WEBVTT 00:00:03.497 --> 00:00:10.497 This is the Convergent Science Network podcast. Leading researchers in the domain 00:00:10.497 --> 00:00:16.777 of neuroscience, brain theory and technology are interviewed by Paul Verschoor and Tony Prescott. 00:00:26.177 --> 00:00:31.217 This is Paul for sure with the Convergent Science Network podcast and I'm speaking 00:00:31.217 --> 00:00:37.977 with Aldo Genovesio who is here as a speaker of our summer school and Aldo you 00:00:37.977 --> 00:00:43.757 you study the prefrontal cortex in the monkey so why why what's so interesting 00:00:43.757 --> 00:00:45.857 about prefrontal cortex why are looking at this structure. 00:00:46.457 --> 00:00:52.477 Yeah, the prefrontal cortex is very interesting for many reasons and my interest 00:00:52.477 --> 00:00:57.737 in the prefrontal cortex is now related to all the integrative capacity that 00:00:57.737 --> 00:00:59.837 the prefrontal cortex has and, 00:01:01.537 --> 00:01:07.517 in the ability of the prefrontal cortex to generate goals based on different 00:01:07.517 --> 00:01:11.877 type of computations and I was always interested in the computations that an 00:01:11.877 --> 00:01:14.257 area like the prefrontal cortex and the form. 00:01:14.337 --> 00:01:18.997 They are very interesting and our information can be combined in the prefrontal 00:01:18.997 --> 00:01:20.577 cortex in order to reach a goal. 00:01:22.037 --> 00:01:26.537 So the prefrontal cortex in primates, 00:01:27.126 --> 00:01:29.906 is in some sense related to complex 00:01:29.906 --> 00:01:33.306 behavior plans executive control working memory 00:01:33.306 --> 00:01:37.086 attentional selection integration associative learning 00:01:37.086 --> 00:01:40.306 rule learning that's quite a list of functions 00:01:40.306 --> 00:01:47.166 so are there some underlying underlying dimensions to all these different functions 00:01:47.166 --> 00:01:54.846 i think that what we we do in neurophysiology we try to understand first the 00:01:54.846 --> 00:01:59.146 role of the upper frontal cortex for each of these functions independently. 00:01:59.806 --> 00:02:05.486 And after we tried to understand if we can explain the same data, 00:02:05.606 --> 00:02:07.326 for example, this is the case of a working memory. 00:02:07.766 --> 00:02:14.046 For a long time, we thought that many tasks, many task activity was related 00:02:14.046 --> 00:02:15.946 just to a working memory of a stimulus. 00:02:16.166 --> 00:02:23.606 And after we discovered that a lot of this activity could be explained by just an attentional effect. 00:02:23.606 --> 00:02:31.766 So we try to understand each of these functions separately, 00:02:31.906 --> 00:02:39.086 but also together to understand which functions come to a play in a specific task. 00:02:39.986 --> 00:02:47.446 But one common ground could be found in the role of the prefrontal cortex in 00:02:47.446 --> 00:02:48.406 the generation of goals. 00:02:48.406 --> 00:02:53.686 And this is a difference that the prefrontal cortex can have also compared to 00:02:53.686 --> 00:02:58.926 other areas like the parietal cortex that represents pace and time, for example, numbers, 00:02:59.266 --> 00:03:04.526 different magnitudes, but doesn't have this capacity that the prefrontal cortex 00:03:04.526 --> 00:03:08.326 has to be connected to the premotor cortex in a special way to generate goals. 00:03:09.006 --> 00:03:17.546 Okay, so how do you study this ability to generate or maintain goals and strategies in prefrontal cortex? 00:03:18.349 --> 00:03:27.029 We designed a task that we called a strategy task in order to try to understand 00:03:27.029 --> 00:03:32.309 how the representation of future and previous goals are represented in the brain. 00:03:33.249 --> 00:03:42.429 And this task was requiring the monkey to make a choice based on the repetition 00:03:42.429 --> 00:03:44.809 or not of the cue from the previous trial. 00:03:44.809 --> 00:03:50.309 And the task was organized in a way in which the monkey could maintain in memory 00:03:50.309 --> 00:03:55.889 the information about the future goal, and we could study, because there was 00:03:55.889 --> 00:03:58.709 a delay before a monkey could plan a movement, 00:03:58.889 --> 00:04:01.409 and this way we could study, for example, in that period, 00:04:01.629 --> 00:04:05.009 the neural correlate of maintaining in memory a future goal. 00:04:05.149 --> 00:04:09.929 So this is an approach to correlate the neural activity with the behavior. 00:04:09.949 --> 00:04:17.429 You have a monkey that cannot make a choice. You study the neural activity in 00:04:17.429 --> 00:04:20.909 that moment, and you look at how the information is maintained. 00:04:21.089 --> 00:04:23.229 We can study the representation of future goals. 00:04:23.529 --> 00:04:27.869 If we ask the man in the strategy task to remember also what the monkey did 00:04:27.869 --> 00:04:31.469 before in order to perform the current trial, 00:04:31.689 --> 00:04:40.049 we can also study what is the representation for a monkey of what the monkey did in the past. 00:04:40.609 --> 00:04:45.669 But now, so what you also emphasize there that monkeys already come to the task 00:04:45.669 --> 00:04:50.909 pre-equipped, if you want, with some standard strategies that they use in these tasks. 00:04:50.909 --> 00:04:56.109 Yes, it was discovered by Betsy Burra in the National Institute of Health, 00:04:56.169 --> 00:04:57.609 working with Steve Wise, that 00:04:57.609 --> 00:05:03.829 when monkeys learn to associate a set of stimuli to a set of responses. 00:05:04.429 --> 00:05:10.009 They do something more than just learning the association. 00:05:10.009 --> 00:05:16.029 But because they understand the underlying structure of the task, 00:05:16.189 --> 00:05:18.689 they are able to apply two strategies. 00:05:19.029 --> 00:05:23.249 One we call the repeat-stay strategy. Every time that the stimulus repeats, 00:05:23.569 --> 00:05:26.509 the monkey needs to stay with the same response. 00:05:26.829 --> 00:05:31.569 And the other strategy is a change-shift strategy in which the monkey shifts 00:05:31.569 --> 00:05:34.789 goal every time that the stimulus is different. 00:05:35.089 --> 00:05:38.769 This is because each stimulus is associated only with one response. 00:05:38.769 --> 00:05:40.529 So the monkey can apply these two strategies. 00:05:40.769 --> 00:05:46.109 They found the lesions of specific areas of the prefrontal cortex is able to 00:05:46.109 --> 00:05:50.249 damage the use of this strategy. 00:05:50.389 --> 00:05:57.049 It compromises the use of the strategies while the learning is still going on. 00:05:57.229 --> 00:06:00.029 The monkeys are able to learn the association, but they cannot use the strategy. 00:06:00.309 --> 00:06:05.589 It's something different just learning the association, but it's something of a higher order level. 00:06:06.049 --> 00:06:12.369 Right. So, but so now what you found in a sub-region of prefrontal cortex. 00:06:13.156 --> 00:06:17.016 In some of the found cells that are responding to these goals, 00:06:17.176 --> 00:06:18.916 but also that are responding to these strategies. 00:06:19.096 --> 00:06:23.476 So what's exactly the kind of physiology that you encountered there? 00:06:23.656 --> 00:06:29.116 Yeah, we found a combination of signal at the level of a single cell. 00:06:29.236 --> 00:06:33.556 This is the power of our methodology of behavioral neurophysiology because we 00:06:33.556 --> 00:06:40.256 can investigate what a neuron does, if the same neuron combines different information or not. 00:06:40.256 --> 00:06:44.836 And we find neurons that are able to combine a very abstract representation 00:06:44.836 --> 00:06:51.796 such as a strategy with information about the goal, so about what will be the 00:06:51.796 --> 00:06:54.296 goal that the monkey is going to choose or with the previous goal. 00:06:54.416 --> 00:07:00.736 So they encode the conjunction of goal and strategy. 00:07:01.056 --> 00:07:05.676 And this is also another difference between prefrontal and parietal cortex, 00:07:05.776 --> 00:07:13.316 the ability of encoding conjunctions of factors, or in this case a strategy or a goal. 00:07:13.556 --> 00:07:18.216 Because in this test, the monkey is trained to, let's say, reach for certain 00:07:18.216 --> 00:07:21.616 locations or look at certain locations in order to get a reward, 00:07:21.836 --> 00:07:23.256 and that's also the association it learns. 00:07:23.356 --> 00:07:27.836 There's a cue, the cue is placed in space somewhere, and then dependent on the 00:07:27.836 --> 00:07:32.576 correct response, it will get a juice reward or something of this kind. 00:07:32.816 --> 00:07:36.156 So the goal is now a location in this task space. 00:07:36.536 --> 00:07:41.096 And what I found interesting is indeed he has these conjunctive cells now. 00:07:41.136 --> 00:07:43.476 So these are cells that would respond to, let's say, the cue, 00:07:43.696 --> 00:07:46.136 the goal, or maybe the strategy. 00:07:47.096 --> 00:07:50.916 But you found different combinations here, right? Because some of the cells 00:07:50.916 --> 00:07:55.596 you measured from in this part of prefrontal cortex were specific to location, 00:07:55.856 --> 00:07:58.236 to goal. Exactly. While others... 00:07:59.067 --> 00:08:04.047 Were more mixed in their response. What's the kind of regularity that you extract from that? 00:08:04.207 --> 00:08:08.547 Yes, it's difficult to know how the information combines, not to understand 00:08:08.547 --> 00:08:12.987 the way, why the information combines in a way rather than another. 00:08:13.987 --> 00:08:21.067 But we found that the higher order information, the strategy was not only encoded 00:08:21.067 --> 00:08:24.007 together with the goal, but also with the stimulus feature. 00:08:24.167 --> 00:08:29.547 For example, you could have a neuron that was encoding the stimulus A that was 00:08:29.547 --> 00:08:37.007 leading to a strategy to a repeat-stay strategy but not to the change-shift strategy. 00:08:37.327 --> 00:08:42.687 So both stimuli and goals could be associated to strategy in a very particular way. 00:08:42.847 --> 00:08:47.607 But what I was saying today is that I thought this is a common rule. 00:08:47.727 --> 00:08:51.927 We have exception and when we have an exception we try to understand why we have an exception. 00:08:51.927 --> 00:08:59.147 We found that looking at the data, we found that neurons that were encoding 00:08:59.147 --> 00:09:01.407 the future goal, what we are going to do next, 00:09:01.667 --> 00:09:09.227 when we looked at the same neuron, we didn't see a signal related to the previous goal. 00:09:09.467 --> 00:09:12.727 So these neurons are encoding what you want to do in the future, 00:09:12.827 --> 00:09:15.667 but they don't know what you did in the past. 00:09:15.847 --> 00:09:20.107 Why did you expect that it might reflect your past goal? 00:09:21.567 --> 00:09:28.087 This task is a task that require to the monkey to remember the goal. 00:09:28.247 --> 00:09:34.947 So we are in an experimental condition where we require the monkey to know the 00:09:34.947 --> 00:09:37.407 goal, to remember the goal. 00:09:37.547 --> 00:09:41.127 Because based on the previous goal, the monkey will decide the next goal. 00:09:41.627 --> 00:09:44.707 But one thing that I don't understand, because on the one hand it's an association 00:09:44.707 --> 00:09:48.427 task, right? You get the cue and now you have to go to certain location. Okay. Yes. 00:09:48.607 --> 00:09:53.067 You learn these pairs. You're trained, and as a monkey you're trained for many months on these pairs. 00:09:53.347 --> 00:09:57.227 Okay? Yes. So then how does your previous goal figure into this? 00:09:59.927 --> 00:10:04.587 In our experiment we had two tasks. We had one learning task where the monkey 00:10:04.587 --> 00:10:08.187 were applying also the strategy, and one only strategy task. 00:10:08.587 --> 00:10:13.947 The strategy task actually didn't ask the monkey to associate stimuli to responses, 00:10:13.947 --> 00:10:17.727 It was just asking to implement a strategy. 00:10:18.527 --> 00:10:25.507 So in this case, the monkey just needed to remember the previous goal in order 00:10:25.507 --> 00:10:28.267 to choose the next. Okay. Because today…, 00:10:28.730 --> 00:10:33.970 I simplified a little bit the task. So there's a dependency then between these goals. Exactly. 00:10:34.170 --> 00:10:38.050 If the first goal is left lower corner, then the next one might be right upper 00:10:38.050 --> 00:10:40.070 corner. Some regularity of that kind. Exactly. 00:10:40.270 --> 00:10:45.250 Depending on the next cue. If the cue is the same, the monkey has to use the 00:10:45.250 --> 00:10:47.790 repeat stay, otherwise the change shift. Exactly. 00:10:47.890 --> 00:10:51.170 So the cue is informing the monkey about the rule it should follow. 00:10:51.750 --> 00:10:55.010 The fact that it is the same or different. Right, exactly. But then, 00:10:55.090 --> 00:11:01.030 so now here with this task, We understand the manipulation. We understand the role of memory. 00:11:01.670 --> 00:11:05.150 And now you find two things. And why don't we have sort of a bag, 00:11:05.350 --> 00:11:08.790 if you want, a collection of cells with variable responses, right? 00:11:08.830 --> 00:11:10.170 Some are strategy-specific. 00:11:10.870 --> 00:11:14.310 Others are conjunctive. They mix different aspects of the task. 00:11:14.630 --> 00:11:18.070 But in some sense, they're not reflecting the memory of the task. 00:11:18.370 --> 00:11:22.850 So that would give you a rather incomplete representation. So in this case, 00:11:22.870 --> 00:11:26.950 where do you think then the memory of the previous goal resides? 00:11:27.370 --> 00:11:30.370 Is it still within the system and you were just unlucky? You didn't see it? 00:11:30.690 --> 00:11:35.070 We found it. We found a signal related to the memory of a previous goal, 00:11:35.190 --> 00:11:41.230 but that signal is never in the same neuron that represents the signal that 00:11:41.230 --> 00:11:42.310 represents the future goal. 00:11:42.550 --> 00:11:47.210 So it is there, the representation of a previous goal, or a previous spatial 00:11:47.210 --> 00:11:52.510 goal, but in neurons that are different from the neurons that represent the future goal. 00:11:52.530 --> 00:11:55.530 You see this as a principle? That's really an organizational principle? 00:11:55.870 --> 00:12:00.670 What I think is that, for example, if you go to buy food, for example, 00:12:00.670 --> 00:12:04.890 in a grocery store, you have a list of objects or food that you need to buy. 00:12:05.690 --> 00:12:09.430 For example, you see a milk, you take the milk, this is your future goal, 00:12:09.530 --> 00:12:12.430 and you put your milk with you. You buy it. 00:12:12.670 --> 00:12:17.910 If you see the milk again, you don't buy it again because the milk now, 00:12:18.050 --> 00:12:20.570 it was a future goal, before, but now it's a previous goal. 00:12:20.850 --> 00:12:28.350 So there is probably the need for us to perform what we call output monitoring function. 00:12:28.530 --> 00:12:35.350 That is to, in each instance, to know which goal is still pending and which goal is accomplished. 00:12:35.670 --> 00:12:39.110 Because if a goal is still pending, we need to accomplish it. 00:12:39.350 --> 00:12:45.590 And we can do it even in very simple tasks like buying food in a grocery store. 00:12:46.430 --> 00:12:52.630 But what the results say, showing a separation of representation within the 00:12:52.630 --> 00:12:56.090 prefrontal cortex between future and previous goal, is that, 00:12:58.198 --> 00:13:04.918 Having two separate representations may facilitate an operation of monitoring 00:13:04.918 --> 00:13:06.118 on these two representations. 00:13:06.338 --> 00:13:10.358 If we want to know if something was done or not, we look where the information is. 00:13:10.638 --> 00:13:15.558 I don't know from the computational perspective how this sounds reasonable, 00:13:15.958 --> 00:13:22.158 but it looks like that at least in our data we see that we have this distinction 00:13:22.158 --> 00:13:24.258 between the two, at least in our task. Right. 00:13:24.438 --> 00:13:32.298 But then is this representation of the previous goal only pertaining to the 00:13:32.298 --> 00:13:35.478 previous trial or does it have a variable depth? 00:13:35.658 --> 00:13:39.438 Is the monkey also representing the goal 10 trials back? Okay. 00:13:40.318 --> 00:13:47.518 I actually, I don't think that, I never did a very sophisticated analysis of two trials back. 00:13:48.738 --> 00:13:53.458 Maybe there could be an effect, but it was so small to be identified with normal statistic analysis. 00:13:53.578 --> 00:13:56.638 Maybe at the population level, there could have been a small effect, 00:13:56.798 --> 00:13:59.398 but it was nothing visible, let's say. 00:13:59.758 --> 00:14:06.178 So I think that if there is an effect, it's small. Do you have the sense that 00:14:06.178 --> 00:14:11.398 behaviorally the monkey keeps a memory depth that's larger than one trial? 00:14:12.838 --> 00:14:18.178 In this task, we don't know because of this task. 00:14:18.558 --> 00:14:26.618 There are other studies that show that prefrontal cortex can maintain more than one trial in memory, 00:14:26.778 --> 00:14:33.198 but maybe this trial was not the trial it was asking the prefrontal cortex to 00:14:33.198 --> 00:14:34.858 do. to maintain for two trials. 00:14:35.298 --> 00:14:38.338 But now the other thing is that of the cells that you measured, 00:14:38.378 --> 00:14:42.658 you have about, let's say, you then try to classify your different cells, right? 00:14:42.698 --> 00:14:45.398 So some are conjunctive and some are strategy-related and so on. 00:14:45.438 --> 00:14:48.098 But it's interesting that these classifications, 00:14:49.111 --> 00:14:54.311 stop at around 30 percent right so you always have a large subset of cells that have no. 00:14:55.391 --> 00:14:59.671 Interpretation exactly so in your mind what are those cells doing let's say 00:14:59.671 --> 00:15:03.971 that when i say that 30 percent of cell are representing the future goal we 00:15:03.971 --> 00:15:09.891 had to think that maybe other 10 percent are representing the previous goal 00:15:09.891 --> 00:15:13.291 or another 10 percent can represent the stimulus feature sure. 00:15:13.431 --> 00:15:19.651 So I don't know, I cannot tell you now what is the number of cells that don't 00:15:19.651 --> 00:15:21.451 represent any variable of a task. 00:15:21.571 --> 00:15:26.091 But this is an interesting question, to know of all the variables that we studied, 00:15:26.191 --> 00:15:30.311 how many cells were out of a task. So I don't know. 00:15:30.931 --> 00:15:33.611 Because there's another aspect to this, right? That in some sense, 00:15:33.631 --> 00:15:40.691 we are interpreting the way in which the monkey brain is describing a task, right? 00:15:40.771 --> 00:15:43.451 So we say, okay, there are goals, there their cues, their rules. 00:15:43.631 --> 00:15:45.231 That's it. That's what we're going to look for. 00:15:45.651 --> 00:15:50.871 But it's not impossible that that monkey brain actually is introducing other 00:15:50.871 --> 00:15:54.171 aspects to a task description that we're not looking for. Exactly. 00:15:55.151 --> 00:15:59.551 So do you have any idea what these additional factors could be that could help 00:15:59.551 --> 00:16:04.031 you to interpret these unclassified cells? Okay, yes. 00:16:06.111 --> 00:16:11.551 We, by chance, we looked at the activity because in the exploration phase of 00:16:11.551 --> 00:16:17.051 the data, We looked at the activity after the delay, after the period of presentation of a stimulus. 00:16:17.111 --> 00:16:19.671 It could be one second, one second and a half, and two seconds, for example. 00:16:19.991 --> 00:16:25.651 And we saw that there was, I think, 15% of cells that were modulated by the 00:16:25.651 --> 00:16:27.891 duration of the previous cue. 00:16:28.531 --> 00:16:33.651 So that was an irrelevant information, and we were able to look at, 00:16:34.159 --> 00:16:39.639 this because we by by looking at the raster we noticed that there was an incredible 00:16:39.639 --> 00:16:45.039 effect in some cells so we said oh what is this and we we now we know that this 00:16:45.039 --> 00:16:49.339 is a an encoding of elapsed time that could not be explained by the reaction 00:16:49.339 --> 00:16:52.479 time so we did several analysis to um, 00:16:53.219 --> 00:17:01.799 factor out for example the reaction time effect and so we we by chance we looked 00:17:01.799 --> 00:17:06.399 at this cell so So you never know if a cell doesn't do anything because maybe 00:17:06.399 --> 00:17:08.139 you didn't look at the right variable. Right, exactly. 00:17:08.419 --> 00:17:13.639 So you're right. So you would say mixed in to these features that are encoded 00:17:13.639 --> 00:17:15.459 are also temporal aspects of the task. 00:17:15.619 --> 00:17:21.499 But I cannot tell you how many of these cells are encoding because this analysis may not have an end. 00:17:21.639 --> 00:17:26.059 So you never know when to stop. Of course. To analyze it, to look at combination. 00:17:27.179 --> 00:17:32.319 So at this moment, I don't know if these cells are encoding some other factor. Very nice. Okay. 00:17:33.099 --> 00:17:38.859 So now we have an understanding of what this prefrontal cortex does. 00:17:38.999 --> 00:17:42.979 It represents parts of the task, but also some linking in time, 00:17:43.019 --> 00:17:49.519 like the previous goal, the current goal, or a possible future goal. And now... 00:17:50.561 --> 00:17:54.381 You interpret these representations in terms of output monitoring. 00:17:54.581 --> 00:17:59.281 Yes. So why do you think output monitoring is a good way to describe these properties? 00:17:59.281 --> 00:18:07.941 I think that the separation of goals in the two networks is not output monitoring per se, 00:18:08.081 --> 00:18:11.541 but can facilitate output monitoring if we assume 00:18:11.541 --> 00:18:19.741 that some other cell will look at this activity to decide if an external observer 00:18:19.741 --> 00:18:23.441 that we can think that can be another area or another group of neurons will 00:18:23.441 --> 00:18:29.241 decide if a goal was accomplished or not looking simply at the present. 00:18:29.281 --> 00:18:33.681 Of activity in one of these two networks. This is a possibility, but it may be also wrong. 00:18:35.401 --> 00:18:39.981 But we know that we have failure on monitoring in prefrontal patients, 00:18:40.241 --> 00:18:43.161 so it's one of the problems of people with prefrontal damage. 00:18:43.321 --> 00:18:48.781 For example, they are not able to accomplish a serious goal, 00:18:48.961 --> 00:18:51.901 they get confused, they don't know if they did something or not. 00:18:52.061 --> 00:18:56.941 So we know that also lesions of epiphytonal cortis or patients with dementia 00:18:56.941 --> 00:18:58.641 have problems in this task. 00:18:58.841 --> 00:19:04.741 So it's also related to some disease, psychiatric or neurological disease. 00:19:05.181 --> 00:19:08.801 So, but monitoring in this case means something very specific, 00:19:08.921 --> 00:19:15.081 such as goal achievement, or is goal achievement an operationalization to measure 00:19:15.081 --> 00:19:15.921 something that's broader? 00:19:16.181 --> 00:19:23.461 Could be broader. Could be other, we know that we, but I think that we need 00:19:23.461 --> 00:19:25.701 to monitor many variables in what we do. 00:19:25.781 --> 00:19:31.781 For example, maybe we will talk later about my last studies, 00:19:31.921 --> 00:19:35.761 but I can anticipate that we see that goal is important. 00:19:36.021 --> 00:19:40.781 And it's not only prefrontal cortex important to encode what we are going to 00:19:40.781 --> 00:19:46.621 do, but it's also important to monitor what we are doing, what we are finishing to do. 00:19:46.901 --> 00:19:54.201 So golf may be a special relevance not only when we look at the future but also when we look at. 00:19:54.693 --> 00:19:57.573 What we did and even when this was 00:19:57.573 --> 00:20:00.513 this is not important anymore for what 00:20:00.513 --> 00:20:04.013 we are doing now but would you say this frontal area is 00:20:04.013 --> 00:20:06.893 sort of imposing an intentionality on on the world because 00:20:06.893 --> 00:20:11.173 you could argue look at this frontal area is just filtering everything that 00:20:11.173 --> 00:20:16.013 this animal is engaged with with respect to goals and just saying look okay 00:20:16.013 --> 00:20:19.933 which of my goals did i achieve where am i vile where am i failing to achieve 00:20:19.933 --> 00:20:24.593 a goal so it's really this massive goal filter is that is that a reasonable way to look at it? 00:20:24.793 --> 00:20:28.673 Yeah, I think that there is a, gold can be the main function. 00:20:28.833 --> 00:20:34.473 And I give you another example, and that we have looking at the correlation between neurons. 00:20:34.573 --> 00:20:38.453 So we, okay, we can, we recorded several neurons simultaneously. 00:20:39.193 --> 00:20:43.593 And because we recorded for many, many days, and with several electrodes on 00:20:43.593 --> 00:20:47.533 time, we can have two conditions, one in which we have two future gold cells 00:20:47.533 --> 00:20:52.113 recorded simultaneously, and another in which we have two previous gold cells 00:20:52.113 --> 00:20:53.233 recorded simultaneously. 00:20:53.653 --> 00:20:58.673 We can look at the correlation between future and future cell and past and past cell. 00:20:58.833 --> 00:21:03.533 We see that they're correlated only the neurons that represent the future goal, 00:21:03.673 --> 00:21:06.073 but the neurons that represent the previous goal are not correlated. 00:21:07.733 --> 00:21:12.133 So, it's true that prefrontal cortex encodes goals, but maybe the future goal 00:21:12.133 --> 00:21:16.953 has a special relevance that we can see from the correlated activity that can 00:21:16.953 --> 00:21:21.633 be a way of driving the premotor cortex in a choice. 00:21:21.933 --> 00:21:25.893 Right, exactly. In the selection of one course of action. So the correlation 00:21:25.893 --> 00:21:28.433 you observed is only between these goal cells? 00:21:28.773 --> 00:21:31.333 The future goal cells. Only the future goal cells. Yes. Okay. 00:21:31.373 --> 00:21:35.213 And when we look at the cells that are encoding what you did before, 00:21:35.413 --> 00:21:40.033 and you find by chance the same record in two of these cells, 00:21:40.093 --> 00:21:40.833 they are not correlated. 00:21:41.113 --> 00:21:43.693 So they don't have anything. So this is still mysterious for us. 00:21:43.733 --> 00:21:45.453 We have these results, but we still don't know. 00:21:45.453 --> 00:21:53.513 But are you interpreting this like future goals have to be included in monitoring in the future, 00:21:53.613 --> 00:21:57.873 so I have to sort of maintain these representations and has to make them part 00:21:57.873 --> 00:22:00.473 of a possible task set or something of that kind? 00:22:00.693 --> 00:22:05.913 Well, these old goals in some sense don't matter that much anymore. 00:22:05.993 --> 00:22:08.553 I can slowly forget about them, so I don't have to maintain them. 00:22:08.833 --> 00:22:12.013 Maybe they don't need to activate simultaneously other neurons. 00:22:12.013 --> 00:22:16.573 They don't need mechanisms of temporal summation, for example, 00:22:16.573 --> 00:22:20.133 to activate neurons in a competitive way. 00:22:20.800 --> 00:22:25.180 To generate a behavior in the premotor cortex and the cortex later. 00:22:25.880 --> 00:22:32.220 Okay, so that means you see these goal cells as also organizers of neural activity, 00:22:32.920 --> 00:22:37.720 for instance neural activity pertaining to cues that you might see in actions and so on. 00:22:37.920 --> 00:22:42.800 Yeah, for now we know about the future goal. We don't know if this correlated 00:22:42.800 --> 00:22:46.460 activity can apply to other couples or pairs of neurons. 00:22:46.700 --> 00:22:51.700 We don't know if the dichotomy is between past and future. Now it looks like that. 00:22:52.360 --> 00:22:56.180 Unless we may discover one day there was another reason. 00:22:56.660 --> 00:23:03.200 But now it looks like it's past future. So that means that you have in the response, 00:23:03.380 --> 00:23:05.120 the neural response in prefrontal cortex. 00:23:06.020 --> 00:23:09.560 You find that these future goal cells are tightly coupled in their activity 00:23:09.560 --> 00:23:12.920 while past goal cells are not coupled. 00:23:13.100 --> 00:23:18.760 And also does it imply that these future goal cells are also strongly coupled to future cues? 00:23:19.500 --> 00:23:23.180 Could be, could be, yeah. But there's no data on that. There is no data, 00:23:23.220 --> 00:23:27.220 but you're right. Could be, we don't know. Would you predict that? 00:23:29.960 --> 00:23:34.460 If I, it's difficult to know. It depends on the task. We need to imagine a task to think that. 00:23:34.660 --> 00:23:38.540 No, but look, if it's the same task, wouldn't you suggest, look, I have a future goal. 00:23:38.840 --> 00:23:42.940 Then I use that. I use that state to also already predict, ah, 00:23:43.080 --> 00:23:47.660 if that's my goal, then I would expect this Q and I expect whatever, this kind of response. 00:23:47.660 --> 00:23:55.600 In this task, the goal that the monkey is achieving doesn't predict the future, 00:23:55.760 --> 00:23:57.860 the next goal after the future. 00:23:57.940 --> 00:24:00.860 So it's difficult to know if, so it stops there. 00:24:01.420 --> 00:24:06.140 Right, I understand. So now we have a bit of an idea about the representation 00:24:06.140 --> 00:24:12.260 of a task in prefrontal cortex organized around goals, future and past. 00:24:12.260 --> 00:24:14.660 And then in some sense, you also 00:24:14.660 --> 00:24:18.760 followed this notion of a hierarchical structuring of the frontal lobe. 00:24:20.296 --> 00:24:23.116 You moved to the frontal pole, which is sort of really the structure all the 00:24:23.116 --> 00:24:27.556 way. Yeah, we moved at the end of the brain. Yeah, exactly. There was nothing else. 00:24:28.036 --> 00:24:34.236 Beyond that, that's it, right? There's only skull. So what did you expect to find there? 00:24:34.596 --> 00:24:41.196 I think that we're considering that this area is in a special location that 00:24:41.196 --> 00:24:48.636 is farther compared to the dorsolateral prefrontal cortex from the premotor cortex. 00:24:48.636 --> 00:24:52.736 Context, we were thinking, and also considering that this is a primate innovation, 00:24:53.056 --> 00:24:56.876 is an area that should solve new problems maybe in evolution, 00:24:57.276 --> 00:25:02.236 I don't know, maybe a trivial idea was to find more complexity and that we thought 00:25:02.236 --> 00:25:04.696 to be more integration of information. 00:25:05.116 --> 00:25:07.936 Which also would be consistent, I think, with the literature, 00:25:08.016 --> 00:25:14.696 there's some hierarchy, it gets more and more abstract, but all the features 00:25:14.696 --> 00:25:17.516 are sort of abstracted. Exactly, exactly. 00:25:17.776 --> 00:25:25.076 Yes, yes. And so we found that the frontal pole cortex had a completely different function. 00:25:25.336 --> 00:25:33.556 It is a monitoring function, so it has just at least, we could study the cells in only a few tasks, 00:25:33.776 --> 00:25:40.376 but we found that these cells are active only when the monkey receives the feedback 00:25:40.376 --> 00:25:46.396 about the correctness or not of the behavior and it represents the goal that the monkey achieved. 00:25:47.676 --> 00:25:50.736 We don't find any other signal of 00:25:50.736 --> 00:25:53.356 the signals that I described previously like the representation of a 00:25:53.356 --> 00:25:56.916 stimulus or the representation of the past goal or 00:25:56.916 --> 00:26:00.836 a prospective representation of a future goal nothing but when I say nothing 00:26:00.836 --> 00:26:06.656 really nothing no example so not a few was really impossible to find the example 00:26:06.656 --> 00:26:11.536 and we found this very clear signal in 30% of the cell around the monitor but 00:26:11.536 --> 00:26:15.476 it's very surprising right because is there anything special about let's say say, 00:26:15.476 --> 00:26:20.256 the anatomical organization of this structure compared to other parts of prefrontal cortex? 00:26:20.376 --> 00:26:26.056 Yeah, the spatial position is farther from the premotor cortex, 00:26:26.276 --> 00:26:31.096 and so it speaks more with other prefrontal area than with the premotor cortex 00:26:31.096 --> 00:26:33.576 compared to the dorsolateral prefrontal cortex. Right. 00:26:33.776 --> 00:26:41.696 So this is kind of spatial, and so it's more distant from the behavior. 00:26:41.716 --> 00:26:46.016 Yeah, but this also means this is very odd, right? Because we have this cortical 00:26:46.016 --> 00:26:52.096 circuit, the different parts of the frontal cortex will be tightly interconnected in this. 00:26:53.200 --> 00:26:56.520 Earlier area you measured from closer to the motor cortex itself, 00:26:56.800 --> 00:27:02.380 you find cells related to future goals, present goals, past goals. 00:27:02.640 --> 00:27:04.880 So there's a clear temporal window. 00:27:05.200 --> 00:27:06.940 You would believe, given the anatomy, 00:27:07.040 --> 00:27:09.560 that this should all percolate up towards the pole. Yeah, I agree. 00:27:09.860 --> 00:27:13.540 And it's not there. Yeah. It's interesting because, you know, 00:27:13.640 --> 00:27:16.080 these two areas are connected to each other. 00:27:16.080 --> 00:27:20.340 They have reciprocal pronation, but the information that maybe comes up to the 00:27:20.340 --> 00:27:25.500 frontal pole doesn't activate the frontal pole the way in which we think about. 00:27:25.740 --> 00:27:27.540 Yeah, but how could you get this selectivity? 00:27:27.800 --> 00:27:32.900 Because, I mean, the prefrontal cortex deals with monitoring and goals. 00:27:33.100 --> 00:27:37.160 It operates in a certain temporal window with future and past included. 00:27:37.580 --> 00:27:42.600 Now, you go all the way in the front to the frontal pole, and it lives in the 00:27:42.600 --> 00:27:46.640 now, right? It's completely just now, completely unlocked to the task. 00:27:46.720 --> 00:27:49.960 Yeah, we have this activity even in the dorsolateral prefrontal cortex, 00:27:50.000 --> 00:27:51.720 but it's not the only activity that we have. 00:27:52.060 --> 00:27:58.260 So this is what I was saying, that it looks like a filtering function is applied 00:27:58.260 --> 00:28:01.320 where everything else is gone and it remains. 00:28:03.120 --> 00:28:07.020 But actually, we need really a model to understand the function, 00:28:07.200 --> 00:28:08.900 why we need cells like these. 00:28:08.900 --> 00:28:15.120 But would you say that, does that imply that the brain, all the way in the front, 00:28:15.880 --> 00:28:24.220 the most exclusive spot of the cortex, is actually completely preoccupied with 00:28:24.220 --> 00:28:30.120 just assessing how well we do with our real-time performance now? 00:28:31.120 --> 00:28:35.960 Would you buy that? Maybe this is, as we were discussing today, 00:28:36.380 --> 00:28:41.560 this looks not so important in a simple task like this, in a more complex task, 00:28:41.780 --> 00:28:46.500 we might understand better the function of these cells. 00:28:46.560 --> 00:28:49.180 We might think, okay, without these cells, how could we do? 00:28:49.520 --> 00:28:55.000 But in this task, it is more difficult to understand their role. 00:28:55.120 --> 00:28:59.700 But what I was thinking is that this signal can help. 00:29:00.520 --> 00:29:04.800 We don't know why this signal can do that, but can help the transfer of information. 00:29:04.840 --> 00:29:07.980 But is this observation consistent with the clinical literature, 00:29:08.180 --> 00:29:11.420 like patients with lesions only to the frontal pole? 00:29:11.440 --> 00:29:13.680 The problem is that it's never only the frontal pole. 00:29:13.780 --> 00:29:21.460 The lesions are very broad, especially frontal pole is part of a larger area that is damaged. 00:29:22.780 --> 00:29:27.120 So it's very difficult to know now. This is why we study also monkey because 00:29:27.120 --> 00:29:33.660 neuropsychology in humans is very limitative on that. Right. 00:29:35.740 --> 00:29:40.380 If we look at this hierarchy of processing in the frontal lobe, 00:29:40.440 --> 00:29:42.820 then what you were saying. 00:29:45.566 --> 00:29:50.726 Areas like orbital frontal cortex, so on, might have a good understanding of 00:29:50.726 --> 00:29:54.486 the task and the strategies you use, but the only area that actually really 00:29:54.486 --> 00:29:59.886 know whether you succeeded, really knows success, is the frontal pole. 00:30:00.106 --> 00:30:05.906 In a very unique way, yes. That doesn't do anything else and it has a pure signal 00:30:05.906 --> 00:30:09.446 until we find something else. Of course. 00:30:09.786 --> 00:30:13.686 But it is that. Orbital frontal cortex, we studied also orbital frontal cortex 00:30:13.686 --> 00:30:14.986 with a second strategy task. 00:30:15.566 --> 00:30:20.166 More simple and we found that there is an activity related to the goal at the 00:30:20.166 --> 00:30:24.386 moment of the feedback, but this activity doesn't depend on the success. 00:30:24.786 --> 00:30:30.626 So these neurons are just representing the goal as left or right, 00:30:30.706 --> 00:30:34.346 but they don't consider the success of the. 00:30:37.786 --> 00:30:44.766 Task. And then you see this, so frontal pole monitors now the task, success, 00:30:45.526 --> 00:30:49.846 we did it, and then do you believe it's that information that percolates back 00:30:49.846 --> 00:30:53.926 into other areas of the frontal cortex to represent the previous goal? 00:30:55.606 --> 00:31:00.466 I may, okay, let's assume that there is already the representation about the 00:31:00.466 --> 00:31:03.206 previous and the future goal, about the future goal. 00:31:03.326 --> 00:31:08.386 And this activity about the future goal need to move from a group of neuron 00:31:08.386 --> 00:31:11.746 to another group of neuron. And let's assume that there is already a connection 00:31:11.746 --> 00:31:15.486 between the future goal cells and the previous goal cells. 00:31:15.686 --> 00:31:19.166 But this connection is not enough to activate. This is an hypothesis. 00:31:19.266 --> 00:31:22.946 It's not enough to activate the previous goal activity. 00:31:23.266 --> 00:31:26.866 It may need an additional input. This is a possibility. 00:31:27.086 --> 00:31:32.306 It may give an additional input that together with the input from a future goal 00:31:32.306 --> 00:31:41.346 cell may activate the previous goal cell. But this is just a scheme just to reason about. 00:31:41.666 --> 00:31:44.886 Right, exactly. Because the difficulty of that scheme is that in some cases 00:31:44.886 --> 00:31:48.186 you really start to think about it as different modules performing specific operations. 00:31:48.686 --> 00:31:52.286 I don't think that there could be a dissociation. 00:31:52.366 --> 00:31:57.326 We found future and ghost cells in all the penetrations. So I don't think about 00:31:57.326 --> 00:31:59.046 the spatial segregation of these cells. 00:31:59.566 --> 00:32:06.546 So this is not my idea. Yeah, but there could be, because these neurons belong 00:32:06.546 --> 00:32:12.326 to different population of cells, we know already that there are two different 00:32:12.326 --> 00:32:13.786 networks, partially overlapped. 00:32:13.966 --> 00:32:17.246 The overlapping is very small. We have a few hybrid cells. 00:32:17.646 --> 00:32:24.626 So we have already a segregation in the same area of these neurons. 00:32:24.926 --> 00:32:30.906 Right. But would you say there is any critical anatomical difference between 00:32:30.906 --> 00:32:36.966 the circuits in the frontal lobe and, or pole, sorry, and let's say premotor? 00:32:38.186 --> 00:32:41.986 Yeah, but motor cortex has this connection with the motor cortex, 00:32:42.006 --> 00:32:44.686 so it's very close to the… But in terms of the local circuit, 00:32:44.806 --> 00:32:46.606 really how the cells are wired up together? 00:32:49.366 --> 00:32:52.746 Within each area, it's difficult to say. Right. Yeah. 00:32:53.346 --> 00:32:58.066 That's interesting, right? Because even… so you find these rather variable response 00:32:58.066 --> 00:33:03.106 patterns in a neural substrate that at the local level is relatively uniform. 00:33:03.606 --> 00:33:08.666 Exactly. We don't know. We don't know what is the difference in activity of 00:33:08.666 --> 00:33:10.506 cells in different layers. 00:33:10.626 --> 00:33:14.866 This is often a limitation of a neurophysiology that we don't know where we're recording. 00:33:15.106 --> 00:33:18.966 Often we don't know if it's a pyramidal neuron that's an output outside or an 00:33:18.966 --> 00:33:20.206 interneuron. Right, exactly. 00:33:20.666 --> 00:33:26.466 We will do that. Probably there are now new electrodes with multi-contacts. 00:33:26.666 --> 00:33:32.566 And probably this will be the answer to these questions to understand the microcircuit. 00:33:32.746 --> 00:33:37.226 Of course. Of course. So how many cells in the frontal pole could you actually 00:33:37.226 --> 00:33:38.626 characterize in this way? 00:33:38.686 --> 00:33:41.986 We recorded hundreds of cells. 00:33:42.326 --> 00:33:46.886 And how many could you classify? Let's say it was 30%. So significant. 00:33:47.086 --> 00:33:54.646 So do you believe that this argues against this very hierarchical view on the 00:33:54.646 --> 00:33:58.646 frontal lobes and this hierarchical abstracting? 00:33:59.268 --> 00:34:06.188 Of information towards the frontal pole? This very specific signal can be a 00:34:06.188 --> 00:34:07.288 result of an abstraction. 00:34:07.788 --> 00:34:12.908 It can be something because alpha is a very specific signal that can be the 00:34:12.908 --> 00:34:18.848 result of an abstraction from many other signals in a different way from what we are used to think. 00:34:19.168 --> 00:34:24.328 So the abstraction is not just integration but it's producing something new that could be a monitor. 00:34:24.608 --> 00:34:29.248 Of course, abstraction is also to remove information, right? and to become specific. 00:34:29.568 --> 00:34:31.528 Exactly, yes. Okay. So maybe that. 00:34:33.128 --> 00:34:37.688 But for sure we need to study more of the frontal pole cortex. 00:34:37.788 --> 00:34:40.388 We still have, you know, this is the first study. Sure, of course. 00:34:40.588 --> 00:34:45.708 So our speculation in the case of the frontal pole cortex is very limited to our results. 00:34:46.088 --> 00:34:52.948 And this is a difference with other areas where we know we can think and compare 00:34:52.948 --> 00:34:56.788 the results from different studies and to spend our days comparing. 00:34:57.088 --> 00:35:02.668 Exactly. Here we have only this study and some neuroimaging study where we know 00:35:02.668 --> 00:35:07.928 that also only a part of the frontal pole cortex can be an homologous area with 00:35:07.928 --> 00:35:09.848 the monkey by frontal cortex. Right, exactly. 00:35:10.108 --> 00:35:13.588 So we have this additional difficulty with the frontal pole cortex that we don't 00:35:13.588 --> 00:35:18.028 know exactly if there is an homology. Right, exactly. 00:35:18.508 --> 00:35:21.668 But that's not necessarily bad. That means there's still quite some experiments 00:35:21.668 --> 00:35:24.368 to be performed. Yeah, it's still a frontal pole. I feel that we are still studying 00:35:24.368 --> 00:35:29.828 the frontal pole with a common origin from both human and man. 00:35:29.968 --> 00:35:34.108 The other thing is that in this frontal area, the monitoring area, 00:35:34.288 --> 00:35:39.568 goal-oriented if you want, the question is of course also what's the kind of 00:35:39.568 --> 00:35:43.528 information it receives from other areas, from other modalities? 00:35:43.708 --> 00:35:48.948 If you have a visual cue, what information of the visual cue really enters in 00:35:48.948 --> 00:35:53.928 this area? and you could imagine that this again may be symbolic and fairly abstract and for that. 00:35:54.551 --> 00:35:58.651 You performed a number of experiments where you start to look at how really 00:35:58.651 --> 00:36:05.931 very specific aspects of visual stimuli would engage with these decision-making circuits. 00:36:06.091 --> 00:36:09.611 So what was really the idea there? Here we use different modalities. 00:36:09.891 --> 00:36:14.611 In the frontal pole, we use different cues like orientation, color, and reward. 00:36:14.831 --> 00:36:17.611 Also, we have this connection with the orbital frontal cortex, 00:36:17.651 --> 00:36:23.011 so the number of drops of reward could have been an appropriate cue. 00:36:23.011 --> 00:36:30.251 Queue, but we didn't find any representation of the queue, either as a color, 00:36:30.531 --> 00:36:33.251 an orientation, or a number of reward. 00:36:33.471 --> 00:36:39.351 This is very interesting because we found an increase of activity that was selective 00:36:39.351 --> 00:36:41.711 for the goal in the monitoring phase. 00:36:41.971 --> 00:36:46.091 It was not just selective for the goal, but was just a ramping up of activity. 00:36:46.091 --> 00:36:55.631 So, without using as a cue a reward, we might have thought that this was a reward 00:36:55.631 --> 00:36:58.051 signal with an appetitive meaning. 00:36:58.331 --> 00:37:04.011 But interestingly, when we look at the presentation of a reward as cue, 00:37:04.231 --> 00:37:07.771 we don't see any change of activity. The activity remains flat. 00:37:08.651 --> 00:37:14.511 So this is very interesting that the reward per se is not able to activate this 00:37:14.511 --> 00:37:20.391 neuron. So it is something that has to do with the monitoring of a success independently 00:37:20.391 --> 00:37:23.011 of getting something good. 00:37:23.631 --> 00:37:29.511 So we were lucky to think about having the reward as a cue. 00:37:29.611 --> 00:37:33.671 Of course. Otherwise we would think about doing that experiment now. 00:37:33.851 --> 00:37:36.071 At least one experiment less is done. 00:37:36.751 --> 00:37:41.471 That's very good. So what I was thinking about is in the next set of experiments 00:37:41.471 --> 00:37:44.191 where you sort of backtrack again out of the frontal pole. 00:37:44.811 --> 00:37:49.351 Where you start to look really at, okay, this is a rather puzzling question, right? 00:37:49.491 --> 00:37:53.911 The responses, so here we have the frontal cortex. 00:37:54.071 --> 00:37:55.331 It responds to different aspects 00:37:55.331 --> 00:37:59.891 of a task, but these responses seem normalized in some sense, right? 00:37:59.951 --> 00:38:05.231 So even though you might have the physical salience of the stimulus vary, 00:38:06.071 --> 00:38:11.091 that might be small or big, for instance, or close or far, the neural response 00:38:11.091 --> 00:38:16.791 doesn't seem to be really strongly modulated by these differences in just the 00:38:16.791 --> 00:38:19.851 visual aspects or the properties of the stimulus. 00:38:20.251 --> 00:38:24.331 You're talking about the distance task? Yeah, exactly. 00:38:25.231 --> 00:38:31.211 I showed you SELITAR encoding the decision today, so the decision about which 00:38:31.211 --> 00:38:33.431 stimulus is farther, which stimulus is longer. 00:38:34.331 --> 00:38:37.631 So I focused on this aspect. 00:38:37.871 --> 00:38:46.671 But they still encode the feature of the stimulus and the small details of the task. 00:38:46.931 --> 00:38:49.371 Yeah, but they don't seem to be strongly modulated. 00:38:50.151 --> 00:38:55.471 Look, if a stimulus might be more salient because it's nearby or it's bigger, 00:38:55.571 --> 00:39:00.851 the response you find in frontal cortex is not that different as to another 00:39:00.851 --> 00:39:02.811 stimulus that might be further away and smaller. 00:39:04.191 --> 00:39:08.751 Or do you find modulations there? In this case, in the distance task, 00:39:09.091 --> 00:39:16.471 we have stimulants that have a different distance from the center, not from the animal. 00:39:16.651 --> 00:39:18.551 Maybe from the animal it could have been different. 00:39:19.431 --> 00:39:25.571 And we see actually pretty much the same number of cellular encoding with a 00:39:25.571 --> 00:39:31.511 greater activity that the first stimulus is farther than the second or vice versa. 00:39:31.511 --> 00:39:36.331 So we I don't know if you meant that we don't see an unbalanced representation of, 00:39:37.402 --> 00:39:40.362 Looking at the relative representation of which stimulus is farther. 00:39:42.502 --> 00:39:47.642 But what you're looking at is really the physical organization of the display 00:39:47.642 --> 00:39:52.682 and how this percolates into the neural response in the frontal cortex. 00:39:52.922 --> 00:39:56.742 And it appears that it seems to be normalized. These cells actually don't really 00:39:56.742 --> 00:40:00.602 care about, let's say, saccadic distance or anything like this. 00:40:00.722 --> 00:40:05.642 It's just saying, no, this is a cue and the cue is in the display and has to 00:40:05.642 --> 00:40:11.322 represent it. Yeah, maybe you want to say that this is what I'm presenting is 00:40:11.322 --> 00:40:12.202 an abstract representation. 00:40:12.602 --> 00:40:16.802 It doesn't depend on the actual position. So the position can be very different, 00:40:16.902 --> 00:40:18.262 but the signal is the same. That's right. 00:40:18.462 --> 00:40:21.462 So the important thing is that there is a relationship with one distance. 00:40:21.882 --> 00:40:24.082 Right, well, you could argue it's like a symbolic representation, 00:40:24.302 --> 00:40:28.862 right? Because it's not modulated by the analog properties of the cue. 00:40:29.042 --> 00:40:32.262 Yeah, but we still have, also in this case, cells with mixture properties. 00:40:32.262 --> 00:40:35.922 For example, a very interesting cell is a cell that is encoding, 00:40:36.182 --> 00:40:40.322 for example, that the second stimulus is farther than the first, 00:40:40.442 --> 00:40:42.922 but only when the second stimulus is on the top. 00:40:43.742 --> 00:40:48.622 So we have cells that are an intermediate level of abstraction. 00:40:49.062 --> 00:40:54.622 These are very interesting, these cells, that really it depends not only... 00:40:54.622 --> 00:40:56.542 So this looks like an intermediate representation. 00:40:57.102 --> 00:40:59.322 Why are you saying that? Because you could also argue it's a rule. 00:40:59.962 --> 00:41:01.522 This cell extracted the rule. 00:41:02.262 --> 00:41:09.122 The final one that represents just which stimulus is farther independently of the position. 00:41:11.122 --> 00:41:20.042 Or if you say a certain cue only when another cue is in a certain position. 00:41:20.162 --> 00:41:21.182 So you have a conditional. 00:41:22.622 --> 00:41:26.802 I don't think that this has to represent a rule because the rule is always the 00:41:26.802 --> 00:41:29.822 same. The rule is choose the second stimulus. 00:41:31.242 --> 00:41:35.482 That's your behavioral rule. Yes. But you could also argue that the display 00:41:35.482 --> 00:41:40.582 of the stimuli is following certain rule-like regularities. 00:41:42.750 --> 00:41:49.250 So, Fred, you just described a cell that gave a specific response to a certain 00:41:49.250 --> 00:41:53.710 cue, but conditional on another cue being in a certain position. Yes. 00:41:54.070 --> 00:41:57.810 So, I could say that's a rule that defines that little part of the display. 00:41:58.230 --> 00:42:02.230 Yeah, you can say that. So, these are the rules. You can also think in this term. Exactly. 00:42:02.350 --> 00:42:06.810 So, it's not really feature encoded, but it's rule encoded. 00:42:06.950 --> 00:42:11.210 It could be. You could think, yeah, it's conditioned to the appearance of a 00:42:11.210 --> 00:42:12.930 stimulus in a special location. 00:42:13.230 --> 00:42:18.990 Or you may think it more simply is strictly associated to space and it cannot 00:42:18.990 --> 00:42:20.390 generalize more than that. 00:42:20.650 --> 00:42:23.590 I don't know. It's difficult to know. But it's an important distinction because 00:42:23.590 --> 00:42:29.610 in your interpretation, so you're saying, well, prefrontal is not, 00:42:29.690 --> 00:42:35.530 if you want, really encapsulated and segregated from this dirty, noisy outside world. 00:42:35.530 --> 00:42:38.670 It's not just the abstract and the beauty. Exactly. 00:42:38.990 --> 00:42:44.370 Right? Because it's sort of also the physical properties of the world and of 00:42:44.370 --> 00:42:46.430 the display percolate into its representation. 00:42:46.630 --> 00:42:48.630 You were saying like, that's where you talked about the mixture. 00:42:48.910 --> 00:42:52.310 Yeah. Because we know that neurons in prefrontal cortex care about space, 00:42:52.470 --> 00:42:54.990 they have receptive fields, you know, really associated. Sure. 00:42:55.290 --> 00:42:58.630 But some neurons go beyond that. 00:42:58.850 --> 00:43:03.390 Others, maybe they are representing in a conditional way, as you say, 00:43:03.490 --> 00:43:06.050 more in a rule-like, but For me, 00:43:06.070 --> 00:43:13.670 I see that activity more related to the physical world, but I don't know. 00:43:13.750 --> 00:43:16.410 Yeah, but wouldn't this be… It's difficult to know. As a physiologist, 00:43:16.670 --> 00:43:19.270 wouldn't this be, or from a computational point of view, this would be tremendously 00:43:19.270 --> 00:43:21.630 annoying because here I have this neuron. 00:43:21.690 --> 00:43:25.730 It sits all the way behind a huge hierarchy of processing of, 00:43:25.750 --> 00:43:26.830 in this case, visual stimuli. 00:43:27.230 --> 00:43:30.870 But where we know we go through areas, like if you're a temporal cortex, 00:43:30.970 --> 00:43:33.550 where you have very invariant representations to the world. 00:43:34.351 --> 00:43:37.851 And now I'm receiving, I'm in the frontal cortex, I'm receiving this kind of information. 00:43:38.151 --> 00:43:42.491 And I can just do my job so you would believe that, of course, 00:43:42.511 --> 00:43:46.811 you take the most abstract representation so you reduce noise and so on, right? 00:43:47.371 --> 00:43:50.791 So in that sense, in this perspective, you could argue, well, 00:43:50.851 --> 00:43:56.471 there's a lot, we lose a lot when we drop the very clear segregation between 00:43:56.471 --> 00:43:59.231 more analog representations of the world. 00:43:59.351 --> 00:44:02.451 That means they vary with properties of the world. Yeah, with the distance in 00:44:02.451 --> 00:44:06.551 this case would be the distance, but the first level of analysis would be the 00:44:06.551 --> 00:44:07.471 distance from the center. 00:44:07.611 --> 00:44:10.751 Right. But you seem to be much more permissive than I am. 00:44:10.891 --> 00:44:13.831 You're more like, well, okay, maybe they vary properties of the world, 00:44:13.891 --> 00:44:17.191 but doesn't that mean that your whole concept of how this part of the brain 00:44:17.191 --> 00:44:21.271 is organized and relates to other parts of the brain falls apart in some sense? 00:44:21.591 --> 00:44:26.891 My idea is that we have a basic representation is similar to the parietal cortex of a metric. 00:44:26.971 --> 00:44:30.371 For example, the distance or if the stimulus is on the top or on the bottom. 00:44:30.371 --> 00:44:33.051 And after we have a representation it is 00:44:33.051 --> 00:44:37.711 very abstract we have something in the middle and unfortunately when I looked 00:44:37.711 --> 00:44:42.391 at these cells that we don't know what they are it is very difficult to understand 00:44:42.391 --> 00:44:48.791 based on the temporal profile of their activation where they what is their role 00:44:48.791 --> 00:44:51.891 so it's I think that also in this case. 00:44:52.383 --> 00:44:57.883 We need a model to understand because otherwise very difficult to understand, 00:44:58.003 --> 00:45:02.563 especially why we need certain computation instead of others. This is my point. 00:45:02.783 --> 00:45:08.003 Why we need, for example, I'm describing cell that are encoding which stimulus 00:45:08.003 --> 00:45:10.523 is farther based on the order first or second. 00:45:10.563 --> 00:45:14.003 But this is not a requirement of the task because the monkey has just to say 00:45:14.003 --> 00:45:16.063 the blue or the red is farther. 00:45:16.503 --> 00:45:20.903 So we have also we were talking at the beginning of the interview about this. 00:45:20.903 --> 00:45:26.123 Do we have a neuron that represents something that the monkey doesn't need to know? 00:45:26.303 --> 00:45:32.663 The order is something like that, because we could avoid having a representation 00:45:32.663 --> 00:45:34.883 based on the order, but we still have it. 00:45:35.203 --> 00:45:40.563 And so maybe everything needs to be organized in an ordered way. 00:45:40.803 --> 00:45:44.803 Right. And this is even when the task doesn't require that. 00:45:44.803 --> 00:45:47.923 But do you think there's something to be gained, 00:45:48.483 --> 00:45:51.683 to try to reinterpret the responses of 00:45:51.683 --> 00:45:57.563 these mixed cells as you call them in a rule in terms of rules would that help 00:45:57.563 --> 00:46:01.363 you you think to understand how the system works or I don't know it's a futile 00:46:01.363 --> 00:46:05.963 exercise I don't know I don't know if it can help to conceptualize that as a 00:46:05.963 --> 00:46:10.863 rule or as half a rule or I don't know it's very difficult to okay but now, 00:46:11.523 --> 00:46:14.123 so the neurons that you looked at. 00:46:16.083 --> 00:46:20.203 Would you say that in the end, they operate in a similar encoding space? 00:46:20.363 --> 00:46:24.323 Like, for instance, they all normalize their responses in some way. 00:46:25.083 --> 00:46:28.983 Like, for instance, if I would have, let's say, a stimulus that's further away 00:46:28.983 --> 00:46:33.563 or nearby, do I try to normalize all these responses to say, 00:46:33.623 --> 00:46:37.343 look, no, the only thing that matters really are my decision variables that 00:46:37.343 --> 00:46:38.423 tell me something about evidence. 00:46:38.543 --> 00:46:41.763 I don't want to know anything about, let's say, physical organization of the 00:46:41.763 --> 00:46:43.183 scene. Or do you really see this mixed? 00:46:43.363 --> 00:46:48.743 No, no, I see this mixed. The cells that I showed you are the pure cells, 00:46:48.863 --> 00:46:55.703 because otherwise, in a talk, if you start to show a mixed cell, people get confused. 00:46:57.163 --> 00:47:02.643 They say, for example, a common idea is that, we had also a problem with some 00:47:02.643 --> 00:47:06.263 referee in the past, is that how is it possible this cell is representing this, 00:47:06.403 --> 00:47:10.803 if this cell is representing also some other variables. 00:47:10.803 --> 00:47:14.803 But we know that a cell can represent more variables at the same time. 00:47:14.863 --> 00:47:18.003 But if you present in a talk, for example, a mixture of cells, 00:47:18.103 --> 00:47:22.463 many people, you know, we need to explain very well what is going on. 00:47:22.523 --> 00:47:27.903 So if we have poor cells and we want to make the point that these neurons make spatial computation. 00:47:28.263 --> 00:47:31.203 But now, doesn't it raise another problem? Because here we go. 00:47:31.343 --> 00:47:37.523 Now you allow these mixed cells. So that means if I'm reading out a mixed cell 00:47:37.523 --> 00:47:41.903 and the mixed cell is active, I have uncertainty because I don't really know 00:47:41.903 --> 00:47:46.043 if it's the physical display or is it because there's some important decision information. 00:47:46.443 --> 00:47:48.143 So that would mean…. 00:47:49.358 --> 00:47:53.298 If I have an integration-based model of decision-making, which is a popular 00:47:53.298 --> 00:47:56.498 way to think about it, that you say, okay, I just integrate my evidence over 00:47:56.498 --> 00:48:00.638 time and whatever reaches threshold first is what I do, then your mixed cells 00:48:00.638 --> 00:48:02.578 would mess up that whole scheme. 00:48:02.978 --> 00:48:06.678 Exactly. So you need to… So how are you going to solve that? You need to have, 00:48:06.798 --> 00:48:09.978 we can only think about a distributed 00:48:09.978 --> 00:48:14.578 way of representing an information that is mixed to other information and can 00:48:14.578 --> 00:48:20.838 be still extracted because all the other information will be filtered out because 00:48:20.838 --> 00:48:29.038 they would be on the opposite side. Right. 00:48:30.538 --> 00:48:35.178 But the point is, of course, this is why I was sort of hoping to find some sort of symbolic encoding. 00:48:35.338 --> 00:48:39.578 You say, look, I just care about informational aspects that pertain to my task, 00:48:39.698 --> 00:48:42.778 and I'm not biased by salience and so on. 00:48:42.778 --> 00:48:45.798 But now so that's not what you find no right 00:48:45.798 --> 00:48:48.598 and we are already in part of the prefrontal cortex that are 00:48:48.598 --> 00:48:51.318 really close to the motor cortex where we're going to 00:48:51.318 --> 00:48:55.058 our premotor where we're going to execute our actions so i 00:48:55.058 --> 00:49:00.678 have to get to if you want a pure informational representation in my frontal 00:49:00.678 --> 00:49:04.858 cortex before i can really make optimal decisions exactly so where is that happening 00:49:04.858 --> 00:49:11.058 then we it could happen in prefrontal but at a certain time and after a certain 00:49:11.058 --> 00:49:12.878 computation is performed formed, we don't know. 00:49:13.738 --> 00:49:16.798 Also it would be interesting again to look at the correlation of these cells 00:49:16.798 --> 00:49:20.438 like we were saying, if the cell that are representing a pure signal are more 00:49:20.438 --> 00:49:21.538 correlated than the cell. 00:49:22.958 --> 00:49:29.798 But have you ever found in any of your experiments a majority of responses being 00:49:29.798 --> 00:49:32.498 driven by these, let's call them, pure cells? 00:49:32.798 --> 00:49:34.878 No, it's always a minority. 00:49:35.338 --> 00:49:37.558 Right. Yeah, it's always a minority. Well, this is interesting, 00:49:37.638 --> 00:49:42.758 right, because this might be telling us that also this idea of just pure rate-based 00:49:42.758 --> 00:49:46.838 encoding of decision-making might actually not be the reality of the frontal cortex. 00:49:46.958 --> 00:49:52.798 Because if you have these mixed encodings with conjunctive cells further modulated 00:49:52.798 --> 00:49:57.018 by properties of the display, these responses, these rates of responding, 00:49:57.178 --> 00:50:00.758 are not informative on the goals you have to pursue. 00:50:01.098 --> 00:50:05.638 Exactly. So we don't know if the information will be extracted from these cells 00:50:05.638 --> 00:50:09.618 or if these cells are just an intermediate step for the final computation. 00:50:09.618 --> 00:50:11.278 So this is what we don't know. Right. 00:50:11.718 --> 00:50:16.838 Is the frontal pole better at that? The frontal pole, for now we know that there is just one signal. 00:50:17.138 --> 00:50:19.418 So we- Right, exactly. So also that cannot help you, right? Exactly. 00:50:19.478 --> 00:50:22.538 So it's on the opposite extreme, 00:50:22.778 --> 00:50:29.098 but also I was saying today that we don't have to expect every signal present 00:50:29.098 --> 00:50:32.238 in the prefrontal cortex, for example, in the strategy task in the dorsolateral 00:50:32.238 --> 00:50:35.918 prefrontal cortex, the monkey needed to remember the previous stimulus. 00:50:37.116 --> 00:50:43.856 We didn't find any evidence in thousands of cells of the representation of the previous stimulus. 00:50:44.096 --> 00:50:49.296 While the new task that I was describing, the distance task, 00:50:49.356 --> 00:50:50.456 just required a monitoring. 00:50:50.676 --> 00:50:55.956 That doesn't require really a monitoring, but we show that the cells show a monitoring activity. 00:50:56.256 --> 00:51:00.136 We see that when a stimulus was a goal, it's still represented, 00:51:00.396 --> 00:51:05.676 but even if it's not necessary to represent it, but when it was necessary to 00:51:05.676 --> 00:51:07.816 represent it, it wasn't. 00:51:07.996 --> 00:51:12.056 Right, exactly. In dorsal hyperfrontal cortex, obviously. Yes, this is amazing, right? 00:51:12.616 --> 00:51:21.776 So the dominant view on decision-making in the cortex is… The dominant view 00:51:21.776 --> 00:51:26.936 is very much integration-based, right, according to this race model of bounded diffusion models. 00:51:27.156 --> 00:51:30.176 So your data doesn't really fit that model. 00:51:30.996 --> 00:51:35.256 So in your mind therefore these bounded diffusion models are not an accurate 00:51:35.256 --> 00:51:39.896 description of this system or is it just something that you haven't looked at sufficiently yet? 00:51:40.376 --> 00:51:43.576 No, I don't think that it's in contrast with this model. 00:51:43.776 --> 00:51:50.196 It's just a more complex task where we can maybe look at a raised model where 00:51:50.196 --> 00:51:54.116 we can study the activity based on a raised model. 00:51:54.116 --> 00:52:01.116 Model, but still we may have competing goals, like right and left or different 00:52:01.116 --> 00:52:05.656 objects that compete with each other, even in my task. 00:52:05.856 --> 00:52:11.136 But the moment in which that happens is very fast, so it's very difficult for us to study. 00:52:12.136 --> 00:52:16.976 Unless we maybe study an interpopulation of cells recorded simultaneously, 00:52:17.476 --> 00:52:20.796 we may understand more what is going on in the moment of a decision, 00:52:20.916 --> 00:52:22.036 but the decision is very fast. 00:52:22.036 --> 00:52:27.776 Yeah, but still you talk about what, dozens if not hundreds of milliseconds, right? 00:52:27.996 --> 00:52:31.316 Yes. So it's not that fast, is it? 00:52:31.536 --> 00:52:36.476 But maybe to understand in our task what a single neuron does, 00:52:36.716 --> 00:52:37.916 I think that is difficult. 00:52:38.176 --> 00:52:44.096 Because we don't see the representation of competing alternative because immediately, 00:52:44.716 --> 00:52:48.896 the monkey can make a decision about one goal compared to the other. 00:52:48.896 --> 00:52:53.196 So there is no, maybe the uncertainty that we can have, you can have with a 00:52:53.196 --> 00:52:56.496 random dot experiment where you can manipulate the uncertainty. 00:52:56.596 --> 00:52:58.216 Here we cannot manipulate the uncertainty. 00:52:58.496 --> 00:53:01.976 Okay. So it's difficult without manipulating it. Okay, but, but. 00:53:02.691 --> 00:53:09.011 Wouldn't your data suggest that there must be other modes of integrating information 00:53:09.011 --> 00:53:14.451 in prefrontal cortex that is more sophisticated than just adding up? 00:53:14.471 --> 00:53:18.111 Yeah, information like the movement of the dots. 00:53:18.171 --> 00:53:22.091 So I think that the computation that is required here is not too much. 00:53:22.451 --> 00:53:26.491 I don't know. I cannot see it in the context of a race model. 00:53:27.311 --> 00:53:30.831 Right, exactly. So it's more difficult for me to see it within the paradigm. 00:53:30.831 --> 00:53:34.991 Now which is good because i think the race model is is not necessarily it's 00:53:34.991 --> 00:53:38.491 only the beginning of a story on decision making and certainly not the end right 00:53:38.491 --> 00:53:41.831 um so now in your last set of experiments. 00:53:42.531 --> 00:53:47.931 You talked about the encoding of irrelevant information yes this is so why would 00:53:47.931 --> 00:53:51.111 that be interesting because in some sense it's like i know these people that 00:53:51.111 --> 00:53:54.211 that you know you go shopping you get a shopping list but first they tell you 00:53:54.211 --> 00:53:57.471 all the things you do not have to get that, which seems very inefficient. 00:53:57.791 --> 00:54:00.751 So why would you worry about encoding irrelevant information? 00:54:01.171 --> 00:54:05.131 Yeah, I think it's important to encode irrelevant goal. We found that mainly 00:54:05.131 --> 00:54:13.211 we encode relevant goal because we may find better way of reaching a solution. 00:54:13.411 --> 00:54:18.311 For example, think about a task in which you need to associate A and B to different 00:54:18.311 --> 00:54:24.291 position, but you perform this task and after a while you understand that you 00:54:24.291 --> 00:54:25.511 are required only to go right. 00:54:25.671 --> 00:54:29.591 So you can avoid taking care of or paying attention to the stimulus. 00:54:29.791 --> 00:54:33.271 Without monitoring you would continue to perform the task in a more complex 00:54:33.271 --> 00:54:37.591 way. So I think we have a lot of situations like that where we can find a shortcut. 00:54:37.971 --> 00:54:42.411 Okay, so it's like an incremental pruning of information in some sense. 00:54:42.431 --> 00:54:43.291 Yes, I think so, yeah. Okay. 00:54:43.491 --> 00:54:47.371 So what's the mechanism there? How does this play out in prefrontal cortex? 00:54:47.971 --> 00:54:54.631 Now my study doesn't allow us to understand the way in which this information is used. 00:54:54.891 --> 00:54:56.931 So we don't have a use of this information. 00:54:57.611 --> 00:55:01.131 And so it would be nice to design a task where we can see all this information. 00:55:01.471 --> 00:55:05.911 But how rapidly does... Because here comes the display. I perform my task, okay? 00:55:06.251 --> 00:55:09.351 So it would expect I have all these cells, I have my conjunctive cells, 00:55:09.451 --> 00:55:12.931 I have whatever. So all these cells get allocated to describing this task. 00:55:13.331 --> 00:55:16.871 But now I'm going to figure out that the subset of these descriptions are irrelevant. 00:55:17.291 --> 00:55:20.631 Yes. So then you would expect these cells start to drift again and their response 00:55:20.631 --> 00:55:23.651 is in some way... We don't know if these cells were more before. 00:55:23.651 --> 00:55:29.211 Because we may be, you know, in learning the paradigm, the cells were double 00:55:29.211 --> 00:55:31.051 than the cells that we are having now. 00:55:31.151 --> 00:55:36.471 And now we are just looking at the survivals of a mechanism that is broader 00:55:36.471 --> 00:55:38.791 than this with much more neurons involved. 00:55:39.491 --> 00:55:45.171 And also an interesting point that should be studied is that now we have cells 00:55:45.171 --> 00:55:49.791 that are encoding the previous goal when it is irrelevant. But what if we start 00:55:49.791 --> 00:55:54.171 to give wrong messages to the monkey and the monkey gets confused? 00:55:54.431 --> 00:56:00.431 Will the monkey start to monitor other information different from gold and will be confused? 00:56:01.017 --> 00:56:03.937 Dorsolateral prefrontal cortex the right area where to find that 00:56:03.937 --> 00:56:06.977 information so i don't know okay so there are many questions open but 00:56:06.977 --> 00:56:10.677 you do see you would suggest that there is always a 00:56:10.677 --> 00:56:13.637 goal-driven monitoring of the task to sort 00:56:13.637 --> 00:56:16.917 of focus on the relevant information there's 00:56:16.917 --> 00:56:19.977 a continuous selection process going on there is that what you have in mind 00:56:19.977 --> 00:56:25.477 i think that yes but not only that because we have also we have relevant goals 00:56:25.477 --> 00:56:28.517 that are represented but But the interesting thing is that this representation 00:56:28.517 --> 00:56:35.197 is not so much bigger than the information about the previous goal that we found when the goal, 00:56:35.377 --> 00:56:39.037 we might get the requirement to maintain in memory the goal. 00:56:39.117 --> 00:56:44.417 That was kind of interesting because we don't find big difference when we need 00:56:44.417 --> 00:56:47.017 to remember something from when we don't need at all. 00:56:47.257 --> 00:56:49.437 Okay. So that… How do you explain that? 00:56:49.717 --> 00:56:55.317 It's very difficult to understand that. Could this be a nonspecific kind of memory response? 00:56:56.117 --> 00:57:01.017 Let's say whatever you shoot into this prefrontal cortex, you have a memory field, right? 00:57:01.097 --> 00:57:05.317 So it will be sort of automatically maintained, non-specifically with respect to the task. 00:57:05.517 --> 00:57:08.997 If it's a goal, yes. Would you buy that interpretation? 00:57:10.557 --> 00:57:16.917 The problem now is that what I think is that now we need to be more cautious 00:57:16.917 --> 00:57:20.837 when we study, when we say that the signal is associated to a behavior. 00:57:20.837 --> 00:57:25.157 For example, if we can find neurons that encode a previous goal, 00:57:25.237 --> 00:57:28.357 even when the previous goal was not a requirement. 00:57:28.557 --> 00:57:33.617 Now when I have a requirement in a task, I don't know anymore if this is a monitoring 00:57:33.617 --> 00:57:36.117 activity or a memory activity. 00:57:37.332 --> 00:57:40.372 As a function of reaching a goal. 00:57:40.452 --> 00:57:46.252 So now this is my idea that we need to think about this. 00:57:46.332 --> 00:57:51.192 So that means that now after all these experiments, one important conclusion 00:57:51.192 --> 00:57:57.092 is that to just only interpret activity in frontal cortex in terms of goal monitoring 00:57:57.092 --> 00:57:58.932 is not sufficient anymore. 00:57:59.212 --> 00:58:02.532 There might also be more non-specific memory effects playing out. 00:58:02.572 --> 00:58:03.472 This is what you're saying. 00:58:04.292 --> 00:58:10.472 I'm saying that when we know that there is a task requirement and we see a representation, 00:58:11.072 --> 00:58:14.952 that we think, okay, we have this representation because there is that task requirement. 00:58:15.132 --> 00:58:19.152 Maybe it's not because we have the task requirement. Maybe even without it, 00:58:19.232 --> 00:58:20.932 we would have that representation. Exactly right. 00:58:22.232 --> 00:58:26.712 But would you have any difference in the physiological signature of these responses? 00:58:26.712 --> 00:58:32.252 Like a nonspecific memory response, so that it's not dependent on a specific task requirement. 00:58:32.252 --> 00:58:35.212 Requirement might for instance have a different latency a different amplitude 00:58:35.212 --> 00:58:37.852 are there any differences there that helps you because we are 00:58:37.852 --> 00:58:40.832 comparing different tasks different monkeys just a proportion of 00:58:40.832 --> 00:58:43.552 cells that are selective so it's very difficult now to 00:58:43.552 --> 00:58:46.452 say i would say that the proportion of cells is not so different i 00:58:46.452 --> 00:58:49.952 can and the task is different as different requirements so it's very difficult 00:58:49.952 --> 00:58:55.112 to compare latencies would be interesting i think to continue this line of research 00:58:55.112 --> 00:58:59.932 with chronic recording and to see and manipulating the experiment and to understand 00:58:59.932 --> 00:59:04.072 this is the way to do instead of doing multiple experiments, comparing them. 00:59:04.512 --> 00:59:08.832 So now we know we have sort of different pools of cells, at least around sort 00:59:08.832 --> 00:59:10.612 of goal monitoring and memory. 00:59:10.972 --> 00:59:14.852 Yes. And so now we have an evolving task, right? So I have one trial, 00:59:15.032 --> 00:59:17.992 I succeed or not, I go to the next one and so on. 00:59:18.212 --> 00:59:21.092 What's the information transfer between these trials? 00:59:23.072 --> 00:59:27.652 What do I carry over in information? These last experiments you described, 00:59:27.892 --> 00:59:31.452 you try to assess, indeed, look, I have a task configuration. 00:59:31.452 --> 00:59:37.092 I have a goal, I have an action, I get a reward or not. And now I get my next trial. 00:59:37.612 --> 00:59:41.472 So the question is, okay, of this trial at T is one, 00:59:42.842 --> 00:59:47.142 What's the information that I really carry over to the trial at T plus 10? 00:59:47.662 --> 00:59:52.902 Yeah. Right? We, yeah, we, I think that we need to design an experiment to see 00:59:52.902 --> 01:00:00.162 under which condition we, we have a memory activity that goes over the monitor activity. 01:00:00.182 --> 01:00:03.262 It is something more than a monitor. But I thought you were suggesting today 01:00:03.262 --> 01:00:08.062 that actually only information on the previous goal really moves over. 01:00:08.182 --> 01:00:10.562 So for example, you can introduce random mistakes. 01:00:10.762 --> 01:00:15.442 For example, sometimes you don't give a reward to a monkey. and what happens next in the next trial? 01:00:15.562 --> 01:00:20.082 Is the monkey starting to represent something more than this because it needs 01:00:20.082 --> 01:00:25.662 to reconsider the rule of the task and maybe we start to monitor even the characteristics 01:00:25.662 --> 01:00:26.922 of the second stimulus? Right. 01:00:27.142 --> 01:00:33.662 Yeah. So this would be my, if I could do 10 experiments at the same time, 01:00:33.682 --> 01:00:34.902 this would be one of them. 01:00:35.222 --> 01:00:40.022 So now we looked at prefrontal cortex. So, but in terms of territory, 01:00:40.722 --> 01:00:45.402 How big a percentage of the neocortex do we call prefrontal cortex? 01:00:45.882 --> 01:00:51.282 We have to distinguish granula and agranula prefrontal cortex. 01:00:52.222 --> 01:00:55.322 The granula prefrontal cortex, we can study only in primates, 01:00:55.482 --> 01:00:58.322 and this is a primate innovation of the granula prefrontal cortex. 01:00:58.522 --> 01:01:02.422 Also the rodents of the prefrontal cortex, but it's just a granula. 01:01:02.522 --> 01:01:07.342 So it depends on what we consider, and by the granula prefrontal cortex, probably it, 01:01:08.262 --> 01:01:11.102 it has a role an important role in 01:01:11.102 --> 01:01:14.122 this would be about one third of the cortical sheet in 01:01:14.122 --> 01:01:20.222 primates it could be around maybe one third I can't tell exactly it's quite 01:01:20.222 --> 01:01:25.842 a chunk of cortical so then the question becomes how many subdivisions of this 01:01:25.842 --> 01:01:33.902 should we really consider we can consider a lateral part of the frontal cortex and a more, 01:01:34.522 --> 01:01:40.182 orbit orbitofrontal part that sometimes is associated to a ventrolateral. 01:01:40.871 --> 01:01:44.851 Or to the more medial dorsomedial prefrontal cortex. 01:01:44.991 --> 01:01:49.791 So it depends on the scheme of connection that you consider and so the classification 01:01:49.791 --> 01:01:56.171 is still depends on what you consider important if it is the connections to 01:01:56.171 --> 01:02:01.871 have a common input a common output or you look at more of the functions or 01:02:01.871 --> 01:02:03.591 different functions of the prefrontal cortex. 01:02:03.751 --> 01:02:10.311 Is there consensus on this in the field? There is a consensus about macroscopic divisions, yes. 01:02:10.531 --> 01:02:18.371 But when we look at each microscopic division, you can have more or less subdivision 01:02:18.371 --> 01:02:23.391 based on how much you want to be specific, how much you want to divide. 01:02:23.631 --> 01:02:26.471 So in anatomy, we never have a perfect 01:02:26.471 --> 01:02:29.571 number of areas because also the 01:02:29.571 --> 01:02:33.591 neurophysiology cannot follow because we just 01:02:33.591 --> 01:02:36.551 now with neuroimaging with new with a 01:02:36.551 --> 01:02:39.611 make with a possibility to do a magnetic 01:02:39.611 --> 01:02:44.971 resonance to a monkey we know more about the location but it's still difficult 01:02:44.971 --> 01:02:49.011 to target specifically a part of the orbital frontal cortex so when you see 01:02:49.011 --> 01:02:53.351 for example as results about orbital frontal cortex they don't distinguish too 01:02:53.351 --> 01:02:59.211 much there all the areas that Carmichael and price is divided by orbital frontal cortex. 01:02:59.351 --> 01:03:05.691 So sometimes we have the tendency to divide more, to be more precise, 01:03:05.771 --> 01:03:12.691 to see differences, but sometimes when it is too much, we cannot use it to understand the neurophysiology. 01:03:12.771 --> 01:03:17.051 So it's very difficult to understand which is the right level in making divisions. 01:03:17.471 --> 01:03:21.371 So would you say your monkeys do learn this task the same way humans would? 01:03:23.137 --> 01:03:29.697 We know that the problem is for a monkey is to understand the rule of the games as usually. 01:03:29.837 --> 01:03:34.117 So it's very difficult, you know, to understand, to compare. Okay. 01:03:34.557 --> 01:03:39.957 Because they need to understand at the beginning that touching the screen can produce something. 01:03:40.197 --> 01:03:43.537 And so it's very difficult for 01:03:43.537 --> 01:03:49.157 also for the monkey to eliminate all the potential alternative to a task. 01:03:49.157 --> 01:03:53.517 For example, the monkey can focus on the fact that it needs to choose the stimulus 01:03:53.517 --> 01:03:57.617 on the top always, and after try the stimulus on the bottom, 01:03:57.777 --> 01:04:01.957 and maybe the monkey find the solution and go back again. 01:04:02.197 --> 01:04:07.497 So it's an exploration of possibilities, the training. So for your test, 01:04:07.597 --> 01:04:09.277 how long do you train these monkeys usually? 01:04:09.617 --> 01:04:15.297 Can be, let's say, rarely less than six months, one year. So I can reach two 01:04:15.297 --> 01:04:17.157 years. Right, a lot of time. 01:04:17.557 --> 01:04:22.757 So Aldo, to finish up our conversation, there are two things. 01:04:22.897 --> 01:04:28.157 So you're reworking now on monkey physiology for quite a while. 01:04:28.357 --> 01:04:32.297 You have gained also amazing insights in how the system works, 01:04:32.397 --> 01:04:36.397 despite all the unclarities, but this is also the research aspect of it. 01:04:36.897 --> 01:04:43.037 But now, if we would like to follow your example in the study of the brain, 01:04:43.117 --> 01:04:44.137 what would be Aldo's law? 01:04:45.951 --> 01:04:59.111 I think that what I tried to do was always to be kind of at the frontier of the field, 01:04:59.171 --> 01:05:01.471 like with the frontal pole we were the first, 01:05:01.611 --> 01:05:04.551 with the time we were among the first. 01:05:04.791 --> 01:05:09.491 So I like to be one of the first to do something. So that was kind of a rule 01:05:09.491 --> 01:05:15.491 that is on the other opposite. you may be interested in only one subject to 01:05:15.491 --> 01:05:17.671 be a specialist of only one thing. 01:05:17.811 --> 01:05:20.051 So I chose the other. Right. No. 01:05:20.791 --> 01:05:24.431 Novelty. The novelty. I like the novelty and get bored after a while. 01:05:24.731 --> 01:05:28.171 So I like some new challenge. Great. So, so far it was like this. 01:05:28.371 --> 01:05:32.731 So Aldo, so five years from now, we're going to come visit you wherever you are. Now you're in Rome. 01:05:33.211 --> 01:05:35.731 And I'm going to confront you with the prediction you're going to make today. 01:05:36.071 --> 01:05:39.311 So I'm going to ask you, look, Aldo, you predicted X. 01:05:39.631 --> 01:05:43.891 Did it happen or not? not. So what's the one prediction you would like to make 01:05:43.891 --> 01:05:47.931 today that you're most… I would like to be able to record, let's say, 01:05:47.951 --> 01:05:49.631 more neurons and more areas together, 01:05:50.251 --> 01:05:54.951 simultaneously to understand really how a circuit made of different areas, 01:05:55.091 --> 01:05:59.411 like we were saying, the frontal pole and the dorsolateral upper frontal cortex work together. 01:05:59.611 --> 01:06:03.351 So I would like next time to answer more questions, some of the questions that 01:06:03.351 --> 01:06:05.211 I was… No, I don't let you get away so easily. 01:06:05.491 --> 01:06:10.551 I want a hypothesis on prefrontal cortex. Yeah, now I'm very interested in social 01:06:10.551 --> 01:06:14.351 cognition, so I think that the next step will be talking about novelty. 01:06:15.411 --> 01:06:19.411 So I'm not continuing studying time, but I'm moving to social interaction, 01:06:19.591 --> 01:06:23.431 I want to understand really what all these signals, if all these signals that 01:06:23.431 --> 01:06:29.451 I found previously apply to the representation of a different agent that is interacting with that. 01:06:29.551 --> 01:06:32.211 So the new challenge is social, I think. 01:06:32.411 --> 01:06:36.111 So in five years time, you think you have shown that the prefrontal cortex is 01:06:36.111 --> 01:06:39.091 the substrate for social cognition? For some aspects, I hope, 01:06:39.171 --> 01:06:40.131 at least for some aspects. 01:06:40.371 --> 01:06:43.411 Great. So, Aldo Genovinozzi, thank you very much for this conversation. 01:06:43.791 --> 01:06:45.391 Thank you, Paul. Thank you for inviting me. 01:06:45.680 --> 01:06:51.920 Music. 01:06:51.951 --> 01:06:57.511 The CSN Podcast was produced by the Convergent Science Network of Biometrics 01:06:57.511 --> 01:07:03.931 and Biohybrid Systems, a project funded by the European 7th Research Framework Program. 01:07:05.411 --> 01:07:10.771 For more interviews, recorded lectures, or upcoming conferences in the field 01:07:10.771 --> 01:07:17.011 of biometrics and biohybrid systems, go to csnnetwork.eu. 01:07:17.840 --> 01:07:25.520 Music.