WEBVTT 00:00:00.017 --> 00:00:05.477 So I'm here with, this is Paul Vachure with Kathy Roepland at the Catherly Center 00:00:05.477 --> 00:00:06.797 for Theoretical Physics. 00:00:07.157 --> 00:00:11.017 We're both attending a workshop on network architecture and the brain. 00:00:11.477 --> 00:00:17.837 And Kathy, you gave us a presentation yesterday where you really gave us quite 00:00:17.837 --> 00:00:21.317 a broad overview of the anatomy of mainly cerebral cortex. 00:00:21.677 --> 00:00:29.817 Yes. So, what do you think are the key properties of a neocortex that we should 00:00:29.817 --> 00:00:32.177 be aware of? Of the neocortex? 00:00:37.237 --> 00:00:45.797 Well, let me start with what may seem backwards and say I think the distributed property. 00:00:47.257 --> 00:00:52.357 Most people are not going to say this first thing. But if you look at these 00:00:52.357 --> 00:00:56.217 axons, and you look at the divergence, and you look at the collateralization. 00:00:57.977 --> 00:01:04.137 The first word, English word, that comes to your mind is distributed or divergent. 00:01:04.377 --> 00:01:09.997 Now we can proceed a little bit more. That immediately gets you into questions 00:01:09.997 --> 00:01:15.677 of probably random or specific, which we can come back to. 00:01:15.677 --> 00:01:19.877 I'm not saying random, I'm just saying simple fact of distributed. 00:01:21.077 --> 00:01:26.657 And so that's one thing we can come back to. Probably what most people would 00:01:26.657 --> 00:01:34.017 then talk about the cortex is because it's been promoted and promoted and promoted, 00:01:34.137 --> 00:01:36.557 a uniform organization, 00:01:36.977 --> 00:01:38.677 this is what we were talking about at lunch. 00:01:38.677 --> 00:01:45.657 But that really, if you look closely at the anatomy, that's not the first thing that comes to mind. 00:01:45.777 --> 00:01:49.537 I could show you pictures later. We can develop that theme also. 00:01:50.437 --> 00:01:57.417 Having said that, I would then introduce the idea that, which many people are 00:01:57.417 --> 00:02:01.837 saying, that the anatomy and the function is a little bit discrepant. 00:02:02.597 --> 00:02:06.677 We've been, again, trained to say structure-function correlation. 00:02:08.317 --> 00:02:12.597 But now it's very clear that there can be functional aspects and structural 00:02:12.597 --> 00:02:16.137 aspects, and sometimes there's an apparent discrepancy. 00:02:16.817 --> 00:02:20.137 Cortex, of course, can deal with it, but for us it looks... Can you give an 00:02:20.137 --> 00:02:24.297 example of discrepancy? Yeah, I think, and we might have talked about this yesterday also. 00:02:25.117 --> 00:02:30.417 One of my favorite examples, and you can pick this apart, please, 00:02:30.537 --> 00:02:34.157 is ocular dominance columns in the monkey. 00:02:34.157 --> 00:02:40.437 So you are shown ocular dominance columns with beautiful 2DG, 00:02:40.577 --> 00:02:46.257 so an eye is blocked and you see the 2DG pattern more or less through the layers. 00:02:47.391 --> 00:02:53.251 It's very convincing. It's very real. Then you are told what is the truth, 00:02:53.471 --> 00:02:58.071 that the basis of this is thalamocortical connectivity. 00:02:59.651 --> 00:03:03.531 Now, there are actually several things to say about that. One is, 00:03:03.711 --> 00:03:09.591 if you look at the scale of the anatomy and the functional column, 00:03:09.731 --> 00:03:11.471 there's an immediate mismatch. 00:03:11.471 --> 00:03:20.131 So in layer 4C, the main layer, parvosellular axons are smaller than 500. 00:03:20.231 --> 00:03:25.891 I should say the ocular functional columns are about 500 microns in diameter. 00:03:26.791 --> 00:03:30.191 But the parvosellular LGN is smaller, 250. 00:03:31.131 --> 00:03:35.231 Magnocellular is larger, with two or three clusters. 00:03:35.751 --> 00:03:40.711 Then 4A, you have another population of thalamocortical axons. 00:03:40.711 --> 00:03:43.751 It's very small, 100 microns, center to center. 00:03:45.311 --> 00:03:48.831 Sometimes they're collaterals in layer 6. These tend to be smaller. 00:03:49.271 --> 00:03:54.811 The cytochrome oxidase population is the axons are less than, 00:03:54.831 --> 00:03:57.131 equal or less than 100 microns. 00:03:57.171 --> 00:04:01.551 And in layer 1, whatever goes up there from the LGM is very divergent. 00:04:02.711 --> 00:04:08.131 So I was puzzled by this for a long time, and my own explanation is there must 00:04:08.131 --> 00:04:13.751 be some operation, something, some operation, which is making those axons converge, 00:04:14.131 --> 00:04:20.091 and the key word here would be maybe something operation, and converge in a 500 micron space. 00:04:20.451 --> 00:04:24.151 And the two candidates would be something molecular. 00:04:26.299 --> 00:04:29.639 Going back to development presumably, maybe extracellular matrix, 00:04:30.679 --> 00:04:33.739 and activity-related, or and or both. 00:04:33.999 --> 00:04:39.379 Okay, so here, let me summarize. The functional column is unambiguously bicarbonate 00:04:39.379 --> 00:04:44.119 microns, but the anatomical basis of this, the primary anatomical basis, 00:04:44.939 --> 00:04:48.819 is all, I'll use the same term probably several times, all around the block. 00:04:49.519 --> 00:04:55.859 Now, of course, the reconciliation, we think, would be from the intrinsic processing, But that gets you in. 00:04:55.879 --> 00:04:59.759 It's not the same thing as saying thalamocortical connections are the basis 00:04:59.759 --> 00:05:01.339 of ocular dominance columns. 00:05:01.519 --> 00:05:07.759 It means that the cortex is doing, the intrinsic cortex is doing something on those connections. 00:05:08.239 --> 00:05:13.239 But how do you then define this functional scale of organization of 500 micron? 00:05:14.579 --> 00:05:20.799 What you would see, well, so my first response is operationally that you can link. 00:05:20.799 --> 00:05:25.239 I think if you have a functional marker like 2-D-alcyclicose or CFOS, 00:05:25.339 --> 00:05:30.519 and you can link it to a behavior, you will tend to see a columnar, 00:05:30.619 --> 00:05:34.099 often a columnar organization of this amount. 00:05:35.019 --> 00:05:38.179 Does that, does that want to go a little bit further? No, go ahead, go ahead. 00:05:38.319 --> 00:05:43.379 But having said that, there is some beautiful work, and I'm sorry I forget the 00:05:43.379 --> 00:05:48.919 authors here, but it was within the last five years on the CFOS of... 00:05:49.759 --> 00:05:53.799 Might have been Japanese group Yamamori where 00:05:53.799 --> 00:05:56.719 they were looking at different 00:05:56.719 --> 00:06:00.479 conditions of ocular deprivation in 00:06:00.479 --> 00:06:09.279 the monkey in CCOS which is a better resolution than the 2DG and they would 00:06:09.279 --> 00:06:14.499 see they were able to see several laminar patterns depending on the time interval 00:06:14.499 --> 00:06:18.459 and what they had actually done with the ocular dominant. 00:06:18.599 --> 00:06:23.739 So even there, there is a suggestion that when you're looking at average techniques. 00:06:26.439 --> 00:06:30.099 It's telling you part of reality, but not the whole. 00:06:30.319 --> 00:06:36.639 Okay, so one bottom line of this is that structural and functional organization 00:06:36.639 --> 00:06:38.559 does not necessarily matter. 00:06:39.099 --> 00:06:46.559 Not to the investigative view. So it works in biology, but not in the way we 00:06:46.559 --> 00:06:49.419 might think. There's no simple mapping between the two. Often not. 00:06:49.519 --> 00:06:53.019 In fact, when it occurs, it's almost the exception. 00:06:53.219 --> 00:06:57.119 And one exception would be, for example, corticospinal tract, 00:06:57.339 --> 00:07:02.899 where you have the difference of primate with the fine pincer movement versus 00:07:02.899 --> 00:07:06.119 the carnivores that cup. And there is an anatomical basis. 00:07:06.239 --> 00:07:11.339 Or even the development of the corticospinal tract, so babies can't walk until 00:07:11.339 --> 00:07:14.839 the corticospinal tract does something, myelination, maybe other things. 00:07:14.839 --> 00:07:17.659 But I've come to think it's almost the exception. 00:07:19.367 --> 00:07:22.807 Okay. But now in your presentation, you also went a bit further than that, right? 00:07:22.867 --> 00:07:28.047 Because you also pointed out that the sort of standard notions of feed forward 00:07:28.047 --> 00:07:32.567 and feed back that people have been using to describe the structure of the cortex 00:07:32.567 --> 00:07:35.227 might not be that clear-cut either. 00:07:35.807 --> 00:07:38.667 Absolutely. And I think in several ways. 00:07:41.147 --> 00:07:47.247 One, even if you take the schematic, the strong view with layers and so on, 00:07:47.247 --> 00:07:51.567 There are many exceptions, really abundant exceptions. 00:07:51.767 --> 00:07:57.707 And when you go to the rodent, it's been well known for years that the laminar-based 00:07:57.707 --> 00:08:00.547 scheme breaks down in a big way. 00:08:00.767 --> 00:08:06.167 You have a layer one is favored maybe by what could be feedback. 00:08:06.287 --> 00:08:09.727 Layer four is favored, but there's a tremendous blurring. 00:08:11.267 --> 00:08:16.467 But even within primate, you have a lot of exceptions. But what are the rules, 00:08:16.527 --> 00:08:17.927 really, on which we have these exceptions? 00:08:20.627 --> 00:08:24.407 What do you mean by that? Well, as you say, you have many exceptions in rodent 00:08:24.407 --> 00:08:28.447 and also in monkey, but then there's a rule to which you have an exception, 00:08:28.667 --> 00:08:30.547 right? So what's the rule for it? 00:08:30.727 --> 00:08:35.107 All right. Well, the old rule, which has some basis, in fact, 00:08:35.227 --> 00:08:42.727 in static anatomical fact, is that… Okay, feed forward. 00:08:43.187 --> 00:08:49.067 See, it's really based on a relay view of the brain, and I don't think that's very useful. 00:08:49.527 --> 00:08:54.127 But if you accept that relay view, the visual information goes to the eye, 00:08:54.247 --> 00:08:56.467 goes to the LGN, goes to V1, and so on. 00:08:57.927 --> 00:09:03.467 Then, and of course, V1 goes to LGN, but also other things go to LGN. 00:09:03.607 --> 00:09:06.347 So none of these are pairwise. Absolutely none. 00:09:07.227 --> 00:09:14.427 When you get to V1, and you say, okay, let's start the relay from there. and if you do V1, V2, V3, 00:09:15.361 --> 00:09:19.361 let's just put in MT, extra striate, V4, infratemporal cortex, 00:09:19.661 --> 00:09:31.481 there is, you can plausibly say layer 3 goes to layer 4, an input layer, in your feedforward. 00:09:32.401 --> 00:09:37.981 And the feedback is actually usually two layers, layer 6 and layer 2, 00:09:38.161 --> 00:09:42.081 feeding back to layer 1 and sometimes layer 6. 00:09:42.081 --> 00:09:50.001 But there is a dissociation, one can present a dissociation of a Layer 4-based 00:09:50.001 --> 00:09:54.141 feed-forward and a Layer 1-based feedback. 00:09:54.501 --> 00:09:57.741 And there also is, this was the data, these were the data I was showing yesterday, 00:09:58.001 --> 00:10:00.861 a dissociation in spatial organization. 00:10:01.261 --> 00:10:05.161 Feed-forward tends to have smaller arbors, sometimes multiple, 00:10:05.321 --> 00:10:08.121 feedback divergent in Layer 1. 00:10:08.121 --> 00:10:16.181 But I would still, I suggested yesterday and would still like to suggest that 00:10:16.181 --> 00:10:22.901 you can rephrase this in terms of a Layer 1 organization or a Layer 1 biased 00:10:22.901 --> 00:10:26.201 connection and a Layer 4 biased connection. 00:10:26.381 --> 00:10:30.181 There's no real need to say feed forward and feed back in that. 00:10:30.971 --> 00:10:35.731 And it has a very bad assumption of saying, well, something happened, 00:10:35.851 --> 00:10:38.131 and then there was a feedback. 00:10:38.511 --> 00:10:41.491 But we don't know when the film started. 00:10:42.271 --> 00:10:49.111 So there is a strong assumption in the terminology of this temporal or relay 00:10:49.111 --> 00:10:52.751 pattern, which I think is not, it can be useful to an extent, 00:10:52.831 --> 00:10:54.991 but it cannot be the whole truth. 00:10:55.251 --> 00:11:03.751 Right. Okay. Okay, but then another issue then becomes this so-called uniformity of cortex, right? 00:11:03.811 --> 00:11:05.751 So, for instance, you would say, well, we have a layered structure. 00:11:06.951 --> 00:11:12.111 And pseudo-architectonically, it sort of roughly remains the same if we sort 00:11:12.111 --> 00:11:14.951 of go from the front to the back, right? 00:11:15.391 --> 00:11:20.171 And also there would then be this idea that maybe in terms of their functional 00:11:20.171 --> 00:11:25.491 organization, things will change in the sort of occipital areas and parietals, 00:11:25.491 --> 00:11:28.111 more sensory, and if we go more frontal, we go more motor. 00:11:28.791 --> 00:11:31.971 Is that then at least a pattern we can still sort of adhere to? 00:11:33.211 --> 00:11:36.331 Yes, by all means. But of course, I'm going to say but. 00:11:37.851 --> 00:11:45.371 And you have to keep in mind the threshold problem, which is glaringly obvious in fMRI. 00:11:46.031 --> 00:11:50.371 So if you set your, if I understand correctly, if you set your threshold low 00:11:50.371 --> 00:11:51.851 enough, you're going to see lots of stuff. 00:11:53.291 --> 00:11:54.691 And it's apparent... 00:11:56.331 --> 00:12:02.331 I think, well, V1, your sensory areas in the adult, in the primate, 00:12:04.751 --> 00:12:10.691 probably do function strongly unimodal. But there are these cross-modal connections. 00:12:11.211 --> 00:12:17.571 Even in monkey, I'm going to say it's in the peripheral visual field. 00:12:19.131 --> 00:12:26.431 Maybe a little bit less in the fovea. So this, again, would substantiate the threshold phenomenon. 00:12:26.811 --> 00:12:33.651 But in monkey V2, to some extent V1, you have, in V2 it's quite respectable. 00:12:33.751 --> 00:12:40.051 You have auditory input, you have sensory input, in the rodent you have.