on the scale of milliseconds. Now, brain lesions, the reason we started working with that in conjunction with Andy Onalinas when we were at UC Davis, They tell you about the necessity of a brain region, and one of the things that had always bothered me with fMRI is it is fundamentally correlational, and there's actually several examples of, um, very strong fMRI responses in a brain area, or even single neuron responses, but when you take out that brain region, it has no effect on the behavior whatsoever, right?[01:04:00] 

And so that cued us to think about looking at patients with focal brain damage, um, to try to get a better handle on the necessity of a brain region for a function. So that's how we got into that line of research and scalp EEG is one of these techniques which has very poor spatial resolution But is really the only technique where you can get this time resolved neural activity outside of patients, right?

So we could do this with healthy subjects So that's how we got into that but we tailor every question toward the strengths and limitations of the method

Benjamin James Kuper-Smith: But is there, uh, I'm curious, like from a practical perspective, I mean, is there, is, uh, do you have different people in your lab? Like, is every person in your lab working on a different techno, technique or something like that? Or, uh, does that cause problems then? Or, yeah, I'm just curious kind of how, uh, Because obviously you, I get, well, maybe not obviously, but it feels to me like it's nice as a lab to have a, a shared, uh, you know, methods you can exchange questions and that kind of stuff, right?

I'm just curious [01:05:00] when, when you have four different like imaging techniques going on, or not imaging, but you know what I mean? Um, does that, how do you deal with that basically?

Arne Ekstrom: That's a great question. That gets to a larger question about groupthink in labs. And I think, you know, we like to think in science we're not like religion, we're not like cults, but every lab has groupthink. It's unavoidable. And you as a PI can have a strong influence on that in good ways and bad ways.

And I personally think groupthink could be very toxic to science. Working on multiple methods, it teaches you to think that the problem is not as simple as what your method is telling you. And in my lab, we have enough people where we usually have at least Um, two people working with a method, if, if not more than that.

Um, right now I only have one student who's working on intracranial EEG, but I'm hoping to have more in the future. And in general, students come in, you know, with some, some background, uh, in an area or, you know, willingness to [01:06:00] learn, um, that they can usually get up to speed on it. But I generally, I, I find that having multiple methods in the lab is a strength because it teaches people to appreciate the complexity of a problem and to avoid what I think is just.

We often kind of fall into in science, just thinking in terms of simple answers to complicated problems. And if we only did fMRI, I think we, it would produce a sort of echo chamber of way of thinking about problems that would ultimately not be productive to solving those problems. Um, I guess I would also mention, I don't know how it is in Germany, but our funding agency here, the National Institutes of Health, really wants people to avoid just using one, technique.

They want insight from multiple different ways of looking at a problem, because that, that improves generalizability, right? And so we, we're very much in tune with that kind of way of thinking.

Benjamin James Kuper-Smith: I was assuming it would be like more one lab attacks it this way, the other lab attacks it that way, and then, I guess you could do it in a lab. Uh, just, uh, the, the lab, the, the, [01:07:00] The brain region necessity point you brought up earlier. Um, I'm just curious about that one because in, what is it, in two weeks actually, exactly, I'll be starting a postdoc where I'll be doing fMRI and TMS in this case to kind of get a similar question.

Just out of curiosity, um, maybe don't ruin the method for me entirely, but why are you not using TMS but instead using brain lesions?

Arne Ekstrom: Great question. We worked with invasive stimulation about 10 years ago in collaboration with a neurosurgeon, Nitin Tandon, who we published a couple of papers with, including the Nature Neuroscience paper I mentioned, where we looked at connectivity across the brain. And so we did another project where we stimulated different brain regions and we were curious what it would do.

And I think there's been, been the idea around for a long time that if you stimulate in the right way, you can improve what a brain region does. Um, and, you know, we just always, no matter what we did, we got the opposite, that when you stimulate a brain region, you disrupted function. Um, [01:08:00] and the more I read in the literature, the more I came convinced that, That was the typical response that you disrupted because you're imposing an abnormal signal on the brain.

It's not the endogenous signal. Now, TMS has the advantage that, or disadvantage depending on how you look at it, you target superficial areas with the hope that you could get the endogenous activity into deeper brain areas. But to me, that's essentially its fundamental limitation. You can't directly target deep brain areas, so you really have no idea what's getting in there.

And if you try to use fMRI to determine that, well, as we talked about, you have this Rosetta Stone problem. You don't know exactly if an activation or a deactivation corresponds to changes in electrophysiological activity. It's very difficult to know if what you've done with TMS has had any effect on any brain structures other than the one you're directly targeting.

And the most common response that I think one would predict, and I think this is probably most robust in the TMS literature, [01:09:00] is some kind of disruption, right? You're disrupting the normal activity of a brain area. And so, I think on the other side of it, there's a strong desire to use TMS to improve cognitive function.

And I'm personally just very doubtful about whether it can do that. Um, and I think it leads down a potential path which can be dangerous, which is you try to look for improvements in behavior, but these can often be very difficult and subtle to find, um, if they're even there. So that, since you asked, that's, you got my opinion.

Benjamin James Kuper-Smith: yeah, I mean, so my assumption was, uh, I mean, I haven't read that deeply into TMS yet, but my assumption is that it's, yeah, more a disruptor than an enhancer for precisely the reason you mentioned. Uh, but I'm curious why, um, I mean, just, just from, from your personal, uh, perspective, like why go with patients?

Is it because you can get deeper regions or whatever, because, yeah, I mean, it's, it's a, it's a. [01:10:00] Um, what are the advantages, I guess, for, for your research for patients rather than

Arne Ekstrom: Ultimately, our goal as basic researchers should be to help solve neural disease like stroke and like epilepsy, right? So even though we're doing basic research and trying to reveal brain mechanisms. We ultimately want to come up with therapies, you know, either stimulation or cognitive or pharmacological that can improve the health and well being of these individuals.

And so I agree with you that TMS is a good technique if you want to perturb the healthy brain. But ultimately a patient with brain damage is unlikely to look anything like The 20 year olds who you stimulate with TMS. These are people who had strokes or hypoxia or something that damaged the brain. And there's been some period of recovery to the brain that's changed how the brain is wired fundamentally from a healthy individual.

And if our goal is to try to help these individuals in some form or [01:11:00] another, we want to try to get the best of both worlds. We want to ask basic research questions about what it means to have damage to that brain area and what the resulting behavior is like. But also, are there ways we could use that information to inspire therapies that could potentially help them recover from their brain damage?

Benjamin James Kuper-Smith: uh, I see. Yeah, that's a difference. I have no interest in that. I don't help people. I'm not as

Arne Ekstrom: Fair enough.

Benjamin James Kuper-Smith: Um, uh, so yeah, at the end of each episode, I ask the same three recurring questions. Uh, and, uh, yeah, first one is, uh, what's a book or paper you think more people should read? This can be famous, not famous, old, new.

Uh, just something you think more people should read.

Arne Ekstrom: I mean, the list is extremely long. Um,

Benjamin James Kuper-Smith: Yeah.

Arne Ekstrom: I mean, I would say from the scientific standpoint, I've been very inspired by some of the work with network mapping in lesion patients. That shows [01:12:00] that, um, patients with focal brain lesions, their cognitive deficits can better be described by patterns of white matter and or, um, functional resting state connectivity than what would be contributed by just the damage to a single brain region.

Um, so it's the idea of the brain as a network that I think is something that really inspires me and I think that some of those, those papers, some of which I could mention would, would be interesting and

Benjamin James Kuper-Smith: just, yeah, do you have a paper in particular that you think encapsulates this particularly well, or

Arne Ekstrom: Sure. Um, it's a paper, the first author is Argeopoulos, um, I don't know him personally, correspond with him a little bit. And it's a paper in 2019 called Network Wide Abnormalities Explain Memory Variability in Hippocampal Amnesia. And they analyzed a large sample of patients with amnesia and showed that the connectivity patterns rather than the focal brain, uh, damage best explain their cognitive deficit.

So it suggests that there's a lot more to amnesia than, than the [01:13:00] hippocampus. Um, so that's a paper that I continue to think about and continue to be inspired by. I'd say another set of papers, uh, Yusepi Aria has done some work with patients who have extreme difficulty navigating, it's called topographical developmental disorientation.

And when he looks in their brains, they have no hippocampal abnormalities. They have actually very little obvious abnormalities at all. And the current thinking is that there's some kind of network connectivity problems that underlie their function, but it's still up in the air. But it's fascinating to kind of approach the problem from the other angle.

Instead of what brain region you have damaged, what's What behavior do you have damaged? And it turns out when you look under the hood, it's, it's much more complicated than we thought. Um, so I think those would be two examples of papers, uh, and, and, and research that to me has been very inspiring and interesting.

Benjamin James Kuper-Smith: Yeah, second question is, uh, something you wish you'd learned sooner. Um, so this could be from your personal life, from your work life, whatever you want. [01:14:00] Uh, just something you think like, you know, if you, if you'd learned that maybe, maybe your life would have turned out a little bit nicer or a little bit easier.

And, uh, yeah, depending on what it is, you can share how you learned it or what you did about it or, yeah.

Arne Ekstrom: Yeah. I mean, I think going back to earlier that there's many different paths one can take in life and one can learn from failure. Failure is not always a bad thing. It's extremely frustrating to deal with, but, um, sometimes you can, you can learn more from that than just from success. And I guess, I guess that was a lesson that I, to some extent, wish I had appreciated earlier.

Um, and that there's, there's not always a right answer to everything that we're thinking about. Um, yeah,

Benjamin James Kuper-Smith: Okay. Uh, and final question is, um, yeah, I mean as I mentioned earlier, I will be starting a postdoc soon. Uh, this is a kind of open question, any advice for PhD students, postdocs, people kind of maybe in that [01:15:00] transition period. Uh, yeah, just anything you want to give people like me. Okay,

Arne Ekstrom: be frustrating. I think there's an advantage both to being Very focused on a specific experimental question and, and almost myopically focused in terms of what papers you read, what techniques you try to learn, because it's a huge world cognitive neuroscience, and you can only reasonably learn so much in a five or six year Ph.

D. But at the same time, I mean, it's really important to have an open mind and dream about things and engage in the intellectual debate and find the concepts interesting. Because I think both paths are needed to. Be successful and in academics, if you just do the methods and you're only excited about the methods, then you should be an engineer and you know, you're, you're not going to get excited enough about many of the big debates we have.

If you're only interested in the conceptual debates, then maybe journalism is a better path. What's difficult about academics is you need to have some combination of [01:16:00] both. And, um, I think my suggestion is when you're doing a PhD, be open to both possibilities. But if you find yourself moving in one direction or another, That's totally fine.

And, you know, like we were talking about earlier, um, you can learn a lot from what doesn't work for you. And, um, we need people who want to be engineers, technicians, coders. I've had several students go on to work for Facebook and Amazon, uh, and they're, they're very happy with it. And conversely, we need Journalists, communicators who can explain to politicians, science, all the pressing issues now in the world related to climate change and things like that.

We need those people. And so be proud of your strengths and don't, don't see it as weaknesses, but at the same time, be realistic with yourself in terms of what, what excites you.

Benjamin James Kuper-Smith: great. Um, yeah, I mean those were my questions. So, uh, thank you very much for the time.

Arne Ekstrom: Yeah, thank you again. Really appreciated it. And feel free if you have any follow up questions, you can always [01:17:00] drop me an email. I can clarify things.