Matthias Nau is a cognitive neuroscientist at the Kavli Institute for Systems Neuroscience in Trondheim, Norway. He finished his PhD recently in Christian Doeller's group at the Kavli, where he currently works as a postdoc. Whenever the current pandemic cools down, Matthias will start a position as a postdoc at NIH with Chris Baker (this position was supposed to start in early 2020).
In this conversation, we talk about a variety of topics, from Matthias's recent research (the link between vision and high-level spatial coding principles in the brain (e.g. grid cells), the development of cognitive maps in humans, and a novel form of using fMRI for eye-tracking he co-developed with Markus Frey), to the relationship between electrophysiology studies in animals and fMRI studies in humans, education in neuroscience, and science communication.
Timestamps:
0:00:15 MR-based eye-tracking
0:22:50 Switching to Python
0:26:20 Grid Cells and vision
0:39:59 Development of the cognitive map in humans
0:45:10 Electrophysiology and fMRI
1:02:25 The interdisciplinary education of neuroscientists
1:20:17 Twitter, science communication, and this podcast
1:35:38 Matthias's plans for the future, complicated by COVID
Links:
Podcast website:
Matthias's links:
Ben's links:
Grid cells info:
Papers mentioned:
Killian, N. J., Jutras, M. J., & Buffalo, E. A. (2012). A map of visual space in the primate entorhinal cortex. Nature.
Nau, M., Schröder, T. N., Bellmund, J. L., & Doeller, C. F. (2018). Hexadirectional coding of visual space in human entorhinal cortex. Nature neuroscience.
Nau, M., Julian, J. B., & Doeller, C. F. (2018). How the brain’s navigation system shapes our visual experience. Trends in cognitive sciences.
Wills, T. J., Cacucci, F., Burgess, N., & O'Keefe, J. (2010). Development of the hippocampal cognitive map in preweanling rats. Science.
-
[This is an automated transcript that contains many errors]
Matthias Nau: yeah. Yeah. And then, um, deep MRI is, uh, is basically the main project that we're working on now that we spend most time on, and we just presented it at, at Division Sciences Society, at the annual meeting of the Vision Science Society, VSS. And, uh, I felt, um, it was received very well and people liked it, liked the overall idea, and we ourselves liked the idea a lot.
And I personally think this is actually a really important topic because it's basically a new eye tracking framework. That's also why we spell it deep. Uh, MRI as, uh, EYE. Um, so, um. For the listeners, uh, Benjamin just asked me to spell it. Just exactly. Um, exactly. I feel like that was half of the [00:01:00] attention on Twitter.
It was just the name. It was a name. Yeah. Like, I mean, seriously, I think the, the name and overall the, the package as a whole, of course, um, helps to get the message across, right? If, let's say, uh, I dunno if, if the name is not catchier, so then people, it, it doesn't get the say, it doesn't get the attention that I think it deserves.
Benjamin James Kuper-Smith: Yeah.
Matthias Nau: If you have some
Benjamin James Kuper-Smith: weird acronym that's hard to remember
Matthias Nau: exactly. People just forget about it. Yeah. So I think we're lucky that we actually found one that seems to be catchy. Yeah. But anyway, it's not the, it's not important in a way. At the end of the day, what's important is that, um, it's an eye tracking framework for FMRI experiments that allows you to do eye tracking actually without camera and even in the data that already exists.
So you can actually go back in the data that you acquired, let's say a year ago. And actually reconstruct viewing behavior from these data. And you can reanalyze those. And I think this is really important because for ma, many reasons. One is of course you get an additional behavioral readout for your experiment.[00:02:00]
Uh, frankly, uh, that's really cool. The other one is if, let's say you have two conditions that you wanna compare and the viewing behavior differs between these con these two conditions, then uh, also the difference in brain activity is really hard to interpret. But because maybe it's because, I don't know, let's say you have a memory paradigm, maybe the activity you see in the brain is related to memory encoding or so, or it's actually just the difference in viewing behavior.
So just to interpret, to tease apart these different options, it's just important to do eye tracking. And then another one is, uh, uh, and there are also a couple of actually scanning artifacts associated with, uh, eye movements. Um, and then of course the mo most important one is that a lot of people are simply interested in studying view behavior with FMRI, um, and ocular motor systems.
And, uh, you can use it to build better task-based models, for example. And again, this is, uh, working in existing data. So that's, that's the cool thing here. So it, um, in short, it's a convolutional neural network, [00:03:00] or it's, the whole framework is centered on a convolutional neural network that takes the boxes of the eyeballs as a, basically like a classical regions of interest, uh, analysis and uses these v to classify gaze position.
And then in addition to gaze position, it gives you a predicted error score te telling you, um, yeah, how certain you can be that the, that the decoded gaze position is correct. So this is really important. If you go back to existing data, you can now reconstruct viewing behavior, but you also get a sense of how reliable this decoding is.
Because let's say if you have a free movie watching data set and you don't actually have eye tracking, you don't have any test labels, you don't know what the ground truth gaze position is, the user kind of probably wants some measure of certainty for this decoding before. Analyzing the actual FRI data with it and, uh, yeah, both of this you actually get from this network.
So you can, let's say, um, you can download a resting state data set, uh, I dunno, a [00:04:00] thousand people. Um, and reconstruct viewing behavior from these data, filter out the participants or samples that did not work for, for which the decoding did not work well. And then analyze the rest. Uh, maybe accounting for viewing or for eye movement, um, artifacts and eye movement related activity in the brain.
And then just clean up your data or actually analyze it, analyze it in some other meaningful way if you want.
Benjamin James Kuper-Smith: Yeah, I mean, it would be interesting for like, I mean these, like for example, the resting state networks or is that the same as the default network? No, that's something else, right? Or is it? I
Matthias Nau: mean, you see the default mode network in during resting state.
Benjamin James Kuper-Smith: Okay. Right. Yeah.
Matthias Nau: Often.
Benjamin James Kuper-Smith: Yeah. I mean like, it would be interesting there just to see when people like with their eyes closed, just. Do they have their eyes closed anyway, just to analyze, like
Matthias Nau: mm-hmm.
Benjamin James Kuper-Smith: How their eyes move. But yeah, I mean, sounds,
Matthias Nau: it's also actually interesting and resting to say that I think the, uh, most experiments or a lot of experiments use a fixation cross so people actually have their eyes open.
Uh, but a, a, a ton of other [00:05:00] experiments just have the eyes closed. And there are a couple of studies showing that the resting state networks you uncover in these two different types of rest, if you want, are completely different, or at least they're not exactly the same, at least. Yeah. So that's also, which makes sense.
Which makes sense. Exactly. Eyes open versus eyes open. It is a different task. Yeah,
Benjamin James Kuper-Smith: yeah.
Matthias Nau: Yeah. That's also cool thing about DRI is that it works when, while the eyes are closed, you can actually do eye tracking, um, while the eyes are closed, for example, during REM sleep, which just opens up. Also, can people sleep in the scanner?
I think they, uh, I mean, uh, I don't know. I, I've never scanned, uh, people, uh, sleeping or. Slept in the scanner. Um, I think it's doable. So at, at least there are papers even showing that during resting state, for example, people, I think thir, I dunno, 30% of the people or so fall asleep within the first couple of minutes of the resting state.
Or you find at least evidence for sleep, uh, within the first three minutes or so. In a large part of the people.
Benjamin James Kuper-Smith: Yeah. Yeah. No, actually, if I remember like people saying that a lot of people, not [00:06:00] a lot, but like they will have participants who will just fall asleep, I guess. I mean, you're lying down, it's kind of warm.
Mm-hmm. And it's a kind of like monotonous noise often in the background.
Matthias Nau: Yeah. And you are, you are in this capsule kind of protected environment. Nothing, especially during resting state, nothing really disturbs you. You're just in this, I dunno, sound box. Mm-hmm.
Benjamin James Kuper-Smith: You
Matthias Nau: can't do much. You can't move either. Uh,
Benjamin James Kuper-Smith: so one question I had, so earlier you mentioned that, um, it doesn't work or you can, you can take out people for whom it didn't work.
Matthias Nau: Mm-hmm.
Benjamin James Kuper-Smith: Uh, or for whom it doesn't work as well. So do you know what that depends on?
Matthias Nau: Yeah, I mean, what we do is, uh, we take the data, uh, we pre-process it minimally actually do like realignment, so we correct head motion and so on in this data. And um, yeah, really minimal pre-processing. And then we register everything to our own template space, which is an average.
Brain for which we know of, of [00:07:00] like, I think 27 people of which we know where they are looking. So they're fixating the screen center. So we know if the eyeballs look like this, they fixate at the screen center. Mm-hmm. And we register everything using non-linear transformations to this template space. And it's, it's a little bit of a complex registration procedure, but at the end of the day, it might not work exactly equally well for everybody.
Right. So in some people the mask might be a little bit shifted or, um, they just buy and from because of anatomical reasons, they might have different distortions in the eyeballs or so, um, some people just didn't do the task well or so. Right. So, um,
Benjamin James Kuper-Smith: okay. Yeah.
Matthias Nau: I don't know. Some if, let's say if you have a viewing task, some people might just close their eyes actually.
Some, literally some, some people might just literally fall asleep in the scanner and you kind of, you could in theory catch those people in, in your dataset. You could, in theory, you account for them somehow. Using eye tracking.
Benjamin James Kuper-Smith: Yeah, it seems, I mean, it's almost like I remember when you [00:08:00]presented, um, I mean, so when you were at our institute in Hamburg, you presented, I guess like most of your PhD and the deep MRI was just five to 10 minutes or something at the end, I think.
Matthias Nau: Yeah.
Benjamin James Kuper-Smith: Um, it did sound like the thing, which is like too good to be true. Almost like where you have like, uh, you know, you start with one thing, like, hey, we can just use, um, you know, you can, you can use MRI to get, uh, eye tracking. It's like, oh, that's really cool. And it's like, even if the eyes are closed, okay, well if that makes sense.
Like, you don't need eyes open. And it's like, well, even for old data sets, like any MRI that you have, and then it just, you like, you just stacked it more and more. It's like, wow, this,
Matthias Nau: yeah. No, that's why we are so excited. And it, it, we also explore, um, scanning parameters. So we scanned a subset of people with eye tracking, with camera based eye tracking and nine different imaging protocols.
So, uh, it's a three by three design for. Repetition time. So basically temporal resolution of your, uh, of your FMRI scanning sequence [00:09:00] and, uh, Voxer size. So spatial resolution. And, uh, yeah, we have vox from 1.5 millimeter to 2.5 millimeter and, uh, repetition times between 800 milliseconds and 2.5 seconds.
Right. So we have overall, actually the entire dataset consists of 267, I believe people, or 270 whatever. Um,
Benjamin James Kuper-Smith: so
Matthias Nau: the training
Benjamin James Kuper-Smith: dataset
Matthias Nau: for the, the train Exactly. It's cross, it's a cross validated, uh, training test dataset. So all the data goes into training and into test in a cross validated fashion. Yeah.
Benjamin James Kuper-Smith: Mm-hmm.
Matthias Nau: Um, exactly. In a subset of people we scan with nine different imaging protocols, but overall we have 14. And for all of those 14 that we have, it works really well. So even for a large vox of 2.5 millimeters, or for very long TRS of 2.5 seconds. Or repetition times. It still works quite amazingly actually.
Um, I think one, one thing that [00:10:00] um, should also emphasize is that in order for it to work really well, you kind of need training data for your own. I mean, if the best would be that you, let's say if you have a data set of 30 people lying around and you wanna reanalyze these data, it might make sense to scan another five to 10 people with an iTracker calibration script where, you know, while people fixate at certain locations, you just present a fixation cross on the screen, use these data to train the network using the same scanning sequence.
So same distortions, same kind of slice, similar slice package, and so on. Uh, and then decode from the data that already exists. So you can't just, you in theory, you could out of the box, take our model weights that we will also share with the code and decode from a dataset, but that's not gonna be. It's not gonna work extraordinarily well.
Um, it's really best if you have your own training data.
Benjamin James Kuper-Smith: Okay. So like the best. So it's, [00:11:00] how should we say it? I mean, like, one thing that sounded, um, to me super cool about this is that, so I've never used eye tracking, um, in or outside an MRI scanner, but a lot of people here do. And you fairly regularly, we see emails of people saying iTrack is broken again or something because they seem to be Yeah.
Or it's, they couldn't calibrate it properly or whatever. It seems to be a, um, at least the one we're using here tool that, uh, is not as reliable as other tools we're using, let's say.
Matthias Nau: Mm-hmm.
Benjamin James Kuper-Smith: Um, and it seemed to be like super cool that you could just like get rid of the tool almost, basically. But it's, but you're suggesting basically it would be still be good too.
Just do it with a few people and then you don't, but you don't need to do it for all, at least.
Matthias Nau: Yeah. You don't really need eye tracking for that. You can scan people without an actual eye tracker, but just have fixation crosses on the screen. Oh, I see. So you still don't need the camera. Yeah. So you can save these, like $50,000 on the camera, for the camera and just basically scan a five minute, uh, calibration script for every person.
So you Oh, okay. [00:12:00] This is actually how it is. Yeah. So you don't, oh, well that's then
Benjamin James Kuper-Smith: not much a problem then. Yeah, I mean,
Matthias Nau: exactly. So you scan like a five minute scan for, for calibration. That's it. Um, but of course, if you already have an eye tracker, you might just as well use it. But then even then, in my experience, the calibration sometimes doesn't work for people.
Eye tracking in the scanner is often really noisy. It depends a bit on the setup, but then for, for a lot of people, you will actually in the end up, uh, at the end, end up losing data, eye tracking data. For those people. In theory, you could also, even if you have an eye tracker, you could then use deep eye to recover the viewing behavior in those participants.
Or let's say, um, your eye tracker broke during the scan and you have two runs with eye tracking and one run without, you can just use it for the other run, whatever. But overall, I think the big and the, um, the big potential is really to, um, have your own, if you have your own training data of [00:13:00] in a few participants and just decode from everybody else, and that works for existing data, but also of course for future experiments, let's say you plan a new scanning study, you just add this five minutes snippet of calibration script to your scanning session for every single participant, and you will have plenty of training data for your entire, uh, for your entire experiment.
Benjamin James Kuper-Smith: That's, so how accurate is it then? Like if you, I'm assuming you're comparing it to standard, um, eye tracking or,
Matthias Nau: yeah. It's really accurate. Uh, so, um, we have a couple of measures, of course. So model performance, if you want, you can measure in multiple ways. I think the, not the best, but one way of doing it is piercing correlation between real and predicted gaze path.
And, uh, the median piercing correlation between those two, real and predicted is across all these people that we have, across all data sets is at 0.9 Pearson correlation. So that's extraordinary, right?
Benjamin James Kuper-Smith: Yeah.
Matthias Nau: Uh, but, and even if you, and so [00:14:00] Pearson correlation doesn't take into account the scaling of the data.
So you could have basically co varying signals, but let's say the scaling is completely off. Um, I mean, co varying gaze positions, let's say one is at 10 degree visual angle. The other one is at one degree visual angle. So in theory you have a, a large EU ian error. The fluctuations are the same. So you will get up, you will end up having a good, a high appealing correlation, but even if you take these scaling issues into account, you take the euan error and so on, you get a really, really good model performance.
So I think for many FMRI settings, um, yeah, I wouldn't feel comfortable putting like a number now on it because we're still working on that.
Benjamin James Kuper-Smith: Yeah, yeah.
Matthias Nau: But, um, it's definitely very accurate for, for the large majority of FMI studies, it will be enough. So it will definitely get a different, I, uh, like, uh, where people are looking.
Uh, plus, minus maybe two degree visual angle, one degree visual angle. It depends a little bit on your training data and on, but [00:15:00] I mean, the predicted error score will tell you what the predicted accuracy is, um, given your data, your training data. So that's also pretty cool.
Benjamin James Kuper-Smith: Yeah. So just to go, um. Um, to ask a more, a question about the resol, like the precision or reso the resolution of the thing.
So I'm just curious, you mentioned earlier that this could be cool for if you have like a, um, what's it called? Like natural vision. Not natural vision, but like, let's say you're watching a film or something, right? Mm-hmm. Um, like let's say you are, you want, I don't know, you watching let, you're letting people watch some movie and with people's faces or whatever.
Like, I'm just curious, I'm wondering like, would you be able to tell like whether people are looking at people's eyes or, I mean, it probably depends, like, obviously how big the thing is on the screen, but um,
Matthias Nau: yes. I, I think so. You, you will be able to do that. You will be able to, during the movie, tell where people are looking and we, we also compare that actually during [00:16:00] movie in a, in a different project.
I think that's also something that, um, I presented in Hamburg when I visited you guys. Um, it was a, a, a study on, on visual grid cell-like activity in the human trinal cortex and how it progresses or how it develops, uh, um mm-hmm in development over so over age. Um, and there we also decode basically eye movements from the FMRI data, uh, and compare it also directly to camera based eye tracking during free movie watching.
And also there it's really a great match. So you will, if even if you overlay those two real eye tracking and deep MRI onto the movie, you will get a really nice result.
Benjamin James Kuper-Smith: Yeah. That's really cool.
Matthias Nau: Yeah.
Benjamin James Kuper-Smith: I almost wish I was doing visionary research now. I could use it for the stuff I'm using. I don't think it's gonna be much use anytime soon, but
Matthias Nau: Yeah, I agree.
Yeah. Um, there's so much to do with it. I think. Uh, however, I wanna emphasize that this is not only a tool for vision researchers, right? This is [00:17:00] obviously cool for, for people interested in vision, but the whole viewing behavior. Also the viewing behavior confound issue is, is present in a large part of cognitive neuroimaging.
I would say. Uh, it's not only, it's not limited to vision and I personally, in my own research, I'm interested actually in yeah, the interface between perception and memory and the human brain or how you derive, uh, like a, a map like representation of the environment from your visual input. So I'm really working on this interface between more higher level, um, your memory processes and really low level vision, let's say in V one, how motion is processed and also how this in turn affects how you remember things and where you remember things and so on.
So for me, this is really a, a great tool because you can, um, not only from the vision perspective, but because I'm interested in eye movements and gaze behavior in let's say, uh, and signals in the hippocampus, which is not typically associated with viewing behavior, or at least currently it's not. [00:18:00]
Benjamin James Kuper-Smith: So it, do you almost think that this, I mean, you mentioned like, um, eye movement artifacts, that kind thing.
So do you think this, it would make sense to make deep MRIA standard part of a pre-processing pipeline to say like, okay, for every trial, like take out the eye, the, the neur activity related to eye movements?
Matthias Nau: Yeah. Or, or account for it. Uh, so account for it or study it, right. You can. Um, yeah, for sure. I think it, I mean this is a, this is a big question of course.
Uh, or a, that would change things quite drastically. It would, basically what you're suggesting is adding a, basically a seventh realignment parameter or so, like a seventh, uh, nuisance regressor to the realignment realignment parameters. I think that's definitely possible with deep MRI and we, we discussed that too at the end of the day.
Of course, that depends on the user and what they want, but yeah, that's, that's definitely possible. Yeah.
Benjamin James Kuper-Smith: Have you tr found anything where you. I know in either your [00:19:00] like public data sets or your own stuff where you, I don't know, found that like part of your results were explained by eye movement rather than some experimental thing you thought it was related to.
Matthias Nau: Uh, so in my own research, I'm, I'm doing eye tracking basically always.
Benjamin James Kuper-Smith: Right, okay. Yeah. Okay. That's a vaguely stupid question, but,
Matthias Nau: but um,
Benjamin James Kuper-Smith: of course. Do you know what I mean? Like, is there, like, have you tried that or
Matthias Nau: Yeah, we did. Yeah. So we didn't try to explain like published results and try to Yeah.
Benjamin James Kuper-Smith: I mean,
Matthias Nau: find like confounds in other people's stuff.
That's not what, that's also not the point. Absolutely not.
Benjamin James Kuper-Smith: No, of course, of course. But it would, it would just show like how important it is. That's what
Matthias Nau: I mean. Yeah, exactly. But I think that we do, so we take the predicted, the actually camera based eye tracking and to predict it. Eye tracking data, so deep MRI derived eye tracking data and just regress it against brain activity in a just whole brain fashion.
Just see where in the brain does viewing behavior predict any activity. And we use, um, [00:20:00] data from Russell Epstein and, and Joshua Julian, so that they scan people while they perform the visual search tasks or very classical, uh, search for the L among the T or like a distractor, uh, set, um, task. Yeah. And we, um, use eye tracking and DMRI and request it against brain activity.
And we see this entire, this, this gigantic brain network that is actually explained by the viewing behavior, just by basically how far the eyes moved during a tr. Um, it's amazing. You see, of course V one, you see parietal and frontal eye field region. So this attention network, um, you see, um, human motion process, uh, complex.
You also see retro nia cortex, medial parietal lobe. You also see hippocampus, you see ventral prefrontal cortex. Um, many regions that are not typically associated with ocular motor function. So, um, that's, yeah, I think that's what makes it so important. You, [00:21:00] I think this ana, this set of analysis, it's also on the poster that we presented at VSS really demonstrates that it is important to do eye tracking if you wanna study an account or account for these effects.
Um, and also that deep MRI can be used to do so.
Benjamin James Kuper-Smith: Yeah. Yeah, it is really cool. I'm just thinking about like my own stuff. Like you, I mean, I'm not doing anything that's, yeah. As I said, like directly related to vision and it's like the, what people see is fairly boring often. Mm-hmm. Just like a few numbers on the screen or something.
But you could at least, um, then it seems like, look, by people's attention is what they, what are they really focusing on? Um, well, I guess at least using eye gaze as a proxy for attention.
Matthias Nau: Mm-hmm.
Benjamin James Kuper-Smith: Um. Yeah. It really, it sounds really cool.
Matthias Nau: Yeah. Or, or or if they are awake, if they're
Benjamin James Kuper-Smith: doing Yeah, I'll just check for that.
They're just like randomly pressing something while sleeping.
Matthias Nau: Exactly.
Benjamin James Kuper-Smith: Oh God. Yeah.
Matthias Nau: No, yeah, exactly. I think, I think there are so many data sets and [00:22:00] applications for it. It's, we are excited about it and we are glad that, at least at VSS people were excited about it as well.
Benjamin James Kuper-Smith: Yeah. And so you're using your, like, remaining time and time, time right now to just to get that as far as you can or
Matthias Nau: Yeah.
Deep MRI and, um, exactly this other project I explained earlier with Ignacio Pal.
Benjamin James Kuper-Smith: Yeah.
Matthias Nau: And, uh, yeah, I'm also planning just new stuff, of course, so for my future lab, but also, uh, with Christian Doll, uh, with whom I'm currently a postdoc with. Um, I'm also planning another study, actually another scanning study.
Benjamin James Kuper-Smith: Ah, okay. Yeah,
Matthias Nau: so, so it's basically in a conceptual state at the moment, but still it's, it's fun to play around with these ideas, maybe write some, um, psych toolbox scripts and so on. I'm also trying to transition to Python. Um,
Benjamin James Kuper-Smith: yeah,
Matthias Nau: yeah.
Benjamin James Kuper-Smith: How's that going?
Matthias Nau: Uh, that's good actually. Yeah, I mean, I'm, so far I use matlab, uh, and it's, of course the commands are different [00:23:00] and some of the concepts are a little bit different how you use it.
But I mean, if you know how to program in one language, it definitely, it, it, it speeds up how you learn the other one.
Benjamin James Kuper-Smith: Yeah, definitely. I mean, uh, yeah, we also, well, we wanted to kind of transition to Python a while ago, but
Matthias Nau: mm-hmm.
Benjamin James Kuper-Smith: Haven't, it's one of those things I think a lot of people want to do at some point, but then it's always kind of, you know, just slows everything down for a while.
Yeah. 'cause you have to like learn all the new functions and.
Matthias Nau: Yeah, exactly. That's the big problem actually, because you always are, in a way, in a hurry, you also always wanna get stuff done, and it's always easier in the language that you already speak or, or program in. Right. And then you, even if you try to do it in Python and, but you are really comfortable in MATLAB at some point when things are getting a bit tight or the deadline is coming up and you switch back to the platform that you know best.
And that's, that's the problem. Actually, I have the same, I definitely have the same issue. So if things need to go fast, I then switch back to matlab and then I spend way too much time in MATLAB again. Which,
Benjamin James Kuper-Smith: what is the [00:24:00] reason for you for wanting to switch the Python? Is it like, yeah,
Matthias Nau: I think it's the future, right?
Uh, matlab, um, fewer and fewer people are using matlab. Um, this entire deep learning community is centered on Python. Uh, and as, and also for me, I mean, looking forward in a few years from now, I feel like if you're. If you're not speaking any python, then I feel like a dinosaur somehow. I think it's what the future will hold.
Yeah.
Benjamin James Kuper-Smith: Yeah. I mean, like, one thing that just annoys me about MATLAB is the license thing. Like in theory it's very straightforward, but it's happened to me a few times where like, I know you didn't have access to some toolbox somewhere or something. And then,
Matthias Nau: yeah.
Benjamin James Kuper-Smith: So yeah, I mean, yeah. I, I agree. I think it's,
Matthias Nau: it's open access.
Benjamin James Kuper-Smith: Exactly.
Matthias Nau: And this is, this is the main reason, and this is also the main reason why people use it that much. Right. And this is amazing. It's, it's a great reason. [00:25:00]
Benjamin James Kuper-Smith: Yeah. I really like, yeah, I just, I really like the idea behind it. Just like everyone anywhere in the world can just, you know, you have open code, open open data, then anyone like, I don't know, some smart 10-year-old somewhere with internet connection can theory download your data set and analyze it.
Python it if he wants to.
Matthias Nau: Exactly.
Benjamin James Kuper-Smith: Yeah.
Matthias Nau: That's amazing. Yeah, I like that too.
Benjamin James Kuper-Smith: And then find errors that you didn't find or something.
Matthias Nau: Yeah. I mean, that's really important. Exactly. Yeah. Yeah.
Benjamin James Kuper-Smith: I was, I'm wondering, um, so one thing, um, so one thing that course of my supervisor once mentioned that he found that, like, one thing he found a bit tricky was that you often have this carryover phase in a new position where you're doing lots of old projects.
Matthias Nau: Mm-hmm.
Benjamin James Kuper-Smith: And that, I think for him in particularly, he has like, I think now he has like, still like three, four things or something. He basically wants to finish all the time, but, you know, it's difficult to, when, when you have three PhD students [00:26:00] at your feet, uh, it could be, you know, you have other stuff to do.
So is, I was curious, like is in your new position whenever it will start mm-hmm. Um, is that closely related to what you're doing now or is that kind of a bit of a step? Um, I mean, it, it sounds like it would be pretty similar.
Matthias Nau: Yeah, maybe just a short recap of what I've been working on. Um,
Benjamin James Kuper-Smith: yeah.
Matthias Nau: I started basically in a, in a vision lab and in a neuroimaging lab with Andreas Bartels and worked on how visual cortices integrate visual and motor information, eye movement information in a visual motion task.
So I, I was wondering, um, how the visual system stabilizes perception during self motion. That's basically how I entered this entire neuroimaging field. And back then I remembered, or I re I remember thinking that, um, knowing how the visual system might be able to stabilize [00:27:00] that, stabilize perception, I felt like I'm really understanding a piece of the puzzle.
How, how we just, um, experience of our entire visual experience, how we experience the world and so on. But then. In the course of that study, I read a lot about, um, play cells, grid cells, head direction cells, so things that, or cells that you would find in, um, the hippocampal formation and medial temporal lobe in humans.
And they, I always felt they are really related to what I was working on because they basically derive a stable representation of the world from these visual inputs that I've been working on. But at the same time, most of these data were, uh, acquired in rodents, navigating, navigating in boxes. And it was, um, electrophysiological recordings and I was working on visual motion FMRI, and at the same time I've simultaneously, I felt somehow related to, uh, this work, but at the same time, very distant, right?
Benjamin James Kuper-Smith: Yeah.
Matthias Nau: Also methodologically. And that changed when I read a paper by Beth Buffalo from [00:28:00]Seattle who recorded in the interal cortex of monkeys, and they showed that these grid cells and just, uh, really, uh, I'll come back to what grid cells actually are. Um, but they recorded grid cells in, in a monkey just looking at images.
So suddenly these, these coding principles in the, in the hippocampal formation were just much closer to what I was working on, which is they played out in visual space, which, which is what I felt comfortable working with. And then another paper by my current PI Christian, uh, Christian Doah, who showed that these grid cell like signals in, you can pick them up in FMRI, while people navigate in virtual environments or at least, uh, something that we think is a proxy of grid cell activity.
And these two papers combined really, um, sealed the deal for me and I noticed, wow, I can actually work on with the tools I already have with the skills I already have. I can actually work [00:29:00] on this topic as well and have a more, a bigger perspective on how visual cortex and these hippocampal cortices somehow in Yeah, how the interface.
Um, so we ran a study with Christian and then also a couple of follow-up, uh, studies. But, oh, and this is the core of my PhD if you want, along with the paper that came out two weeks ago in in Nature Communications, where we show head direction modulation in FMI as well, uh, using an encoding model. And overall my PhD, I felt, was really focused on, um, this forward sweep of information from visual cortex to the hippocampus, how you, how you derive a, a stable representation.
But now looking forward into my postdoc, um, I feel I actually wanna go the other route. I want to know what these, what these higher order visual spatial coding principles actually allow you to do. And I [00:30:00] think. This is, there's two things. One is it they help to integrate visual experiences into memory and retrieve them when you need them, and they allow you to plan your behavior in space.
And in particular, in a, in a text review published in 2018 with Josh Gillian, I proposed that the human, also the human hippo formation, um, represents this coordinate system of visual space that allows you, that allows you to do these things, um, memory recall and formation and planning of viewing behavior in this case, and this is what I'm gonna study in my postdoc as well.
Benjamin James Kuper-Smith: So maybe could you briefly describe your nature neuroscience paper?
Matthias Nau: Mm-hmm. Yeah. So the in, in the entire paper builds on the idea that there are grid cells in the brain, and they were discovered in rodents by the Moser lab, who is basically the, the core of the institute. I'm working, uh, at now, at the, the KA Institute for for Systems Neuroscience and Runtime Norway.[00:31:00]
And they discovered these cells that represent the environment, uh, in a, basically with a grid, grid-like code. So they, these cells, if you record the activity of a single cell, it would fire at distinct locations in the different locations in the room while the rodent is running around. And these locations are not just randomly distributed, they actually tile up the environment in a hexagonal grid-like fashion.
So it's, it's amazing for me, this is, this is definitely among the most beautiful discoveries in, um, in neuroscience. Really. I mean, this, the pattern is absolutely striking and I was just hooked by it when, when seeing this. I thought, why is that? This is insane. Yeah. Right. And I think a lot of people felt that way when they saw these grid cells.
And also the discovery was, um, uh, was rewarded with a Nobel Prize together with to O'Keeffe, who discovered a related cell type called the place cell, which fires only at one location in the room. And, uh, those cells can be found in the adjacent and hippocampus. But [00:32:00] I was focusing on these grid cells, and I tried to find evidence for these grid cells that, that also Beth Buffalo found in monkeys as a function of where, where people are, uh, where, where the non-human prime at the macca was looking on the screen.
They found that a similar grid-like pattern emerged, uh, in this viewing task. And I was wondering, can we find evidence for these visual grids cell-like, uh, codes also in the human andal cortex? So to study that, I went back to what I knew already from Andreas Baral lab, which is this visual tracking task.
So back then I, I used a, a smooth pursuit task where people, uh, fixate at a fixation target that's just smoothly moving across, uh, the screen. Yeah. So in my case, um, it was always a very pre, a very clear predefined trajectory that allowed you to sample all directions equally and in a, uh, in a, um. With like constant speed and all these factors matched, [00:33:00]
Benjamin James Kuper-Smith: but like the participants didn't know where, when it was gonna change or anything like that, or,
Matthias Nau: well, I mean, it's, I think it, after a while it does become a little bit predictable, I guess.
Um, but they, the actual task they performed was memorizing locations on the screen, so they, they fixated with the eyes this moving sation targets. But the actual task that we got a behavioral readout from was, uh, memorizing locations on the screen that just popped up some, there was like targets popping up on the screen and they would memorize the location of these targets on the screen and report them in the course of several trials.
So they basically formed during this smooth pursuit task, they formed a spatial memory representation of the visual field and where these locations are. So that, that was the task. Then we analyze the Trinal FMRI signal as a function of eye movement direction. And now you would think, why would you, I just told you that grid cells encode the location of the animal [00:34:00] in space.
Why are we now looking at direction of eye movements? It's quite a stretch, right? So for that, you need to understand that this grid like, or this grid pattern that these cells express is, uh, built off, um, built up from hexagonal patches. And if you rotate this entire pattern, this firing rate map, if you want, if you rotate it, it would look very similar every 60 degrees of rotation.
So it's modulated by 60 degree or, uh, sixfold rotational, e symmetric. And the other thing is that if you walk, um, on the main axis of these grid, you would cross over more firing fields than if you walk off or orthogonal of these ex with these exes. So overall the prediction was that if there are a lot of grid cells in the brain in, in these FMI boxes, then you should see a sixfold rotationally symmetric signal as a function of movement direction as well, because the location signal for various [00:35:00] reasons that are too complex here now, but the, the positional signal, uh, we didn't expect to see in the FI signal, but the directional signal should come out, and that's what we looked for.
So we looked specifically for the sixfold rotationally, uh, rotational symmetry in eye movement direction. And that's exactly what we observed, uh, in the, in this task, in this viewing task. So there was no, I mean, think about it, in, in rodents, these cells were discovered while, um, the rodent was running around in a box.
And now we can measure these things in FMRI while people look at the screen. So that's pretty, I thought this was pretty cool. And, uh, most beautifully it was also published back to back with a completely independent paper from a, from a different group, from Russell Epstein's group and led by, uh, Josh Julian, who later became a postdoc in our group.
And we also collaborated on, on various projects. So that was also really cool. But these two projects are completely independent and they made exactly the same observations. So giving you a lot of confidence, also putting it out. That [00:36:00] was for me also just a beautiful feeling. I mean, in the beginning, of course, I think Josh and I both were a little bit like, holy shit, somebody is doing the same thing as me.
It's just like it is for a young PhD student knowing, I mean, you get a heart attack when you know that somebody else is, is doing remotely the same as you. Some other, some other person. This is my brilliant idea. No, no, no. But I mean, this is, it might be your brilliant career, right? Yeah, yeah, of course. Who knows?
Uh, anyway, or just, I mean, you spend so much thought and so much effort into a project, especially in the beginning. It takes you. So much dedication to get a project going and finishing it. And then of course you get a heart attack if anybody remotely looks, looks at an, at the same FI machine as you, then you think like, oh holy cow.
They're doing the same.
Benjamin James Kuper-Smith: Yeah. So in the beginning it was, so when did you find out that they were doing the same thing? Was it just when you submitted it to the journal or was it before?
Matthias Nau: Uh, no, we, we found out before actually at a conference. Yeah. At SFN.
Benjamin James Kuper-Smith: Ah, okay. So did you like coordinate this and say, let's submit it together?
[00:37:00] Or was it, I'm always curious, like, because there's been a few papers I've seen recently, uh, like the development of the, um, internal, like the mm-hmm. On place, so development that, that was also back to back publication by I think John O'Keefe's group and the Moses group.
Matthias Nau: Mm-hmm.
Benjamin James Kuper-Smith: And I always wonder like whether that's like partly intentional or just complete coincidence.
Matthias Nau: I mean much of this is probably, um, agreed upon beforehand and I think I'm actually a big fan of that just from the get go. Speak as transparently as possible about what you're doing. I think nobody tries to step on your feet, or at least very few people really try to screw you over. Or then, um, I think most people, if you are transparent, we'll also be transparent.
Benjamin James Kuper-Smith: Yeah. Yeah.
Matthias Nau: And in case of, uh, uh, Russ and, and Josh, we just talked at the conference the first time actually at SFN and then, um, back then they were really far ahead of us, I think, in this project. And they couldn't wait for us basically. [00:38:00] But then during the submission process, so they submitted and um, the paper was.
Sent to us for review. So we were invited to review it. Yeah. Yeah. Or Christian was invited to review it, but then we said we, we cannot possibly review this paper. We are obviously not, um, naive or, um, yeah, we have a conflict interest. We have a conflict interest. Yeah, exactly. So we responded to the editor, we are really sorry, this is, this looks like great work, but we cannot review this.
But we have, we have very similar results that we also would like to submit very soon. So is it okay if we submit? And then he invited us to submit it as well. And then during the review process, these papers actually synced up and came out together.
Benjamin James Kuper-Smith: Okay.
Matthias Nau: So, uh, that, yeah, it worked out beautifully in the end, but I mean, in the beginning, of course, everybody, I think, yeah, it was just a lack of communication.
It was not really miscommunication or so, or, uh, nobody tried to screw anybody over, but it was just a, a lack of communication in the beginning between the [00:39:00] two of us or both groups. And, um, yeah, in the end it worked out and this was the most beautiful thing of all, I think for me.
Benjamin James Kuper-Smith: Yeah. It's, I mean, it must, like, from the outside, it seems like it must be amazing to have like a, a really cool paper immediately, um, independently kind of, uh, um, what's the word?
Replicated or confirmed.
Matthias Nau: Yeah. Replicated each other basically. Yeah,
Benjamin James Kuper-Smith: exactly.
Matthias Nau: Exactly. Yeah. And it, and the, I mean, the signal itself, the sixfold rotational sym, uh, rotational symmetry was then also replicated, but to be AAU Eagle, who is now a PI in, in M Munich, in MET, actually
Benjamin James Kuper-Smith: ME,
Matthias Nau: so he's, so he also looked at a very similar viewing tasks.
He looked at Atal
Benjamin James Kuper-Smith: cortex.
Matthias Nau: Yeah. I mean, it's source reconstructed to the medial temporal lobe. You cannot really say It's like, this is Rin cortex. Yeah, exactly. But, but, uh, it's the medial temporal lobe. It's the really, the signal is coming from where you would expect it to come from. Mm-hmm. Also in MEG.
Benjamin James Kuper-Smith: Okay. That's cool.
Matthias Nau: Yeah. And now [00:40:00] we also replicated it in, uh, yet another data set that I presented in Hamburg. In which, in this aging or in this developmental data set. Yeah. Can we talk about that briefly? Yeah, sure. Yeah. Uh, so we downloaded a, a publicly available data set of people between five and I think 20 years old, 21 overall, I think for the oldest, uh, group of people.
We don't have too many subjects, so we kind of cut off the oldest group. Um, but yeah, it's a really range of, it's, it's a nice age range in the, in the participant group. And we run these grid cell like, uh, analysis, this what we call the hexa directional analysis because it's, um, sixfold rotation as symmetric.
This hexa directional analysis. We ran on all of these people in a viewing task. It's a movie watching. So I think people watch, uh, finding Nemo. No, despicable, despicable me. Of course, finding Nemo is the other different dataset. Sorry. This is the crucial [00:41:00] difference for the data. Yeah, it's the crucial difference.
Yeah. Both great movies. Uh, exactly. So Despicable Me, they watch like a short snippet and we see that the grid cell-like signal actually builds up in the course of development. So it's, um, we see that the amplitude of the signal, it's a bit complex, but the amplitude of the overall signal, the, this modulation in the FMRI signal is more or less the same in all but the distinct basically, um, where these signals peak or which directions have a high activity in which directions have a low activity.
This is more stable in the old than in the young. So in the young it's is as if these visual grid cells, if you want, or visual grid cell, like signals are not anchored to the screen as much as in the adults. Even though the overall modulation of the signal is similar, it's just more stable in the adults.
So we, we interpret it. As the cognitive map being more anchored to external cues in the [00:42:00] adults versus the children because it's more, because it's relative to the movie, it's these, um, signal modulations become more stable. So they are aligned to the screen if you want.
Benjamin James Kuper-Smith: And Chris, if I remember correctly, in the rodent literature about the developmental stuff, um, as in like how the place cell grid, cell border cells, these things, how they're developing rodents.
If I remember correctly, the, it seems like the rodents, they kind of start out with proximity to borders. So the border cells are very strong. Mm-hmm. And then, I can't remember like the exact, all of them, but it seems like basically you have like some sort of, you look for external cues first, um, like distance to walls and that kind of thing.
And then as they get a bit older, they, they develop this map where they can move kind of free yet, like kind of more independently of these landmarks. Yeah. Um. Do, can you look for that at all in or is it, I dunno whether you can like, look for example, look for boundary vector cells or something like that, or is that just not, we don't have the [00:43:00] analysis for that.
Matthias Nau: We discussed that a lot actually, and we act, we even ran some boundary analysis, but they were not really, they didn't really show anything. I mean, one thing we did is just we tiled up the entire screen in different bins and just compared boundary bins versus center bins, but you don't really see a difference between those two.
At least so far we don't see anything interesting popping up. And it's always, I mean, and f i's always complicated to think about. Where I think the beauty of, of this kind of research, or at least what, what brought me into this FMRI research, this specific question for example, was, um, that you have a very clear prediction taken from electrophysiology and you test it.
You think about how would this translate to a population level signal. Wow. Then you have a prediction that you can now test an fm. This nice taking a, a hypothesis from efis and testing it on population level in FMRI. That's, yeah. Was so, what was so, uh, attractive for me in a way, in [00:44:00] this project that, that's what I thought was really beautiful in these previous studies also from Christian, for example, it was this beautiful single cell inspired finding, um, that he put forward and for boundary, uh, vector cells or boundary cells or border cells, um, at least to me it's not very clear how the population level prediction would look like.
Maybe just comparing boundary versus cent bins is actually not the best thing to do. I think in these, uh, so in, in Beth Buffalo's data, at least they see that there's something like a visual border cell that does fire along the boundaries of the visual display, but not at boundaries within that display.
So the, of course. The movie itself that they watch, for example, or the image itself might, might have some different edges and stuff inside the picture. And these cells do not fire at those edges. At least what they, in the data that we, that I, I've seen these cells do not fire at these [00:45:00] edges within the image, but at the edge of the image.
And that, that's what we looked for, but we didn't see it. Yeah. But maybe, maybe our prediction is just wrong. Who knows?
Benjamin James Kuper-Smith: Yeah, but I mean, but as you said, like it, it is very, it is a kind of a, a big step to go from single cell recordings in rodents to population based analysis and FM and humans. I mean, it's just, yeah, it is.
I mean, that it, you know, Casola was a nature paper, right? From being able to show that it's possible with grid cells, right? Like that by itself is a major achievement.
Matthias Nau: You need to, of course, emphasize that. Also, I don't wanna sound like I'm suggesting that. FM I measures grid cells. This is not what I'm saying.
I'm just saying we took a prediction from efis literature and tested it on fm i and it was confirmed what the actual link is between these two worlds. I have no idea about. Right. There might be it, it might end up actually that none of the FRI signal is explained [00:46:00] by, like grid cells spikes in itself.
But maybe it's the inputs to grid cells that might actually be not perfectly correlated with the spiking of a grid cell. It might also be, it's actually some, uh, head direction cell signal that some somehow is modd by six degree. It's, um, there's some evidence also for, um, also in our group and, uh, to be as analyzing LFP data, seeing, uh, and also in MEG and, and recording in other people's recordings data.
You see this sixfold symmetry also in, in the LFP. What really drives it, whether it's spiking activity off grid cells is still kind of unknown. We, we like to believe that, but that link is not fully closed, I would say.
Benjamin James Kuper-Smith: Yeah,
Matthias Nau: that's a good point. Don't, don't interpret my, I need
Benjamin James Kuper-Smith: your web shoe. It's grid cells.
Yeah, you're right.
Matthias Nau: Of course. I mean the, it's, it's hard. At least, I don't know any other explanation for these [00:47:00] results. What other, other stuff than grid cells would lead to a sixfold rotational symmetry? I don't know. And the prediction comes from grid cells and it was confirmed. That's, that's where we are at now.
Benjamin James Kuper-Smith: But I guess the more important point, isn't it also that like it's more about the computational mechanisms behind it rather than whether a single cell does it or networks of cells? I think to some extent, to me that's maybe not even the most interesting thing, but the, the thing that you can see, like the same computational patterns is I think to me more interesting anyway.
Matthias Nau: Yeah, for me it always, when I think about FMRI and my own data, I always think about, um, yeah, the, especially for the computation, if you, if you wanna discuss that level, I mean, FMRI is sensitive to synaptic processing, which usually reflects the inputs to the system. And of course local processing, right?
Just wherever you, wherever there is, um, transmitter re-uptake, you have a lot of metabolic [00:48:00] activity somehow. And, um, you might find an FMRI signal or an in change in FM RI signal. Um, but does that really mean that there is a computation going on? If let's, if there are inputs coming into this system, let's, let's assume there's a cell, uh, somewhere downstream, there's a receiver cell and it receives some synaptic inputs, but none of these inputs is enough to actually make that cell spike.
Then you would still have potentially an MI signal increase. But in that, in that receiver cell, potentially there's not a single spike. So would you call that a computation then? Or you could say, well, there's synaptic computations, but what are they good for if the receiver urine is not spiking? So I always think about finding these results.
Does not finding FMI uh, effects does not necessarily mean that the computation is going on in that area where you find it in at least, um, at least there's a chance that it, that it might actually origin [00:49:00] some somewhere else.
Benjamin James Kuper-Smith: That's a very good point.
Matthias Nau: How, how do you see that?
Benjamin James Kuper-Smith: To be honest, I haven't really thought about FMI that much in detail in a sense because I mean my, yeah, most of my stuff has been in, I mean the, the two master's projects I did do were with MEG and EEG and only now I'm kind of, I mean I've done behavioral stuff so far in since starting my PhD.
Matthias Nau: Mm-hmm.
Benjamin James Kuper-Smith: Um, so this is kind of something that's still ahead of me. Something I'm, I guess, well, I've been saying I'll start soon for like a year now. Um, to be honest, I don't think I have a particularly qualified point on that. Like, for example, the, I don't exactly like, the other thing is like the specifics of where the bolt signal comes from.
Matthias Nau: Mm-hmm.
Benjamin James Kuper-Smith: I'm not that, I don't really know that much about it to be honest. Um, yeah, I'm not even sure how much is known about it.
Matthias Nau: I mean, there, yeah, I think there are plenty of studies on it, but there's, there are still many [00:50:00] unknowns I would say. And of course, sp let's say spiking in, uh, in a, let's say the receiver urine dust spike, then the chance of neighboring synapses picking up that spike is also very high.
So of course, let's say if there is activity in a region, let's say spiking even output activity, the chance of this spike receiving. Neighboring neurons is also fairly high. So this intrinsic connectivity is usually very strong in many regions. So you might end up seeing it in the same, you might see the FRI signal in the area where the spiking happens.
All I'm saying is you cannot really make that causal link, I think, in the data. And just being aware of that I think is important. At least for me, I, I try to remind myself always.
Benjamin James Kuper-Smith: Yeah. Yeah. Right. I'm really glad you did. Yeah. Yeah. I mean, it's just a, so, one question here. So you, I mean, you said you started off in vision itself, right?
Or you studied biology even at the beginning, right? In your mm-hmm. Diploma or bachelor's or [00:51:00]whatever it was. Um, I'm curious why, like, why do it, why do vision in humans if you can do it in animals and have, you know, much more precise knowledge about what you're actually measuring? Um, yeah, because, yeah. So like, what, what made you interested in studying vision in humans specifically rather than.
Um, doing it in animals.
Matthias Nau: Well, ultimately for me, most of the animal research is, I think it depends a bit on what your motivation is to do science at all. Uh, for me, I, at least, I, I came into science also for personal reasons, because I just wanted to know what this, uh, I mean, it sounds a little bit, I don't know, traumatic, but I mean, I, I came into, uh, this business because I was curious.
I wanted to know how this world works and how, who I am and, uh, how I perceive stuff and studying biology, I, I simply thought, um, well I could now [00:52:00] study, I dunno, the, some formation of trees or some, some specific animal or so. Then at the end of the day, I would go home and know a bit more about this tree.
How it works. Yeah. Would be amazing. I'm not saying this is not, this is really important science and I, I, I love reading about this and so on, but I felt, for me personally, it was not as satisfying as knowing about how I perceive all of those things. Let's say understanding how I perceive color. Made me, made me go home at the end of the day of university thinking, wow, I learned now something, how I see this tree, this animal, this car.
Basically how I interface with reality or what I, what I can interface with at least. So it's more, for me at least, it was more like a philosophical, I think, motivation to get into science. And it's basically a self discovery in the beginning. And from that point of view, I just wanted to study vision in humans because ultimately that's who I am, [00:53:00] right?
And, uh, who all of us are. And it's also, I think, the closest, the methodology actually, also I liked a lot. So just kind, I like the methods. Also. I like FMRI, I like neuroimaging. I like having whole brain resolution. A lot of questions you can simply ask only on network level. Uh. I mean, now of course with calcium imaging and neurop pixel probes and so on, you can ask a lot of these questions also in animals.
But when I started off a couple of years ago, of course people did calcium imaging back then, but it was not the thing that it is now. It's, it was still, you studied a little patch of cortex and oftentimes you studied, let's say, fewer than 50 cells, let's say, right? Yeah. In most of these recordings. Um, or in a lot of them at least.
Yeah. I just wanted to go for humans and, um, have a bigger perspective on things, on, on, on the whole brain, how different areas interact. And I think this question that I'm, in a [00:54:00] way posing also is in my PhD, was how do you derive a stable representation of the environment from visual inputs? That you cannot, that entire question you cannot address by looking at V one alone in isolation or hippocampus in isolation or retro screen neo cortex in isolation.
You need to understand the entire pathway. And that's, for me at least a convenient way of doing that is just a tool that allows you to give access to, that gives you access to all of these things simultaneously.
Benjamin James Kuper-Smith: I was about to say, it sounds like you, I mean, that is exactly what you're doing, right?
You're looking at how, let's say more, uh, more visual areas and hippocampus, for example, interact, I guess you said like in your, your PhD you looked from, sorry, from V one to hippocampus, and now you wanna look at the other way, basically. Mm-hmm. Um, so yeah, yeah. I see how, how
Matthias Nau: it's a network, right? For me, in my understanding, the, the entire brain is a network, no single area evolved in isolation.
Also, the function of an [00:55:00] area. It can only be understood in the context of what all other areas are doing. It's not, you cannot isolate, hippocampus and understand what it is doing. You cannot cut off neocortex. Of course, you can learn a lot of, a lot about the hippocampus doing that, right? You can, of course understand it's anatomy, intrinsic connectivity, maybe it's cell types, all this sorts of stuff.
But ultimately, all the areas are embedded in a network that, that comprises all areas and they are functioning only they, they, in my understanding, at least, they function the way they do because this entire network is there. And yeah, I, I find this network level perspective, um, is necessary and it's at least FMRI.
It does give you access to that. Of course, there's a lot of criticism about FMRI also, because it's imprecise, it's not clear what it actually measures. Uh, temporal resolution is often poor, even though, um, people like [00:56:00] Niko Shook for example, put forward that actually the resolution might be much higher than people, uh, thought so far, temporal resolution.
But yeah, it's, it gives you this global perspective on the brain and it measures really everything that goes on in a voxel, let's say there is a spike. It will also trigger a lot of other spiking, it'll also trigger GL cells, for example, that reuptake the transmitters and all this sorts of stuff. So it measures at least everything that goes on in that voxel, not, let's say these few percent of the metabolic demand that's associated with the actual spike.
So you're in a way Yeah. Gets,
Benjamin James Kuper-Smith: is that, is that a good thing if you're measuring everything? I mean, doesn't that just, if you are, you know, you're measuring stuff you're not interested in, can't that just increase your, uh, signal to No, decrease your signal to noise ratio? I
Matthias Nau: don't think it's really, it's good.
It's neither good nor bad, I would say. It's just important that you know about it, right? You just [00:57:00] need to know what you're measuring. It's not inherently bad or good to measure everything that goes on in a box. And then of course, there are all sorts of other problems, multiple comparisons and all sorts of stuff which need to be addressed, of course, and are being addressed.
And this is really important. Um, but from just from a, yeah, how you say from a, you need to understand, you need to be aware of what you're measuring. That's just all what I'm saying. Yeah. And it's neither good or bad. It's, it's a signal that you get. It tells you something about what the brain does. Yes. It might not be individual spiking neurons.
Uh, it's still for me, very interesting. And I truly believe that it does measure. Indirectly neuron activity. I mean, look at, for example, uh, retinal topic in, in V one, for example. I mean, it's such an amazing strong signal that you get in v in V one, how visual visuals, uh, stimuli are represented in the brain.
You can [00:58:00] actually draw letters into the cortex using different stimuli. You can, let's say if you present an M on the screen, you will see that m reflected in on the cortical surface. I mean, how would that even be possible if it doesn't measure anything meaningful? And I mean, if, let's say the fact that there are a lot of clear cells in there, I'm not an expert on that.
Really not. But maybe it even helps you pick up the signal because it amplifies, uh, the signal if you want. Because let's say there's one neuro spiking, then a lot of cells would take reuptake the transmitter and overall potentially boost the metabolic demand in that box. So in a way it could actually, for FMI be beneficial that there's a lot of stuff happening because it will amplify the neural signal.
Of course it will add noise as well. Um, but yeah, it's,
Benjamin James Kuper-Smith: yeah, and like I, yeah, I dunno much about clear cells either or, [00:59:00] I mean, I had one like small module on neuron glia interactions and, um, I mean the, the, the big picture message I remember from that is that, you know, Lia cells are involved in computations and all these kind of things.
Matthias Nau: Mm-hmm.
Benjamin James Kuper-Smith: They really take up calcium and whatever they release it. Um, at least in some situations. So, I mean, yeah, you would assume that, that there's a decent chance it has something to do with the neuro activity also.
Matthias Nau: Yeah, I think so too. Yeah. I mean, may, it might not be in all of the cases, but in some, and I think there's beautiful evidence that it does.
So in some. Yeah, visual cortex. That's another thing actually in the beginning, because you, you asked me why I studied vision in humans. Why not in animals? Um, just seeing retina toppy in humans and, and, uh, something called population receptor field mapping, uh, just blew my mind as a, as an undergrad, just seeing in your, in basically in your own data, in your [01:00:00] own brain.
I was also part a frequent participant in, in studies in back in tubing and where I studied of course, basically that's how I, uh, how I financed my bachelor's and my masters, if you want. Of course, my parents, uh, supported me as well, but also in addition, I got some, some money on the side for, from just participating in every single project there is, and just seeing basically my own brain activity in this data as in response to these different tasks that I, that I was performing was beautiful for me.
Benjamin James Kuper-Smith: Was that in Andreas Barass group, or was that in,
Matthias Nau: in all sorts of groups? Yeah. But also in, in the group as a, I was, back then, I was basically the pilot subject for most of the studies.
Benjamin James Kuper-Smith: That's pretty good. Going from, from participant to you did your master's project there, or what was it?
Matthias Nau: I did, yeah, my master's, and then I worked as a research assistant in the lab as well for quite a while.
Benjamin James Kuper-Smith: So I'm, I'm not particularly familiar with the, you know, biology curriculum, but I'm somewhat surprised that you'd have like FMRI, Tino [01:01:00] topic images in undergraduate biology. Was that like as part of a, I know, vision module, or how did, how did, how did you come across those first?
Matthias Nau: Actually, I think in the actual, uh, university studies, the only FM RI study we discussed was the dead settleman, which you might be aware of.
Of course. I'm not sure if everybody at the, at the biology department was a big fan of FMRI. Um, yeah, FMI was not a popular topic overall I think I felt at, at university. Uh, but I was always, I had friends, uh, who studied psychology also and who just got into contact with the Max Blank Institute there. Um, yeah, I think I actively reached out to human researchers who then were much closer to FM RI and this kind of thing.
So in the actual studies, I think it gave me a, a, a good understanding of the neurophysiology and overall the physiology, um, is, is what [01:02:00] reeled me in if you want. Right. Understanding how, how stuff maps onto neurons and even how the kidney works, this kind of Yeah. You understand how, how you work and by doing so, you understand, at least I feel, I understand how I.
What I can even know about the world if you want.
Benjamin James Kuper-Smith: Do you think that's, um, so I'm always like, one thing I always, oh, not always, but I occasionally think about is like, what, um, or sometimes when, like even some people ask me who maybe are doing a bachelor's project or something like that, or like, in their bachelor's, what they should do for their masters.
So I'm always curious like what path in a sense people should take if they want to do, let's say something like what you're doing. Um, so I'm curious, what do you, so was your, your bachelor's was in biology and then your master's also, or was that in, what was the topic there? Officially?
Matthias Nau: Uh, neurobiology, neuro
Benjamin James Kuper-Smith: biology.
So I'm [01:03:00] again like a very like cellular kind of course, or was that also, um,
Matthias Nau: cellular and I would say behavioral also. So an amazing, uh, prime course on, on PRI biology and behavior with Andreas nida. And tubing in. That was amazing also, um, I remember I gave a talk on, uh, vocalization behavior, just basically yeah, something like rudimentary, like rudimentary, uh, language if you want.
Um, yeah, I would say it's mostly cellular and behavioral studies that we discussed, probably.
Benjamin James Kuper-Smith: Okay. So I'm curious, like what do you think is the advantage of having that background? Like, is there, is it something like you said earlier, like you, you know, what you can, that you are, they may be more aware of maybe asking what you can find out with certain [01:04:00] signals or, um, what you can even know with these kind of things or what, what do you think is maybe the advantage of having studied biology versus psychology, mathematics, physics, computer science, whatever.
Matthias Nau: Yeah, yeah, it's difficult. It's a difficult question because I didn't study all these other things, but for me, uh, every, my entire thinking is inspired by my knowledge about neurons and single cells and electrophysiology and just biology overall. This is where I come from, and at the end of the day, I use FMRI to study the brain.
And the brain doesn't compute in voxels, right? It, it, it does, it doesn't compute in signal to noise ratios, or actually it might do, but, uh, not an FMRI scale. Um, at the end of the day, what I'm interested in is, is the biology, right, and, and the [01:05:00] cognition that it gives rise to. And for that, at least my thinking is, uh, is based on my bio biological background.
I wouldn't want to miss it. Also, I would study biology again, maybe with a, maybe actually more physics as well. If we could just start over again, I would probably study like a combination of your or yeah, general physics. So the two, kind of, the two, two approaches to reality. So basically studying how the world out there likely is, and then also how I perceive it.
Knowing about my own senses, knowing about how I interface, knowing about how I built, how I acquire knowledge about the world. But in a way, these two things go hand in hand for me. You, if I wanna know what's there and, and out there, I kind of need to understand how I measure stuff, how my body is measuring stuff, how I work, how my brain works.
[01:06:00] And I feel at least, I mean, privately I like to read about, I dunno, physics, also astrophysics and so on. Always mind blown about what they are able to show. Maybe I also don't understand it enough to know what's wrong with it, but, uh, I mean, if people, I don't know if I read a, if I read, actually I do read some papers even sometimes on just some off topic stuff, uh, at least for my own research, irrelevant stuff.
Um, and if I read a paper about how different atmospheres are measured on distant planets, and I feel like this is insane. Um, and I would love to, at least in my, in my studies, I would've loved to have more of that, I think.
Benjamin James Kuper-Smith: Do you think that kind of stuff is, I always wonder whether that stuff is easily teachable or, you know, like, because then it's a module you have to take and then they tell you certain things, whereas.
[01:07:00] If it's kind of something you just explore on your own, I feel like often that can, uh, maybe not enhance the sense of wonder more, but, um, I don't know. Maybe it depends on like what kind of student you also are. Um
Matthias Nau: hmm. I di yeah, I di I didn't quite get that. What, what do you mean exactly?
Benjamin James Kuper-Smith: Well, um, so it sounded to me as if you, uh, meant that it would be cool if in your bachelor's or masters you would've had maybe a module on, well, I'm not entirely sure what, like was it just on like a different topic?
Like something that's like completely left field and has nothing to do with biology or what exactly? Um, would you have liked to have had
Matthias Nau: basically a, a full, uh, full study in physics?
Benjamin James Kuper-Smith: Okay.
Matthias Nau: If you want. Okay. You just wanna do physics. Okay. In addition to a full study in biology? Yeah. Okay,
Benjamin James Kuper-Smith: I see. Okay.
Matthias Nau: Actually, if you think of it, [01:08:00] also knowing about the FMRI signal you actually learn a lot about physics or you need to know a bit about physics. In a way, I feel the research I am doing, and also you are doing, um, is really at the interface of many of these disciplines, right? You, we take, you, you take inspiration from maybe biology, from efis or so, um, but it only matters to you because you wanna understand maybe the, the cognition behind it or psych.
Basically from a psychological point of view, what this means to you, all this biology. But to do so, you need to know how, how your um, how your set, how your measuring device works. If you want, what it measures, at least in fm RI, you need a decent understanding of physics if you wanna do it right, if you develop your own sequences and so on, you are basically forced to dig into more of the details.
Which is good. I think you also need a, from a statistical point of view, uh, need a lot of back [01:09:00] or strong background. Basically, you need to understand all these tests that you're doing, all these data, clean up signal processing, you need to understand programming. So there's some computer science coming in as well.
Um, yeah, it's, I feel like it's not one or the other. You need this bigger perspective and interdisciplinary, um, yeah, all these interactions and I was always lucky that I've been working, um, with people from very diverse backgrounds. For example, uh, psychologists, uh, my lab, my, my office mate here, Marcus Fry, who, um, co-developed Deep MRI with me, he's a computer scientist by training, so machine learning and ai, basically that's where he comes from.
I'm a biologist using FM I and eye tracking and the two of us in the same room. Um, is amazing because you, we try to, or we, we address the same problems, but from a completely different perspective, right? [01:10:00]And the fact that we both have a, a background in coding at least, or some, some programming skills at least gives us a language that we can talk in.
And I think we both have a decent understanding in of, of signal processing, right? All these basics. How, I dunno, how does a filter work? What does, what is convolution? All of these things, if you know about this stuff and a bit of statistics, you can at least interface with these other disciplines a bit easier.
I feel, at least for me, this was the most fruit or very fruitful interaction, um, with Marcus and overall with computer scientists. Then psychologists of course, um, are a lot Christian Doah. My, my PIs a psychologist by training. Um, I had a, a. I supervised a master student who actually pursued a physics study here in, in runtime also, again, basically coming from a different perspective.
Yeah.
Benjamin James Kuper-Smith: So what
Matthias Nau: was many, many,
Benjamin James Kuper-Smith: what was he doing with physics? Like was [01:11:00] it a MRI physics thing or what was his
Matthias Nau: Yeah, he analyzed, we analyzed some data. Yeah, so data analysis project, so 70, 70 data. But I think for him it was also interesting because at the end of the day, this is the, the analysis part, the analytical part of the project and identifying the problem, trying to solve it, um, writing code to solve it, interpreting the results, visualizing it, all this sorts of stuff is basically the same no matter what you do, what kind of field you are in.
And as I said, I mean, FMRI is a physics heavy field if you want. Um, and for him, I think that gave him a natural entryway also. So he came from physics and wanted to know a bit more about the brain, but entered the field via this. How does FM RI work? Kind of, this was his approach to a field, I think, and then he learned more and more about the brain as well.
And I went the other way, right? I, I came from, this is a cell, [01:12:00] um, the middle country, the powerhouse of the cell. So that's a, that's a classic two. How does the scanner work? I dunno what statistics, what is programming?
Benjamin James Kuper-Smith: Yeah, I'm curious about, like, one thing I always find a bit difficult is because it's so interdisciplinary and you can go like, you know, it seems like, uh, like in any direction you want to go, there's just so much stuff you can look at.
Like even if you want to, I mean, if you just do a very simple. Experiment with very simple statistics. You know, just in the statistics you can, you know, I mean, there's people doing PhDs and statistics, right. So I'm curious, how do you decide how much to look into a field? Like how much you need to know for a field, um, let's say for MRI physics, like there's a, you know, there's a point where you have to stop looking into how the physics of it works, otherwise you're doing [01:13:00] physics.
Um, so I'm curious, how do you kind of handle this? Um, well, at least what for me is quite difficult to figure out when I've like, learned something at enough detail and when I should move on with my main stuff.
Matthias Nau: Mm-hmm. Yeah, that's a very good question. Um,
I don't, I don't think I have a very good answer for it. I would think that, when do you know. When you know enough about some basically site side topic?
Benjamin James Kuper-Smith: Yeah. I mean, I, I think I might sometimes go, uh, a bit too far some things where you realize like, okay, reading one paper probably would've been enough.
Like you did need to go like two, three citations into that paper, like references into that paper.
Matthias Nau: Mm-hmm.
Benjamin James Kuper-Smith: I don't know whether this is something you have to like [01:14:00] learn by experience and see like, okay, I don't know. The last few times I, I learned lots of stuff that is just, I just can't use in this, or Yeah.
But then I always find like sometimes you're also maybe, you know, you're missing out on some sort of serendipitous finding because you just don't understand the thing well enough.
Matthias Nau: At the end of the day, I think it's probably a question of practicalities. You need to move forward with your project, right?
At some point you're just restricted by time, limited by time. In theory. I would love to know. Everything about everything that's clearly not possible. Um, and I, I dunno, I don't, I don't think I, I, I don't feel I lose myself too much in some side topics. Maybe I do, but I don't experience it that way. Uh, at some point it just becomes too detailed for me, or I think this is not relevant for my own work.
I think you, you might, I think one good advice could be [01:15:00] to, um, think about how this actually informs what you are doing in your own work, and if you feel like this is actually, or, or maybe some, or how it speaks to some private interest of yours. But at some point you notice, well, this is actually, if you ask yourself what I'm just reading, does this get me anywhere in my private search for knowledge or in my work?
And if the answer is not really, then you might have gone too, too far down the rabbit hole.
Benjamin James Kuper-Smith: Yeah.
Not
Matthias Nau: sure. Yeah,
Benjamin James Kuper-Smith: but have, so I like, like one example that I'm curious like how much, um, for example, I think this is a question that many people, especially those who come from psychology, ask themselves, like, how much math do you need to know? Um, I think, well I think knowing none is probably not a good idea and doing an entire maths degree is, may be a bit much, but like, uh, so how, how have you, I dunno, did you have [01:16:00] any in your undergraduate post, like bachelor's, masters, or did you have to like, learn all of that yourself?
Or how did you go about that?
Matthias Nau: Yeah, so I had some in, of course, in, in bachelor's and masters. Uh, that was also very useful. Um, I wouldn't call myself like a math expert or so, um, I think I know the stuff for me at least. I think it's again, a practicality issue. Um, I don't really like copying code from other people, so I try to implement it myself.
Usually, sometimes I get a script from somebody, but I would never use it. I would actually then look at how it works and then implement my own script based on it. And to implement it yourself, you kind of need to dig into the math, but then you're already, uh, limited to what is relevant to you.
Benjamin James Kuper-Smith: You mean because you have like an outline of what the other person did, or,
Matthias Nau: I mean, you have a, you have an OB objective and you kind of search for the maths that get you there for the math that [01:17:00] get you there and, uh, I don't know.
I mean, overall I feel like I could, I could, I would love to have a much better or more solid understanding of math, um, than I, than I do, but I feel like I do have enough to pursue my work and in the end, at, at the end of the day, a lot of the stuff that you would learn, let's say in a, in a math, pursuing a math degree.
Um, is not directly relevant to your practical work as a biologist, it can inform it. I think if you have a solid background in math, it's amazing. Um, your, your basically, your starting position is a completely different one. Um, but it's not absolutely necessary and you will develop the skills that you need to pursue your work along the way.
If you're dedicated at the, at the end of the day, it's all a question of dedication for me, or maybe if you wanna bring in this word commitment for me, it's all about commitment. Um, [01:18:00] you need to be committed to what you're doing, and if you are committed to neuroscience, then you're automatically committed to learning the skills you need.
And that includes math, some basic knowledge of physics, some psychology, biology you need. Yeah, I think this is the way forward for, for us as a overall, as a discipline. Maybe that's a big, big question now, but overall as a discipline. It's, it's the way forward is to grow in all of these directions and have people that, at the interface of these disciplines and make them talk
Benjamin James Kuper-Smith: ate.
Do you them see yourself see as like a person at the interface, like someone connecting people from other areas or, um, someone who is like, very heavily based on, um, you know, very heavily based on biology and FMRI, whatever. And then you are kind of, do you see what I'm trying to get at? Mm-hmm. Like, I feel like there's some people who are maybe really good at [01:19:00] communicating with lots of different kinds of people and bringing them together and others maybe.
Who are the other kinds of people who then say, okay, well let's say for example, you, uh, come from biology background and you say, okay, I just want to, now I'm collaborating with, um, what's his name, Marcus High. Mm-hmm. Um, a computer scientist. Is, is that kind of how you see your role in this or.
Matthias Nau: I mean it way it might go too far.
I dunno. Um, I'm just a guy like everybody else and, um, I'm doing what I love. I think I, I do bring people together, but I am also brought to other people by other people. Uh, it's an interaction. I wouldn't, I dunno, definitely I do bring people together and I'm really grateful for a lot of people in my surroundings, in my environment that, that connect me to other people.
Uh, so I guess the quick, the, the easy answer is yes, [01:20:00] but so are a lot of other people, and I think this is, this is, again, this is a network issue. It's like the brain, right? Every, everybody co-evolved in a, in a big gigantic network. And uh, yeah, everybody is a connector. If you want, I
Benjamin James Kuper-Smith: have a. If I can change topic a bit.
Yeah. I have one question I'm curious about, and this is very dumb topic almost, but it's the use of Twitter because I think over the last, I don't know, since whenever it was started, I think over the last five, six years or something, I've occasionally started a Twitter account and then never knew what to do with it, and then basically deleted it the next three years later.
I was like, I should maybe try and do Twitter again. And it seems to me as if you are doing, you're quite, I mean, it's not like you're tweeting all the time, um, but it seems to me you, you have quite a few followers for, for a scientist on Twitter. Right. Um, if I remember correctly. And I, I find like you looking at your tweets is [01:21:00] pretty informative, curious.
Thank you. So I'm curious, like how are you, like, how do you, like, what do you put on there? How do you decide what to put on there, what not to put on there? Um.
Matthias Nau: I mean, I, I try to limit it to science. I, I don't really engage with political discussions on Facebook, uh, on, uh, Twitter or Facebook. Um, I'm not active on Facebook at all, actually, but on Twitter, I am active indeed, no politics.
Um, I stay, I try to stay to, um, with science and evidence and papers. Um, it's, I think it's important to do science communication, which is, uh, uh, one of the reasons I'm doing that. The other one is just I love to share stuff that I like. Uh, that's, that goes for science, but that also goes for music. I don't post music on Twitter, but I do spam my, my friends all the time with like, look at this, uh, YouTube video.
This is amazing. This is, this is a great song. So I would also [01:22:00] send these recommendations to people that do not ask for it. Thanks for that, Matt. Exactly. Thanks for that song. But I mean, on Twitter, at least the people who follow you. Yeah.
Benjamin James Kuper-Smith: Yeah.
Matthias Nau: I mean, they could always unfollow you, right? Yeah. Uh, yeah, so I, these are the two reasons I think science communication is important and I love sharing stuff.
I love,
Benjamin James Kuper-Smith: so it's like two fellow scientists. And is it also then like for the general, I mean, whoever the general audience is, um, like you said, science communication?
Matthias Nau: Yeah. So I did, I did for example, for Real scientist that's also a Twitter account website.
Benjamin James Kuper-Smith: So
Matthias Nau: what is, what do they do? Yeah, so they share. So that's, um, you can, different people host that, that Twitter account for let's say a week and then they tweet about their work.
Um, daily, daily stuff, how the work is, uh, kind of about the practicalities of the work. This is my desk. This is my, oh,
Benjamin James Kuper-Smith: sorry. Like
Matthias Nau: literally your work. This is my lunchroom. I see, I see. Exactly. But then also, so your [01:23:00]workplace, but also about your science or that actually might be anything. Um, could be, there could be people that, um.
Work in industry, for example, or do academic research. So it's, it's a, it's a broad scope. I was one of them, one of the hosts speaking to the German public about science and tried to explain to them what grid cells are and me, and, um, play cells and memory formation, episodic memories, all this sorts of stuff.
And I enjoy, enjoyed that a lot. It's also a lot of work, of course, if you wanna do that properly. It's a, it's a lot of work. And also finding the good, a good German phrasing of all these, uh, papers is some, is is specifically also a lot of work in a way, in few. It always feels a bit in, in very few characters.
Indeed. Yeah. I mean, you can make these threats. So I just made threats.
Benjamin James Kuper-Smith: Okay. Yeah, yeah, of course. Yeah.
Matthias Nau: And then actually most of the time I spent searching for figures that I would be able to use, because of course you can't use licensed figures from the actual [01:24:00] papers. You can't just tweet them on Twitter.
Really? I, I don't think so. I think need license it the though, right. Was it just people own
Benjamin James Kuper-Smith: papers
Matthias Nau: or, yeah. Um, I am not a hundred percent sure if that even is allowed, but I will at least, I'm also not saying it's, it's not allowed, but I didn't feel comfortable sharing anything that I don't have a license for.
So I actually, I ended up just, I ended up just searching images that I would be able to use that, that the general audience would understand that, that I could show my, also my, I dunno, my parents who do not have a neuroscience degree and they would be able to understand. And I think that's, again, that's important in the end.
In the end, this is where all this tax money is coming from. This is, uh, the society that we, that we as a scientists, as a scientists also serve in a way you, um, overall for a society to have basic research is great and, um, at least to some degree also luxurious right to [01:25:00] be able to afford this kind of science.
If you are, if you, um, are a society that has a lot of problems, you end up cutting. Often these societies cut the science funding, um, sadly, uh, because the money is just needed elsewhere more immediately. Whereas basic science is often, often the long term project that you need additional money for, and to explain why you need that and why it's important and why your research is important, and why overall science should be funded and why science also basic science is critical.
This depends on the acceptance and the motivation of the, of society as a whole. Um, yeah, chipping in and funding all of this stuff. So I feel, yeah, this is important.
Benjamin James Kuper-Smith: Yeah, I mean, to me that's like, I mean, I think there's a few reasons why I want to do this podcast, but one of them is also that just, you know, there's a lot of people who seem to be interested in these kind of topics and [01:26:00] who, um.
You know, just have a completely different life and don't have the time to, you know, spend hours reading papers or whatever. Um, and I think, yeah, most stuff I think can be communicated fairly easily if you try. Um, at least I, I think especially from psychology, because most of the stuff you study is to some extent about stuff people talk about anyway.
Um, I know like hexa directional symmetry is a bit harder to explain using just words. Mm-hmm. But, um,
Matthias Nau: yeah, the, the funny thing about that signal is if you see it, it's very clear. And if you see a grid cell firing map,
Benjamin James Kuper-Smith: it's the most obvious
Matthias Nau: signal. You know what it is, it's very obvious, but yeah, putting it in words can be overly complicated.
Yeah. And also, I mean, you probably noticed in that PO in this podcast as well. Podcast as well, is that I didn't find the best phrasing yet for grid cells, the most concise one. It's always like, oh, this is how it, [01:27:00] but
Benjamin James Kuper-Smith: yeah. Yeah. I think, I don't know. Yeah. I've also wondered like how to, yeah, I mean maybe I think like one way might be to like link it to stuff that people know, right?
Let's say like a honeycomb is a hexagonal pattern.
Matthias Nau: Mm-hmm.
Benjamin James Kuper-Smith: Um, I dunno how clearly people know that. Uh, like, um, yeah. Yeah. It is, it is a weird thing where I feel like if you just say like, just look at this photo, like, then people will know. Um, there must be like some good real life examples where you can just say like, it's like that.
Matthias Nau: Mm-hmm.
Benjamin James Kuper-Smith: Um,
Matthias Nau: yeah. There are of course. Yeah. Yeah,
Benjamin James Kuper-Smith: yeah.
Matthias Nau: So maybe I can ask you a question. You already. Um, you already men mentioned it a little bit, but so the motivation for you to, to establish this podcast, by the way, I find this amazing that you do that. 'cause you mentioned, um, before this, [01:28:00] before this recording, you also mentioned that this was a long term dream or ambition of yours.
Benjamin James Kuper-Smith: Yeah. I've been thinking about it
Matthias Nau: for a while. To start this, right? Yeah. Or you. Yeah.
Benjamin James Kuper-Smith: So basically I, so what is, I mean, I think I just always, I mean one of the things is I just always like listening to podcasts. So I've been, I don't know, for five years or something. And I think, I think that's at least when I became aware of podcasts and I think before then, you know, I listened to whatever, like long form interviews on TV or whatever, whenever the, they were available.
Um, and I dunno, I just always enjoyed listening to usually. Two people talk. I think in general, I'm not someone who's a huge fan of groups. I, I like, like a dialogue or like two or three people at most. And so I think I've always just always enjoyed listening to people talk about something that they and I find interesting.
Um, and so yeah, so I, I think, I dunno, I think in general when I see someone do something that's cool, I just want to do it also, which is, I think a very [01:29:00] natural habit. I think I have it maybe more than others and it leads to me starting like various different projects, um, which can be a bit overwhelming than sometimes.
But, um, yeah, so then I think for the last four or five years I've been like vaguely thinking about doing something like this, but never really had the, um, I mean one thing is I think four or five years ago, I dunno how good the technology was to do this kinda stuff. I mean, you could, like, you can use Skype and that kind of stuff, but I always felt like the audio quality, they just wasn't that good.
Um, maybe it was, I don't know. It seemed to me like that. And then, I mean, also I felt like I didn't quite have enough stability in sense of like knowing what I wanted to do. Um, and like, you know, like when you're doing your masters or whatever and you're move, I mean, I've been pretty much moving around like between cities almost every year for the last 10 years.
Um, well, not exactly, I had like three years of bachelor's, but apart from that, it's been like a new city every year. [01:30:00] Um, and, but I feel like now I have a bit more stability in the sense that, you know, I know like I have, I have like an income for a few years where I know like I can afford the software or something and the whole thing.
Um, and um, yeah, I thought I'd just try it. I mean, another thing was also, so, I mean, as I said, one thing I think is, um, it's like science. I mean, I'm not like into science communication per se, but, um. I, you know, I do enjoy like, popular science books about topics I know nothing about. Um, but another thing is also just that, you know, it happens to me so often or to everyone properly that you, who's a scientist, you read a paper, you think something's really cool, and you have a few questions about it, and then usually you just kind of have those questions and you try and answer them yourself, or you maybe ask someone who, um, might know something about it.
Uh, but I feel like in most cases, [01:31:00] you, you know, you don't just go up to the author and ask them, at least I don't, um, unless you happen to know them or something. Um, so part of the hope is, I mean, yeah, part of the hope is that I can just talk to people whose stuff I find cool and like ask them about it.
Right? It's just, you know, like, I mean, another thing is maybe, so we have these talks in humble. And, um, like, well, you know about it 'cause you, you were one of the speakers. Um, and then people, you know, they usually come from somewhere else to Hamburg, then they stay like a day or two or something, I dunno.
And then they give a talk at like, well, some point in the day. And then they usually have like, um, they'll send like an email around saying like, Hey, this person's coming, does anyone want to talk to them? And like, um, and that's cool and I've done that before, but it's always so limiting because you usually have like someone who's like spoken about the same things with like four people before you [01:32:00] or something for the last few hours.
They may be a bit tired or jet lacked or whatever. And then, um, it's just not the most efficient way of doing that. So my hope is just, I'll just contact the people I want to talk to and then, um, just learn something about their work and also how they approach stuff.
Matthias Nau: Yeah, it's amazing. Yeah. And I, if I understand you correctly, you also don't only invite scientists, right?
It could be any, anybody.
Benjamin James Kuper-Smith: Yeah, I mean, it's
Matthias Nau: that you think has something interesting to tell.
Benjamin James Kuper-Smith: Yeah, I mean, I'm, I'm just interested in like, loads of stuff. Um, and I mean, yeah, so I, there's just, I mean, I, I read lots of books. I, you know, enjoy film, music, art, whatever, sports, and, um, so there's just, the hope is that, that I can also talk to those people about whatever they do right now.
Um, I mean, like one reason why a lot of my guests will be neuroscientists and psychologists [01:33:00] and whatever is because I know a few of them, for example, it was, you know, much easier to ask you because we've met before, um, than to just like randomly mail someone who's, you know, has no idea who I am and this kind of thing.
So I know a few people in science and. Um, I think it's also easier for me to establish connections there because I can say like, Hey, I read these three papers of yours and I can really show, like I know their stuff. Um, so I think most guests probably will be from that, at least in the beginning. Um, but yeah, the hope is, I mean, I have like a few guests already or a few people who said like they were gonna come on who are, um, not scientists.
Um,
Matthias Nau: maybe eventually you could invite somebody who is running a podcast and discuss basically setting up a podcast.
Benjamin James Kuper-Smith: Yeah. I've already, there's
Matthias Nau: no, I think for people like yourself, people who are planning a podcast, that actually might be very interesting.
Benjamin James Kuper-Smith: Yeah. It's, I mean, these matter things always, I always find 'em slightly annoying.[01:34:00]
Okay. Like a, uh, I mean, like, it, it could be fine, but like, you know, like you read a book about someone who's writing a book, whether. Person in the book, you know, it's like, okay, you're very clever. But, um, but, but in this kind of nonfiction sense, I, uh, I agree. And there, there is already one guy who, oh, I'll not mention the name now just because he might say no.
And, um, but there, there is one guy who I'm gonna ask because he's doing a somewhat similar podcast. Um, yeah, because it's, I mean, yeah, just to, just to see what his experience was like so far. It's, yeah, it would be
Matthias Nau: fun.
Benjamin James Kuper-Smith: Yeah. Yeah. I mean, it is a bit work, putting up the podcast and that kind of stuff, but it's, my, my hope at least is that it's not like a huge amount of work.
We'll see. I mean, like right now I do read like, you know, like I read Your Nature Neuroscience paper, your Takes paper for this. Um, so that, you know, that does require some [01:35:00] effort. Um,
Matthias Nau: of course. Yeah. Yeah. And it makes sense to invite on people that you actually want to read the stuff.
Benjamin James Kuper-Smith: Yeah. Or you often I just have, right.
Like in your case, I hadn't read the Pips before, but I saw your talk, so, um mm-hmm. Yeah. Do, do you wanna start your own podcast?
Matthias Nau: Uh, no. No intentions here at, at this point? No, no time actually. Yeah.
Benjamin James Kuper-Smith: Yeah. We'll see how the time thing works out. Um, yeah, I'm just starting it. We'll see. But I dig myself into hole.
Yeah. I'm not gonna get out of,
okay, so I mean, you mentioned earlier your, is the plan then that you stay in Heim until the end of the year now? Or maybe, maybe can you explain, like, you wanted to go to the us like you finished your PhD, you wanted to start a postdoc in the US in March, or whenever it was, or when did you want to go there first?
Matthias Nau: Yeah, so in March was the initial plan that was postponed to May. Then [01:36:00] this was postponed to June, uh, sorry, July, and now it was postponed to, I don't know, November, who knows? Um, because of the COVID-19 pandemic. Of course. It's, it's really complicated making any plans at the moment. But yeah, so my initial plan was to start a postdoc at the NIH with Chris Baker, was an expert on scene, um, or yeah, on high level visual cortex as a, as a whole.
But I'm very interested in working with him for his work with, uh, on, on scene processing because I think it really connects these low level visual areas to these hippocampal cortices that I'm interested in. So this is the interface, if you want, between the two worlds. Also, between my work with Andreas Bartels and my work with Christian Doah, I feel like Chris Baker is right in the middle.
Bridging all of those levels in a way, bringing those together. Um, yeah. And the plan was to start in, in, in March initially. Now it's, now it's almost August. [01:37:00] So that's a bit frustrating, of course, but the plan still is to move there and I'm, I'm really looking forward to that as well.
Benjamin James Kuper-Smith: Yeah, I mean, it sounds like you are, in a sense, lucky that, you know, that Christian can support you for a bit longer, that kind of stuff, because I feel like there must be a lot of people who are in a similar situation to you right now and who aren't in that position, and then
Matthias Nau: Absolutely.
Yeah. No, I'm also deeply, deeply grateful for that. Absolutely. Yeah. And I mean, Chris Christian and I, we, we also have quite a few projects that I would've worked on anyway, or even after starting at the NIH. So now I just have some more time to actually get those further and. And maybe even fi finish some of those.
Uh, and that's great in a way, right? So I make, I think, I feel, I feel like I'm making good use of that time now, but it was just not how I initially planned it. I mean, this entire world, I mean entire 2020 for everybody. For all of us, it was a mess and confusing, if you want. [01:38:00] And, uh, it was also for me. Yeah,
Benjamin James Kuper-Smith: that's, yeah.
You, you're one of everyone.
Matthias Nau: Exactly. Yeah. The, yeah. I think there are difference a little bit is that I still don't know how it's, when things are moving forward.
Benjamin James Kuper-Smith: I mean, yeah, who knows?
Matthias Nau: I think, I mean, you mentioned that in Hamburg things become a bit more normal now, and also here in Tron time things are back to normal.
In fact, we already scanned here in Tron time again.
Benjamin James Kuper-Smith: Oh yeah. We've been scanning for
Matthias Nau: Oh yeah, okay.
Benjamin James Kuper-Smith: A month or two or something. I can't remember. I mean, basically, I don't know exactly what the trajectory is, but I think for the last two, three weeks now, handbook has, on any given day. Well, on a, on a seven week average, I think they had like 5, 6, 7 cases or something in handbook for the last three weeks.
And, you know, handbook has 2 million people, so it's, it's very limited right now in terms of the number of infections.
Matthias Nau: Mm-hmm.
Benjamin James Kuper-Smith: I know it, it always seems to be like a bit of an ebb and flow right now where you have [01:39:00] like individual outbreaks in Germany. Um, but right now, well I think, I guess general, the northern part of Germany seems to be pretty, like Mechan didn't have a single case, I think for like two weeks or something.
Matthias Nau: Mm-hmm.
Benjamin James Kuper-Smith: Um, but yeah, but who knows how, I mean, I guess you, they like, they're gradually opening stuff again. Um, and then kind of see how that affects transmission rates. Um, but yeah, so far it's, I dunno whether everything, I think pretty much everything is open now in hunt bulk.
Matthias Nau: Yeah. Here in runtime too, I think things are almost back to normal, but then I'm, I always feel like I'm just waiting for something.
I'm waiting to leave and yeah, this is, it's a bit taxing, I must say. And always the short term, uh, this short term postponing is taxing,
Benjamin James Kuper-Smith: like short term. We're talking like, what, like two weeks before it's like, Nope, you're not coming. Or has it sometimes been a
Matthias Nau: I mean, I [01:40:00] always, I mean, I, I, I still don't have the visa and the embassy is still closed and there's no information from the embassy when they will reopen and, uh, they don't know.
So I always end up waiting as long as I can before making any new decisions and new arrangements. That also always means that really on short notice, we just. Change the entire plan. And it's not, it's not a, it's not a small plan. Right? It's, it's, it's moving to the us it's moving our entire lives to the us and I've, I've, I've been planning this for so long now that, um, yeah.
It's, it's about time to actually do it. So
Benjamin James Kuper-Smith: like us means you and your wife or,
Matthias Nau: yeah, exactly. Yeah.
Benjamin James Kuper-Smith: So what does, what is her situation like? Does she, like work-wise or something? Does she have, uh, does she have something there that she's gonna do then there, or, and it's the same thing as you? Or how does that work?
Matthias Nau: Yeah. Luckily she's, she's working here, the, the [01:41:00] hospital. Uh, and, uh, so in the, in the medical genetics department as a, as a tech, how you say? It's a Norwegian, we call it bio engineer. Okay. Yeah. An
Benjamin James Kuper-Smith: engineer or, you
Matthias Nau: know, it's a, um, yeah. Lab, lab tech.
Benjamin James Kuper-Smith: Yeah.
Matthias Nau: If you want, um, running all these analysis that can also be used for, uh, COVID-19.
Um, yeah, diagnosis. So she doesn't have a, a job right now in the US or yet, but, uh, I mean the chances of her finding something are, I think fairly high 'cause of her background and yeah, the immediate application of, of basically her somewhere in the hospital there. And I, I, I mean, I will be working at the NIH.
Benjamin James Kuper-Smith: Sounds like
Matthias Nau: that has a job. Yes. So exactly. It feels like we already know quite a few people who, um, might be interested in hiring, recruiting her as well.
Benjamin James Kuper-Smith: That's good.
Matthias Nau: Yeah. Yeah.
Benjamin James Kuper-Smith: [01:42:00] And okay, so assuming you get there at some point, whenever that might be, do you already have like specific experiments you already, you know, you're gonna do?
Um, or is it still very much like, I don't know, you discussed a general direction with Chris Baker and then you'll see Yeah, you'll figure out the details when you get there.
Matthias Nau: No, I think the experiments are, are laid out, so I have plenty of ideas and, and discuss them with Chris also, I mean, identified the ones that that overlap with his interests most.
And yeah, so these projects are, I think some of the scripts are already done even like, okay, so you i and stuff ready to go. Also, the, also the lab, of course, already has a really, uh, a good foundation in, in, in scripts and a good collection of, of, of original topic mapping stimuli and analysis package, uh, analysis packages and so on.
And also my, i, I myself, have quite a few stimuli that I wrote and, and, and, and never [01:43:00] used and basically just adapted already quite a few scripts. And conceptually these projects, I think are laid out nicely already.
Benjamin James Kuper-Smith: So it's kind of like a, we're
Matthias Nau: just waiting to actually do it,
Benjamin James Kuper-Smith: but you can't do the way you are now or. I dunno, like you have some sort of, you say, well, I guess you say you're already starting other projects now with Christian. Um, but let's say, I know there's a second wave in the US in Autumn, and then it's the, like, you can't come in or you don't want to go in, I don't know, um, for like another half a year or whatever.
Um, is, um, is it like, could you just run the same stuff in one time and it's a collaboration with co group then, or?
Matthias Nau: Yeah, so this is, this is one of the plans for, uh, for that project I mentioned that I'm starting with Christian. It's actually a project that I, that we've been planning for also, for quite a while.
But then I didn't, I, I thought I [01:44:00] wouldn't have time before the US before, before our move to run it here in Heim. Uh, and now, um, I might scan it at the NIH and it will be a collaboration with, um, Chris and Christian. And yeah, I mean, who knows if let's say the US remains closed for the next year or so, then yeah, I might end up scanning it here or, or maybe also in L Christian.