cover john lisman

John Lisman on theta-gamma code and brain oscillations

  • cover play_arrow

    PLAY EPISODE


cover john lisman
Season 2015
Season 2015
Description arrow_drop_down

Description

What if the brain organizes thought not as a continuous stream but as a series of discrete packets, timed by nested brain oscillations? Neuroscientist John Lisman explains how theta and gamma rhythms work together to chunk information into ordered sequences , a coding scheme he proposed 20 years ago that recent experimental breakthroughs have finally confirmed. Subscribe for more from the Convergent Science Network podcast series. John Lisman joins Paul Verschure and Tony Prescott at the BCBT summer school to revisit his influential theta-gamma coding hypothesis, first published with Idiart two decades earlier. The core idea is that within each cycle of the slower theta oscillation (roughly 5–15 Hz), the brain fits approximately six or seven discrete gamma cycles (30–100 Hz), and each gamma cycle carries a distinct piece of information. In the hippocampus, this means different spatial locations are represented at different gamma phases within a single theta cycle , not as a continuous signal, but as an ordered, discretized sequence. The discussion explores what recent data from Foster and colleagues has added to this picture: direct evidence that hippocampal representations jump between discrete positions in space, locked to successive gamma cycles, confirming that the phase code is genuinely discrete rather than continuous. Lisman argues this amounts to a multi-part message delivered in under 100 milliseconds , a compressed movie of a navigational path that downstream structures like the basal ganglia could evaluate for costs and benefits during decision-making. The conversation also tackles deeper questions about whether the brain operates with anything resembling a clock cycle, how pattern completion can occur within a single gamma window, and why the irregularity of gamma timing does not undermine the coding scheme. Lisman, Verschure, and Prescott debate the relationship between episodic and statistical memory, the computational parallels to digital processing, and whether oscillatory codes represent a fundamental organizational principle or just one of many strategies the brain employs. Key topics include the theta-gamma nesting hypothesis, discrete phase coding in hippocampus, working memory capacity, attractor dynamics within gamma cycles, decision-making via sequential replay, and the role of brain oscillations in structuring cognition. Part of the Convergent Science Network podcast series from the BCBT Summer School.

Tagged as:

About the author call_made

CSN Podcasts

Both the triumphs of humanity and its most evil deeds have resulted from collaboration. In a time where humanity is required to aspire to the former and minimize the latter, the question arises of how collaboration arises and why it fails. Surprisingly, this phenomenon, so central to who we are, is not well understood. Hence, a collaborative effort is required to understand collaboration in its full biological, psychological, sociological, cultural, and economic complexity and to translate this understanding into operational impact. This series of podcasts is one step toward achieving these complementary goals. The Collaboration Podcast presents interviews with people who are central orchestrators of collaboration in various domains including business, government, science, art, health, sustainability, and the military. The discussions were conducted by Prof. Dr. Paul F.M.J. Verschure and members of the Program Advisory Committee of the Ernst Strungmann Forum on Collaboration (https://www.esforum.de/forums/ESF32_Collaboration.html) during 2021 and had the goal to sketch a map of opportunities, challenges, and obstacles in human collaboration. The forum took place in May 2022, and now we would like to share this series of interviews with a broader audience. The full report of the Forum will be published in 2023 by MIT Press. The podcast was produced by the Convergent Science Network (https://www.convergentsciencenetwork.org/). Context: The stability of social systems depends critically on realizing sustainable methods of “collaboration,” yet how and by which means collaboration is achieved is not clearly understood; neither are the conditions or processes that lead to its breakdown or failure. Collaboration can be understood as cooperation between agents toward mutually constructed goals. Part of the reason for our lack of understanding is that the phenomenon of collaboration is, by nature, a highly multidisciplinary problem, and effective research into its complexities has been difficult to achieve across the broad range of scientific and technical disciplines involved. The need for a fundamental understanding of collaboration, however, has become increasingly important. Not only does humankind demand answers as it attempts to address critical challenges at multiple scales (e.g., climate change, migration, enhanced automation, social and economic inequality), but ever-increasing technological and economic means of interconnecting people and societies are disrupting long-established, familiar patterns of how we interact. Radical technological changes that are ongoing have the potential to reshape collaboration in ways that are currently hard to predict or influence (e.g., by altering configurations in interaction, information creation, and modes of communication). On one hand, such changes could disrupt hitherto stable forms of collaboration by affecting critical communication channels and traditional roles, as can be observed in the rapidly changing patterns in governance, commerce, and social interaction. Conversely, technology could lead to the emergence of novel, successful forms of collaboration that deviate from traditional “hierarchical” architectures. Evidence of this can be seen in areas as diverse as highly automated manufacturing plants, the open science movement, collaborative software repositories, user-centered services, and the sharing of economy-based modes of organization. Without a fundamental understanding of the mechanisms, processes, and boundary conditions of collaboration, it is not possible to evaluate or predict which of these possible scenarios are sustainable or even plausible. The Forum “How Collaboration Arises and Why it Fails” (May 8–13, 2022, Location: Frankfurt am Main, Germany) Chairs: Andreas Roepstorff and Paul Verschure Program Advisory Committee: Jenna Bednar, Julia R. Lupp, Bhavani R. Rao , Andreas Roepstorff, Ferdinand von Siemens, and Paul Verschure

More posts

Timestamp

  • fast_forward00:00:00 - If this works out, then Tony owes me a lot of beers.
  • fast_forward00:00:08 - This is the Convergent Science Network podcast.
  • fast_forward00:00:13 - Leading researchers in the domain of neuroscience, brain theory,
  • fast_forward00:00:17 - and technology are interviewed by Paul Verschure and Tony Prescott.
  • fast_forward00:00:22 - All right. Okay, it's Paul Verschure with Tony Prescott for the Convergence
  • fast_forward00:00:29 - Science Network podcast.
  • fast_forward00:00:30 - And today we're here with John Lisman, who was speaking at our summer school.
  • fast_forward00:00:35 - And John, actually, this is a really special year for you because it's exactly
  • fast_forward00:00:39 - 20 years ago that you published your paper with IDIARD on tetragammacycle.
  • fast_forward00:00:44 - And 20 years later, you're as excited about it as you were 20 years ago.
  • fast_forward00:00:49 - So why is this such a big deal?
  • fast_forward00:00:53 - Well, ideas are nice, and it's much nicer when you know that they're true.
  • fast_forward00:01:01 - And I think that that's what motivates me.
  • fast_forward00:01:06 - Actually, some people often will say, oh, you know, that's such a great idea,
  • fast_forward00:01:12 - and I think it's so beautiful, irrespective of whether it's true.
  • fast_forward00:01:17 - And I understand that some people feel that way about their work,
  • fast_forward00:01:21 - but I don't feel that way.
  • fast_forward00:01:23 - I feel that it's only beautiful to me if it's true.
  • fast_forward00:01:27 - So if it takes 20 years and it turns out to be true, I'm very happy.
  • fast_forward00:01:32 - Okay, but now just your happiness is not necessarily convincing others that it is true.
  • fast_forward00:01:41 - I guess there's an empirical base for that, or not?
  • fast_forward00:01:48 - I don't know whether other people are convinced. I haven't polled them.
  • fast_forward00:01:54 - Maybe you guys will tell me what you see as the holes.
  • fast_forward00:01:58 - And, I mean, I could talk about some of the holes myself.
  • fast_forward00:02:02 - But in terms of representation, that is, in terms of what you can record,
  • fast_forward00:02:08 - it seems to me completely convincing that different information is held at different
  • fast_forward00:02:14 - theta phases and that it's not a continuous phase spectrum,
  • fast_forward00:02:18 - that it's a discrete phase spectrum, and the discreteness is organized by gamma.
  • fast_forward00:02:25 - So that seems to me experimentally clear. Now what,
  • fast_forward00:02:30 - you know, somebody could argue is that I don't even believe that this is a code
  • fast_forward00:02:34 - until you show me that some downstream network or behavioral system is reading
  • fast_forward00:02:42 - this, and that therefore,
  • fast_forward00:02:46 - if you put some altered information in a particular gamma slot that the animal
  • fast_forward00:02:51 - will behave differently.
  • fast_forward00:02:52 - That would be a reasonable next step. And so maybe...
  • fast_forward00:02:58 - One would really have to show that before one could claim victory.
  • fast_forward00:03:04 - But I wouldn't go that far. Perhaps I think for listeners we need to say what we mean a bit more.
  • fast_forward00:03:10 - Okay. Because not everyone will have watched the talk.
  • fast_forward00:03:15 - So let me try and briefly say what I think you mean.
  • fast_forward00:03:19 - So we have these slower brain rhythms and these faster brain rhythms.
  • fast_forward00:03:24 - Theta is between 5 and 15. Okay. Is that right, roughly?
  • fast_forward00:03:29 - Most people would. We don't really know exactly how to define these things. Okay.
  • fast_forward00:03:35 - But that's not so bad. Let's take that. And then the gamma range is roughly
  • fast_forward00:03:41 - three times that, four times that?
  • fast_forward00:03:43 - Well, maybe from 30 to 100, let's say. Okay. So really quite broad bands.
  • fast_forward00:03:49 - And the proposal is that these are fundamentals to processing in different areas of the brain,
  • fast_forward00:03:56 - and that they combine together so that perhaps gamma is maybe,
  • fast_forward00:04:05 - when you say that we're thinking in discrete ways, we're thinking in the gamma
  • fast_forward00:04:09 - cycle peaks and that the slower theta waves is modulating what can happen in the gamma spikes.
  • fast_forward00:04:18 - Is that right? I think the fundamental idea is that some item,
  • fast_forward00:04:24 - whatever you want to call it, it could be in the hippocampus,
  • fast_forward00:04:27 - a place, is represented by the set of cells that fire in a gamma cycle.
  • fast_forward00:04:33 - So we have a window there of, let's say, five milliseconds.
  • fast_forward00:04:37 - And now we come, and we can call that the first gamma cycle within a theta cycle.
  • fast_forward00:04:44 - And that general concept is called nesting.
  • fast_forward00:04:47 - That is, within a theta cycle, you have room for maybe six or seven gamma cycles, the faster rhythm.
  • fast_forward00:04:55 - The faster oscillations are nested within a cycle of the slower oscillations.
  • fast_forward00:05:02 - So now that I've said that one place would be represented by spatial code in the first gamma cycle,
  • fast_forward00:05:10 - we can say, well, is there evidence that different information is represented
  • fast_forward00:05:18 - during the second gamma cycle?
  • fast_forward00:05:22 - And the answer is, I think now, completely clear that the answer is yes.
  • fast_forward00:05:27 - And so what you wind up having, you know, is some sort of data formatting system
  • fast_forward00:05:33 - such that within one theta cycle, which lasts, you know, in the order of 150 milliseconds,
  • fast_forward00:05:39 - you get to send a multi-part message.
  • fast_forward00:05:43 - There's seven, roughly seven parts to this message.
  • fast_forward00:05:48 - And possibly more, actually. And conceivably more and conceivably less.
  • fast_forward00:05:52 - So, I mean, we don't know the exact frequencies and how they might vary.
  • fast_forward00:05:57 - So you're proposing that the oscillations is providing a computational role,
  • fast_forward00:06:03 - for instance, of chunking or packaging up information.
  • fast_forward00:06:08 - Maybe in distant parts of the brain, you can form things in a package where
  • fast_forward00:06:13 - if they fall within the same theta cycle, they're in some sense part of the same package.
  • fast_forward00:06:19 - Well, if they fall in the same gamma cycle, they're part of the same package, I would say.
  • fast_forward00:06:23 - Right, but then there's a hierarchy here. So there's the theta forming.
  • fast_forward00:06:28 - And so then you can form a much bigger message, like all that would be contained within one theta cycle.
  • fast_forward00:06:35 - So what's a specific example of this? This would be an example of the whole
  • fast_forward00:06:40 - information of a transverse path.
  • fast_forward00:06:45 - So, a transverse path would involve multiple locations between where you are and the goal.
  • fast_forward00:06:51 - And now we have clear examples of a whole path being played out within a theta
  • fast_forward00:07:00 - cycle. But it's not a continuous process.
  • fast_forward00:07:03 - It's discretized by gamma. But that's the more recent work on hippocampus. Yeah.
  • fast_forward00:07:08 - So, and I think your original inspiration was not really linked to hippocampus.
  • fast_forward00:07:13 - The original inspiration is much more linked to ideas about working memory,
  • fast_forward00:07:16 - about if you want a sequencing of memory, that memory has a certain capacity.
  • fast_forward00:07:22 - Yes, you're absolutely right. I mean, the original work was quite confusing
  • fast_forward00:07:29 - to people, and that was fair,
  • fast_forward00:07:30 - because the data that inspired this was data about theta and gamma observed
  • fast_forward00:07:39 - in the field potential of the hippocampus.
  • fast_forward00:07:41 - And then all of a sudden I started applying this to principles of working memory.
  • fast_forward00:07:48 - Most people would not think of working memory as a hippocampal function.
  • fast_forward00:07:53 - So there was kind of a sleight of hand in that original work.
  • fast_forward00:07:58 - And, you know, maybe it was lucky that it turned out not to be wrong.
  • fast_forward00:08:04 - But what was really the first empirical breakthrough that you think you found to support that idea?
  • fast_forward00:08:12 - Well, I think the first strong suggestion that different information was probably
  • fast_forward00:08:19 - encoded in different gamma cycles
  • fast_forward00:08:23 - was John O'Keefe's discovery in the hippocampus of the theta phase code.
  • fast_forward00:08:31 - So you say, well, why am I so excited about this recent work if O'Keefe had
  • fast_forward00:08:39 - already shown known that the phase of firing of hippocampal cells within the
  • fast_forward00:08:47 - Theta cycle was so important.
  • fast_forward00:08:52 - What he didn't show was that the phase was discretized.
  • fast_forward00:08:59 - And so that's what this recent Foster paper shows so beautifully.
  • fast_forward00:09:04 - So now we know that you can't just have any old phase within theta.
  • fast_forward00:09:09 - You can only have certain discrete phases within theta.
  • fast_forward00:09:14 - So that, in a sense, really proves that you have to have something.
  • fast_forward00:09:19 - And, and the, the foster proved that the discreteness arises from gamma.
  • fast_forward00:09:24 - So that really settles the issue. You have, you know, on the order of six or
  • fast_forward00:09:29 - seven discrete chunks of information thrown at you.
  • fast_forward00:09:35 - So in the, in the foster case, they were back in hippocampus.
  • fast_forward00:09:38 - Yeah. And, and you, you jumped forward then 20 years again, because this actually
  • fast_forward00:09:42 - was published this year.
  • fast_forward00:09:44 - Right. Right. And what you show there, so I have a rat or the rat navigates
  • fast_forward00:09:48 - through an environment and what they've managed to do is really very precisely
  • fast_forward00:09:53 - align the play cell response while I'm moving through this environment with
  • fast_forward00:10:00 - the TET and the gamma cycle.
  • fast_forward00:10:02 - Well, they weren't moving through the environment, but this was the readout
  • fast_forward00:10:07 - while the animal was still.
  • fast_forward00:10:10 - But yeah, so in a sense, this very precise work was made possible by the fact
  • fast_forward00:10:16 - that they had so many cells that they were able to decode position on a very fine timescale.
  • fast_forward00:10:24 - And it's only because they could decode position on a very fine timescale that
  • fast_forward00:10:30 - they could find out that actually the animal is thinking about position X and stuck in position X.
  • fast_forward00:10:37 - And then 30 milliseconds later, suddenly it's thinking about position x plus 1.
  • fast_forward00:10:43 - That's the amazing result, that it jumps. And so that's what discrete means.
  • fast_forward00:10:50 - It's not a continuous process. And so that's supporting this idea of a discrete
  • fast_forward00:10:56 - phase code, of a theta-gamma code.
  • fast_forward00:10:58 - I mean, would you, so when you say discrete codes as opposed to continuous codes,
  • fast_forward00:11:04 - That makes me think of a standard computer CPU, which obviously goes in clock
  • fast_forward00:11:09 - cycles, which are determined by an oscillator.
  • fast_forward00:11:13 - Are you thinking this is the brain's clock cycle?
  • fast_forward00:11:18 - Well, let's talk about the concept of a word in computer framework, right? Right.
  • fast_forward00:11:27 - So what is exactly a word in computer?
  • fast_forward00:11:31 - Well, I mean, it's a certain number of bits, and the bits are ordered, right, and, well,
  • fast_forward00:11:43 - automatically we have to say that, and they're ordered in time sometimes, right?
  • fast_forward00:11:53 - So, what are the analogies to the theta-gamma code?
  • fast_forward00:11:56 - Well, one huge non-analogy is the information content.
  • fast_forward00:12:05 - So, this elementary unit in the brain, in the computer, is just zero or one.
  • fast_forward00:12:12 - But what is the information content within one gamma cycle?
  • fast_forward00:12:19 - Well, it's almost infinite. it well i
  • fast_forward00:12:22 - would say slightly different if you think about a standard cpu your
  • fast_forward00:12:26 - clock cycle is how many instructions you
  • fast_forward00:12:29 - can do one instruction during a clock cycle and then
  • fast_forward00:12:32 - you can do millions of those within a second in a modern cpu
  • fast_forward00:12:35 - but the instruction set also changes so that a modern cpu has a much bigger
  • fast_forward00:12:41 - instruction set so rather than just say uh adding one one together you can now
  • fast_forward00:12:48 - do some very sophisticated numerical computations in an instruction.
  • fast_forward00:12:53 - So that makes it faster.
  • fast_forward00:12:56 - And if there is any analogy at all to what's happening in the brain,
  • fast_forward00:13:01 - and I just want to push you on this to see if you think there is,
  • fast_forward00:13:04 - then the clock cycle in the CPU is really saying that,
  • fast_forward00:13:10 - the way we're going to work is that we're going to do instructions one after
  • fast_forward00:13:14 - another sequentially, and we're going to do them at this speed and within that
  • fast_forward00:13:19 - each time step we can just do one instruction,
  • fast_forward00:13:23 - maybe apply one operator to something that's in memory.
  • fast_forward00:13:27 - And in the brain, presumably what you would, if you're going to agree with that analogy,
  • fast_forward00:13:33 - which you might not want to, you would want to say that in certain parts of
  • fast_forward00:13:36 - the brain at certain times there appears to be something like a clock cycle
  • fast_forward00:13:41 - which is organizing computation into these sort of discrete steps.
  • fast_forward00:13:46 - Okay, well, you know, let me.
  • fast_forward00:13:49 - Talk about computation for a minute, and then you can tell me whether you see an analogy.
  • fast_forward00:13:53 - So first of all, let's look at what happens within a theta cycle.
  • fast_forward00:13:59 - Well, we have the first gamma cycle, and we have an ordered set.
  • fast_forward00:14:05 - So already this buys you an interesting analogy.
  • fast_forward00:14:09 - You have an ordered set over 60 or 70 or 80 milliseconds.
  • fast_forward00:14:13 - You have this gamma cycle, the second, the third, and fourth.
  • fast_forward00:14:17 - And so you get order out of that one. Without theta, you wouldn't have order.
  • fast_forward00:14:22 - Now let's look at what happens in a gamma cycle.
  • fast_forward00:14:26 - So first of all, as I was saying before, there's an enormous amount of information,
  • fast_forward00:14:32 - represented within a gamma cycle.
  • fast_forward00:14:33 - This is the brain's way of doing things, is that I'm going to coordinate the
  • fast_forward00:14:38 - firing of the million neurons in the dentate gyrus during one gamma cycle,
  • fast_forward00:14:44 - and some percentage of them, let's say one
  • fast_forward00:14:46 - percent are going to be active and that's going to represent the
  • fast_forward00:14:50 - information that i'm representing in that gamma cycle and so
  • fast_forward00:14:54 - that's you know how that's you know probably a million bits of information because
  • fast_forward00:15:02 - you've got the spatial pattern you know so many different spatial patterns that
  • fast_forward00:15:06 - you can form given a million cells um so this is you know just the opposite of a digital computer,
  • fast_forward00:15:14 - which maybe in one step is zero or one, here we have such a high-dimensional.
  • fast_forward00:15:21 - So the next question is, you know, what kind of computation can you do within a gamma cycle?
  • fast_forward00:15:29 - Is it just the cells of fire that represent X? No.
  • fast_forward00:15:34 - In a paper that Paul and Cesar and I published,
  • fast_forward00:15:40 - we showed that actually it's not unreasonable within that gamma cycle to do
  • fast_forward00:15:48 - what could be called pattern completion or could be called an attractor operation. Why is that?
  • fast_forward00:15:55 - Because, let's say that the network receives input about a pattern,
  • fast_forward00:16:04 - and that pattern is slightly corrupted and incomplete.
  • fast_forward00:16:10 - So, let's say 10,000 cells should fire, but of the 10,000 cells that should
  • fast_forward00:16:16 - fire, in fact, you know, only 5,000 are triggered.
  • fast_forward00:16:22 - Well, it only takes two or three milliseconds for those cells through their connections,
  • fast_forward00:16:29 - with other cells in the network and specific synaptic weights that connect these
  • fast_forward00:16:37 - axons selectively to the cells that are part of that 10,000 pool.
  • fast_forward00:16:43 - And as a result, two or three milliseconds later, bang, the cells that didn't fire.
  • fast_forward00:16:52 - Will fire. I mean, the cells that are part of the pattern will fire,
  • fast_forward00:16:57 - even though they weren't fired by the external input.
  • fast_forward00:16:59 - So here's an example of pattern completion occurring within two or three milliseconds, i.e.
  • fast_forward00:17:07 - Within the 20 or 30 milliseconds that is allotted to each gamma cycle.
  • fast_forward00:17:15 - So yes, you can do computation within a gamma cycle.
  • fast_forward00:17:22 - So, okay, now we can step back from this and you can give me your opinion.
  • fast_forward00:17:27 - How would you compare this to the way a computer works?
  • fast_forward00:17:31 - Well, I think the analogy might be with something like a graphical processor,
  • fast_forward00:17:37 - which can do lots of parallel operations, for instance, on an image in one clock cycle.
  • fast_forward00:17:43 - So that the the only thing
  • fast_forward00:17:47 - that i'm really trying to say is is does the brain have a clock cycle
  • fast_forward00:17:50 - because as you say some people will will will argue definitely not and people
  • fast_forward00:17:55 - will talk about if they want to use a computer computer analogy they will talk
  • fast_forward00:17:59 - about event-based codes so it's a continuous time system and then you get spikes
  • fast_forward00:18:04 - and we do all the processing on those spikes and it's actually,
  • fast_forward00:18:10 - you're losing information when you say, let's have time steps.
  • fast_forward00:18:15 - Well, let me tell you one more piece of information and get your opinion.
  • fast_forward00:18:18 - So there's been a big debate about the anti.
  • fast_forward00:18:24 - There are a group of people who are sort of anti-oscillations.
  • fast_forward00:18:28 - And one of the things that they've pointed out, which is absolutely true,
  • fast_forward00:18:32 - is that the periods of oscillations are not clock-like.
  • fast_forward00:18:37 - The periods, if you look at, let's say, even during a theta cycle,
  • fast_forward00:18:43 - and you look from one gamma cycle to the next and ask the question,
  • fast_forward00:18:46 - you know, how regular is the gamma period?
  • fast_forward00:18:51 - And the answer is, not very regular.
  • fast_forward00:18:55 - And it's not only empirically true, but given current models about how gamma is generated,
  • fast_forward00:19:03 - this sort of ping model that's just a negative feedback between the principal
  • fast_forward00:19:10 - cells, inhibitory cells,
  • fast_forward00:19:12 - then feeding back onto the principal cells, those models also do not predict
  • fast_forward00:19:17 - that this is going to be a regular oscillation. So.
  • fast_forward00:19:26 - If you use the word clock, you'd have to say, oh my God, this is a terrible
  • fast_forward00:19:30 - clock because gamma period is highly fluctuating.
  • fast_forward00:19:35 - So a downstream, one kind of operation that maybe in a computer you could do,
  • fast_forward00:19:40 - but in the brain, according to the theta gamma code, you can't do,
  • fast_forward00:19:45 - do is to say, given that I have seen the output of cells during one gamma cycle,
  • fast_forward00:19:53 - I can predict that exactly 25 milliseconds later, I'll get another big bang, right?
  • fast_forward00:19:59 - That is not true. You cannot do that.
  • fast_forward00:20:03 - But maybe there's also another aspect to this that makes
  • fast_forward00:20:06 - a computer metaphor maybe less helpful because
  • fast_forward00:20:10 - in a computer it you
  • fast_forward00:20:13 - know it largely operates also on
  • fast_forward00:20:15 - a segregation of memory and program or memory and processing right these are
  • fast_forward00:20:22 - strongly separated and to keep the content and the operation synchronized you
  • fast_forward00:20:26 - have to clock the whole thing well but if you look at this hippocampal process
  • fast_forward00:20:30 - it's not so obvious that let's say the memory and the process are actually differentiated.
  • fast_forward00:20:35 - And maybe the problem you will have much more if you are a hippocampus or a
  • fast_forward00:20:40 - brain is to impose some sort of order on these continuous streams of events
  • fast_forward00:20:45 - that you're bombarded with.
  • fast_forward00:20:47 - So maybe more this event-based interpretation that also Tony mentioned earlier.
  • fast_forward00:20:50 - So another way maybe to think about the teta-gamma cycle is that first you try
  • fast_forward00:20:58 - to get to some sort of a temporal segmentation of these continuous input streams through Teta,
  • fast_forward00:21:04 - but you rely very much on these local competitive processes that give rise to
  • fast_forward00:21:08 - Gamma, so the excitatory-inhibitory interactions that are fairly local,
  • fast_forward00:21:11 - to then fill in within those short segments that Teta gives you what the most,
  • fast_forward00:21:17 - let's say, dominant features are in that local process.
  • fast_forward00:21:21 - Because the other thing we shouldn't forget about, it's not that Teta Gamma
  • fast_forward00:21:24 - as we now see it in the hippocampus is all there is to it, because actually
  • fast_forward00:21:29 - gamma is something that is playing out in lots of local volumes in a completely
  • fast_forward00:21:35 - desynchronized fashion,
  • fast_forward00:21:36 - because it is just the range of an interneuron, you could say,
  • fast_forward00:21:40 - that dictates what the sizes of a population of excitatory cells that play a
  • fast_forward00:21:44 - role in generating gamma.
  • fast_forward00:21:47 - So if you go, let's say, a few millimeters further down, you have a very different
  • fast_forward00:21:50 - kind of, let's say, gamma encoding taking place.
  • fast_forward00:21:52 - It has nothing to do anymore with the first one, but still they're all being
  • fast_forward00:21:57 - then, if you want, aligned in time through the slow cycle of theta.
  • fast_forward00:22:02 - And then if you talk about the clock, maybe the clock is much more at the level
  • fast_forward00:22:06 - of the theta cycle than gamma.
  • fast_forward00:22:08 - Gamma is much more asynchronous and local, but the real clock,
  • fast_forward00:22:11 - if you want to talk about the clock, is then the system that dominates the theta cycle.
  • fast_forward00:22:16 - So how regular is theta compared to gamma?
  • fast_forward00:22:21 - I'm not really sure, but my guess would be that theta is maybe even more irregular.
  • fast_forward00:22:27 - In fact, you know, we don't really even know how to distinguish theta in the cortex from alpha.
  • fast_forward00:22:36 - And this is all very confusing right now. In other words, where do these names come from?
  • fast_forward00:22:42 - And the answer is they they
  • fast_forward00:22:45 - came originally from the
  • fast_forward00:22:49 - eeg field and they had
  • fast_forward00:22:52 - zero intellectual basis that is
  • fast_forward00:22:55 - it was just arbitrary they said okay let's just define this range by one greek
  • fast_forward00:23:01 - letter and this range by another greek letter and so you know we've just all
  • fast_forward00:23:07 - been very confused by you whether to make a big deal out of the difference between
  • fast_forward00:23:11 - 8 Hz in the hippocampus, 7 Hz in the hippocampus,
  • fast_forward00:23:15 - and maybe that's the same kind of process as 10 Hz in cortex. I mean, who's to say?
  • fast_forward00:23:21 - But we have maybe a handle on that, right? Because we know the origins of these oscillations, right?
  • fast_forward00:23:29 - So gamma is seen to arise out of a local circuit where excitation inhibition interacts, right?
  • fast_forward00:23:36 - I think there's consensus on that, whether it's cortex or hippocampus.
  • fast_forward00:23:38 - Fair enough. This is seen as a source of gamma.
  • fast_forward00:23:40 - Okay. That means then that rhythmicity and regularity will depend on the local
  • fast_forward00:23:45 - properties of that circuit.
  • fast_forward00:23:47 - Right. Theta and hippocampus is seen as arising from the septum.
  • fast_forward00:23:51 - Okay. It's really as a driver of a slow, of this slow oscillatory response.
  • fast_forward00:23:55 - For the cortex, it might be the thalamus that dictates, or it will be the thalamus
  • fast_forward00:24:00 - that dictates the slow rhythm.
  • fast_forward00:24:02 - Right. So, if you now look at the sources of theta and gamma.
  • fast_forward00:24:07 - Well, as you said, I think you were on the right track when you said maybe what's
  • fast_forward00:24:12 - really important in terms of interactions between brain regions is the slow oscillation.
  • fast_forward00:24:19 - And that does maybe have to be synchronized between areas that communicate.
  • fast_forward00:24:24 - Maybe the gamma's not so important. It's all local.
  • fast_forward00:24:27 - I think that the notion of a clock, although in a computer it's going to be a metronome.
  • fast_forward00:24:34 - It doesn't have to be, because the key word for me that John said was discrete.
  • fast_forward00:24:40 - So you have a cycle, and it may be varying in its timing, but as long as the
  • fast_forward00:24:45 - different parts of the system that need to understand each other are on the same point in the cycle,
  • fast_forward00:24:52 - then the order that you're imposing by saying, I'm only going to.
  • fast_forward00:24:57 - Treat things within a gamma cycle as being together
  • fast_forward00:25:00 - and i'm going to take these bunch of events that happen with the
  • fast_forward00:25:03 - gamma cycle and treat them as one event in some
  • fast_forward00:25:06 - way and process that so that seems
  • fast_forward00:25:09 - to me the strong claim you're making that the the brain buys some uh some usefulness
  • fast_forward00:25:16 - from this binding of everything together into discrete chunks and that this
  • fast_forward00:25:22 - is maybe makes more sense than just having all these continuous processes.
  • fast_forward00:25:27 - Because if I'm a bit of the brain over here, listening to this bit of the brain
  • fast_forward00:25:31 - over here, I'll have to listen continuously and integrate everything that's
  • fast_forward00:25:35 - happening in order to know what's going on.
  • fast_forward00:25:39 - But you're saying, I get a series of snapshots with the gamma cycle,
  • fast_forward00:25:43 - or maybe with the theta cycle, and I can just pay attention to what's in that snapshot.
  • fast_forward00:25:48 - I don't like that metaphor either.
  • fast_forward00:25:52 - That no i i like it but i want to i want to
  • fast_forward00:25:54 - i want to propose that we use another word okay instead
  • fast_forward00:25:58 - of snapshot movie right and why
  • fast_forward00:26:01 - do i like that because let's say
  • fast_forward00:26:04 - in the experiments that we've been discussing today in various forms the johnson
  • fast_forward00:26:09 - reddish experiment where the animal in a sense gets a view of what happened
  • fast_forward00:26:16 - if it goes down one path uh of the maze so what do I mean by view?
  • fast_forward00:26:23 - Well, what the data show is that each point along the path is represented, and that is a movie.
  • fast_forward00:26:32 - That's not a snapshot. It's a sequence of snapshots. It's a sequence of snapshots, which is a movie.
  • fast_forward00:26:39 - And I mean, that's really impressive in terms of decision -making,
  • fast_forward00:26:46 - because you're getting that movie, you know, within less than 100 milliseconds,
  • fast_forward00:26:51 - and that's very valuable information.
  • fast_forward00:26:54 - And I was discussing, Paul, I was discussing before with Tony one thing I really,
  • fast_forward00:26:58 - really like about now imagining how the basal ganglia evaluates this movie.
  • fast_forward00:27:05 - Um the movie can be rich the movie can say you know i i turned last time i turned left,
  • fast_forward00:27:13 - and uh you know the first thing i came to was you know some breadcrumbs and that was okay,
  • fast_forward00:27:21 - and then i went on and uh and i smelled some cat urine and i thought for sure i was going to get eaten,
  • fast_forward00:27:27 - but finally I got to the reward site and there was some really yummy milk.
  • fast_forward00:27:37 - Okay, I mean, so you have costs and benefits about the left choice.
  • fast_forward00:27:42 - Well, all of that information is passed within less than 100 milliseconds to the basal ganglia.
  • fast_forward00:27:50 - And I think that current current models of the basal ganglia would allow you to actually integrate,
  • fast_forward00:27:58 - to evaluate each of those steps within, let's say, one gamma cycle,
  • fast_forward00:28:03 - get the value or the costs and the benefits of.
  • fast_forward00:28:09 - And wind up at the end with some sort of integrated number which told you the
  • fast_forward00:28:15 - sum of the costs and the benefits.
  • fast_forward00:28:18 - And that's wonderful. I mean, and we know that that's the way we want to run
  • fast_forward00:28:23 - a railroad is not just to look at what's at the end of the path,
  • fast_forward00:28:28 - but to know that, you know, what we're going to risk if we take that path along the way.
  • fast_forward00:28:35 - It was sort of the point there is that Another element I think that you have
  • fast_forward00:28:41 - not emphasized yet, even though you're dealing with it,
  • fast_forward00:28:45 - what's maybe more important is that also if you are a brain,
  • fast_forward00:28:49 - which we all are in some way, you have to progressively throw away more information
  • fast_forward00:28:55 - and distill things down to what really matters.
  • fast_forward00:28:58 - And also the way you think about
  • fast_forward00:29:01 - the gamma cycle, it's also a progressive deletion of noise, if you want.
  • fast_forward00:29:09 - Because also what you're saying, in the gamma cycle, I can generate,
  • fast_forward00:29:14 - what, seven plus or minus responses within a theta cycle.
  • fast_forward00:29:20 - But these are all the result of a local winner-take-all. So what I'm going to
  • fast_forward00:29:24 - report in this tata cycle is the stuff that really stands out for me among possibly
  • fast_forward00:29:29 - hundreds of possible responses.
  • fast_forward00:29:32 - So if I'm in this maze, like we take the Johnson and Reddish experiment,
  • fast_forward00:29:36 - I'm at the T crossing in the maze.
  • fast_forward00:29:41 - I imagine I'm going to the left. So now I have a sweep through my hippocampal
  • fast_forward00:29:46 - representation, right?
  • fast_forward00:29:48 - Within this gamma cycle, I have 100 milliseconds, and now what pops out are
  • fast_forward00:29:57 - actually the most relevant spots I might visit there. It's not all possible spots.
  • fast_forward00:30:01 - So it's all this incremental selection that might be an important role there,
  • fast_forward00:30:05 - and not only how the information is processed further downstream. Right.
  • fast_forward00:30:10 - Okay, well, now we get into a really complicated area of episodic memory.
  • fast_forward00:30:13 - So, I mean, strictly speaking, you know, what you would recall,
  • fast_forward00:30:18 - if we really think in terms of episodic memory, is one of the times that I went down that path.
  • fast_forward00:30:25 - If you take the point of view that the hippocampus is really the episodic store
  • fast_forward00:30:34 - and that what Johnson and Reddish is seeing is a recall of one previous traversal,
  • fast_forward00:30:41 - not the statistical properties of all previous traversals of that path,
  • fast_forward00:30:46 - but one of them, then we get into this whole area of,
  • fast_forward00:30:51 - you know, is this the way you want to railroad? road?
  • fast_forward00:30:54 - Do you want to, you know, do you want to use your hippocampus to remember individual events?
  • fast_forward00:31:01 - And then if so, which ones, which exact ones do you recall?
  • fast_forward00:31:04 - Or do you want your hippocampus to be statistical?
  • fast_forward00:31:09 - I take the point of view that, you know, the great thing about the hippocampus
  • fast_forward00:31:15 - is that it recalls an individual, at least it hopes to recall an individual trial.
  • fast_forward00:31:22 - And that has pluses and minuses. It says, you know, that time on July 4th when
  • fast_forward00:31:27 - I went down this path and, you know, I thought there was going to be a cat,
  • fast_forward00:31:31 - but, you know, there was actually a buzzing going on and there wasn't a cat.
  • fast_forward00:31:37 - So this time, if there's not a buzzing, if there's a buzzing,
  • fast_forward00:31:41 - I'll take the risk because maybe there's not a correlation.
  • fast_forward00:31:44 - I mean, all that detail could be important in making your next choice,
  • fast_forward00:31:48 - but that's not statistical.
  • fast_forward00:31:51 - And I mean, I love the idea that people have shown that actually.
  • fast_forward00:31:57 - You know, if you use your episodic memory, you will often make worse choices.
  • fast_forward00:32:05 - So my favorite example, and I've heard that these experiments are a little controversial,
  • fast_forward00:32:13 - they actually, they did the following. They showed people two houses,
  • fast_forward00:32:19 - and they went through each house. And this was done probably on a computer.
  • fast_forward00:32:25 - And they said, oh, here's the kitchen. It's a wonderful, modernized kitchen,
  • fast_forward00:32:28 - but here's the bath. It's not really ever been modernized. And here's the bedroom.
  • fast_forward00:32:33 - You know, it's really large, but here's the living room. It has a great view.
  • fast_forward00:32:36 - And the costs and benefits of the two houses were very different.
  • fast_forward00:32:41 - And in fact, they had arranged it so that they thought that most people would
  • fast_forward00:32:46 - pick house one as the better house.
  • fast_forward00:32:48 - Okay, now they had two groups. After showing them these two houses,
  • fast_forward00:32:54 - one group was told, please do this arithmetic problem.
  • fast_forward00:32:58 - And the other group was told, you know, think about this house.
  • fast_forward00:33:07 - Then they asked them, which house is better?
  • fast_forward00:33:11 - And, you know, one can argue about whether it's possible to decide objectively
  • fast_forward00:33:16 - that one house is better. But anyway, the people who did not think about it were better.
  • fast_forward00:33:22 - Now, why would this be? Well, there's actually a very interesting and,
  • fast_forward00:33:26 - I think, useful explanation.
  • fast_forward00:33:28 - Probably what happens when you're using your thinking is that you get stuck in silly examples.
  • fast_forward00:33:36 - You say, you know, that kitchen reminded me of my least favorite aunt.
  • fast_forward00:33:42 - And, you know, she, damn it, you know, I went to her house and she gave me this
  • fast_forward00:33:46 - porridge that was just disgusting.
  • fast_forward00:33:49 - And you're spending all your time thinking about this aunt, so you actually
  • fast_forward00:33:52 - aren't thinking about all the other rooms.
  • fast_forward00:33:56 - And so it turns out that your basal ganglia which is sort of more statistical
  • fast_forward00:33:59 - actually winds up doing a good job because even you know your episodic memory
  • fast_forward00:34:04 - isn't sampling very well anyway.
  • fast_forward00:34:10 - Look i i think i think there's a problem here uh i'm sorry to to bring the bad
  • fast_forward00:34:15 - news but earlier you said and you referred to a paper we published together
  • fast_forward00:34:20 - that that memory is is based on attractor dynamics.
  • fast_forward00:34:25 - And by virtue of that, you can do pattern completion, for instance.
  • fast_forward00:34:30 - But an attractor dynamic has, by necessity, statistical properties because it
  • fast_forward00:34:36 - will pull input states towards some average state.
  • fast_forward00:34:40 - So in that sense, to then declare that attractor state as a factual,
  • fast_forward00:34:45 - accurate representation of a single sampling, I think would be a really long shot.
  • fast_forward00:34:52 - How are you going to defend your attractor from drifting and that's one of the
  • fast_forward00:34:55 - things we showed we explained we explained also in that paper that the drift that people saw,
  • fast_forward00:35:02 - in the kind of say a three memory when you change the environment so now we
  • fast_forward00:35:06 - are actually visiting the houses we were slowly changing them that there is
  • fast_forward00:35:10 - a drift in that memory and that drift is also an expression of a statistical
  • fast_forward00:35:14 - property that you are averaging over experience right so.
  • fast_forward00:35:20 - Isn't the idea of an attractor dynamic based memory actually forcing you to
  • fast_forward00:35:25 - also commit to a more statistical interpretation of episodic memory well let
  • fast_forward00:35:31 - me try to see if i can argue the other way um i mean you know.
  • fast_forward00:35:41 - Why is the attractor relevant when you you know you think about the kitchen
  • fast_forward00:35:45 - in that house you You were shown, and it was because it was really very similar
  • fast_forward00:35:51 - to your aunt's kitchen. So there's the attractor.
  • fast_forward00:35:54 - So we've utilized the attractor.
  • fast_forward00:35:58 - But if it was just generalized kitchen, right?
  • fast_forward00:36:05 - You're saying it's really a kitchen attractor. It's not my aunt's kitchen attractor.
  • fast_forward00:36:10 - Then it would bring up associations with all kitchens, which might be very positive on average.
  • fast_forward00:36:16 - But in fact what happened was that it.
  • fast_forward00:36:22 - Was not generalizable that kitchen just got you stuck in your aunt's kitchen
  • fast_forward00:36:28 - i think we're possibly looking at a false dichotomy here because you what you
  • fast_forward00:36:33 - want for your episodic memory,
  • fast_forward00:36:35 - is to do pattern completion so you want to be able to fill out what's happening
  • fast_forward00:36:42 - now based on things which have happened in the past which are relevant which
  • fast_forward00:36:47 - may help you interpret that situation.
  • fast_forward00:36:49 - So you're in that kitchen, you're reminded of your aunt's kitchen.
  • fast_forward00:36:53 - And that may also make you think that, well, maybe this kitchen has got some
  • fast_forward00:36:57 - of the properties of that other kitchen I was in, and that might be useful for
  • fast_forward00:37:00 - understanding that scene.
  • fast_forward00:37:02 - So the pattern completion is going to partly depend on many possible past events
  • fast_forward00:37:08 - which are relevant to interpreting this one.
  • fast_forward00:37:12 - And then the other thing you want to do is pattern separation.
  • fast_forward00:37:15 - So you want to, and this is you want
  • fast_forward00:37:18 - to be able to say but this isn't my aunt's kitchen so although i
  • fast_forward00:37:22 - remember it as being similar it's not
  • fast_forward00:37:25 - the same thing so your episodic memory has to be able to distinguish
  • fast_forward00:37:28 - this kitchen from previous ones which is what you're
  • fast_forward00:37:31 - you're saying is that we needed to do both these things
  • fast_forward00:37:34 - and the fact that i think it can do
  • fast_forward00:37:36 - both but it doesn't do either of them perfectly is the
  • fast_forward00:37:39 - thing we want to try and explain about episodic memory that sounds
  • fast_forward00:37:43 - like a very nice compromise i'll go i'll go
  • fast_forward00:37:46 - for that good at
  • fast_forward00:37:50 - least john will not demolish the studio now that's great
  • fast_forward00:37:53 - but so now so so now we have an idea about this episodic memory in the hippocampus
  • fast_forward00:37:58 - and we want to use that as a way to understand coding in the brain right and
  • fast_forward00:38:03 - what we see there is that this episodic memory that's playing out on on the
  • fast_forward00:38:07 - slow oscillation of theta in the gamma cycle is um.
  • fast_forward00:38:13 - Is telling us, let's say, what's the most relevant aspects of the processing
  • fast_forward00:38:18 - going on within this piece of volume of the hippocampus.
  • fast_forward00:38:22 - But now if you go back to the data of Foster that you talked about earlier,
  • fast_forward00:38:27 - where they have, with very great temporal precision, unpacked the response in the gamma cycle,
  • fast_forward00:38:34 - in some sense it showed, again, a further complexification classification because
  • fast_forward00:38:38 - it's not that on every single gamma cycle there's just a single spike.
  • fast_forward00:38:43 - You have multiple spikes riding within the gamma cycle.
  • fast_forward00:38:48 - So can you imagine that the nesting would go further than only theta gamma,
  • fast_forward00:38:52 - but it also go, let's say, low and high gamma?
  • fast_forward00:38:58 - Well, first of all, you know, we don't really understand anything about the
  • fast_forward00:39:03 - differentiations between low and high gamma.
  • fast_forward00:39:06 - And there's one really, you know, accepted place where you see these simultaneously,
  • fast_forward00:39:12 - or at least overlapping,
  • fast_forward00:39:15 - and that's in the high gamma that comes into CA1 from the cortex and the low
  • fast_forward00:39:21 - gamma that comes into CA1 from CA3.
  • fast_forward00:39:28 - And it's really just unclear to me what this all means. I don't think anybody knows.
  • fast_forward00:39:35 - One surprising thing is you might say, let's look at the output and see who
  • fast_forward00:39:41 - the output is listening to?
  • fast_forward00:39:44 - And the answer is neither. The output just has its own gamma,
  • fast_forward00:39:50 - not tied to either of them.
  • fast_forward00:39:53 - So, I mean, one very simple way of looking at it, which I favor, you know, is that,
  • fast_forward00:40:01 - you know, inputs are coming in and the cell is integrating them on its own time
  • fast_forward00:40:06 - constant and deciding how to output the information on its own.
  • fast_forward00:40:12 - That's not tremendously satisfactory, but that seems to be the best description,
  • fast_forward00:40:19 - that I can come up with of what has actually been found.
  • fast_forward00:40:22 - But that seems reasonable, right? Because let's say I'm in CA3,
  • fast_forward00:40:26 - I'm packing up information in my gamma cycle based on a local competitive process.
  • fast_forward00:40:31 - Now, what I'm telling CA1 about, the next station upstream or downstream from me.
  • fast_forward00:40:37 - Well, that would be the subiculum. The downstream from CA1 is the subiculum.
  • fast_forward00:40:41 - I was just talking about CA3, right? No, I was talking about CA1.
  • fast_forward00:40:45 - Okay, right. So downstream of CA1. No, but I want to start at CA3 because I
  • fast_forward00:40:48 - want to say, basically CA3 makes a preselection and said, okay,
  • fast_forward00:40:52 - these are the few items that I care about.
  • fast_forward00:40:55 - So it has filled out a lot of noise.
  • fast_forward00:40:58 - Now CA1 will do something similar because it now gets this whole barrage of
  • fast_forward00:41:02 - inputs from CA3 still and it has to make a selection.
  • fast_forward00:41:04 - So again, it selects within its own gamma cycle using the same competitive process
  • fast_forward00:41:09 - to tell the subiculum what it really cares about.
  • fast_forward00:41:11 - So it's really also a hierarchy then of competitive processes.
  • fast_forward00:41:16 - So this is how you think about it. I don't know. I think this is just going to...
  • fast_forward00:41:21 - I personally would think that, you know, a lot of effort should now be put into
  • fast_forward00:41:26 - understanding CA1 because it's just so fascinating and what's happening and
  • fast_forward00:41:31 - we have so much data and something, some principle is going to emerge,
  • fast_forward00:41:35 - but I'm not, I honestly have no idea what it is, but I actually think I'd like to spend some time.
  • fast_forward00:41:41 - Seeing if I can figure it out.
  • fast_forward00:41:43 - So then early in the week, we also had quite some information presented to us
  • fast_forward00:41:51 - by Alfred Moser on the grid cells.
  • fast_forward00:41:52 - And the grid cell story actually for the last 10 years looks so clean and nice
  • fast_forward00:41:58 - because here we had these cells in the entorhinal cortex input to the hippocampus,
  • fast_forward00:42:02 - have a very nice grid-like response to space driven by velocity.
  • fast_forward00:42:06 - Velocity and this might then
  • fast_forward00:42:09 - be a great substrate for place cells and hippocampus to
  • fast_forward00:42:12 - learn to learn about space right now that
  • fast_forward00:42:15 - story has become more complicated because it looks like place cells
  • fast_forward00:42:18 - can develop place fields in hippocampus even if you have no grid cells available
  • fast_forward00:42:22 - to you so that the place cells are way more promiscuous in that sense than initially
  • fast_forward00:42:27 - anticipated so that might give us some space to also speculate about what these
  • fast_forward00:42:34 - grid cells really are for if they're not really,
  • fast_forward00:42:38 - the only source of spatial information.
  • fast_forward00:42:41 - So it was up to you. How else would you abuse the grid cells?
  • fast_forward00:42:45 - Well, we very much favor a new view of grid cells.
  • fast_forward00:42:51 - And the question is, you know, what functions are there out there that you have
  • fast_forward00:42:56 - to account for and which ones do you want to assign to grid cells.
  • fast_forward00:43:00 - So classically, what people have said is, well, you have sensory information
  • fast_forward00:43:05 - about the outside world, you know, here I am.
  • fast_forward00:43:08 - People have said, path integration, I integrate velocity, so if I knew where
  • fast_forward00:43:13 - I was, I integrate my velocity, I know where I am now.
  • fast_forward00:43:18 - And then this interesting phenomenon, which we've talked about earlier,
  • fast_forward00:43:23 - on the basis of Johnson and Reddish, which is I'm not moving,
  • fast_forward00:43:28 - but my mind is going to move down one of the paths. And we call that mind travel.
  • fast_forward00:43:35 - So in the end, we have these three functions.
  • fast_forward00:43:39 - How are we going to build a system that does all three? That's what we've got to do.
  • fast_forward00:43:45 - And by my way of thinking, the grid cells are really well suited for doing this mind travel.
  • fast_forward00:43:53 - So they move you in an imaginary way through space.
  • fast_forward00:43:58 - They can integrate, but what they integrate is not your actual velocity,
  • fast_forward00:44:04 - but some artificial velocity which part of the brain generates saying,
  • fast_forward00:44:10 - I'm curious about what happens if I move down in this direction with this velocity.
  • fast_forward00:44:15 - And then the grid cell system can give you the answer.
  • fast_forward00:44:20 - It's a coordinate system that says, well, given this velocity in this direction,
  • fast_forward00:44:27 - this is where you will be, and it gives you that path.
  • fast_forward00:44:32 - The beauty then is that it can force that movement through space on the hippocampus.
  • fast_forward00:44:40 - What can the hippocampus contribute to that?
  • fast_forward00:44:44 - It can contribute associations that happened with those places.
  • fast_forward00:44:50 - So if there was reward or a cat associated with some place that you mind travel
  • fast_forward00:44:56 - to, that's incredibly valuable information.
  • fast_forward00:44:59 - So I don't think that the grid cells know about cats or food,
  • fast_forward00:45:06 - but they do know about the coordinate systems of space.
  • fast_forward00:45:09 - So you move through space and you tell the hippocampus, here's how we're moving through space.
  • fast_forward00:45:17 - And then the hippocampus says, given that you've made me move through space,
  • fast_forward00:45:23 - this is what I found associated with that.
  • fast_forward00:45:27 - That so then we
  • fast_forward00:45:31 - would say well great if if if that's
  • fast_forward00:45:34 - what grid cells are good for don't we
  • fast_forward00:45:38 - also have to do path integration well we
  • fast_forward00:45:41 - there we can make some suggestions that it
  • fast_forward00:45:44 - might occur earlier in the structure and there's
  • fast_forward00:45:47 - some interesting candidates about where path integration might occur
  • fast_forward00:45:50 - so that's the integration of real velocity so um
  • fast_forward00:45:53 - i have a bit of a worry about this story
  • fast_forward00:45:56 - because you know edward moser presented this
  • fast_forward00:45:59 - beautiful data set and his interpretation
  • fast_forward00:46:03 - or my interpretation of his interpretation is that
  • fast_forward00:46:06 - we have these modules at different levels of
  • fast_forward00:46:10 - granularity which are giving us this wonderful
  • fast_forward00:46:13 - metric map of space so that when you come into this room that metric is laid
  • fast_forward00:46:21 - out over the space so that every point in space now has a unique code assigned
  • fast_forward00:46:27 - to it across these different modules of grid cells.
  • fast_forward00:46:31 - And it's anchored possibly at the door or some other point. All the grid cells
  • fast_forward00:46:36 - always seem to have an anchor at the place that that is introduced.
  • fast_forward00:46:40 - Now I've got this metric map of space.
  • fast_forward00:46:45 - I can navigate in it. But your suggestion seems to be that you're going to take
  • fast_forward00:46:52 - my map of space and you're now going to start moving it around.
  • fast_forward00:46:55 - So I'm no longer on firm ground knowing what my coordinate frame is because
  • fast_forward00:47:01 - my coordinate frame is shifting.
  • fast_forward00:47:04 - You're moving it around. Or have I misinterpreted what you're saying?
  • fast_forward00:47:08 - Maybe misinterpreted because when you described Edward Moser's proposal,
  • fast_forward00:47:14 - I found myself nodding, not nodding to sleep, but nodding in agreement.
  • fast_forward00:47:19 - And so the answer is, I think that I would agree with everything you said.
  • fast_forward00:47:26 - It's a coordinate system which allows navigation.
  • fast_forward00:47:31 - The question is, you know, what do you use that coordinate system for?
  • fast_forward00:47:35 - And the nice thing about it, it's a pre-wired coordinate system.
  • fast_forward00:47:40 - So you can use it in any environment. you just have to
  • fast_forward00:47:43 - attach it yeah to a given environment and
  • fast_forward00:47:46 - now you have the means for saying well if
  • fast_forward00:47:49 - i want look this way you know where will
  • fast_forward00:47:52 - i wind up or if i go this way where will i wind up which
  • fast_forward00:47:55 - is what i'm saying right so i don't think we're i
  • fast_forward00:47:58 - don't think we're at odds uh the only um no
  • fast_forward00:48:04 - i i guess i don't see them i don't see them as yeah as country
  • fast_forward00:48:07 - okay i guess you're you're putting the ability to
  • fast_forward00:48:10 - move within the map in the grid cells which are themselves the
  • fast_forward00:48:13 - coordinate frame map and i worry a little bit whether you can do that whether
  • fast_forward00:48:17 - the the grid cells just have to form the coordinate frame and then some other
  • fast_forward00:48:21 - system can represent movement within it okay so i think you did say one thing
  • fast_forward00:48:25 - that really i'm forced to to come back to and admit that's a a problem and suggest a solution.
  • fast_forward00:48:33 - And that is, if we use the grid cell system for.
  • fast_forward00:48:40 - These imaginary movements, then how do you reset it at the end of a theta cycle
  • fast_forward00:48:47 - so you're back to current position?
  • fast_forward00:48:51 - Because we've sort of mucked with this. Right? And my,
  • fast_forward00:48:58 - proposal is that that we do have another integrator which is integrating true
  • fast_forward00:49:05 - velocity and which is, in a sense, doing what everybody thought the grid cell system would do.
  • fast_forward00:49:13 - So, you know, you could say, well, I've made life more complicated now because
  • fast_forward00:49:18 - now I have a whole other system which is keeping track of actual position,
  • fast_forward00:49:23 - another grid-like system that's keeping track of actual position,
  • fast_forward00:49:27 - and that can, since it knows where I am, can reset the grid cell system to current
  • fast_forward00:49:36 - position after I muck with it and make it do this imaginary movement.
  • fast_forward00:49:41 - You were absolutely right. You have to, after you do the imaginary movement,
  • fast_forward00:49:46 - you have to move back to current position.
  • fast_forward00:49:47 - So, somebody must know current position.
  • fast_forward00:49:51 - But, I mean, it's not, I mean, this complexity is demanded by the fact that
  • fast_forward00:49:58 - we see these imagined movements through space.
  • fast_forward00:50:02 - So, somebody's got to be doing this.
  • fast_forward00:50:05 - So, if you take another model, they also have to solve of these dual requirements.
  • fast_forward00:50:11 - I think both of you are much too pessimistic about this because on the one hand.
  • fast_forward00:50:18 - Why would I even need a parallel system for that?
  • fast_forward00:50:21 - Because I just need to replace my velocity vector, right?
  • fast_forward00:50:25 - Either the velocity vector comes from my vestibular system or optic flow,
  • fast_forward00:50:31 - whatever, that tells about physical movement in space.
  • fast_forward00:50:33 - And with that, I'm driving my grid cell security response.
  • fast_forward00:50:36 - This signal is arriving driving in the entorhinal cortex over the thalamus.
  • fast_forward00:50:40 - So I have a number of stages in this pathway where I can basically hijack the
  • fast_forward00:50:45 - signal. I can pump in anything else I want.
  • fast_forward00:50:48 - I drive my grid cells around again.
  • fast_forward00:50:50 - Now you have to, and the point is, so far the data, at least on rats,
  • fast_forward00:50:54 - has shown that if they might travel, like in the sweeps we've seen Johnson and
  • fast_forward00:51:00 - Reddish or Pfeiffer and Foster, the animal's standing still.
  • fast_forward00:51:05 - The animal's not moving, right? I'm not aware of any data where the animal is
  • fast_forward00:51:10 - actually doing this kind of mind travel as it is moving.
  • fast_forward00:51:13 - So that means as I now start to move, I take over this whole channel again to
  • fast_forward00:51:18 - pipe a velocity signal into my grids.
  • fast_forward00:51:20 - Let me stop you, Paul, because I'm afraid I have to tell you that the phase
  • fast_forward00:51:24 - precession, the interpretation of the phase precession that I think everybody
  • fast_forward00:51:28 - would accept at this point is that while the animal is moving, it is looking ahead.
  • fast_forward00:51:32 - But not in this extreme way, as is shown in, let's say, the Johnson and Reddish
  • fast_forward00:51:39 - sweeps or the Pfeiffer and Foster paper. But it nevertheless is.
  • fast_forward00:51:43 - And I mean, the interesting thing is now when we come to experiments, right? So….
  • fast_forward00:51:48 - Why do I think I'm on the right track? Because this looking ahead that occurs
  • fast_forward00:51:53 - while the animal is moving, which is reflected in this phenomenon called phase precession,
  • fast_forward00:51:59 - that does disappear when you get rid of the grid cells or when you get rid of
  • fast_forward00:52:03 - the medial entorhinal cortex.
  • fast_forward00:52:05 - So there is some experimental basis for this, but the play cells survive.
  • fast_forward00:52:11 - So we're beginning to get experimental support for this dissociation.
  • fast_forward00:52:15 - In fact, it's this sort of dissociation that was the demise of the previous theories.
  • fast_forward00:52:24 - We're starting to get a lot of data which now begins to differentiate between different models.
  • fast_forward00:52:29 - This model that we're putting together now, I think, is at least consistent
  • fast_forward00:52:33 - with all the existing data. It may fall apart when new data comes,
  • fast_forward00:52:35 - but this is where we're at.
  • fast_forward00:52:37 - I think I'm more happy with your model if I'm understanding it correctly.
  • fast_forward00:52:42 - Which we have this metric map
  • fast_forward00:52:45 - which is anchored to the room and all you're
  • fast_forward00:52:47 - saying is that the grid cell activity in
  • fast_forward00:52:51 - the grid cell can represent the fact i'm moving around in
  • fast_forward00:52:54 - an imaginary way in the room and that different patterns
  • fast_forward00:52:57 - of grid cells are firing to represent those different
  • fast_forward00:53:01 - points i'm visiting that will then feed through
  • fast_forward00:53:04 - to ca3 ca1 to represent the the non-spatial features of those locations in space
  • fast_forward00:53:12 - what it would look like if i was standing there looking around and yeah i'm
  • fast_forward00:53:17 - fine with that because that my metric map is anchored and.
  • fast_forward00:53:22 - When i when i open my eyes and look then yeah i'll be my my grid cell map i'll
  • fast_forward00:53:27 - be back at the point my grid cell map where i really should be so i'm not sure
  • fast_forward00:53:31 - there is a problem that you're describing i i thought you were co-opting the
  • fast_forward00:53:35 - grid cells to do something other than have a stable, wet graph of space.
  • fast_forward00:53:40 - No, I was not. It's only the velocity factor that's changing, actually.
  • fast_forward00:53:43 - Right. And, you know, McNaughton has tried to generate models in which all you
  • fast_forward00:53:51 - do is you have one sort of grid cell system which is responding both to real
  • fast_forward00:53:57 - velocity and to artificial velocity.
  • fast_forward00:54:02 - So the idea is that.
  • fast_forward00:54:05 - I know where I am because I've integrated real velocity, but now I'm going to
  • fast_forward00:54:09 - add an additional velocity and move imaginary through space.
  • fast_forward00:54:14 - But, and now what he does is has a speculation about how you can subtract out
  • fast_forward00:54:19 - that artificial velocity and
  • fast_forward00:54:23 - it says make it a negative, and so that puts you back to where you are.
  • fast_forward00:54:27 - So here you could deal with two kinds of velocity in the same network.
  • fast_forward00:54:32 - And my only difficulty with that is that it's computationally difficult to have
  • fast_forward00:54:41 - two velocity terms and then subtract out the effect of only one of them.
  • fast_forward00:54:46 - It's not impossible, but it's not elegant.
  • fast_forward00:54:50 - And therefore, I favor the view that you just keep it elegant and have one guy
  • fast_forward00:54:55 - who's dealing with true velocity and another guy who's dealing with the artificial velocity,
  • fast_forward00:55:02 - but who can be reset after all is said and done by the guy who's keeping track
  • fast_forward00:55:06 - of real velocity. It seems more elegant to me.
  • fast_forward00:55:09 - But also what you would be saying is that in my look ahead with this mind travel.
  • fast_forward00:55:17 - I have a finite capacity to look ahead.
  • fast_forward00:55:20 - I can only do this for how many teta cycles?
  • fast_forward00:55:24 - It's one full teta cycle with the seven odd gamma responses I have within that?
  • fast_forward00:55:30 - Yes. So if you want to now look ahead, let's say if you are in an environment
  • fast_forward00:55:35 - that you know very well, like Woods Hole,
  • fast_forward00:55:37 - you might be able to look ahead basically to any kind of street,
  • fast_forward00:55:41 - any possible trajectory, with much higher precision and much further depth in
  • fast_forward00:55:46 - space than you can do in a novel city like here, Barcelona.
  • fast_forward00:55:49 - So how then can I change the spatial scale and my resolution on which I can
  • fast_forward00:55:54 - look ahead if everything has to happen within a single tether cycle?
  • fast_forward00:56:01 - Well, we strongly suspect, based on some evidence, that the whole system is just replicated,
  • fast_forward00:56:10 - along the long axis of the hippocampus at larger and larger scales.
  • fast_forward00:56:19 - So the dorsal part of the hippocampus, which is the one that's almost exclusively
  • fast_forward00:56:24 - studied these days, keeps track of very small distances. I mean,
  • fast_forward00:56:29 - we're talking about distances, you know, of a foot or so.
  • fast_forward00:56:34 - In contrast, the ventral hippocampus, when you look at the size of the place
  • fast_forward00:56:40 - fields, when you look at the phase precession, you know, is looking over meters.
  • fast_forward00:56:43 - I'm changing units here, but that's okay.
  • fast_forward00:56:46 - I'm in Europe. Yeah, don't worry about it. Confusion is expected.
  • fast_forward00:56:53 - So you know and and you know we
  • fast_forward00:56:56 - were talking with david reddish the other day about you know why do we
  • fast_forward00:56:59 - know so little about the ventral hippocampus is because it's
  • fast_forward00:57:02 - so hard to study it's hard to get an electrode in there um and you know it's
  • fast_forward00:57:09 - so easy to generate lots and lots of data because you can get in the dorsal
  • fast_forward00:57:14 - hippocampus you not only get an electrode in there easily but you get a zillion
  • fast_forward00:57:18 - electrodes in there easily.
  • fast_forward00:57:20 - So you get really rich data sets, and it just isn't true.
  • fast_forward00:57:24 - But, you know, at this point, you know, some data is available about the ventral
  • fast_forward00:57:31 - hippocampus, and then I think we'll have, you know, more answers to your question at that point.
  • fast_forward00:57:35 - But the available data does show that the scale changes.
  • fast_forward00:57:39 - So then, do you see this as a unique feature of hippocampus and toralocortex,
  • fast_forward00:57:44 - this ability for mind travel?
  • fast_forward00:57:46 - Travel it's unique or can cortex also mind travel hmm that's a very interesting
  • fast_forward00:57:52 - question maybe this is a good point to end i have no idea about that.
  • fast_forward00:57:57 - No no we're not going to end here john so um
  • fast_forward00:58:02 - so it's okay let's let's let's
  • fast_forward00:58:05 - see let's see what the future brings us but now so this
  • fast_forward00:58:07 - is amazing this is a great moment you're going to all buy his
  • fast_forward00:58:10 - beers later on 20 years of teta gamma coding right
  • fast_forward00:58:14 - and we still could not talk it out of your head even not in this this podcast
  • fast_forward00:58:17 - interview i think you're still going strong on that thank you having having
  • fast_forward00:58:22 - you know fought this battle now for so long and still standing and smiling what
  • fast_forward00:58:29 - is john's law that we should adhere to to study and understand the brain,
  • fast_forward00:58:33 - to study the rest of the brain everything what's john's law what what assures
  • fast_forward00:58:39 - progress and understanding the brain.
  • fast_forward00:58:41 - Wow. You're on the right of the wall here. John's law. Right,
  • fast_forward00:58:44 - John's law. You just have to define it.
  • fast_forward00:58:48 - Yeah, that it's not as complicated as you think.
  • fast_forward00:58:52 - That is to say that, you know, if we think it through carefully,
  • fast_forward00:58:58 - you know, we'll see organization principles fall in.
  • fast_forward00:59:02 - So I definitely think that people have underestimated how quickly we will come
  • fast_forward00:59:10 - to understand the brain.
  • fast_forward00:59:11 - And I think that understanding the brain is a lot like solving,
  • fast_forward00:59:15 - you know, other kinds of puzzles.
  • fast_forward00:59:18 - And when you look at the process, let's say of a jigsaw puzzle,
  • fast_forward00:59:21 - you know, at first you don't see the patterns, you know, you don't know how
  • fast_forward00:59:26 - to recognize certain, you know, uniformities of colors and patterns.
  • fast_forward00:59:31 - And you also don't have many constraints but
  • fast_forward00:59:35 - everything changes you know when you get near
  • fast_forward00:59:38 - the end you understand the rules you have constraints
  • fast_forward00:59:41 - from other um you know pieces of the puzzle and everything just goes very fast
  • fast_forward00:59:47 - as you get near the end and i think we you know we are getting to that point
  • fast_forward00:59:52 - okay so don't get confused by complexification right so now the The other thing is, John,
  • fast_forward01:00:00 - as you know, Tony likes traveling and he likes hamburgers and Captain Kidd.
  • fast_forward01:00:06 - So five years from now, he wants you to take him to Captain Kidd in Woods Hole.
  • fast_forward01:00:10 - And then he will confront you with a prediction you're going to make today that
  • fast_forward01:00:15 - you will have proven right or wrong by then.
  • fast_forward01:00:17 - So what's the most important prediction you would like to commit yourself to
  • fast_forward01:00:22 - that needs to be verified by that time? Oh, well, that's easy.
  • fast_forward01:00:27 - Because, you know, even older than the theta-gamma idea is the molecular basis of memory,
  • fast_forward01:00:36 - KAM kinase, that now we have refined that model to say that it's the complex
  • fast_forward01:00:43 - of KAMK2 with the NMDA channel at the synapse.
  • fast_forward01:00:48 - That is the actual molecular memory. And in my own lab,
  • fast_forward01:00:53 - we've done the critical test to see whether we could erase LTP by attacking
  • fast_forward01:00:59 - the CAMK2-NMDA complex, and we could.
  • fast_forward01:01:04 - So that was great. And so now we have set up upon the next phase of our work,
  • fast_forward01:01:11 - which is to erase a behavioral memory.
  • fast_forward01:01:15 - So we have everything set up. We have a nice learning task of spatial memory.
  • fast_forward01:01:23 - We have a virus which contains a dominant negative form of CAMK2.
  • fast_forward01:01:29 - And so what we're doing is to put the virus into CA1 and see if we can make the animal forget.
  • fast_forward01:01:36 - And we've even arranged this to be a spectacular form of virus which does just
  • fast_forward01:01:44 - what we want. It turns out to be an HSV virus.
  • fast_forward01:01:47 - What's so spectacular about it is that it only expresses for two days.
  • fast_forward01:01:53 - And so what we can do is teach the animal something, inject the virus,
  • fast_forward01:01:58 - let it do its thing, but then it's gone.
  • fast_forward01:02:02 - Now, if the memory is gone, when we measure 10 days later, all those negative,
  • fast_forward01:02:08 - nasty reviewers, they can't say, well, the virus is just mucking things up,
  • fast_forward01:02:13 - because then we can say, no, the virus is gone.
  • fast_forward01:02:16 - And we really, and the only reason that, you know, the memory is not there anymore
  • fast_forward01:02:23 - is because we've erased it.
  • fast_forward01:02:25 - That's what we want to say. We want to say we were able to erase it.
  • fast_forward01:02:29 - Cool. And so we think we have the, so if this works out, then Tony owes me a lot of beers.
  • fast_forward01:02:38 - Great. John Lissman, thank you for this conversation. Okay. Thank you. Thank you.
  • fast_forward01:02:44 - Music.
  • fast_forward01:02:49 - The CSN podcast was produced by the Convergent Science Network of Biometrics
  • fast_forward01:02:54 - and Biohybrid Systems, a project funded by the European Sevens Research Framework Programme.
  • fast_forward01:03:02 - For more interviews, recorded lectures, or upcoming conferences in the field
  • fast_forward01:03:07 - of biometrics and biohybrid systems, go to csnnetwork.eu.
  • fast_forward01:03:14 - And thank you for listening.
  • fast_forward01:03:20 - That's great, on the anniversary of your paper. Yes, let's see. How old do I have to get?
  • fast_forward01:03:28 - Before I prove the KMK2 hypothesis, if it has to be in years of five,
  • fast_forward01:03:33 - I'm not going to live that long.
  • fast_forward01:03:36 - Well, that sounds like a fantastic experiment. Thank you for generating a prediction,
  • fast_forward01:03:41 - on a topic that we hadn't discussed at all.
  • fast_forward01:03:46 - That was very helpful, John. It was fun. Yeah, it was very good.

Be the first to leave a comment

Leave a comment

Your email address will not be published. Required fields are marked *

Convergent PozitivLinia Logo

Exploring the convergence of neuroscience, robotics, and AI through conversations with leading researchers since 2010.

A project of the Convergent Science Network Foundation.

© CSN Podcasts. Developed by IMCreativeWEBC

0%

Login to enjoy full advantages

Please login or subscribe to continue.

Go Premium!

Enjoy the full advantage of the premium access.

Stop following

Unfollow Cancel

Cancel subscription

Are you sure you want to cancel your subscription? You will lose your Premium access and stored playlists.

Go back Confirm cancellation