WEBVTT 00:00:03.517 --> 00:00:10.497 This is the Convergent Science Network podcast. Leading researchers in the domain 00:00:10.497 --> 00:00:16.737 of neuroscience, brain theory and technology are interviewed by Paul Verschur and Tony Prescott. 00:00:20.537 --> 00:00:26.397 This is Paul Verschur, Convergent Science Network podcast together with Tony Prescott. 00:00:27.297 --> 00:00:29.417 And we're here with Francesca Caccucci. 00:00:30.977 --> 00:00:35.557 Who was speaking today about her work on the development of the hippocampus. 00:00:37.177 --> 00:00:42.637 Where actually you gave a great overview of how we can think these hippocampal 00:00:42.637 --> 00:00:45.117 circuits to develop in the rat. 00:00:46.037 --> 00:00:49.757 But if we now look at this system from a developmental perspective, 00:00:50.417 --> 00:00:53.397 what are the key features we want to understand? 00:00:54.777 --> 00:01:00.557 Okay, so what motivated me to start looking at the development of the spatial 00:01:00.557 --> 00:01:02.397 signals in the hippocampus. 00:01:02.577 --> 00:01:07.617 So try and understand a bit about the interrelationships and the relationships 00:01:07.617 --> 00:01:10.257 between the different components of the spatial system. 00:01:10.437 --> 00:01:13.237 So we know about place cells that encode location. 00:01:13.477 --> 00:01:16.597 We know about head direction cells, which give a sense of direction. 00:01:17.137 --> 00:01:21.417 And we know about grid cells, which may encode distance traveled. 00:01:21.517 --> 00:01:27.537 But what we didn't know, and we still, I argue, don't know, is how much they 00:01:27.537 --> 00:01:30.397 interact, these three components, in order 00:01:30.397 --> 00:01:34.557 to give a complete percept and representation of a low-centric space. 00:01:35.297 --> 00:01:42.757 And also, not only the interactions, but also how they are actually, 00:01:42.877 --> 00:01:48.517 what kind of signals feed into these single representations to give rise to these complex signals. 00:01:49.697 --> 00:01:56.337 Okay, so to anchor that, you also defined if you want a working hypothesis on function, right? 00:01:56.377 --> 00:02:02.817 So you emphasize very much this notion of episodic memory as a key function of the hippocampus. 00:02:04.072 --> 00:02:07.192 But how should we think about that in terms of a rat? 00:02:07.372 --> 00:02:12.312 I mean, because also the paradigms we're going to discuss, do you feel that 00:02:12.312 --> 00:02:17.152 these really uniquely probe this sort of rat construct of episodic memory or 00:02:17.152 --> 00:02:18.412 are we actually looking at something else? 00:02:18.952 --> 00:02:25.252 Okay, so my work doesn't speak at all towards episodic memory development or otherwise. 00:02:25.452 --> 00:02:29.652 And episodic memory, as I mentioned in the lecture, is very difficult to probe 00:02:29.652 --> 00:02:37.012 outside the human domain. Okay, so there have been notable exceptions. 00:02:37.172 --> 00:02:41.392 So Richard Morris has looked at this, Nicky Clayton has looked at episodic memories. 00:02:41.612 --> 00:02:48.292 Memory in animals, but it's difficult to do so because most of what we understand 00:02:48.292 --> 00:02:54.452 about episodic memory can be understood from verbal reports. 00:02:54.452 --> 00:03:01.752 So even when you ask humans about their episodic memories, that's already fraught with difficulties. 00:03:01.972 --> 00:03:08.272 So my work doesn't speak towards the development of episodic memory because 00:03:08.272 --> 00:03:12.492 I think it's too difficult in the animal model. 00:03:12.492 --> 00:03:16.812 And that's why what I talked about was more about the associative properties 00:03:16.812 --> 00:03:24.472 of memory in hippocampal circuits, and I stayed away from the episodic kind of specific. 00:03:24.472 --> 00:03:27.532 Specific but then so you gave us 00:03:27.532 --> 00:03:30.692 as a starting point an overview of the developmental trajectory 00:03:30.692 --> 00:03:33.612 of rats right where so in 00:03:33.612 --> 00:03:38.472 this development of of the rat pup to become an adult rat what do you see as 00:03:38.472 --> 00:03:42.232 the main steps that we should keep in mind when we now start to look at the 00:03:42.232 --> 00:03:47.192 development of these circuits okay so what is surprising to me so first of all 00:03:47.192 --> 00:03:51.592 as i said the rats are developing develop slowly like human infants so they 00:03:51.592 --> 00:03:53.332 go through certain steps that 00:03:53.452 --> 00:03:56.152 kind of are paralleled by human infants, 00:03:56.332 --> 00:03:58.872 both in terms of sensory and motor development. 00:03:59.272 --> 00:04:04.572 But what is interesting for me in terms of development are when you start seeing 00:04:04.572 --> 00:04:06.152 sudden transition in development, 00:04:06.352 --> 00:04:12.292 because development can be thought of a kind of successive refinement of, 00:04:12.492 --> 00:04:17.412 I don't know, sensory information or motor planning and execution, 00:04:17.712 --> 00:04:19.632 and that's kind of gradual generally. 00:04:19.632 --> 00:04:25.312 But then you see there are some things that happen during development and they happen. 00:04:26.479 --> 00:04:32.619 Quickly, sudden transitions. And one of them is the emergence of exploratory drive in the rat pup. 00:04:32.819 --> 00:04:36.259 And this was studied by Linna Dell, amongst others. 00:04:36.539 --> 00:04:41.179 And it seems quite interesting that this spatial exploration drive emerges within 00:04:41.179 --> 00:04:46.879 each pup all of a sudden around when they are 19 or 20, 21 days old. 00:04:47.019 --> 00:04:51.399 And so why is it? What is the critical process that makes this transition happen? 00:04:51.759 --> 00:04:55.359 And another interesting and parallel transition we have found in our own work 00:04:55.359 --> 00:04:58.739 is the emergence of the grid cells, which is really kind of abrupt. 00:04:59.059 --> 00:05:04.059 Right. So what is the key ingredient that is missing up until, 00:05:04.159 --> 00:05:08.359 let's say, the animals are 20 or 21 days old, and that kicks in and makes this 00:05:08.359 --> 00:05:12.399 representation all of a sudden stable to the point and extent that you can see 00:05:12.399 --> 00:05:15.079 it, you can detect it, for example, for grid cells. 00:05:15.419 --> 00:05:19.439 But now, you talked about this analogy with human development, 00:05:20.339 --> 00:05:23.539 but it's like in humans, an order of magnitude slower, right? 00:05:23.539 --> 00:05:27.159 So what happens in days in the rat happens in years in humans. 00:05:27.219 --> 00:05:29.699 Yes, but also we have a different lifespan, of course. 00:05:31.439 --> 00:05:36.259 But from that perspective, you would say they're still the same maturational 00:05:36.259 --> 00:05:38.379 processes. Humans are just slower. 00:05:40.119 --> 00:05:44.839 Possibly, yes. And also there is in terms of the parallel between brain maturity 00:05:44.839 --> 00:05:49.359 at birth in the human versus the rodent is quite different. 00:05:50.339 --> 00:05:55.739 But again, drawing direct parallels is always a bit controversial and difficult to do. 00:05:55.939 --> 00:05:59.319 So I don't want to be… But for you, the main anchoring point is really this 00:05:59.319 --> 00:06:00.319 transition to exploration. 00:06:00.599 --> 00:06:03.099 That's really for you the main… I think that's quite important. 00:06:03.819 --> 00:06:09.599 I think the big difference between humans and rats is that we are born with 00:06:09.599 --> 00:06:11.359 our eyes open. Yes, absolutely. 00:06:13.318 --> 00:06:16.958 When they open their eyes, they're still not exploring very much rats, 00:06:17.198 --> 00:06:22.778 but they're just beginning to be able to move around outside the nest. 00:06:22.938 --> 00:06:29.718 So you have this cluster of sensory and motor systems which are allowing exploration, 00:06:30.038 --> 00:06:35.758 but they can begin to explore outside the nest from 10, 11 days later. 00:06:36.178 --> 00:06:38.958 Yes, it really depends on the environmental conditions. It doesn't sound like 00:06:38.958 --> 00:06:43.938 it's, I mean, the 19-year-old rat pup is really quite mature in its sensory and motor capability. 00:06:45.198 --> 00:06:49.158 Yes. So what I'd like to distinguish is between the emergence from the nest, 00:06:49.258 --> 00:06:52.998 so when they leave the nest for the first time, and if you make the environment 00:06:52.998 --> 00:06:54.678 very warm, for instance, 00:06:54.778 --> 00:06:59.458 you can get rats that come out of their huddle, their group of siblings, 00:06:59.618 --> 00:07:01.278 their nest, very early on. 00:07:01.378 --> 00:07:05.978 But in general, especially in the wild, it won't happen until they are even 00:07:05.978 --> 00:07:08.938 25, 26, 27. seven days old. 00:07:09.218 --> 00:07:14.398 And one thing is just emerging from the nest. So this just coming out of the nest. 00:07:14.518 --> 00:07:17.298 And the other thing is organized behavior, exploratory behavior, 00:07:17.458 --> 00:07:22.818 which is this thing of systematically looking and orienting towards objects 00:07:22.818 --> 00:07:26.498 and things that are around and systematically sampling the environment. 00:07:26.838 --> 00:07:31.338 And that's what Nadel was looking at when he said that it tends to happen around 00:07:31.338 --> 00:07:33.658 when they are 19 to 20 days old. 00:07:33.698 --> 00:07:36.938 And these are laboratory razia, of course, that we're talking about. 00:07:36.938 --> 00:07:41.958 And that's always, we need to qualify this because we don't know what happens in the wild. 00:07:42.218 --> 00:07:47.938 And there is very little work, as far as I am aware of, in the wild. 00:07:48.838 --> 00:07:53.638 Yeah, we had a talk last week from Lea Krubitzer, who's actually doing some 00:07:53.638 --> 00:07:59.358 experiments with semi-wild rats, laboratory rats that have been left in a wild 00:07:59.358 --> 00:08:01.558 place and exhibiting behaviors. 00:08:01.918 --> 00:08:05.478 So it might be interesting for you to follow what she's doing. Yeah. 00:08:05.900 --> 00:08:08.640 This is exactly the kind of work that I'm very interested in. 00:08:08.860 --> 00:08:16.380 Tell me, so we see a correlation and we can look now at the emergence of direction cells, play cells, 00:08:16.680 --> 00:08:23.560 and then grid cells, and see if we can correlate that with these behavioral 00:08:23.560 --> 00:08:26.000 transitions in the developing rat. 00:08:26.720 --> 00:08:32.420 What you showed us is that direction cells are there all along this postnatal 00:08:32.420 --> 00:08:36.760 day 16 and until the rat decides to step out of life. 00:08:36.760 --> 00:08:40.960 Even before, even when they are 12 or 13 days old. Right, so they're really 00:08:40.960 --> 00:08:42.240 the earliest ones. Yeah. 00:08:42.700 --> 00:08:47.500 Then you suggested play cells emerge gradually in that period, 00:08:47.820 --> 00:08:57.240 while the grid cells would then emerge very rapidly around postnatal 20, more or less. 00:08:57.240 --> 00:09:02.260 And that seems to align rather nicely with this exploration behavior. 00:09:02.700 --> 00:09:04.600 So do you really see those as coupled? 00:09:05.640 --> 00:09:10.540 Yeah, but I can only speculate. Yes, it's tantalizing this kind of temporal 00:09:10.540 --> 00:09:16.820 coincidence between the emergence of stable grid cells and the emergence of 00:09:16.820 --> 00:09:20.500 all the kind of both the exploratory, so natural drive towards exploration, 00:09:20.760 --> 00:09:24.460 but also when you test animals on hippocampal dependent tasks that you know 00:09:24.460 --> 00:09:26.280 they're hippocampal dependent in the adult animals, 00:09:27.100 --> 00:09:30.680 the pups can start solving this task only from weaning onwards, 00:09:30.980 --> 00:09:33.140 which is the time when we start seeing these grid cells. 00:09:33.260 --> 00:09:37.360 So it is interesting that these things happen at the same time. 00:09:37.560 --> 00:09:41.780 But at the moment, we haven't done the experiment where you either delay the 00:09:41.780 --> 00:09:47.940 emergence of stable grid cells and you see what happens to the emergence of navigation behavior. 00:09:48.200 --> 00:09:50.820 So until you do the intervention of the experiment, you cannot... 00:09:50.820 --> 00:09:51.880 It's just a correlation. 00:09:51.980 --> 00:09:55.860 It's an interesting kind of... But now do you see these transitions as also 00:09:55.860 --> 00:10:01.220 sort of reflecting the kind of stage-wise transitions that developmental psychologists 00:10:01.220 --> 00:10:02.520 have observed in humans? 00:10:03.451 --> 00:10:07.771 That's a difficult question. So which kind of stage-wise transition? 00:10:07.771 --> 00:10:13.251 Jean Piaget would argue, well, at certain ages, you are not able to perform logical operations. 00:10:13.491 --> 00:10:19.131 Yeah, yeah, yeah. Jean Piaget had, specifically for the special domain, 00:10:19.531 --> 00:10:26.531 he thought that children were, how do you say, trapped into egocentrism, 00:10:26.611 --> 00:10:31.471 so egocentric processing until they were very old, 10, 12 days old, years old, sorry. 00:10:33.451 --> 00:10:39.371 Stuck with my rats. But okay, so for Piaget, human children were egocentric 00:10:39.371 --> 00:10:44.431 and stuck in egocentric processing until they are 10, 12 years old. 00:10:44.971 --> 00:10:48.091 But now we know that that's not the case. 00:10:48.291 --> 00:10:51.851 It was the procedure that he was employing, testing these children. 00:10:52.051 --> 00:10:58.971 But the one thing that stands... What's interesting is that both in terms of for human development, 00:10:59.351 --> 00:11:02.331 studies on human development and also cross-cultural development, 00:11:02.331 --> 00:11:10.271 and comparative development seem to suggest that the default processing of spatial relationships. 00:11:11.431 --> 00:11:12.791 Are allocentric. 00:11:13.071 --> 00:11:17.031 This is the default. It's not egocentric. And this is a huge revolution in the 00:11:17.031 --> 00:11:20.711 kind of thinking about how space is processed generally. 00:11:21.591 --> 00:11:27.291 Because from our perspective of Westerners, it seems natural and intuitive that 00:11:27.291 --> 00:11:32.391 when we discuss space, we discuss it with reference to our location and our 00:11:32.391 --> 00:11:33.771 body in egocentric coordinates. 00:11:34.151 --> 00:11:40.731 But that's possibly because our languages trap us into thinking about space 00:11:40.731 --> 00:11:42.391 in egocentric coordinates mainly. 00:11:42.811 --> 00:11:48.131 If you go and look at other cultures where their language, they talk about, 00:11:48.271 --> 00:11:49.631 okay, where is the water well? 00:11:49.691 --> 00:11:55.891 It's north or northwest, as opposed to say it's in front of you and to the left, 00:11:55.931 --> 00:11:59.391 just take three steps in front of you and then two to the left. 00:11:59.391 --> 00:12:06.211 Then in those cultures, you see that spaces and spatial relations are processed 00:12:06.211 --> 00:12:11.051 allocentrically from the word go. So that's quite interesting. 00:12:12.271 --> 00:12:17.331 Absolutely, because it would also suggest that maybe to get to egocentric declaration 00:12:17.331 --> 00:12:23.131 of spatial relations might be a larger cognitive operation than the allocentric one. Yes, absolutely. 00:12:23.971 --> 00:12:27.931 And it would then oppose Jean Piaget's idea. Absolutely. 00:12:28.442 --> 00:12:30.842 You could argue, well, Jean-Pierre was wrong about many things, 00:12:30.982 --> 00:12:35.302 but that's easy to say afterwards because the big insight was still that these 00:12:35.302 --> 00:12:37.662 qualitative changes and these rapid transitions. 00:12:37.882 --> 00:12:44.222 And so if the rat pup moves into exploration mode on P20, is that for you then 00:12:44.222 --> 00:12:47.342 the rat signature of a transition, 00:12:47.622 --> 00:12:52.742 a cognitive transition and an operational capability of such a rat comparable 00:12:52.742 --> 00:12:54.302 to those of Jean-Pierre? 00:12:54.422 --> 00:12:58.162 Yes. So the idea of the sudden transition is still there. 00:12:58.442 --> 00:13:01.842 And that's what I want to draw the attention on. Yes, that development is not 00:13:01.842 --> 00:13:07.482 just about cumulative, monotonic, incremental, and gradual change. 00:13:07.742 --> 00:13:12.342 There are also these transitions. And what we don't know, we don't have a good 00:13:12.342 --> 00:13:15.682 handle at the neural level on what these transitions are caused by. 00:13:16.082 --> 00:13:19.442 So, for instance, the emergence of grid cells around 20 days old, 00:13:19.522 --> 00:13:22.342 when the animals are 20 days old. I don't think the grid cells are not there. 00:13:22.442 --> 00:13:27.742 The network is there. It's just that all of a sudden it can be anchored to the 00:13:27.742 --> 00:13:32.362 outside, to the cues, and therefore appears stable to us and we can detect it. 00:13:33.022 --> 00:13:35.642 But what we've seen, for instance, for head direction cells, 00:13:35.842 --> 00:13:39.882 these representations are there but they are drifting, they're unanchored from 00:13:39.882 --> 00:13:45.682 the frame of reference of the laboratory, and so we cannot see them as directional cells. 00:13:46.502 --> 00:13:52.582 I just wanted to defend Jean Piaget because his experiment, I don't think it 00:13:52.582 --> 00:13:57.522 was wrong, but he provided a particular tough test of alicentric, 00:13:57.662 --> 00:13:59.762 which was his mountain test. 00:13:59.802 --> 00:14:04.282 You had to imagine what the mountain looked like to an observer from the other side. 00:14:06.462 --> 00:14:13.162 Similar tests have been done in younger children, or I think easier tests in some ways. 00:14:13.302 --> 00:14:16.502 But there's a lot of data that... 00:14:17.135 --> 00:14:21.595 Children find it very hard to put themselves into the shoes of another person. 00:14:21.775 --> 00:14:24.095 Yeah, you need to make it theologically relevant to the child. 00:14:24.195 --> 00:14:28.695 But also, I think there's three different perspectives we have to think about here. 00:14:28.975 --> 00:14:33.975 There's the allocentric view, there's my egocentric view, and then there's my 00:14:33.975 --> 00:14:36.055 ability to think about your egocentric view. 00:14:36.295 --> 00:14:42.835 And then when we're thinking about child development, then people do talk a 00:14:42.835 --> 00:14:47.195 lot about a significant step towards theory of mind to be able to take somebody 00:14:47.195 --> 00:14:48.375 else's point of view. Absolutely. 00:14:48.655 --> 00:14:51.915 And not just theory of mind, but also just the processing of the language when 00:14:51.915 --> 00:14:55.455 you ask a child to put themselves in the shoes of another person. 00:14:55.635 --> 00:14:57.135 So how do you ask the question? 00:14:57.855 --> 00:15:02.715 So if you say to the child, okay, what does the word look like to the doll? 00:15:03.075 --> 00:15:06.895 It's a difficult, linguistically, it's a difficult statement. 00:15:07.515 --> 00:15:11.895 So it's very hard to perform these tasks and these experiments. 00:15:12.255 --> 00:15:16.715 I wasn't blaming Piaget. I'm just saying it was, yeah, it's one of these things 00:15:16.715 --> 00:15:20.375 where you run an experiment and you think that the conclusion is sound and then 00:15:20.375 --> 00:15:22.015 you realize that there was an artifact. 00:15:24.115 --> 00:15:26.995 Now, probably it was me saying something that got up Tony's nose. 00:15:27.175 --> 00:15:30.435 No, no, no, okay, okay. No, but in any case, so the important thing is that, 00:15:30.535 --> 00:15:34.715 again, we put our rats, which are already born in the laboratory, 00:15:34.895 --> 00:15:38.615 grown in the laboratory, in these kind of boxes, okay, featureless boxes. 00:15:39.195 --> 00:15:42.475 And I think we need to stop doing that because the world is not a featureless 00:15:42.475 --> 00:15:45.735 box. So we're doing the same thing as Piaget in a sense now. 00:15:45.835 --> 00:15:51.695 We need to introduce things into the boxes and we need to make this field of 00:15:51.695 --> 00:15:57.535 studying spatial navigation circuits more akin to a naturalistic. 00:15:57.535 --> 00:16:01.395 So we need to interrogate the circuits in more naturalistic environments. 00:16:03.875 --> 00:16:04.355 Sure. 00:16:04.812 --> 00:16:09.712 In your early experiments on these developing spatial cognition circuits that 00:16:09.712 --> 00:16:14.312 we could see as a red analog of cognitive development, there were a number of surprises. 00:16:14.552 --> 00:16:20.252 And we should now inspect those to see whether they really were surprises or actually not. 00:16:20.252 --> 00:16:26.652 So one surprise was that the play cells or something you could call play cell 00:16:26.652 --> 00:16:33.152 responses in hippocampus seemed to emerge already at around postnatal 16 before 00:16:33.152 --> 00:16:37.432 the grid cells emerged postnatal 20. Okay. 00:16:37.872 --> 00:16:41.452 So why was that such a big surprise to you at the time? 00:16:41.472 --> 00:16:46.292 At the time it was an incredible surprise, yes. And that's what made the paper 00:16:46.292 --> 00:16:49.652 become a science paper at the time because people were surprised. 00:16:50.252 --> 00:16:53.432 Because we were just five years from the discovery of grid cells, 00:16:53.712 --> 00:16:55.812 which were discovered in the entorhinal cortex. 00:16:56.252 --> 00:17:00.072 Everybody thought of the grid cells as being at the input, lying at the input 00:17:00.072 --> 00:17:03.592 end of the hippocampus, where you find place cells. 00:17:03.852 --> 00:17:10.252 And it's very easy by summing grid cells to obtain place cells. 00:17:10.592 --> 00:17:14.292 So everybody was thinking that that was the way in which the information was flowing. 00:17:14.592 --> 00:17:18.112 And so seeing that at least during development, and this equation didn't add 00:17:18.112 --> 00:17:20.072 up, then it was surprising. 00:17:20.492 --> 00:17:25.772 Now we know that, of course, with benefit of hindsight, things are never as 00:17:25.772 --> 00:17:27.432 easy as they seem. Okay? Right. 00:17:28.032 --> 00:17:32.952 But on the other hand, we could also argue that the place cells you observe. 00:17:33.552 --> 00:17:38.992 In these early days from P16, P20, so if I'm sort of. 00:17:40.146 --> 00:17:43.906 Putting myself in Tony's egocentric perspective, I could say they don't really 00:17:43.906 --> 00:17:47.406 look like place cells because, you know, they're close to the borders. 00:17:47.826 --> 00:17:51.306 You don't see any response that looks place cell-like in the center. 00:17:51.486 --> 00:17:53.726 The response fields are really very broad. 00:17:54.966 --> 00:18:02.086 They might be orientation invariant. Okay, I give you that. But that's already something, right? 00:18:02.666 --> 00:18:06.726 But it's about it. Yeah, that's about it. But then we need to ask ourselves, 00:18:06.846 --> 00:18:08.886 and these are very important questions, by the way. 00:18:08.886 --> 00:18:11.806 What is a place cell okay what are 00:18:11.806 --> 00:18:14.626 the fundamental characteristics that a 00:18:14.626 --> 00:18:17.366 neural signature needs to display in order 00:18:17.366 --> 00:18:21.386 to be called a place cell a place response exactly and that's very important 00:18:21.386 --> 00:18:25.246 this is a very important question because we see place cells cropping up everywhere 00:18:25.246 --> 00:18:28.506 in the brain you throw an electrode wherever you want and you can find place 00:18:28.506 --> 00:18:33.146 cells nowadays so so what is a place cell and i agree with you that it's not 00:18:33.146 --> 00:18:37.086 just looking at blobs on maps colorful or full blobs, 00:18:37.146 --> 00:18:44.186 we need to understand whether these responses are truly allocentric in the sense that they are. 00:18:44.206 --> 00:18:46.886 So what are the main characteristics of what they place there? 00:18:46.946 --> 00:18:49.526 First of all, that they need to be invariant to directionality. 00:18:49.706 --> 00:18:54.086 And that seems to be the case in the young pups place cells, 00:18:54.166 --> 00:19:01.946 but also that they need to be the result of integrating different cues, 00:19:02.086 --> 00:19:03.666 so responding to different cues. 00:19:03.666 --> 00:19:07.966 So when you do the Q subtraction experiment, you take away one by one the different 00:19:07.966 --> 00:19:12.726 bits of the laboratory that you might think are used, then the place cell must 00:19:12.726 --> 00:19:14.786 remain, the midplace field must remain. 00:19:14.926 --> 00:19:18.126 So that's another kind of litmus test. But what we could argue, 00:19:18.226 --> 00:19:23.606 if you look at the place cell response, is that initially what you have is just 00:19:23.606 --> 00:19:29.686 a broad associative response to head direction cells and some sensory features. 00:19:29.846 --> 00:19:33.466 And the most prominent sensory features in these empty boxes in which you put 00:19:33.466 --> 00:19:35.046 these pups are just surfaces. 00:19:35.666 --> 00:19:39.546 Yes, indeed. Indeed. And that gives you broad associative responses, 00:19:39.746 --> 00:19:45.486 but we can only call it a place cell if we have really this much higher acuity 00:19:45.486 --> 00:19:49.766 in space, much higher localization in space. 00:19:50.126 --> 00:19:53.786 And for that, to reach that stage, you do need the grid cells. 00:19:54.166 --> 00:19:55.586 But I'm not sure about that. 00:19:56.066 --> 00:19:59.486 No, I'm just teasing you with this idea. No, no, no. I'm not sure about that 00:19:59.486 --> 00:20:05.246 because I think in terms of spatial information, having larger or smaller place 00:20:05.246 --> 00:20:07.726 fields doesn't matter. Really doesn't matter. 00:20:08.006 --> 00:20:11.586 I mean, as long as if you have a kind of population code. 00:20:12.506 --> 00:20:14.526 If you have enough of them, you're fine. Yeah, exactly. And that's what we think 00:20:14.526 --> 00:20:18.306 the system might be working. But then you turn the argument around a little 00:20:18.306 --> 00:20:22.106 bit because I thought we were defining place cells on the basis of physiological 00:20:22.106 --> 00:20:26.446 characteristics and behavioral characteristics, not computational ones, right? 00:20:26.506 --> 00:20:30.566 Because we're saying, well, there must be a certain specificity in space and 00:20:30.566 --> 00:20:34.166 there must be a certain invariance to the orientation in which you enter that 00:20:34.166 --> 00:20:35.446 location in space, right? 00:20:35.566 --> 00:20:37.906 So the claim I was making is that…, 00:20:38.679 --> 00:20:45.039 the play cells are formed on the basis of an associative reaction to having 00:20:45.039 --> 00:20:48.719 heading direction and visual features. And this gives you the broad response. 00:20:49.759 --> 00:20:55.919 And what do you think is the extra ingredient? The spatial information coming from the grid cells. 00:20:56.259 --> 00:21:00.839 Okay. So in that story, it's still the grid cells as a necessary requirement 00:21:00.839 --> 00:21:07.519 to sharpen this broad associative response that's only combining features that are around. 00:21:07.519 --> 00:21:12.019 But then when in the adult, you switch off the grid cells pharmacologically, 00:21:12.239 --> 00:21:17.039 you don't see this massive broadening of the place cells that you would expect 00:21:17.039 --> 00:21:19.299 according to this theory. 00:21:19.539 --> 00:21:22.879 No, I can escape from that. I can wiggle my way out of that. 00:21:23.159 --> 00:21:25.399 Okay, okay. Tell me, tell me. That's memory. 00:21:26.479 --> 00:21:33.739 Okay. Now I formed a memory, so I have a strongly ingrained acquired response in my place cell. 00:21:33.759 --> 00:21:37.119 But to sharpen it up, that's where I need my grid cells. 00:21:38.019 --> 00:21:43.219 Well, yes, while I think that, and indeed, and that's why I mentioned the work, 00:21:43.379 --> 00:21:49.979 I think, from Pastrakova's lab about when you switch off the septum with muscimol, 00:21:50.039 --> 00:21:53.159 so you switch off theta and you get rid of the grid cells. 00:21:53.419 --> 00:21:57.559 And if you look at place cell formation in novel environments, 00:21:57.759 --> 00:22:00.859 that tend to happen against the edges of the environment again. 00:22:02.139 --> 00:22:06.939 So, yes, there is definitely something that grid cells confer to the place cell 00:22:06.939 --> 00:22:10.059 maps, but I don't think it's spatial information specifically. 00:22:10.439 --> 00:22:13.839 It's more… It's a sharpening. That's what I'm saying. It's a sharpening of the tuning. 00:22:14.059 --> 00:22:16.979 Yes, but it's selectively when you don't have enough features, 00:22:17.019 --> 00:22:20.539 so you don't have enough precision about the sensory cues because when you are 00:22:20.539 --> 00:22:23.639 away from this… That's exactly the point. Okay, so we're talking about the same 00:22:23.639 --> 00:22:25.039 thing. So, okay, so we are agreeing. 00:22:25.239 --> 00:22:31.539 But what I'm playing with is this idea like, oh, panic, grid cells after place cells. 00:22:31.859 --> 00:22:35.139 But what I'm saying is maybe before the grid cells emerge, 00:22:35.139 --> 00:22:38.239 emerge those cells that will become play cells 00:22:38.239 --> 00:22:41.259 have a broad associative response and then they require the sharpening 00:22:41.259 --> 00:22:43.939 from the grid cells and then they pop out as places this is 00:22:43.939 --> 00:22:47.959 what yes yes and the question is if you were to take away the grid set so what 00:22:47.959 --> 00:22:51.619 what interests me about the grid cells is you take away the grid says what is 00:22:51.619 --> 00:22:55.939 it that the animal cannot do anymore because it doesn't have grid so what is 00:22:55.939 --> 00:23:00.719 the the kind of advantage of having having evolved such a system them. 00:23:00.739 --> 00:23:02.219 And I think that's the real, yeah. 00:23:02.499 --> 00:23:07.759 So Edvard Moser was talking about an experiment in which they put developing 00:23:07.759 --> 00:23:13.759 rats in a spherical container where you didn't really see the walls and compared 00:23:13.759 --> 00:23:17.639 it with an enriched environment and a simple square environment. 00:23:17.819 --> 00:23:22.519 And in the sphere, the rats didn't develop grid cells. 00:23:22.739 --> 00:23:26.419 It took a week or so. You had to then put them into... So, I mean, 00:23:26.459 --> 00:23:31.659 but perhaps the rat pups are in a similar situation that, you know, 00:23:31.659 --> 00:23:33.099 they are in these huddles... 00:23:34.133 --> 00:23:38.513 All the stimulation is kind of, it's very proximal rather than distal. 00:23:38.553 --> 00:23:43.233 You know, what you care about is being close to your siblings who were warm, 00:23:43.373 --> 00:23:45.373 you know, getting nutrition from the mother. 00:23:45.473 --> 00:23:48.733 So you really don't need to know where you are. And it's not until you get out 00:23:48.733 --> 00:23:49.233 and you start exploring. 00:23:49.653 --> 00:23:53.453 But then the puzzle is, why is it that if we think that what grid cells are 00:23:53.453 --> 00:23:57.793 doing is linear integration, so telling you how far you've moved, 00:23:57.853 --> 00:24:01.313 let's say, let's say that you're thinking that that's one of the functions of 00:24:01.313 --> 00:24:04.193 grid cells, you may think that rat pups don't need to do that. 00:24:04.273 --> 00:24:06.813 But then why do they do this wonderful angular integration? 00:24:07.173 --> 00:24:09.033 Why is it that we get head direction cells? 00:24:09.593 --> 00:24:12.913 It's just... They must be able to get home. They do lots of orienting. 00:24:12.933 --> 00:24:13.953 They must get home, you know? 00:24:14.013 --> 00:24:16.873 Well, they need to be able to... But they can do this completely on the basis 00:24:16.873 --> 00:24:20.213 of chemotaxis, maybe, against men. Maybe, you say. 00:24:20.313 --> 00:24:25.513 Maybe it's the operational term. But maybe the difference is that the grid cells have to be learned. 00:24:26.553 --> 00:24:30.673 Yes, so that's it. So why is it that you need to learn about the grid cells 00:24:30.673 --> 00:24:33.313 and not about the head direction cells or the place cells? 00:24:33.393 --> 00:24:37.773 So these are the questions that I'm interested in. But the point I would be 00:24:37.773 --> 00:24:41.913 making is that perhaps if the experience of the rat pod is different, 00:24:42.053 --> 00:24:46.073 then you could shift the developmental timetable and maybe bring it forward. 00:24:46.253 --> 00:24:48.753 Or as an adverse case, you delay it. Yes. 00:24:49.333 --> 00:24:51.833 But it would be interesting to do it the other way. I don't know if anyone's 00:24:51.833 --> 00:24:52.793 tried to bring it forward. 00:24:52.973 --> 00:24:55.493 Yes. I don't think we have successfully done so. 00:24:55.593 --> 00:24:59.193 But also, Tony, from a comparative perspective, we could argue, 00:24:59.193 --> 00:25:04.493 Look, if we go to species that were on this planet before rats emerged or mammals 00:25:04.493 --> 00:25:08.473 in general, or that will still be there also when mammals disappear, 00:25:08.713 --> 00:25:12.413 thanks to us, they use…. 00:25:13.041 --> 00:25:16.521 They do build heading vectors like ants, like desert ants, right? 00:25:16.581 --> 00:25:18.641 They build these heading vectors to find the home position. 00:25:19.061 --> 00:25:23.521 And they just essentially have like an attractor-like integrator that helps 00:25:23.521 --> 00:25:26.281 them just to have this big vector that always helps them to get home. 00:25:26.581 --> 00:25:29.601 Why not generalize that now to your rat pup? But we say, well, 00:25:29.741 --> 00:25:35.001 they use heading direction to have always this big home vector that says, 00:25:35.141 --> 00:25:40.461 not in the simple box in the lab, but if I'm in the wild, in complex environments 00:25:40.461 --> 00:25:41.381 with obstacles, whatever, 00:25:41.561 --> 00:25:43.781 I always know where home is, right? 00:25:43.861 --> 00:25:47.321 Without chemical cues. So that's why that will be more fundamental, right? 00:25:47.421 --> 00:25:50.621 And then I build the rest of the system on that initially through association. 00:25:50.941 --> 00:25:55.281 But then we come back to your question, which we actually should look at after 00:25:55.281 --> 00:25:57.901 we look at a bit more data, but time is running short. 00:25:59.021 --> 00:26:02.141 And you have to follow your homing vector. Yes. Okay. 00:26:03.621 --> 00:26:07.761 Why would we then have the grid cells? And the thing there is the grid cells 00:26:07.761 --> 00:26:12.101 give you a highly accurate metric representation. 00:26:12.661 --> 00:26:15.641 It tells you the direction you're moving into. 00:26:15.821 --> 00:26:18.421 It tells you something about the spatial relations in your environment. 00:26:18.561 --> 00:26:22.661 Grid cells have more spatial information overall than place cells. 00:26:22.661 --> 00:26:25.741 But you have to this works 00:26:25.741 --> 00:26:28.521 because it's a tractor network that follows a certain 00:26:28.521 --> 00:26:31.341 topology the connections in that system have to follow a 00:26:31.341 --> 00:26:34.241 certain topology for this to work and they need to be set up it must 00:26:34.241 --> 00:26:37.221 be a twistotaurus topology otherwise it's not 00:26:37.221 --> 00:26:41.861 going to work from a connectivity perspective so the argument could be look 00:26:41.861 --> 00:26:46.401 I first need to have a heading direction system and some sort of rough space 00:26:46.401 --> 00:26:52.721 estimation using my associative pre-play cells to have a scaffold in which I 00:26:52.721 --> 00:26:55.461 can now fine-tune my grid cells. Once I did that, 00:26:56.117 --> 00:27:00.337 Now I have my metric system. Now I can find you. So I can bootstrap now my play cells. Yes. 00:27:00.617 --> 00:27:04.337 Because now I have my metric system, right? So what's wrong with that story? 00:27:04.617 --> 00:27:07.117 No, no, there is no, nothing is wrong with the story. 00:27:07.737 --> 00:27:11.397 The only problem is that we are done, but we need to prove it. 00:27:11.817 --> 00:27:14.797 We need to prove it. Okay, how are you going to prove that? Tell me. 00:27:14.857 --> 00:27:17.077 I don't know if I'm going to prove it. Why not? 00:27:17.897 --> 00:27:22.077 But what we need to do as collectively as a field, we need to really try and 00:27:22.077 --> 00:27:25.877 understand what this wonderful grid cells that we all love are, 00:27:25.937 --> 00:27:28.497 what kind of advantage they are conferring. 00:27:29.097 --> 00:27:34.637 Okay? So we need to design tasks, and these are behavioral tasks that tap really 00:27:34.637 --> 00:27:37.477 into these properties that we think grid cells. 00:27:37.817 --> 00:27:40.797 Yeah, but look, the tasks that you use, like these empty boxes. 00:27:41.517 --> 00:27:43.537 No, no, I mean a behavioral task. Of course. I'm not doing that, 00:27:43.677 --> 00:27:47.177 yes. No, but right now, right now we make all our inferences, 00:27:47.217 --> 00:27:48.717 or most of our inferences on 00:27:48.717 --> 00:27:54.077 this system, using tasks that sort of natural rats are never exposed to. 00:27:54.177 --> 00:27:57.017 Yeah, absolutely. Isn't that a big distorting factor? 00:27:57.417 --> 00:28:02.417 Absolutely. Yes, it is. And that's why I was saying that we need to start making 00:28:02.417 --> 00:28:05.157 these recordings in more naturalistic environments. 00:28:05.357 --> 00:28:09.337 So I know that Edward has talked about object vector cells. So you start seeing 00:28:09.337 --> 00:28:13.417 new things when you insert objects all of a sudden in these fissureless environments. 00:28:13.417 --> 00:28:19.457 These cells akin to this were discovered also by Nierim, Jim Nierim, quite a while ago. 00:28:19.577 --> 00:28:23.657 So it is important for our field to move away from featureless boxes, 00:28:23.777 --> 00:28:27.037 which were introduced for a very good reason by Bob Muller many years ago. 00:28:27.357 --> 00:28:30.897 Because at the beginning, when John O'Keefe started his studies, 00:28:31.157 --> 00:28:33.237 he worked with very complex mazes. 00:28:33.317 --> 00:28:37.177 And that brought lots of complexity. And it was very difficult to understand 00:28:37.177 --> 00:28:41.577 what the basic phenomenon of place cell or place field was. now we can go back 00:28:41.577 --> 00:28:47.137 to it now that the basics have been so what's the next environment you want to see in the lab 00:28:47.538 --> 00:28:51.658 For me, I want to see a barrel system for the development. 00:28:51.878 --> 00:28:54.978 That's what I want to see. Might be a bit difficult with the electrodes. 00:28:55.158 --> 00:28:56.458 Yes, but we have wireless technology. 00:28:56.738 --> 00:29:00.998 Of course. So that's the thing. Especially in science, yes, that's what we're doing. 00:29:01.338 --> 00:29:04.698 Especially in science, there is the technological advancement that allows you 00:29:04.698 --> 00:29:08.998 to ask the questions that you really wanted to ask. And then you move on and you move on. Of course. 00:29:09.418 --> 00:29:12.838 That's really fantastic. Fantastic. But now, to what extent have you been able 00:29:12.838 --> 00:29:17.598 to also generalize your insights in the system in the RAT to humans? 00:29:18.218 --> 00:29:21.918 Do you think it plays out the same way or is there a transition? Are humans different? 00:29:24.338 --> 00:29:27.298 No, humans are not different. We're just more vicious. 00:29:27.718 --> 00:29:35.318 But I think that now, joke aside, I think the fundamental principles will be very similar. 00:29:36.658 --> 00:29:45.018 And I don't know if we have any specific evolutionary niche that makes us peculiar 00:29:45.018 --> 00:29:48.938 or special in spatial processing. I don't think so. 00:29:49.138 --> 00:29:54.538 All right. But now one of the principles you pointed to was attractor networks, 00:29:54.898 --> 00:29:56.098 right? This notion of attractor. 00:29:56.178 --> 00:30:01.038 But do you think the notion of an attractor has been helpful in the study of this system? 00:30:03.193 --> 00:30:05.933 I think it has been incredibly helpful. Okay, why? 00:30:06.133 --> 00:30:15.333 Yes, because it has made us ask questions about how information is processed, 00:30:15.473 --> 00:30:17.613 encoded, and also stored, 00:30:17.753 --> 00:30:22.133 but also retrieved by the hippocampus in general. 00:30:22.313 --> 00:30:25.713 So I think that's very important. And specifically for the grid cell, 00:30:25.793 --> 00:30:27.093 kind of after the grid cell discovery, 00:30:27.413 --> 00:30:31.273 it was very interesting to see that these two camps of the oscillatory model 00:30:31.273 --> 00:30:37.393 versus the attractor, model of how these grid cells could emerge and now we 00:30:37.393 --> 00:30:40.753 see a convergence between these two models. 00:30:40.973 --> 00:30:43.613 So I think historically as well it's been an interesting journey. 00:30:43.793 --> 00:30:45.533 I thought the interference models are just invalid. 00:30:45.853 --> 00:30:50.173 How do we see convergence of the two? Well, I'm sure you've done a podcast with 00:30:50.173 --> 00:30:53.913 Professor Neil Burgess who will have answered that question. 00:30:55.153 --> 00:30:58.193 But there's one thing we should worry about with these attractor models. 00:30:58.913 --> 00:31:04.173 Attractor models can almost never be wrong right because dependent even the 00:31:04.173 --> 00:31:11.793 interference models I could reinterpret as an attractor model because as long as you define you can, 00:31:12.433 --> 00:31:15.513 arbitrarily define states in the state space of your 00:31:15.513 --> 00:31:19.393 system and you can define the state space in such a dimensionality that you 00:31:19.393 --> 00:31:22.953 have stable states even if they're oscillatory it doesn't matter okay these 00:31:22.953 --> 00:31:28.293 are my attractor so certainly if you talk about memory since memory means there's 00:31:28.293 --> 00:31:32.793 a stability intrinsically it's It's sort of circular to then say, oh, 00:31:32.893 --> 00:31:36.773 an attractor network is fantastic to describe this, because all you're actually 00:31:36.773 --> 00:31:39.113 saying is there are stable points in a dynamical system. 00:31:41.213 --> 00:31:44.753 So shouldn't we then think beyond attractor models? 00:31:45.233 --> 00:31:50.433 Okay, so what would you suggest is a promising avenue to look at? 00:31:50.753 --> 00:31:54.753 One promising avenue might be maybe to bring in more of the specific physiological 00:31:54.753 --> 00:32:00.533 features that we know that this system shows. So attractor models also allow 00:32:00.533 --> 00:32:02.233 us to stick at a relatively abstract level. 00:32:03.382 --> 00:32:06.822 Where we do not necessarily take into account the huge heterogeneity, 00:32:07.002 --> 00:32:09.482 the modular organization, the specific anatomical topology. So let me tell you 00:32:09.482 --> 00:32:12.982 about the specific problem I have with attractor networks, for instance, 00:32:13.042 --> 00:32:14.562 for grid cells, overhead direction cells. 00:32:15.122 --> 00:32:19.642 In order to set them up, first of all, we don't know what kind of wiring really 00:32:19.642 --> 00:32:25.302 at the detailed level could support attractor network topology. 00:32:25.702 --> 00:32:30.122 And the real problem is that in order to set up such wiring, 00:32:30.122 --> 00:32:35.382 wiring you need to invoke quite complex developmental processes and so coming 00:32:35.382 --> 00:32:40.142 from development i want to know how you wire it up in the first place and so 00:32:40.142 --> 00:32:44.582 that's the other thing that people like myself and other people need to really 00:32:44.582 --> 00:32:48.742 work on right yeah absolutely so francesca um. 00:32:49.482 --> 00:32:52.402 So you're in this business for a while you're you're in a very 00:32:52.402 --> 00:32:55.102 rich also as you know rats in rich 00:32:55.102 --> 00:32:57.922 environments do great humans in rich environments do great 00:32:57.922 --> 00:33:01.082 as well you're in a a really rich scientific environment there at 00:33:01.082 --> 00:33:05.282 ucl working with great collaborators um so 00:33:05.282 --> 00:33:08.562 but now given your experience in 00:33:08.562 --> 00:33:13.342 the field what is francesca's law to study the brain what is francesca's law 00:33:13.342 --> 00:33:19.322 to study the brain yeah you got it okay francesca's law to study anything never 00:33:19.322 --> 00:33:26.302 mind the brain is to persist in the face of tragic failure and complete and utter continuous 00:33:26.742 --> 00:33:33.462 failure and persist and just be joyful about it and enthusiastic. 00:33:34.102 --> 00:33:37.782 That sounds more like an autobiographical note. Is that true? 00:33:38.022 --> 00:33:41.182 No, no, no, no, no. I think this is what motivates humans. 00:33:41.482 --> 00:33:44.822 Okay. And this is the positive side about being human, actually. 00:33:45.142 --> 00:33:50.022 Very good. Yeah. And then, okay, look, we're facing all sorts of problems now 00:33:50.022 --> 00:33:54.102 with international travel. Certainly soon I cannot go to UK anymore. 00:33:54.562 --> 00:34:00.402 Yes. Not only because… And Donald Trump is going to make all our boundary cells fire. Exactly. 00:34:00.442 --> 00:34:06.262 But we have an agent in the UK who we can send to your lab in five years from now. 00:34:06.562 --> 00:34:10.242 And that agent sits just next to you here at the other side of the table. 00:34:11.382 --> 00:34:14.642 That's a bit of a train ticket. And Tony is going to come visit your lab five 00:34:14.642 --> 00:34:18.222 years from now to check whether a prediction you made today was falsified or 00:34:18.222 --> 00:34:22.462 verified. So what's the one prediction you would like to see tested in a five-year framework? 00:34:23.342 --> 00:34:27.302 Okay, let me think. What's the five? Where would I? 00:34:28.202 --> 00:34:33.182 One prediction, five years to do it. Five years? That's what he's going to come. 00:34:33.982 --> 00:34:37.682 These tickets are expensive, you know, to save up. 00:34:40.762 --> 00:34:44.782 That hippocampus. I only have very long-term stuff that I'm interested in. 00:34:44.782 --> 00:34:45.782 What do you want, 10 years? 00:34:46.742 --> 00:34:51.942 No, I want like several lifetimes. I'm not sure Tony wants to wait that long. 00:34:52.022 --> 00:34:57.042 I really want to know whether the hippocampus is really just truly about space or not. 00:34:57.783 --> 00:35:00.963 And what about time and the other things? 00:35:01.143 --> 00:35:06.043 So in five years' time, you would like to see the hypothesis tested that hippocampus 00:35:06.043 --> 00:35:07.223 is about temporal processing? 00:35:07.583 --> 00:35:11.243 Yes, they're already starting to test it. But yes, yes, I'm interested in all 00:35:11.243 --> 00:35:16.243 these other things that hippocampus might be doing, yeah, coming out of the space box. 00:35:16.583 --> 00:35:20.963 All right, Francesca Cacucci, thank you very much for this conversation. Thank you. Thank you. 00:35:25.483 --> 00:35:31.383 The CSN podcast was produced by the Convergent Science Network of Biometrics 00:35:31.383 --> 00:35:37.783 and Biohybrid Systems, a project funded by the European Sevens Research Framework Program. 00:35:39.343 --> 00:35:44.663 For more interviews, recorded lectures, or upcoming conferences in the field 00:35:44.663 --> 00:35:50.903 of biomimetics and biohybrid systems go to csnnetwork.eu. 00:35:51.760 --> 00:35:59.120 Music.