WEBVTT 00:00:03.557 --> 00:00:10.517 This is the Convergent Science Network podcast. Leading researchers in the domain 00:00:10.517 --> 00:00:16.797 of neuroscience, brain theory and technology are interviewed by Paul Vershoor and Tony Prescott. 00:00:19.757 --> 00:00:23.057 This is Paul Vershoor with the Convergent Science Network podcast. 00:00:24.397 --> 00:00:27.177 And I'm having a conversation with Moshe Bar, who's a 00:00:27.177 --> 00:00:29.857 speaker at our summer school in in 00:00:29.857 --> 00:00:33.117 barcelona and and moshe you you started 00:00:33.117 --> 00:00:37.317 your presentation with um a view 00:00:37.317 --> 00:00:41.757 on perception where where you brought in let's say a key role for let's say 00:00:41.757 --> 00:00:46.257 prediction and top-down components yes so what's what's the key observation 00:00:46.257 --> 00:00:52.237 and the key insight there well you mean what brought me to this conclusion to 00:00:52.237 --> 00:00:53.317 this presentation yeah Yeah. 00:00:53.397 --> 00:01:00.737 So first of all, it's not only mine, so I can't claim complete pioneering movement here. 00:01:00.857 --> 00:01:05.817 But the idea here is that we've known for decades that most connections, 00:01:06.057 --> 00:01:08.777 if not all, in the brain are reciprocal. They go in both directions. 00:01:09.217 --> 00:01:14.017 And still the idea of feedback hasn't been incorporated into mainstream thinking 00:01:14.017 --> 00:01:15.817 about the brain sufficiently. 00:01:16.017 --> 00:01:20.477 So you see it here and there. But when you just open any textbook on perception 00:01:20.477 --> 00:01:25.177 and cognition, you see some kind of an artificial boundary between perception and cognition. 00:01:25.377 --> 00:01:30.277 There's first perception that you can think of as the analysis of physical signals 00:01:30.277 --> 00:01:32.237 coming through to the brain through the senses. 00:01:32.537 --> 00:01:36.917 And once we understand what is it that we perceive, cognition kicks in. 00:01:37.037 --> 00:01:39.337 That's the old traditional view. 00:01:40.077 --> 00:01:45.717 Cognition kicks in as in memory, attention allocation, executive decisions. 00:01:45.717 --> 00:01:47.697 Decisions, and so forth. 00:01:47.917 --> 00:01:52.477 But there's actually no real reason why there will be a boundary between cognition and perception. 00:01:52.857 --> 00:02:00.777 And our and other's idea is that actually cognition and perception are intertwined 00:02:00.777 --> 00:02:02.477 and they help each other whenever possible. 00:02:02.697 --> 00:02:06.977 And therefore, the bottom line is that what we see and what we perceive from 00:02:06.977 --> 00:02:12.837 our environment is to a large extent affected by cognition and by memory and 00:02:12.837 --> 00:02:16.737 by what we expect and what are our goals So, cognition, 00:02:17.077 --> 00:02:22.717 perception is not purely affected by the senses, but rather also with top-down information. 00:02:23.337 --> 00:02:26.617 But now, would you be able to give a number to that? Would you say, 00:02:26.637 --> 00:02:28.617 look, under normal- 50-50. 00:02:28.877 --> 00:02:32.077 50-50. Yeah. It doesn't vary with task demands. Well, it does vary. 00:02:32.237 --> 00:02:33.417 It does vary. No, of course it doesn't. 00:02:35.967 --> 00:02:40.147 I think zero to a hundred. Okay. So we spoke, if you remember, 00:02:40.187 --> 00:02:45.487 towards the end of the talk, people asked about meditation. 00:02:45.567 --> 00:02:50.347 And I think people that do meditation are actually 100% bottom up. 00:02:50.427 --> 00:02:53.947 I think they quiet down completely the top-down effects. 00:02:56.007 --> 00:03:01.067 And there are interesting findings that I think Similac or Eastwood, 00:03:01.107 --> 00:03:05.907 that's the name of the authors of these interesting studies that we called in the lab the Zen studies. 00:03:05.967 --> 00:03:08.467 They're not ours, but we just really love these findings. 00:03:08.907 --> 00:03:14.647 And in these specific search tasks, I believe, they ask people to just lean 00:03:14.647 --> 00:03:18.207 back, relax, and let the display come to them. 00:03:18.727 --> 00:03:23.327 And people improve their performance just by the fact that they're kind of quieting 00:03:23.327 --> 00:03:24.807 down their top-down expectations. 00:03:24.827 --> 00:03:30.167 So you look for Ts among Els, and you perform better when you just lean back 00:03:30.167 --> 00:03:32.447 and relax and don't think about anything. 00:03:32.447 --> 00:03:35.647 Thing so there are some cases where expectations actually might 00:03:35.647 --> 00:03:39.127 be bad for you especially when you when you have no uh 00:03:39.127 --> 00:03:42.407 basis for your expectations and they're completely uh 00:03:42.407 --> 00:03:46.107 like me in the stock market right if i lose my expertise there 00:03:46.107 --> 00:03:50.907 i'll lose my my house so so whenever you don't have any expertise or whenever 00:03:50.907 --> 00:03:54.627 there is a situation that's completely novel or completely not you not based 00:03:54.627 --> 00:03:58.827 on the past then it's better to quiet off the expectations and this is a case 00:03:58.827 --> 00:04:04.147 where um I would imagine it's 100% bottom-up and 0% top-down. 00:04:06.007 --> 00:04:10.867 Let's think about a case where it's only predictions, or only top-down. 00:04:10.927 --> 00:04:12.807 I think it's when you plan, right? 00:04:12.927 --> 00:04:16.547 When you plan, you're kind of only planning. You don't have the input. 00:04:16.667 --> 00:04:22.307 You just sit down now and plan your dinner. You don't have any input other than from within. 00:04:22.527 --> 00:04:27.507 So I guess, yeah, I think it's roughly between 0% to 100%. Right. 00:04:28.087 --> 00:04:34.167 So here we have the proactive brain. okay now the proactive brain would actually, 00:04:34.770 --> 00:04:38.630 in some form, go from sensor states to predictions. 00:04:38.950 --> 00:04:44.990 But now you used different kinds of techniques and human subjects like fMRI 00:04:44.990 --> 00:04:49.730 and EEG and so on to try to sort of disentangle these pathways a little bit. 00:04:49.850 --> 00:04:56.270 And a surprising result that you presented was that in some sense, 00:04:56.370 --> 00:05:01.470 if you want prediction pathways, that in some sense we jump outside sort of 00:05:01.470 --> 00:05:04.230 the traditional areas of visual processing. 00:05:04.770 --> 00:05:09.130 So, it doesn't seem to be restricted to just, let's say, the ventral stream 00:05:09.130 --> 00:05:12.470 and the temporal lobe, but also to involve frontal areas. 00:05:12.870 --> 00:05:17.830 So, it seems rather surprising to now include more frontal areas in perception. 00:05:18.110 --> 00:05:19.930 So, how should I think about that? 00:05:21.470 --> 00:05:25.570 So, these findings don't necessarily mean that the prefrontal cortex is suddenly 00:05:25.570 --> 00:05:29.290 involved in perception per se, but it just supports what I said at the beginning 00:05:29.290 --> 00:05:33.370 of our conversation, that perception and cognition are not separated. 00:05:33.370 --> 00:05:41.790 And the role of prefrontal cortex in helping perception be accomplished is an 00:05:41.790 --> 00:05:44.070 example of cognition helping perception. 00:05:44.350 --> 00:05:48.670 So something about these high-level areas in the prefrontal cortex and maybe 00:05:48.670 --> 00:05:53.610 anterior temporal cortex and other regions send down initial guesses to help 00:05:53.610 --> 00:06:01.910 the perception of the signal to be accomplished in a quicker and more efficient manner. 00:06:01.910 --> 00:06:05.970 So the involvement of the OFC specifically, orbital frontal cortex. 00:06:06.250 --> 00:06:11.730 As we discussed yesterday, seems to be polysensory. 00:06:11.770 --> 00:06:17.150 So it's not that OFC is a visual area all of a sudden, but it's an area that helps predicting. 00:06:17.350 --> 00:06:21.770 And it helps predicting even though we didn't test it, but I believe that it 00:06:21.770 --> 00:06:25.810 predicts also in other modalities such as olfaction. 00:06:25.810 --> 00:06:29.170 And it also predicts in higher level events 00:06:29.170 --> 00:06:32.210 not only detecting objects but also preparing 00:06:32.210 --> 00:06:35.930 for a new situation or a new script that's coming 00:06:35.930 --> 00:06:38.750 up so an OFC as we know 00:06:38.750 --> 00:06:41.710 is also connected with the limbic system and people think about 00:06:41.710 --> 00:06:45.250 affect and about reward in the context of OFC and 00:06:45.250 --> 00:06:48.370 we think that the reason people see activation in this prefrontal region 00:06:48.370 --> 00:06:51.190 specifically in the OFC in tasks that 00:06:51.190 --> 00:06:54.030 involve reward and affect as 00:06:54.030 --> 00:06:57.410 well as in our own experimental predictions is because what 00:06:57.410 --> 00:07:00.570 unites all these specific all these different 00:07:00.570 --> 00:07:03.890 processes even though they seem separated there's some 00:07:03.890 --> 00:07:07.850 common element to all of them which is a predictive element so 00:07:07.850 --> 00:07:10.550 when you're estimating a reward or an 00:07:10.550 --> 00:07:14.870 effective value you think about the future you think about what would it give 00:07:14.870 --> 00:07:18.510 me or what how would it punish me or what did we do from for me or for anybody 00:07:18.510 --> 00:07:20.150 what would be the outcome of a certain 00:07:20.150 --> 00:07:28.390 choice so i think the ofc um should more wholly uh be seen as a as a. 00:07:29.918 --> 00:07:32.958 Primary player in thinking about the future and predictions rather 00:07:32.958 --> 00:07:36.418 in being specific to vision or to affect 00:07:36.418 --> 00:07:40.198 or to reward okay but then so that means ofc 00:07:40.198 --> 00:07:43.718 which is sort of integrating information from 00:07:43.718 --> 00:07:47.258 many sources include including vision is is 00:07:47.258 --> 00:07:50.038 making predictions as i say a behavioral time scale so it 00:07:50.038 --> 00:07:53.138 would be seconds not milliseconds or both 00:07:53.138 --> 00:07:56.058 both i mean in our experiments it 00:07:56.058 --> 00:07:59.218 was tens of of milliseconds so yeah so uh 00:07:59.218 --> 00:08:01.878 i don't know about the longer range but i 00:08:01.878 --> 00:08:04.718 i'm pretty sure that ofc is recruited also when people 00:08:04.718 --> 00:08:07.658 do what's called effective forecasting when people try to 00:08:07.658 --> 00:08:11.418 predict what would a trip to the um bahamas 00:08:11.418 --> 00:08:15.498 do to you if you like if you're going in half a year from now so it's far away 00:08:15.498 --> 00:08:19.238 but still you can estimate even though people show that effective forecasting 00:08:19.238 --> 00:08:24.318 is something that we're not so good at and um but nevertheless i think that 00:08:24.318 --> 00:08:29.058 ofc is response is responsive or is involved in all time scales of predictions 00:08:29.058 --> 00:08:30.738 not necessarily is there any, 00:08:31.378 --> 00:08:34.398 upper bound to that or that would also go to hours and days, 00:08:35.098 --> 00:08:39.478 as I said with the VK I didn't do an experiment so I'm just speculating I understand 00:08:39.478 --> 00:08:41.798 yeah but I don't see any reason why it would go, 00:08:42.338 --> 00:08:46.458 elsewhere this type of process to depending on time scale it will go to another 00:08:46.458 --> 00:08:50.598 area I think if there's a region that knows how to do this it can be employed 00:08:50.598 --> 00:08:52.878 in all time scales I don't see why not but. 00:08:53.958 --> 00:08:56.718 But I can't say more than speculate yeah but then but 00:08:56.718 --> 00:08:59.798 but there's another element or piece of 00:08:59.798 --> 00:09:03.138 the puzzle that that that you also revealed which is that it's 00:09:03.138 --> 00:09:06.538 not this ofc in some sense gets gets 00:09:06.538 --> 00:09:09.558 a very let's say rough impression of the world right 00:09:09.558 --> 00:09:13.398 it's not that it gets a high resolution kind of impression but it's a fairly 00:09:13.398 --> 00:09:18.198 rough understanding of the world so so what's going on there so in the in the 00:09:18.198 --> 00:09:22.998 case of our experiments which was in vision what you're referring to was a low 00:09:22.998 --> 00:09:26.898 spatial frequency or you can think of as a blurred picture of reality, 00:09:27.178 --> 00:09:33.378 meaning that we indeed, as you said, OFC is not. 00:09:35.920 --> 00:09:40.140 Sensitive to the details, it's not informed of the details, it just gets the gist. 00:09:40.460 --> 00:09:47.480 It gets the gist of a picture, of a scene, of a situation, and it's enough to 00:09:47.480 --> 00:09:50.860 direct the more specified expert 00:09:50.860 --> 00:09:55.300 type of cortex, in this case the visual cortex, how to behave, what to. 00:09:56.260 --> 00:09:57.800 Focus the analysis on. 00:09:57.980 --> 00:10:01.420 So it's enough to say, hey, there is a blob in the upper right corner which 00:10:01.420 --> 00:10:06.100 is not expected, I guess he should go there and analyze it thoroughly because 00:10:06.100 --> 00:10:09.400 it might be something falling on my head or something like this. 00:10:09.860 --> 00:10:15.440 So you're right completely that what we're saying is that it has a gist level understanding. 00:10:15.940 --> 00:10:21.000 But as I showed, I really like this picture of a street with a car in the middle. 00:10:21.060 --> 00:10:24.220 And all of you guys knew that it was a car, even though the picture is severely 00:10:24.220 --> 00:10:32.180 blurred. So our ability to understand these blobs keeps surprising me. 00:10:32.260 --> 00:10:37.500 And I came to the conclusion where give me a context and a low spatial frequency 00:10:37.500 --> 00:10:40.140 image, and I know what everything in the picture is. 00:10:40.340 --> 00:10:45.940 Of course, if you want to know a type of a car or identity of a person or something 00:10:45.940 --> 00:10:48.520 like this, you'll need the details. I'm not saying the details are useless. 00:10:48.900 --> 00:10:53.420 But for everyday quick decisions, low spatial frequency seems sufficient in 00:10:53.420 --> 00:10:57.520 many instances. So your prediction would be that this would also hold for other modalities. 00:10:57.780 --> 00:11:01.460 Like if we look at audition for instance, it would also be, let's say, 00:11:01.460 --> 00:11:06.200 some low-pass filtered version of an auditory world that would enter OFC. 00:11:06.400 --> 00:11:11.240 Yeah, this would be my prediction, but I would love seeing it done by somebody. Of course. 00:11:11.840 --> 00:11:16.360 But then why would it rely on this low-pass filtered version of the world? 00:11:17.234 --> 00:11:22.974 Yeah, because it seems, look, maybe you gain a few dozen milliseconds because 00:11:22.974 --> 00:11:26.474 you can rely on sort of a bit faster pathways to get the information. 00:11:26.694 --> 00:11:31.874 So here we go, OFC, we gain 10 milliseconds as compared to a fast processing 00:11:31.874 --> 00:11:33.714 or a high resolution processing pathway. 00:11:34.074 --> 00:11:38.074 So why would those 10 milliseconds be so crucial? Yes. 00:11:38.694 --> 00:11:42.314 So I used to think like this too, and it puzzled me. 00:11:42.334 --> 00:11:45.374 Why would, you know, it's not 10 milliseconds, a few tens of milliseconds, 00:11:45.494 --> 00:11:51.014 but still, I just recognizing a chair, why would I need any heads? 00:11:51.014 --> 00:11:54.774 I mean, why would I need any advanced warning? warning so um 00:11:54.774 --> 00:11:57.694 there are two things here first it's the most trivial 00:11:57.694 --> 00:12:00.874 question answer which is uh this ability has 00:12:00.874 --> 00:12:03.734 evolved mostly for uh uh 00:12:03.734 --> 00:12:06.874 survival related more crucial uh type 00:12:06.874 --> 00:12:14.514 of recognition like ledoux's example john uh example with a snake in the woods 00:12:14.514 --> 00:12:17.914 that you want to know if it's a snake you don't care if it's a snake or a hose 00:12:17.914 --> 00:12:25.814 or just a type of a of a branch you just when I run away because in some very coarse level, 00:12:25.994 --> 00:12:29.374 it already looks suspicious, so we're safer if we just run away from it and 00:12:29.374 --> 00:12:31.554 analyze their high spatial frequencies later on. 00:12:32.594 --> 00:12:39.534 So in that case, getting half a second head start over the snake might be beneficial. 00:12:40.174 --> 00:12:44.974 But I think there's something unfair about thinking, oh, you know what, 00:12:45.074 --> 00:12:48.234 the HSF will bring all the information anyway, why do I need to rush? 00:12:51.294 --> 00:12:58.674 The thing is that I don't think that there's much that can be done on an image 00:12:58.674 --> 00:13:00.154 with high spatial frequencies only. 00:13:00.274 --> 00:13:04.394 I have some pictures that I didn't show yesterday, but kind of high spatial 00:13:04.394 --> 00:13:08.334 frequency only images that you can make sense out of them. 00:13:08.374 --> 00:13:11.874 You can look at them for hours, and you can't make sense out of them. 00:13:11.954 --> 00:13:16.034 So you need the low spatial frequencies, but actually it goes back to your question 00:13:16.034 --> 00:13:22.134 of why do we need them earlier. So I guess I'll stick to the first answer, which is, yeah. 00:13:22.394 --> 00:13:25.874 But that seems very funny, right? Okay, the results are there, 00:13:25.954 --> 00:13:28.194 so we don't have to debate the data. Right. 00:13:28.614 --> 00:13:31.934 But if you think about the brain as some sort of layered structure, 00:13:32.074 --> 00:13:36.234 where indeed you have structures like the amygdala that Joe Ledoux has been working quite a bit on, 00:13:36.834 --> 00:13:39.794 and there the interpretation, oh, the amygdala is actually really a fast, 00:13:40.014 --> 00:13:43.514 let's say, an alarm detector detector that rapidly prepares 00:13:43.514 --> 00:13:46.534 you for for action to to defend yourself against 00:13:46.534 --> 00:13:49.534 threats in the world and now it seems ofc in 00:13:49.534 --> 00:13:52.814 this interpretation is like rather redundant compared to 00:13:52.814 --> 00:13:58.574 such a really fast responder like the amygdala so i think that they share the 00:13:58.574 --> 00:14:03.154 information as we know they are connected heavily and i think they both share 00:14:03.154 --> 00:14:07.174 i i suspect they both share the same information including low spatial frequencies 00:14:07.174 --> 00:14:09.814 we've shown in other studies where um. 00:14:11.662 --> 00:14:15.682 It was initially an unrated study, but somehow it became linked to it. 00:14:15.742 --> 00:14:19.402 It was a study about first impressions, how people judge other people's faces. 00:14:19.982 --> 00:14:25.062 And we show the faces. We want to see how fast, how first are first impressions. 00:14:25.522 --> 00:14:29.322 And we show them faster and faster and faster. And we found that people, 00:14:29.382 --> 00:14:34.362 even in the masked presentations of faces, in 39 milliseconds, 00:14:34.882 --> 00:14:40.342 were already able to recognize and categorize them as threatening versus non-threatening. 00:14:40.342 --> 00:14:44.862 Well we didn't know what's the actual personality of these people so this was 00:14:44.862 --> 00:14:49.142 just correlated with much longer presentations of a few seconds so 39 milliseconds 00:14:49.142 --> 00:14:51.342 were enough for people to be, 00:14:52.022 --> 00:14:56.182 accurate in their first impressions to the extent that first impressions are accurate. 00:14:57.922 --> 00:15:02.622 Um then the follow-up experiment was to filter those faces because our assumption 00:15:02.622 --> 00:15:08.062 or our hypothesis was that this extraction of features from faces to infer certain 00:15:08.062 --> 00:15:10.782 you know threatness or whether it's threatening or not, 00:15:10.922 --> 00:15:13.422 was based on low spatial frequencies, of course. 00:15:14.322 --> 00:15:20.982 So that's how we feel. That's exactly what we found, that people judge faces 00:15:20.982 --> 00:15:26.322 quickly and they use low spatial frequencies for this judgment and not the high spatial frequencies. 00:15:26.402 --> 00:15:30.362 So it's not only the speed, but also the content that makes them... 00:15:31.022 --> 00:15:35.242 And we found that this activation was in the amygdala. 00:15:35.622 --> 00:15:39.262 So it was... And I think others, I'm pretty sure, actually, that others have 00:15:39.262 --> 00:15:42.642 shown sensitivity to low spatial frequencies in the amygdala. 00:15:42.722 --> 00:15:46.862 So here we're talking about an OFC that's in a way an extension of the amygdala. 00:15:46.882 --> 00:15:50.282 You can say it's redundant, but you can also think about its anatomical connections 00:15:50.282 --> 00:15:54.762 and how it's a polysensory hub that gets information and sends information to 00:15:54.762 --> 00:15:57.562 so many modalities at the same time. 00:15:57.662 --> 00:16:02.922 So I think it's important that if the amygdala has this upper area information that can help... 00:16:04.550 --> 00:16:11.890 Make quick decisions, share this information with this area that can… This is interesting, right? 00:16:11.950 --> 00:16:16.530 Because originally, when you look at the story superficially, 00:16:16.630 --> 00:16:20.670 so from the outside, it looks like it's a story focusing on visual perception. 00:16:20.930 --> 00:16:24.250 So you think, okay, we have sort of a visual hierarchy, and that sort of sends 00:16:24.250 --> 00:16:26.230 information to the orbital frontal cortex. 00:16:26.570 --> 00:16:29.590 And orbital frontal cortex might be then generating predictions back to this 00:16:29.670 --> 00:16:34.350 sort of visual hierarchy to help it sort of resolve all sort of recognition problems. 00:16:34.970 --> 00:16:38.830 But if analyzed in these terms, it sounds much more like, let's say, 00:16:38.850 --> 00:16:45.550 a parallel processing stream that is capitalizing on subcortical processing like by the amygdala. 00:16:45.750 --> 00:16:49.010 So which of these two views would you lean to? 00:16:49.170 --> 00:16:53.270 Would you say both or it's indeed orbital frontal is more driven by the subcortical 00:16:53.270 --> 00:16:56.510 pathways or is it capitalizing other input streams? 00:16:57.530 --> 00:17:00.770 I think, and I'm sure that you'll share this view, that the brain is pretty 00:17:00.770 --> 00:17:04.870 opportunistic in the sense that it will use any information it could to solve a certain problem. 00:17:04.990 --> 00:17:09.370 So I'm not ready to commit to only subcortical or only cortical. 00:17:09.410 --> 00:17:11.470 I think it uses everything possible. 00:17:11.530 --> 00:17:15.090 We know from the anatomy that it gets more of this subcortical information. 00:17:15.590 --> 00:17:26.150 So I'm happy to say that OFC or to hypothesize that OFC indeed benefits from subcortical pathways. 00:17:26.490 --> 00:17:31.810 But at the same time, we know that dorsal pathways that have this Magno-Serac 00:17:31.810 --> 00:17:38.510 type of information project to OFC, and I don't see why OFC has to commit only to subcortical. 00:17:38.810 --> 00:17:40.730 And in a way, now you're bringing me to. 00:17:42.018 --> 00:17:47.298 Think aloud about another advantage of OFC over amygdala. 00:17:47.398 --> 00:17:51.078 Not advantage in saying, I mean, we're not comparing them, but information that 00:17:51.078 --> 00:17:54.778 OFC has that amygdala doesn't, which is more this cortical information that 00:17:54.778 --> 00:17:56.718 it's getting from the dorsal and from other pathways. 00:17:56.998 --> 00:18:01.358 So it can combine, integrate information from more sources than just amygdala. 00:18:01.918 --> 00:18:07.118 Well, in some sense, OFC would be well-placed to modulate the amygdala and its 00:18:07.118 --> 00:18:08.318 responses to the stimuli. 00:18:08.658 --> 00:18:12.938 Because amygdala needs this kind of control. Exactly. Or jumping all the time, 00:18:13.018 --> 00:18:14.578 seeing snakes everywhere. 00:18:16.438 --> 00:18:23.158 But that might mean that this interpretation of you exploiting information in 00:18:23.158 --> 00:18:28.918 the OFC to resolve the perception problem might actually not really be the right emphasis. 00:18:28.918 --> 00:18:32.858 So maybe OFC is just exploiting this low-frequency information, 00:18:33.438 --> 00:18:38.838 accumulating predictions, building up context information about the world for 00:18:38.838 --> 00:18:43.338 more general, let's say, action planning, something along these lines. 00:18:43.338 --> 00:18:50.678 Some people talk about this, and there's a chapter or a paper that I wrote with 00:18:50.678 --> 00:18:56.658 Lisa Feldman Barrett that talks exactly about how these predictions actually prepare the body. 00:18:56.778 --> 00:19:00.478 It's more in the context of affect, but I won't be surprised if these predictions 00:19:00.478 --> 00:19:05.378 that are generated or triggered by the OFC are then disseminated not only to 00:19:05.378 --> 00:19:07.038 perception, but just anybody, 00:19:07.318 --> 00:19:11.718 any taker, any area that wants to benefit or can benefit from it or get it included. 00:19:11.718 --> 00:19:13.398 Of course, preparation for action. 00:19:13.638 --> 00:19:19.538 But now, okay, so now we have OFC. OFC has been building up predictions, right? 00:19:19.578 --> 00:19:22.378 And predictions, as you said earlier, at multiple timescales. 00:19:23.718 --> 00:19:27.818 So, and then the example could be, let's say, the blurry car in the blurry street. 00:19:27.898 --> 00:19:31.078 So now I can say, okay, car in the street, and I have the following predictions 00:19:31.078 --> 00:19:36.258 about the building collapsing and Superman coming out of the phone booth and so on, right? Yeah. 00:19:37.724 --> 00:19:44.064 But now, if you think about context or contextual memory, you very quickly end 00:19:44.064 --> 00:19:45.464 up with structures like the hippocampus. 00:19:47.504 --> 00:19:50.464 So, first we looked at now the comparison with amygdala. 00:19:50.664 --> 00:19:54.244 But in some sense, you could then also argue, if OFC gives me contextual information, 00:19:54.604 --> 00:19:59.104 in some sense, I seem to be repeating a bit the job of my hippocampus, 00:19:59.104 --> 00:20:03.084 which we know is very much dedicated to the formation of these episodic memories. 00:20:03.144 --> 00:20:07.464 So, look, I am in this street and I see the car and the building collapsed and so on. 00:20:07.724 --> 00:20:12.164 So what's the added value now of OFC if I compare it to this episode of memory? 00:20:12.164 --> 00:20:16.424 At some point, you start representing what I'm saying, and you start saying something else. 00:20:16.564 --> 00:20:21.244 And I didn't really claim that OFC does context. It doesn't activate context. 00:20:21.404 --> 00:20:24.364 It triggers predictions. 00:20:24.624 --> 00:20:28.464 I'm not even sure that the predictions themselves are the OFC, 00:20:28.464 --> 00:20:32.164 or rather it sends instructions to the relevant cortex what kind of predictions 00:20:32.164 --> 00:20:33.744 or what kind of representations are relevant. 00:20:33.904 --> 00:20:38.044 I just bring them online and make it a prediction by bringing them online. 00:20:38.044 --> 00:20:44.604 Line so uh in in the studies i showed the second half of my talk uh where i 00:20:44.604 --> 00:20:48.764 talked about context it wasn't the ofc anymore that was involved it was another 00:20:48.764 --> 00:20:52.564 prefrontal region the medial prefrontal cortex that was part of a network that 00:20:52.564 --> 00:20:55.364 included the the the empty the medial temporal lobe. 00:20:55.904 --> 00:20:59.124 Especially the paripocampal cortex and the retrosplenial area 00:20:59.124 --> 00:21:05.264 with a posterior cingulate so there it plays it's different in the role of the 00:21:05.264 --> 00:21:12.024 OFC and we're still studying actually up to late night late last night I was 00:21:12.024 --> 00:21:17.164 corresponding with two people in my lab about a review that we're writing and 00:21:17.164 --> 00:21:18.704 we try to figure out what does the. 00:21:19.344 --> 00:21:21.624 Prefrontal cortex does in this in this. 00:21:23.303 --> 00:21:29.043 What does it do in this type of a network of contexts? There are different hypotheses. 00:21:29.443 --> 00:21:34.743 It's still speculative, so I'm not sure you want to hear them, 00:21:34.843 --> 00:21:39.383 but we think there are different types of contextual activations in the medial 00:21:39.383 --> 00:21:41.363 temporal lobe and in the retrosplenial, 00:21:41.603 --> 00:21:46.743 which some will be more sensitive to the specifics of the context. 00:21:46.863 --> 00:21:50.843 When I tell you the context of a kitchen, and you can think about five, 00:21:50.923 --> 00:21:54.663 seven items, but you're not committing to a specific appearance. 00:21:54.903 --> 00:21:57.583 You know, the fridge can be stainless steel, it can be white, 00:21:57.723 --> 00:22:00.043 it could be on the left, it could be on the right. 00:22:00.403 --> 00:22:05.883 You know some basic stuff like that the sink is concave and it will be on the 00:22:05.883 --> 00:22:10.563 level of the counter, but other specific features you're not committing to because 00:22:10.563 --> 00:22:13.003 you need to see the actual exemplar on the specific kitchen. 00:22:13.103 --> 00:22:17.443 Then these blobs, these slots are being filled with actual specifics. 00:22:17.443 --> 00:22:21.143 Specifics so there is a coarse or abstract representation of 00:22:21.143 --> 00:22:23.783 of a context people in the cognitive psychology in the 00:22:23.783 --> 00:22:26.723 past call it schema for example so there is 00:22:26.723 --> 00:22:31.703 a schema or a frame of of a specific context and it's yet to be filled you know 00:22:31.703 --> 00:22:36.203 there is a kitchen oven and a sink and a refrigerator there but you don't know 00:22:36.203 --> 00:22:39.523 where and you don't know how exactly they look so you give some basic information 00:22:39.523 --> 00:22:44.683 so one part of this network is sensitive to the to the schema and we think it's 00:22:44.683 --> 00:22:45.543 the retrospinal cortex, 00:22:45.743 --> 00:22:48.923 and together with Elisa Eminov and Dan Schachter, we published. 00:22:49.723 --> 00:22:53.883 Some studies that support this, but the specific appearance and the specific 00:22:53.883 --> 00:22:58.183 properties of the context frame is filled up with details, and this happens 00:22:58.183 --> 00:23:01.623 more in the parhypocampal cortex and possibly also in the hippocampus. 00:23:01.643 --> 00:23:05.603 So this is more sensitive to the end and with interactions in the visual cortex 00:23:05.603 --> 00:23:07.243 in the sense of a visual context. 00:23:08.323 --> 00:23:12.903 What the medial prefrontal cortex does here is, again, we think some sort of 00:23:12.903 --> 00:23:15.803 an integration, but we're far from 00:23:15.803 --> 00:23:19.623 being able to make it explicit but the concept would be something like, 00:23:20.296 --> 00:23:23.556 A frontal area gives you more, let's say, a frame for integration, 00:23:23.796 --> 00:23:28.196 a more abstract kind of representational scheme, which indeed you might fill 00:23:28.196 --> 00:23:31.636 in also partially with predictions coming from your orbital frontal cortex, what have you. 00:23:32.076 --> 00:23:37.776 But now to fill in those hooks with concrete information, you have to rely on 00:23:37.776 --> 00:23:41.796 areas like parahippocampal area or even hippocampus itself more than a visual 00:23:41.796 --> 00:23:44.116 hierarchy in case of vision. This is the concept. 00:23:44.436 --> 00:23:48.676 So frontal more, let's say, an abstract framework, there's a frame in which 00:23:48.676 --> 00:23:53.536 you would integrate, and sort of preceding areas really providing you with that content. 00:23:54.216 --> 00:23:57.436 Well, yeah, what I was saying is that the retrosplenial, actually, 00:23:57.556 --> 00:24:01.796 the medial parietal is the one that involves the more abstract representation, 00:24:02.056 --> 00:24:05.976 the more abstract of a context frame, and the prefrontal cortex does some kind of integration. 00:24:06.296 --> 00:24:11.256 And we shouldn't forget also the time dimension, the temporal domain where actually 00:24:11.256 --> 00:24:13.936 things in context don't necessarily happen simultaneously. 00:24:13.936 --> 00:24:16.676 Right, I wanted to ask you about that, because you present some interesting 00:24:16.676 --> 00:24:21.456 results on that, which had to do with how these different areas actually establish 00:24:21.456 --> 00:24:23.776 specific phase relationships in their responses. 00:24:24.376 --> 00:24:30.156 Right. So how is that informative about their interactions? 00:24:30.876 --> 00:24:35.976 That's very good. Just before I get into this, I don't want the previous point to be lost. 00:24:36.156 --> 00:24:39.856 What I was talking about there was that context can be a spatial context of 00:24:39.856 --> 00:24:42.356 things that happen at the same time together in the same environment. 00:24:42.356 --> 00:24:48.636 But there's also a temporal dimension of things happening after things or before 00:24:48.636 --> 00:24:50.696 things. So there's some kind of temporal order. 00:24:51.636 --> 00:24:57.136 So a context is that if you hear a certain sound, you expect another sound afterwards, right? 00:24:57.556 --> 00:25:01.476 So context can be in space. It could be in time. 00:25:01.556 --> 00:25:05.916 And back to your question, this phase lock analysis that we're doing, 00:25:06.036 --> 00:25:08.256 especially with MEG, it's much 00:25:08.256 --> 00:25:12.436 easier to do it because the signal has such better temporal resolution, 00:25:13.336 --> 00:25:20.316 it allows us to infer, we can't really conclude, but suggest and infer patterns 00:25:20.316 --> 00:25:25.156 of connectivity that this can also be done with DCM and other methods, also with MRI. 00:25:25.336 --> 00:25:30.696 But with MEG, we could look for areas that are co-activated with the same phase 00:25:30.696 --> 00:25:32.536 or with a fixed phase difference. 00:25:33.576 --> 00:25:38.736 And from this, we suspect that these areas do something together. 00:25:38.936 --> 00:25:44.116 They're active at the same time in the same pattern. 00:25:44.476 --> 00:25:47.976 And we took it a step further, and I think it relates to a question you asked 00:25:47.976 --> 00:25:50.096 yesterday, and we wanted to test causality. 00:25:50.236 --> 00:25:52.856 So if area A and area B are... 00:25:53.542 --> 00:25:58.702 Activated in the same pattern, does it mean that one affects the other or the 00:25:58.702 --> 00:26:03.802 other way around, or there's a third factor here, or they're just synchronized as an epiphenomenon? 00:26:03.942 --> 00:26:10.602 Using tools such as Granger causality, we could test which area affects the other. 00:26:10.782 --> 00:26:15.122 I really like these demonstrations, even though, again, nothing here is completely 00:26:15.122 --> 00:26:23.982 conclusive because this is highly suggestive, that certain areas speak and affect other areas. 00:26:24.302 --> 00:26:28.422 And it's interesting that in MEG experiment of this context network that I mentioned, 00:26:28.682 --> 00:26:35.582 I didn't present these results yesterday but we find that objects that are highly contextual, 00:26:35.942 --> 00:26:39.842 like the roulette that I showed yesterday or a bowling pin, that are highly 00:26:39.842 --> 00:26:44.522 diagnostic of a specific context, this network of three main nodes is active 00:26:44.522 --> 00:26:48.682 and also is highly correlated, highly phase-locked. 00:26:48.702 --> 00:26:50.282 These three nodes are highly phase-locked. 00:26:50.342 --> 00:26:55.262 If you show an object like scissors or a cherry that is not highly contextual, 00:26:55.502 --> 00:26:59.122 it's as common in our environment, even more common than a roulette, 00:26:59.362 --> 00:27:05.642 but it's as common and as we equate in all dimensions that we can think of, 00:27:05.922 --> 00:27:11.322 this network might be somewhat activated because because all of us have some 00:27:11.322 --> 00:27:12.942 associations with everything else, 00:27:13.582 --> 00:27:15.642 but it's not as diagnostic and as consistent. 00:27:15.802 --> 00:27:18.542 As a result, these nodes are not as synchronized. 00:27:19.702 --> 00:27:24.382 So supporting our suggestion that this network is related to contextual associations. 00:27:24.922 --> 00:27:28.842 But it's essentially a three-node system, right? Yeah. 00:27:29.982 --> 00:27:36.062 Would each node provide a specific component of that context information in 00:27:36.062 --> 00:27:38.982 this case, following the framework we discussed earlier? Yeah. Okay. 00:27:40.582 --> 00:27:44.942 So that's the framework, that's the network that I was trying to ascribe functions 00:27:44.942 --> 00:27:49.682 to each node where we're talking about a schema or an abstract context frame 00:27:49.682 --> 00:27:52.282 with slots that are yet to be filled, 00:27:52.402 --> 00:27:57.762 and then another node that provides the actual features, the actual properties, 00:27:58.042 --> 00:28:01.702 and the third node, maybe the prefrontal cortex, that does the integration of 00:28:01.702 --> 00:28:07.642 some sort and maybe utilizes this information to deploy other areas with predictions. 00:28:07.642 --> 00:28:10.782 But now the synchronization across these three nodes, 00:28:12.015 --> 00:28:15.435 is, let's say, a straightforward form of synchronization? Let's say they all 00:28:15.435 --> 00:28:17.395 start to oscillate in sync, and that's it? 00:28:18.015 --> 00:28:21.975 Or do you see a more, let's say, fine-tuning? 00:28:22.115 --> 00:28:26.195 Let's say you will only see a synchronization within a certain frequency range 00:28:26.195 --> 00:28:29.675 and not in others. And this is, again, node-specific. 00:28:30.735 --> 00:28:36.055 Well, I wish we could be so elaborate in this first analysis, 00:28:36.175 --> 00:28:39.975 but I definitely can tell you that the synchrony was in specific frequency bands, 00:28:40.075 --> 00:28:43.715 which I can't recall now, but it's a PNAS paper that came out a year or two ago. 00:28:47.655 --> 00:28:51.395 So it was specific to frequency band. It didn't just happen all over. 00:28:51.815 --> 00:28:57.775 But I wish I could tell, and maybe in the future we will be able to, 00:28:57.835 --> 00:29:04.455 say how this synchronization is modulated with different information. 00:29:04.655 --> 00:29:09.155 There are some intriguing demonstrations that started, like many other good 00:29:09.155 --> 00:29:11.735 things, started with old cognitive psychology. 00:29:12.315 --> 00:29:17.775 So for example, if I give you the word bank, you can activate two types of context 00:29:17.775 --> 00:29:21.695 frames as we call them context frames. 00:29:21.775 --> 00:29:25.055 One of them is a bank with money and with tellers and all these things. 00:29:25.095 --> 00:29:27.655 The other one is the river bank, right? When you think about fishing, 00:29:27.895 --> 00:29:29.075 jumping in the water, vacation. 00:29:29.455 --> 00:29:34.615 So for a given second or moment, there are two context frames that are active. 00:29:35.055 --> 00:29:41.835 And then if I tell you bank water, then it kind of disambiguates it and you 00:29:41.835 --> 00:29:44.635 know it's the bank with the river, not the other bank. 00:29:44.715 --> 00:29:49.055 So you suppress one context frame, you activate the other, and you're more committed to it now. 00:29:49.295 --> 00:29:53.595 So I bet if you could look, and we did something like this with vision also, with. 00:29:57.593 --> 00:30:01.253 I bet if you looked at the synchrony of this network, within this network, 00:30:01.353 --> 00:30:06.153 during this process, you would see some adjustments being made as two context 00:30:06.153 --> 00:30:09.433 frames are activated, one of them is suppressed, the other one is committed to. 00:30:09.773 --> 00:30:13.313 So I would expect this synchrony, if it's related to the actual function, 00:30:13.553 --> 00:30:16.573 to be changed based on this process. Right. 00:30:16.793 --> 00:30:20.793 But now, would you... Now, there are two interpretations of this, right? 00:30:20.813 --> 00:30:25.033 Because you could say, well, what you call the context frame could be like an 00:30:25.033 --> 00:30:28.933 emergent property, if you want, of the synchronized activity across your three 00:30:28.933 --> 00:30:32.853 nodes, or you could localize it within one of the nodes and say, 00:30:32.933 --> 00:30:35.593 and the other guys are sort of piping information into it. 00:30:36.413 --> 00:30:40.333 So, could you make... Is it possible to distinguish between these two interpretations? 00:30:40.733 --> 00:30:43.113 Not at this point. I think they're equally likely, yeah. 00:30:43.653 --> 00:30:45.993 Do you have a preference for any of these two interpretations? 00:30:47.093 --> 00:30:49.893 I like my interpretation with black 00:30:49.893 --> 00:30:54.713 coffee. No, I don't have any personal preference. I think that they're... 00:30:55.953 --> 00:30:59.953 I mean, given what we know about prefrontal cortex, if I had to put money on 00:30:59.953 --> 00:31:02.853 something, I would put money on the interpretation that says that everything 00:31:02.853 --> 00:31:07.653 feeds into the prefrontal cortex and then it decides what to do and guides other areas. 00:31:07.793 --> 00:31:13.833 But the data doesn't show it yet, so I can't commit to it. Right. 00:31:13.993 --> 00:31:21.473 Okay. So now you also, after dealing with this issue of a context network that 00:31:21.473 --> 00:31:23.113 we now discussed, right? 00:31:23.153 --> 00:31:26.213 So what are the ingredients of context in this? 00:31:26.313 --> 00:31:30.393 Are there boundaries to context information that you would consider in this network? 00:31:31.113 --> 00:31:35.973 For instance, you could say, well, context does not include information about 00:31:35.973 --> 00:31:39.393 self because context is only oriented towards the outside world. 00:31:39.473 --> 00:31:40.673 This could be a boundary. 00:31:42.033 --> 00:31:48.213 Yeah, that's interesting, because for a while we had a little debate with a 00:31:48.213 --> 00:31:54.273 group of people that ascribed specifically to the parahippocampal cortex role in place. 00:31:55.353 --> 00:31:59.333 And initially the criticism was, hey, what you found as contextual activation 00:31:59.333 --> 00:32:02.393 in parahippocampal is actually place in spatial information. 00:32:02.393 --> 00:32:08.673 So for this, which was legitimate criticism, for this we had to design experiments that, 00:32:09.393 --> 00:32:12.873 differentiated between spatial context and context that's more abstract, 00:32:13.133 --> 00:32:18.353 like a picture of Cupid and a heart-shaped chocolate box. 00:32:18.633 --> 00:32:23.313 So both of them has to do with romance, and you hardly see them both together. 00:32:23.373 --> 00:32:24.673 I never saw Cupid in my life. 00:32:24.793 --> 00:32:28.993 So you never see them together in the same place, or justice, 00:32:29.073 --> 00:32:35.173 or other contexts that are not as physically bound to each other and not necessarily 00:32:35.173 --> 00:32:36.933 spatially related to each other. 00:32:37.393 --> 00:32:41.473 So we proved our point that context, and especially in this network, 00:32:41.593 --> 00:32:45.953 is not limited to space-related context, but it did make us think about, 00:32:46.113 --> 00:32:48.393 okay, so what is context if it's not appearing together? 00:32:48.533 --> 00:32:52.573 Because it's so tempting and so easy to think about context as space, 00:32:52.693 --> 00:32:55.253 as things that happen together and... 00:32:57.657 --> 00:33:03.437 And I started to think pretty intensely about, okay, what defines context if it's not space? 00:33:04.957 --> 00:33:12.757 And another construction definition that I'm currently liking the best is that 00:33:12.757 --> 00:33:15.697 actually to contextual relation is, 00:33:15.877 --> 00:33:20.577 or contextual related are all the items that are activated together. 00:33:20.677 --> 00:33:22.877 So rather than looking at the environment, it's looking at the brain. 00:33:22.977 --> 00:33:26.837 All the things that are activated together as a result of one of them appearing 00:33:26.837 --> 00:33:28.817 or occurring are contextually related. 00:33:29.037 --> 00:33:34.597 So if I show you a picture of Cupid, you activate other things that are related to it in your mind. 00:33:34.737 --> 00:33:37.737 And these are the things, that's the definition of context in my mind. 00:33:38.117 --> 00:33:42.357 Now, self can and cannot be, I mean, depending on the instance, 00:33:42.457 --> 00:33:43.857 because if you want to think about. 00:33:46.637 --> 00:33:52.077 The context of a cold shower right now, you would imagine yourself in the shower, 00:33:52.137 --> 00:33:55.997 I guess, and you can... How much more can you do in a cold shower? 00:33:57.457 --> 00:34:00.697 The beach okay so so 00:34:00.697 --> 00:34:04.677 me in a culture okay so then it's not self it's uh yes 00:34:04.677 --> 00:34:07.517 but i think that self is just in that regard might 00:34:07.517 --> 00:34:13.517 be another another uh item that could be in and out but if you define it so 00:34:13.517 --> 00:34:18.677 broadly then context start to coincide with something like working memory i 00:34:18.677 --> 00:34:23.057 guess because you're saying anything that is active in the brain well as a result 00:34:23.057 --> 00:34:25.497 of of uh relation not as a result 00:34:25.837 --> 00:34:28.737 of you know if i give you a telephone number and you put it in your working 00:34:28.737 --> 00:34:31.817 memory that's not because the digits are contextually 00:34:31.817 --> 00:34:34.717 related but just to finish the point about the self 00:34:34.717 --> 00:34:39.897 if you don't mind sure uh we have to be aware also of other people's findings 00:34:39.897 --> 00:34:45.117 here that uh this network of it's a medial network that we call the contextual 00:34:45.117 --> 00:34:49.437 network with the three nodes the prefrontal the medial temporal and the medial 00:34:49.437 --> 00:34:52.437 parietal as i showed yesterday also highly 00:34:52.617 --> 00:34:54.437 overlaps with the default network. 00:34:54.837 --> 00:34:59.017 Right. And when the default network, so we treat them almost as the same. 00:34:59.057 --> 00:35:03.017 And the default network has received a lot of attention recently of people trying 00:35:03.017 --> 00:35:07.717 to explain its function, trying to find a function. And one of the prominent functions, 00:35:08.423 --> 00:35:12.343 theories, other than our own account of saying, oh, the default network is engaged 00:35:12.343 --> 00:35:14.803 in associative activation and the generation of predictions, 00:35:15.023 --> 00:35:16.443 as in planning and simulations. 00:35:17.403 --> 00:35:24.603 But some groups think that the default network does self-referential processes, 00:35:24.863 --> 00:35:26.403 things that relate to self. 00:35:29.063 --> 00:35:33.023 So in a way, they'll answer your question, sure, context is self, 00:35:33.083 --> 00:35:36.363 because it's the same network. It's activated both by context and by self. 00:35:38.443 --> 00:35:46.123 Yeah, but now if we slowly move then to this possible similarity with the default network. 00:35:47.423 --> 00:35:52.723 So is the default network for you an epiphenomenon or it's sort of a real feature 00:35:52.723 --> 00:35:55.063 of brain dynamics that we should explain? 00:35:55.603 --> 00:36:00.343 Oh, it's definitely a real feature and it's intended. I mean, 00:36:00.383 --> 00:36:03.583 you know, you can't have a third of the brain active so vigorously, 00:36:03.603 --> 00:36:07.303 wasting so much energy as an epiphenomenon or something that's useless. 00:36:07.483 --> 00:36:11.623 I'm sure that in my logic, at least, or in my thinking, it has to serve a function. 00:36:11.783 --> 00:36:16.543 And in this case, this actually is what led me to think about the brain as proactive 00:36:16.543 --> 00:36:25.003 and as being continuously on the move to be ready, to be preparing for the future. 00:36:25.003 --> 00:36:29.563 And there are some interesting metaphors that I like to give but they might 00:36:29.563 --> 00:36:34.423 have became cliches by now one of them is for example you know playing I like 00:36:34.423 --> 00:36:37.983 to play squash and you know in squash you have to be always moving even if it's 00:36:37.983 --> 00:36:39.503 your other if your opponent's. 00:36:41.138 --> 00:36:45.118 Turn, you're still moving your feet. It's much easier to move into action from 00:36:45.118 --> 00:36:48.138 a moving, and you know this probably better than I do. 00:36:48.998 --> 00:36:52.798 And in one of the introductions I wrote to either the book on predictions or 00:36:52.798 --> 00:36:57.618 the special issue we had, or one of them anyway, I compared the brain to F-16. 00:36:57.898 --> 00:37:00.998 Showing my past with the Air Force. 00:37:00.998 --> 00:37:08.058 And this type of fighter jets that, because they have to be so agile and so 00:37:08.058 --> 00:37:13.958 maneuverable, their steady state, in a way, is not steady at all. It's to be really wild. 00:37:14.138 --> 00:37:18.638 And it's unlike jumbo that you want to be the Boeing. You want to be steady 00:37:18.638 --> 00:37:21.438 and not to have any abrupt movements. 00:37:21.578 --> 00:37:26.818 Here, in order to be agile, you want to start with a system that's easily transforming 00:37:26.818 --> 00:37:28.438 to one position versus another. 00:37:28.438 --> 00:37:31.798 Other so of course it's a metaphor with a lot of caveats 00:37:31.798 --> 00:37:34.538 so don't take it too seriously but but that's how 00:37:34.538 --> 00:37:37.618 i like to think about the default network as preparing you and 00:37:37.618 --> 00:37:40.298 it's much harder it seems to to go 00:37:40.298 --> 00:37:42.958 into action or into cognition or into some kind of a 00:37:42.958 --> 00:37:45.878 mental operation from a brain dead 00:37:45.878 --> 00:37:48.738 position to rather than something that's already 00:37:48.738 --> 00:37:51.598 preparing always thinking always on the move and of 00:37:51.598 --> 00:37:55.218 course there are also i think the default network involves uh a 00:37:55.218 --> 00:37:58.318 different time scales of planning and of thinking so when 00:37:58.318 --> 00:38:01.178 you stack in traffic you don't your default network is not 00:38:01.178 --> 00:38:04.078 only thinking about the next second because this will be boring 00:38:04.078 --> 00:38:06.998 it will be just like this second but uh it 00:38:06.998 --> 00:38:11.718 also thinks about this afternoon when you arrive home or this evening or or 00:38:11.718 --> 00:38:15.198 your dinner or you're going out with friends afterwards or playing with the 00:38:15.198 --> 00:38:20.118 kids it also involves thinking about the conference you're going to in two weeks 00:38:20.118 --> 00:38:24.278 or something that's so different time scales and And there's something interesting 00:38:24.278 --> 00:38:25.498 that I want to say about this, 00:38:25.538 --> 00:38:28.618 even though I don't think I discussed this in the talk yesterday. 00:38:29.618 --> 00:38:33.638 It's the issue of simulations and the experiences that they, 00:38:33.758 --> 00:38:37.158 experiences in quotation marks, that they afford us. 00:38:37.338 --> 00:38:41.358 So as we all agree, I think, we store in our memory, 00:38:42.101 --> 00:38:44.761 our experiences and we store them with a 00:38:44.761 --> 00:38:48.041 primary reason well there's no proof that that's the primary reason but it seems 00:38:48.041 --> 00:38:52.701 that it's the primary reason or at least a primary reason of being able to use 00:38:52.701 --> 00:38:56.581 this experience in the future right so our behavior in the upcoming seconds 00:38:56.581 --> 00:39:02.521 or minutes or years is based on what we've learned and encoded so in a way experience helps us. 00:39:03.521 --> 00:39:08.061 Store scripts that sometimes can be action plans or motor plans and sometimes 00:39:08.061 --> 00:39:11.101 it could be in in conversations, sometimes it could be in dance, 00:39:11.221 --> 00:39:13.561 in playing basketball, or anything like this. 00:39:13.661 --> 00:39:18.061 So we use our experience, our memories, for the future. 00:39:18.801 --> 00:39:26.021 Now, the default network, this area of just sitting there and stuck in traffic, 00:39:26.121 --> 00:39:29.821 or waiting for your doctor, or being in a shower. 00:39:30.561 --> 00:39:34.081 You create new experiences, again, with quotation marks. 00:39:34.181 --> 00:39:36.321 So you make simulations, You sit on a plane. 00:39:36.501 --> 00:39:40.181 I have a funny example, but what can I say? That's what went through my mind. 00:39:40.201 --> 00:39:43.781 I'm sitting on a plane and reading, reviewing some manuscript. 00:39:44.321 --> 00:39:47.981 And I'm thinking, what would happen? It was a very long flight. 00:39:48.121 --> 00:39:52.261 What would happen if this door opens up and all of us are starting to fall? 00:39:52.761 --> 00:39:58.441 Right? So I'm thinking, oh, I'll take this blanket that's on my lap and I'll use it as a parachute. 00:39:58.441 --> 00:40:01.141 Shoot but oh it might slip from my hands you 00:40:01.141 --> 00:40:03.921 know if i'll be sweating or whatever how do i make holes oh i have this 00:40:03.921 --> 00:40:06.721 pen that i'm holding now so of course this chances of 00:40:06.721 --> 00:40:12.661 this happening is one in a zillion right but let's say less real uh simulations 00:40:12.661 --> 00:40:17.341 we also think about other things now right what happens if uh i don't know what 00:40:17.341 --> 00:40:25.781 the electricity uh um uh yeah so uh we'll be able to manage, right? 00:40:25.861 --> 00:40:29.121 But if, let's say, if we both now simulate this in our mind, 00:40:29.401 --> 00:40:32.581 we think, oh, we'll just use the iPhone for making some light, 00:40:32.761 --> 00:40:35.241 and then we go out and call somebody, right? 00:40:35.281 --> 00:40:38.801 If it happens now, we'll be more ready than people who didn't simulate this. 00:40:38.881 --> 00:40:41.021 They'll probably also do it, but just a little later. 00:40:41.821 --> 00:40:48.101 So these simulations are an amazing way of creating experiences without experimenting. 00:40:48.461 --> 00:40:51.661 So Popper said that he lets 00:40:51.661 --> 00:40:54.981 his hypothesis die in his 00:40:54.981 --> 00:40:57.901 in his behalf so you generate all this 00:40:57.901 --> 00:41:00.581 hypothesis and you choose the right one and that's the one 00:41:00.581 --> 00:41:03.701 you store and now when there's a situation if 00:41:03.701 --> 00:41:06.601 it happens then you're ready for it even though you haven't experienced it 00:41:06.601 --> 00:41:10.161 so i think this is very powerful it's it almost allows you to just sit in your 00:41:10.161 --> 00:41:13.621 couch and experience everything in life and just store it and be ready for right 00:41:13.621 --> 00:41:22.561 but would it be surprising that then this default network is a relatively large 00:41:22.561 --> 00:41:24.501 number of neurons are involved in this, 00:41:24.621 --> 00:41:28.421 but still it's only a relatively small subset of your whole brain. 00:41:28.681 --> 00:41:31.061 And it's not necessarily engaging neurons, 00:41:31.623 --> 00:41:37.323 Let's say all possible areas that might provide you with memory or with contextual information. 00:41:38.203 --> 00:41:41.343 Or for instance, you could think about the so-called mirror mechanisms, 00:41:41.503 --> 00:41:46.123 which again would be running in different systems than this default state network. 00:41:46.703 --> 00:41:54.683 So are you then saying that the brain is running a number of substrates that support simulation? 00:41:55.603 --> 00:41:59.043 Or are they sort of interlinked in some way? 00:41:59.563 --> 00:42:05.623 That's an interesting question. So first of all, the classically defined default 00:42:05.623 --> 00:42:08.243 network does involve the medial temporal lobe. 00:42:08.363 --> 00:42:11.603 So, you know, an area that, but we know that the entire brain is busy doing 00:42:11.603 --> 00:42:13.003 memory. So it's not only that. 00:42:13.383 --> 00:42:19.623 So I would suspect that if your simulation involves smell, it will recruit at 00:42:19.623 --> 00:42:25.923 least momentarily the olfactory cortex or auditory cortex, depending on what 00:42:25.923 --> 00:42:27.143 you're doing in the visual cortex. 00:42:27.143 --> 00:42:36.683 The way the default network is defined is averaged across many subjects and many situations. 00:42:37.043 --> 00:42:43.623 So I would suspect that all these momentary recruitments of specific cortices 00:42:43.623 --> 00:42:47.643 or expert cortices is kind of washed out in the averaging. 00:42:47.843 --> 00:42:51.363 It's a possibility. And what you see is the major mechanism of these simulations, 00:42:51.643 --> 00:42:55.743 but you don't see these extensions that are recruited on and off. Right. 00:42:56.103 --> 00:43:01.323 But then how do you see the role of, let's say, subcortical structures to this. 00:43:02.343 --> 00:43:06.023 Default network, if it's a big simulator of the brain? 00:43:06.503 --> 00:43:11.883 Yeah. So I think it just falls under the category of what we said now about specific cortices. 00:43:11.983 --> 00:43:16.863 So if you simulate a situation that might be scary, I won't be surprised if 00:43:16.863 --> 00:43:18.703 you're recruiting your amygdala during the process. 00:43:18.723 --> 00:43:23.303 But I don't think that traditionally the amygdala will be part of this default 00:43:23.303 --> 00:43:28.323 network only when the simulation or the content of your thoughts pertains to 00:43:28.323 --> 00:43:29.323 anything that the amygdala does. 00:43:29.543 --> 00:43:33.783 Right. So that would mean what we now call the default network is really like, 00:43:33.903 --> 00:43:38.063 let's say, a very rough, let's say, backbone of the simulation structure. 00:43:38.403 --> 00:43:42.463 And then let's say when you are alert and acting, it gets sort of blown up into 00:43:42.463 --> 00:43:47.063 your context network or sort of starts to elaborate into this context network. 00:43:47.423 --> 00:43:50.723 And then when you become inactive and relaxed again, it sort of falls back in 00:43:50.723 --> 00:43:53.443 this default state mode again. And this would be roughly the model. 00:43:53.563 --> 00:43:56.283 Right. Right. Okay. So, but if I. 00:43:57.496 --> 00:44:01.496 So other views on the brain have focused more on, let's say, 00:44:01.516 --> 00:44:05.236 sentence-safe-like theories, where you would say, well, cognition relies very 00:44:05.236 --> 00:44:08.256 much on, let's say, the integrative properties of basal ganglia. 00:44:09.256 --> 00:44:17.596 So in your mind, let's say it's the default network that has this sort of core 00:44:17.596 --> 00:44:22.356 integrative powers that provide us with, let's say, the contents of cognition 00:44:22.356 --> 00:44:23.396 and possibly consciousness. 00:44:23.736 --> 00:44:27.696 Is this really the starting point of that or not? Yeah, and people have talked 00:44:27.696 --> 00:44:31.416 about the default network in the context of consciousness, but I want it to 00:44:31.416 --> 00:44:37.076 be clear that you make it almost sound like we know what's going on. 00:44:37.436 --> 00:44:39.836 No, you're the expert. I'm posting the questions. 00:44:40.476 --> 00:44:44.956 But gradually, we're starting both to believe that we know what we're talking about. 00:44:45.116 --> 00:44:51.296 So it has to be clear that we don't, and that the same type of simulations, 00:44:51.536 --> 00:44:55.276 I won't be surprised if different types of simulations are happening in the 00:44:55.276 --> 00:44:57.076 basal ganglia or other areas. 00:44:57.376 --> 00:45:03.036 So there's a lot of work that still needs to be done in order to better carve 00:45:03.036 --> 00:45:04.816 these different processes. 00:45:05.156 --> 00:45:09.836 What I said was pertaining specifically to this default network, 00:45:10.056 --> 00:45:14.776 but it doesn't exclude similar processes from taking place in other structures. 00:45:18.426 --> 00:45:26.086 What is the overlap, really? So the default network has largely been characterized using fMRI, right? 00:45:26.666 --> 00:45:30.326 Yeah, even though it started with PET. But that was the first demonstration. 00:45:30.466 --> 00:45:34.066 Yeah, you're right. But that means slow signals, right? 00:45:34.886 --> 00:45:40.166 So what do we know about, let's say, the correlates of this default network 00:45:40.166 --> 00:45:42.086 at faster timescales? Right. 00:45:42.346 --> 00:45:48.566 So people are only now starting to look at the default network in high temporal 00:45:48.566 --> 00:45:51.006 resolution modalities such as MEG and EEG. 00:45:51.066 --> 00:45:54.786 And one of the reasons is that its majority is in the medial surface, 00:45:54.946 --> 00:46:00.466 which is hard to kind of disentangle in methods such as MEG. 00:46:00.666 --> 00:46:06.346 But now as methods with MEG improved, we can start differentiating different medial structures. 00:46:07.386 --> 00:46:12.786 But I don't think, at least I'm not aware of the temporal characteristics of 00:46:12.786 --> 00:46:15.566 the default network other than it's active all the time. 00:46:15.986 --> 00:46:19.466 But how does it change over time? 00:46:19.966 --> 00:46:22.086 It'll be interesting to explore, yeah. Okay. 00:46:22.946 --> 00:46:27.306 But then in the last part of your talk, you sort of moved away. 00:46:28.166 --> 00:46:31.506 But okay, there was sort of, if you want, an associative link from the... 00:46:31.506 --> 00:46:33.606 We go back to orbital frontal cortex. 00:46:33.826 --> 00:46:37.526 We go also back to, let's say, the role of orbital the frontal cortex in 00:46:37.526 --> 00:46:41.906 in affective processing possibly mood and 00:46:41.906 --> 00:46:44.966 then you made this step towards depression and 00:46:44.966 --> 00:46:49.866 then the study of depression and that's how this this sounded a bit surprising 00:46:49.866 --> 00:46:53.986 because a bit like okay how does this now relate to this notion of contact so 00:46:53.986 --> 00:46:58.306 so what is the relationship between let's say contextual processing prediction 00:46:58.306 --> 00:47:01.946 and now depression yeah so um. 00:47:03.537 --> 00:47:09.197 It started, this link started by me reading somewhere that people in depression, 00:47:09.437 --> 00:47:13.977 and I had no interest in depression or in psychiatric disorders back in the 00:47:13.977 --> 00:47:20.557 day, to my embarrassment, but I was reading this interesting notion that people 00:47:20.557 --> 00:47:23.577 in depression have a hard time incorporating context. 00:47:23.577 --> 00:47:25.637 They don't analyze the broad context. 00:47:26.017 --> 00:47:31.777 And with all our focus on context, I felt the need and obligation to try to 00:47:31.777 --> 00:47:35.097 explain what in depression is related to this context. 00:47:35.437 --> 00:47:40.157 And we started thinking more and more about symptoms, especially cognitive symptoms 00:47:40.157 --> 00:47:45.877 and symptoms that characterize the pattern of thinking of depressed people. 00:47:45.977 --> 00:47:48.457 So for example, depressed people tend to ruminate. 00:47:48.597 --> 00:47:53.257 So they stuck in the same topic or the same thought over and over and over. 00:47:53.577 --> 00:47:54.957 It's a cyclical type of thinking. 00:47:55.857 --> 00:48:01.437 And this immediately sounded like the opposite of broad associative activation. 00:48:01.697 --> 00:48:06.657 So we expect and we know that the healthy brain goes from one thought to another, 00:48:06.777 --> 00:48:07.957 and it's very associative. 00:48:08.017 --> 00:48:11.917 So unless we're busy and focused on a very narrow task momentarily, 00:48:12.077 --> 00:48:14.677 the brain really goes from one thought to another. 00:48:14.797 --> 00:48:21.257 And here we have a population that is focused on the same topic in a clinical 00:48:21.257 --> 00:48:23.077 manner. They really go on and on and on. 00:48:23.577 --> 00:48:26.177 So this already showed some kind of linking. 00:48:27.937 --> 00:48:30.977 And from there, we started looking for all this. 00:48:31.037 --> 00:48:34.577 We've accumulated all this, some of it's circumstantial evidence and some of 00:48:34.577 --> 00:48:40.777 it is fresh evidence from my lab that shows relationship between mood and associative 00:48:40.777 --> 00:48:43.877 activation, specifically as it relates to context and to predictions, 00:48:44.117 --> 00:48:45.817 associative as in context and is prediction. 00:48:46.897 --> 00:48:50.997 So when I say circumstantial evidence is that when you have a hypothesis like 00:48:50.997 --> 00:48:55.357 this, that this population will suffer from a lack of foresight and you open 00:48:55.357 --> 00:48:58.397 the literature and you see a couple, well not enough, but still a couple of 00:48:58.397 --> 00:48:59.977 demonstrations that people with. 00:49:03.186 --> 00:49:16.546 With depression show deficiency in foresight that supports this idea, but we also show that, 00:49:17.266 --> 00:49:21.446 first of all, we can, together with Malia Mason, a paper that came out in JAP 00:49:21.446 --> 00:49:25.766 a couple of years ago, that shows that if you make even healthy individuals 00:49:25.766 --> 00:49:28.506 think narrowly versus think broadly, 00:49:28.786 --> 00:49:36.826 it can improve their mood significantly, 00:49:37.246 --> 00:49:42.026 not in a major way, but it's definitely statistically significant. 00:49:42.506 --> 00:49:46.366 So that's another way of supporting this idea. 00:49:49.066 --> 00:49:53.426 And overall, we find more and more... So for example, we showed with MRI, 00:49:53.686 --> 00:49:57.566 we haven't published this yet, but we show with MRI that people with depression 00:49:57.566 --> 00:50:01.226 don't recruit the context network with the same efficiency. 00:50:01.606 --> 00:50:05.186 So we're thinking that, or other thing that I showed yesterday, 00:50:05.346 --> 00:50:12.626 that treatment of depression affects regions that we activate also with the context stimuli. 00:50:13.406 --> 00:50:21.226 So there's all this evidence that kind of converge, and we would like to bring 00:50:21.226 --> 00:50:24.146 it to fruition one day and actually help people. 00:50:24.146 --> 00:50:29.186 But for the time being it's more an experimental and it's something that's more in the lab and, 00:50:30.306 --> 00:50:37.846 showing that I mean the crux of this hypothesis is that broad association improves 00:50:37.846 --> 00:50:42.566 mood and narrow associations do not and it also relates to inhibition because 00:50:42.566 --> 00:50:45.766 as I showed yesterday it's hard to show it without a slide but. 00:50:47.066 --> 00:50:52.286 That part of the reason we suspect it's a hypothesis it's not proven yet But 00:50:52.286 --> 00:50:56.866 part of the reason why their brain is so ruminative and so focused and not associative 00:50:56.866 --> 00:50:59.426 is because of hyperinhibition that, 00:50:59.566 --> 00:51:02.406 according to this hypothesis, comes from prefrontal cortex structures. 00:51:02.686 --> 00:51:06.246 So this hyperinhibition makes them less associative. 00:51:06.306 --> 00:51:10.706 It doesn't let their thinking process broaden up like the healthy brain. 00:51:10.906 --> 00:51:14.766 So if we could affect levels of inhibition or if we can train people to think 00:51:14.766 --> 00:51:19.006 more broadly, maybe we can bring about some alleviation of their symptoms in 00:51:19.006 --> 00:51:23.446 depression. But as I said before, it's very far from this and we're still on the basic level. 00:51:23.746 --> 00:51:26.586 What's the causality there exactly? Because in some you're saying like, 00:51:26.686 --> 00:51:31.606 okay, we have, let's say, control from frontal areas on the parahippocampal 00:51:31.606 --> 00:51:34.146 area, which is, let's say, more an associative kind of network. 00:51:34.806 --> 00:51:40.586 And this, let's say, allows more or less, let's say, lateral thinking, 00:51:40.726 --> 00:51:45.286 if you want, or let's say more or less flexibility in an associative process. 00:51:45.886 --> 00:51:50.186 And if you restrict this too much, you get rumination. And then this leads to 00:51:50.186 --> 00:51:51.366 mood disorders like depression. 00:51:51.646 --> 00:51:54.206 This is the causality that you have in mind, right? Yes. 00:51:55.255 --> 00:51:58.075 I could also argue that it might be the other way around because you've got 00:51:58.075 --> 00:52:01.095 to look, mood is related to effective processing, to neuromodulation. 00:52:02.035 --> 00:52:06.635 Neuromodulation has a quite direct impact on, let's say, lateral interactions in cortical area. 00:52:06.855 --> 00:52:12.395 So a mood disorder leads to a deregulation of the neuromodulation that allows 00:52:12.395 --> 00:52:17.235 lateral interactions among, let's say, associative states in the parahippocampal area. 00:52:17.815 --> 00:52:19.615 Could you exclude that interpretation? 00:52:20.375 --> 00:52:24.795 No, actually, I embrace it. I think that we talked yesterday about this chicken 00:52:24.795 --> 00:52:30.335 and egg, as he called it, the issue of it could start from the molecules and 00:52:30.335 --> 00:52:32.375 come all the way up to pattern of thinking. 00:52:32.775 --> 00:52:37.695 But what we do here and bring to the table is the optimism, so to speak, 00:52:37.895 --> 00:52:41.255 of the thought that you can actually start the other way around. 00:52:41.335 --> 00:52:44.015 You can start from the highest possible level of pattern and thinking. 00:52:44.015 --> 00:52:49.135 And maybe by training and changing the pattern of thinking to be more broadly 00:52:49.135 --> 00:52:54.115 associative, you can affect the depressed brain all the way down to the same 00:52:54.115 --> 00:52:56.195 molecules, the same dopamine and serotonin. 00:52:56.375 --> 00:52:58.995 So it might be naive, but it's worth trying. 00:52:59.195 --> 00:53:03.215 And that's what we do. So just imagine that this hierarchy goes both ways. 00:53:03.475 --> 00:53:08.795 But then this also, I guess, takes inspiration from the fact that we know that 00:53:08.795 --> 00:53:10.435 cognitive therapy works pretty well. 00:53:11.027 --> 00:53:14.427 On a large group of depression patients is that 00:53:14.427 --> 00:53:17.427 indeed the case the only reservation i have is with the word large 00:53:17.427 --> 00:53:21.287 so the problem with cognitive behavioral therapy is that it's effective but 00:53:21.287 --> 00:53:25.087 only for a very small percentage of people so you have to understand i think 00:53:25.087 --> 00:53:29.027 i told somebody like yesterday that after i wrote my first paper about depression 00:53:29.027 --> 00:53:33.207 a colleague good friend of mine said i could tell you never had depression because 00:53:33.207 --> 00:53:37.087 uh you have to to know this population better in order to. 00:53:37.647 --> 00:53:40.087 Be able to explain this behavior. 00:53:40.167 --> 00:53:43.987 And you realize that people in severe depression are not really motivated to 00:53:43.987 --> 00:53:48.547 even go to a psychiatrist, let alone sit on a couch and introspect and think 00:53:48.547 --> 00:53:51.227 about their pattern of thinking and now fighting this pattern of thinking. 00:53:51.367 --> 00:53:56.187 So anything that addresses this population or attempts to improve their situation 00:53:56.187 --> 00:54:01.487 has to be minimally demanding because it's a population that the nature of their 00:54:01.487 --> 00:54:04.447 symptoms is not to be involved and not to be active. 00:54:04.447 --> 00:54:08.947 So, a method such as CBT, I think, is on the right track, but what it misses 00:54:08.947 --> 00:54:15.067 is the larger part of the population of depressed that just cannot be engaged. 00:54:15.287 --> 00:54:19.567 So, the question is, can we take elements of CBT and break them down in a way 00:54:19.567 --> 00:54:23.867 that's less demanding? And I think in a way, there are some overlaps between 00:54:23.867 --> 00:54:25.287 what we're saying and what's CBT. 00:54:25.447 --> 00:54:32.167 And I think just based on what we know about the brain, maybe we can just broaden 00:54:32.167 --> 00:54:35.047 their thinking pattern without, you know, no questions asked. 00:54:35.587 --> 00:54:40.427 Don't introspect and don't fight your thoughts and don't change the topics or 00:54:40.427 --> 00:54:42.007 the enemy in your thinking pattern. 00:54:42.367 --> 00:54:44.707 Just be associative. Just, you know, 00:54:44.707 --> 00:54:47.447 play with these games that we'll be developing or something like this. 00:54:47.447 --> 00:54:50.547 And this will be more automatic and less demanding and 00:54:50.547 --> 00:54:53.307 less asking less of the patients and maybe 00:54:53.307 --> 00:54:56.287 it will be efficient for a broader part of the population 00:54:56.287 --> 00:55:04.007 but do you have any evidence today that would suggest that that will work well 00:55:04.007 --> 00:55:08.947 other than showing that the press don't recruit a contextual network well of 00:55:08.947 --> 00:55:13.687 course you can ask who knows if it's a trainable maybe this this is a one -way ticket. 00:55:14.567 --> 00:55:18.507 And the other thing that we've shown is that with healthy individuals, 00:55:18.587 --> 00:55:22.007 we can improve mood by associative thinking. 00:55:22.247 --> 00:55:24.247 What is missing is to show that 00:55:24.247 --> 00:55:27.547 depressed patients are less depressed with broader associative thinking. 00:55:27.747 --> 00:55:30.767 But that's something we're working on, but it's not something that happens fast. 00:55:31.387 --> 00:55:34.687 But now there are also correlations between, let's say, just physical exercise 00:55:34.687 --> 00:55:39.207 and the alleviation of depression, which would argue against your proposal. No, it's not. 00:55:39.627 --> 00:55:45.267 Maybe that's why you and I are so happy because of running, even though you run much more. 00:55:45.447 --> 00:55:51.427 But there's actually a big wave of publications, including in popular media, 00:55:51.587 --> 00:55:57.687 the New York Times likes to write about it, about the effects of running on mood. 00:55:59.658 --> 00:56:03.758 But independently, there's this issue of neurogenesis, the growth of new neurons 00:56:03.758 --> 00:56:06.658 in dentate gyrus within the hippocampus. 00:56:06.718 --> 00:56:11.558 And of course, it's somewhat controversial, or at least still not yet to be 00:56:11.558 --> 00:56:14.438 proven completely, that these new neurons actually assume a function. 00:56:14.578 --> 00:56:20.198 But assuming that neurogenesis does something, there is clear evidence that 00:56:20.198 --> 00:56:25.618 running improves or facilitates the growth of new neurons in the dentate gyrus. 00:56:27.818 --> 00:56:32.738 Interestingly, fluoxetine and SSRIs like Prozac also increase the growth of 00:56:32.738 --> 00:56:34.098 neurons in the dented gyro. 00:56:34.158 --> 00:56:39.978 So we have two things that are parallel and affecting this critical structure 00:56:39.978 --> 00:56:42.118 in the hippocampus in a similar manner. 00:56:42.298 --> 00:56:47.078 So take SSRIs or take running, it almost exerts the same... 00:56:48.018 --> 00:56:50.938 Well, I can't of course commit to this, I didn't do the research and I'm not 00:56:50.938 --> 00:56:56.898 sure they They are comparable in terms of magnitude, but it's tempting to think of how exercise, 00:56:57.158 --> 00:57:04.438 especially aerobic exercise, weights don't exert the same effect on neurogenesis. 00:57:04.938 --> 00:57:09.278 But what's interesting about it, it would be, let's say, the argument earlier 00:57:09.278 --> 00:57:11.838 was to say, look, we can take a cognitive route. 00:57:11.898 --> 00:57:16.698 We can try to change thinking patterns and from there affect mood. 00:57:16.918 --> 00:57:20.458 Right. Well, the exercise example would say it is more, let's say, 00:57:20.478 --> 00:57:26.138 from the bottom up, from the body itself and action itself, so completely non-cognitive, 00:57:26.178 --> 00:57:27.518 that you can also affect mood. 00:57:27.758 --> 00:57:33.478 Right. So we're getting into an interesting topic that I'd like to talk about. 00:57:33.538 --> 00:57:36.398 If we have time, sure, I can expand on this. 00:57:36.458 --> 00:57:42.158 So I think that, see that the SSRIs, like Prozac, affect, if I remember the 00:57:42.158 --> 00:57:45.558 numbers correctly, about 30 to 50% of the patients. 00:57:46.138 --> 00:57:48.598 So you ask yourself, well, these are brains. Brains are brains. 00:57:48.718 --> 00:57:51.978 Why would some respond to this SRS and some are not? 00:57:52.158 --> 00:57:55.918 And I think it goes back to the lifestyle of the press, as I said before. 00:57:56.218 --> 00:58:01.518 And I think this is completely speculative. So nobody writes down anything I say now. 00:58:01.678 --> 00:58:08.498 So the idea there is that all this criticism about neurogenesis, 00:58:08.658 --> 00:58:10.818 how do we know that these neurons assume a function? 00:58:11.038 --> 00:58:14.678 Well, I think that if they grow in the morning because you took drugs or you 00:58:14.678 --> 00:58:20.338 ran, I mean, SSRIs, or you ran, and you have new neurons in your dentate gyrus 00:58:20.338 --> 00:58:24.238 now, they won't connect to any network unless you engage them. 00:58:24.318 --> 00:58:27.918 Otherwise, you go to sleep, they die, and then in the morning you start this process again. 00:58:28.098 --> 00:58:32.598 So you need to engage them. So in order for improvement from SSRIs or from running, 00:58:32.678 --> 00:58:33.978 I think two things have to happen. 00:58:34.178 --> 00:58:37.338 You grow new neurons and you connect them to existing networks. 00:58:37.438 --> 00:58:40.478 And I think they are connected to existing networks by. 00:58:42.158 --> 00:58:45.998 Excuse me, doing something else in addition to just having them grow. 00:58:46.118 --> 00:58:51.178 So learn something, play something, et cetera, do social interactions. 00:58:51.438 --> 00:58:55.158 So I think that only running won't be enough. 00:58:55.278 --> 00:58:58.458 If you just run and then you go back home and you just lie on the couch until 00:58:58.458 --> 00:59:01.738 tomorrow morning when you run again, I don't think you'll have the mood benefit. 00:59:02.018 --> 00:59:05.938 If you just take SSRIs and you sit on your couch and not do anything, 00:59:06.098 --> 00:59:09.178 maybe that's a population that is not benefiting from SSRIs. 00:59:09.178 --> 00:59:12.938 So I think you have to be both socially or intellectually engaged at the same 00:59:12.938 --> 00:59:19.518 time to give these new neurons a better chance of being engaged and being recruited to a network. 00:59:19.738 --> 00:59:25.598 And just one slightly unrelated but still important to emphasize is that I didn't 00:59:25.598 --> 00:59:30.738 talk about the fact that this standard JAR is, I mean, it's part of the default network. 00:59:30.858 --> 00:59:34.738 The hippocampus is part of the same network that's being activated by predictions, 00:59:34.898 --> 00:59:36.918 by the default network, by contextual associations. 00:59:36.918 --> 00:59:42.438 So that's pretty intriguing that this growth of neurons and effect on mood come 00:59:42.438 --> 00:59:49.658 from a structure that both loses its volume and mass with depression. 00:59:49.998 --> 00:59:52.778 And also aging, by the way. 00:59:52.838 --> 00:59:56.498 So who knows, it might affect, might help Alzheimer's in the future, 00:59:56.578 --> 00:59:59.018 or cognitive decline with aging. 00:59:59.018 --> 01:00:05.178 But in any event, with the depression, we hope that this associative thinking, 01:00:05.458 --> 01:00:10.598 combined with maybe running or combined with what we just said, 01:00:10.678 --> 01:00:15.738 that depressed people don't do much, so make them go out and run. Well, it depends. 01:00:15.858 --> 01:00:19.998 It's a spectrum, and some people will do more than others, and I think many 01:00:19.998 --> 01:00:22.678 are interested in alleviating their symptoms. 01:00:22.958 --> 01:00:26.278 But there would be an interesting, let's say, difference within the SSRI case 01:00:26.278 --> 01:00:30.578 and And the running case, because in the running case, you would also be driving 01:00:30.578 --> 01:00:35.958 your grid cells because you're moving in space, providing an input into the dentate gyrus. 01:00:36.078 --> 01:00:38.678 So while with the SSRIs, that would not be the case. 01:00:38.878 --> 01:00:43.498 So that means in the running case, you might actually also facilitate the embedding 01:00:43.498 --> 01:00:44.898 of these neurons in active networks. 01:00:45.198 --> 01:00:48.818 Well, for purely SSRIs, that would not be the case. Yeah, that's an interesting 01:00:48.818 --> 01:00:53.518 line of thought. and I think it has to be tested more rigorously. 01:00:53.858 --> 01:00:57.378 I've been thinking myself about the difference of the benefits for running if 01:00:57.378 --> 01:01:00.818 it's running outside versus running on treadmill. I hate running on treadmill. 01:01:01.078 --> 01:01:06.218 I wasn't thinking like you about the entorhinal and the grid cells, 01:01:06.438 --> 01:01:09.738 which is an interesting line of thought and I have to consider it too. 01:01:10.158 --> 01:01:14.338 I was thinking more about the change of scenery and how it, in a way, 01:01:14.398 --> 01:01:17.578 activates more and more concepts or maybe also associations. 01:01:17.578 --> 01:01:22.358 So it's kind of, it's also intellectually pleasing in a sense that it engages 01:01:22.358 --> 01:01:26.298 more and more items in your cortex than just running in front of CNN, 01:01:26.538 --> 01:01:28.018 you know, on the treadmill. 01:01:28.318 --> 01:01:32.138 So if they run in front of flow fields, basically, and green fields and the 01:01:32.138 --> 01:01:35.058 television. So you can try this in your virtual reality lab, right? Yeah, exactly. 01:01:35.178 --> 01:01:38.198 Just to benefit from treadmill, treadmill with... 01:01:39.205 --> 01:01:42.445 Vr or just running outside of course outside has 01:01:42.445 --> 01:01:45.045 multiple effects it's also fresh air and all 01:01:45.045 --> 01:01:47.945 this absolutely yeah but okay that's 01:01:47.945 --> 01:01:50.685 in the future but now do you believe that how do you 01:01:50.685 --> 01:01:54.125 see the basic science you're doing on this um depression or 01:01:54.125 --> 01:01:57.045 let's say the networks that are correlated with 01:01:57.045 --> 01:02:00.725 with depression like the default network and the context network how 01:02:00.725 --> 01:02:03.485 is that basic science translating in an impact in 01:02:03.485 --> 01:02:06.445 the clinic right now how do you see that you mean beyond 01:02:06.445 --> 01:02:09.705 what we're you're trying to do yeah no well the work 01:02:09.705 --> 01:02:12.965 you're currently doing when do you see that really impacting the 01:02:12.965 --> 01:02:15.925 clinic when it's going to hit the clinic in some form well i 01:02:15.925 --> 01:02:19.105 can't predict you know as much as i would like predictions i 01:02:19.105 --> 01:02:22.385 can't predict this but we're definitely uh 01:02:22.385 --> 01:02:27.905 engaged in collaborations with clinicians and with psychiatric uh um like the 01:02:27.905 --> 01:02:33.225 dcrp at mgh with maurizio fava and other big and important and smart psychiatrists 01:02:33.225 --> 01:02:39.005 that are interested which which also shows you the state of the interaction 01:02:39.005 --> 01:02:41.645 between basic level neuroscience and psychiatry. 01:02:41.665 --> 01:02:46.325 I think that psychiatry could benefit from knowledge that we are acquiring in neuroscience. 01:02:46.465 --> 01:02:50.285 And the other way around, that neuroscientists can make their work more relevant 01:02:50.285 --> 01:02:53.745 and more applicable if they understand more of what's going on in the clinic. 01:02:54.345 --> 01:02:58.725 But I can't commit to how quickly. First of all, we have to show in the lab 01:02:58.725 --> 01:03:03.225 that we can improve the state of depressed individuals with our methods and 01:03:03.225 --> 01:03:04.265 maybe improve our methods. 01:03:06.465 --> 01:03:13.405 There have been attempts to make our ideas commercialized, but I wasn't too 01:03:13.405 --> 01:03:17.225 ecstatic about it just because I want to see more proofs. 01:03:17.305 --> 01:03:22.745 And once we have this in the lab, I think that we would like to make it widely 01:03:22.745 --> 01:03:27.185 available for people who can benefit from this. Okay. So Moshe, to finish up. 01:03:28.754 --> 01:03:33.394 So you made quite a tour through the brain in some sense. Brought the associative. 01:03:33.854 --> 01:03:39.194 Exactly. And also revealing some really core properties of the brain. 01:03:40.314 --> 01:03:44.054 So given your experience in brain research and understanding of the mind, 01:03:44.174 --> 01:03:45.774 what would be Moshe's law? 01:03:46.914 --> 01:03:52.694 Well, it would be make no laws ahead of time. 01:03:52.814 --> 01:03:56.234 I don't know. I mean, if I had a law in mind now, 01:03:56.294 --> 01:04:01.014 it probably would have been um yeah no 01:04:01.014 --> 01:04:03.774 it's this is 01:04:03.774 --> 01:04:06.534 your chance uh well it will be when it's proven 01:04:06.534 --> 01:04:09.374 it will be my chance not when i'm predicting it but 01:04:09.374 --> 01:04:17.454 yeah no love for you really okay so there's no mushes law not now okay and then 01:04:17.454 --> 01:04:21.294 what's what's the key prediction that you feel most strongly about today that 01:04:21.294 --> 01:04:26.034 so if i come to visit you there in the outskirts of tel aviv no that's yeah 01:04:26.034 --> 01:04:27.974 we are or five years from now, 01:04:28.054 --> 01:04:31.934 and say, okay, Moshe, this was your prediction you made in 2012, September. 01:04:33.514 --> 01:04:37.234 What's this one prediction that you feel most passionate about today? 01:04:37.474 --> 01:04:42.674 That the state of psychiatric patients will be much better. Okay, very good. 01:04:43.714 --> 01:04:47.854 Moshe Baer, thank you very much for this conversation. Thank you. It was a pleasure. 01:04:50.034 --> 01:04:55.554 The CSN Podcast was produced by the Convergent Science Network network of biometrics 01:04:55.554 --> 01:05:01.714 and bio-hybrid systems, a project funded by the European 7th Research Framework Programme. 01:05:02.000 --> 01:05:30.253 Music.