WEBVTT 00:00:03.517 --> 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 Verscher and Tony Prescott. 00:00:19.557 --> 00:00:25.157 This is Tony Prescott for the Convergent Science Network podcasts from the Barcelona 00:00:25.157 --> 00:00:28.237 Cognition Brain and Technology Summer School 2012. 00:00:28.237 --> 00:00:35.217 And I'm here with Mitra Hartman from Northwestern University where Mitra leads 00:00:35.217 --> 00:00:37.057 a group looking at perception. 00:00:38.137 --> 00:00:45.697 Mitra, in your talk this morning you gave us a definition of active touch which 00:00:45.697 --> 00:00:51.537 comes from the psychologist James Gibson where he said that active touch was 00:00:51.537 --> 00:00:55.477 what happens when we do touching rather than when we're being touched. 00:00:55.817 --> 00:01:01.057 Can you explain a little bit more about what you understand about the idea of touch being active. 00:01:03.497 --> 00:01:10.017 Well, I think we need to be careful with the definition of active touch. 00:01:10.537 --> 00:01:16.277 And the reason is that you might think of a mechanical definition, 00:01:16.537 --> 00:01:22.817 that is just that the muscles are active and the muscles are used to go touch 00:01:22.817 --> 00:01:29.077 something thing, versus the muscles are not active and the human or animal receives 00:01:29.077 --> 00:01:31.037 the data in and that's passive. 00:01:31.577 --> 00:01:37.697 But the way that Gibson really meant the use of the word active touch is in 00:01:37.697 --> 00:01:40.597 the sense of purposeful exploration. 00:01:41.477 --> 00:01:50.777 And so one possible example is if you use a cloth to grab a hot pan off the 00:01:50.777 --> 00:01:56.397 stove, you would typically say that you grabbed the cloth, you grabbed the pan. 00:01:57.377 --> 00:02:00.797 You wouldn't say that you touched the cloth or you touched the pan. 00:02:01.017 --> 00:02:07.977 You might say that you touched the cloth if you actively explored the texture of that cloth. 00:02:08.257 --> 00:02:13.517 And that sort of exploratory movement would be the type of movement that Gibson 00:02:13.517 --> 00:02:16.617 would have described as active touch. 00:02:17.177 --> 00:02:23.817 But it's clear that in both cases, the muscles are active and you're making use of sensory data. 00:02:24.377 --> 00:02:32.577 CB So you're restricting active touch to be behaviors where you're using touch 00:02:32.577 --> 00:02:35.337 to actually sense the properties of the world around you. 00:02:35.417 --> 00:02:39.637 That's the sense in which Gibson used the word active touch. 00:02:39.977 --> 00:02:44.537 And so if we're going to be consistent with Gibson, that's how it would have to be defined. 00:02:44.877 --> 00:02:51.337 I think it's also useful to talk about what one might call active somatosensation, 00:02:51.817 --> 00:02:54.017 and there you could have a. 00:02:54.827 --> 00:03:02.667 Purely mechanical definition in which the muscles are active in a particular sensory surface or not. 00:03:02.847 --> 00:03:06.667 That is, the muscles are responsible for the sensory data that you receive or not. 00:03:07.027 --> 00:03:11.747 And that would be independent of the intent of the movement. 00:03:12.287 --> 00:03:18.227 So for example, as you're walking, you might get sensory data on your feet, 00:03:18.367 --> 00:03:21.667 and that would be active somatosensation, sensation 00:03:21.667 --> 00:03:24.467 but it would not be what gibson would have 00:03:24.467 --> 00:03:27.547 called active touch yeah i 00:03:27.547 --> 00:03:31.207 agree that it's it's complicated i suppose active sensing 00:03:31.207 --> 00:03:34.167 is now a phrase that people use for many 00:03:34.167 --> 00:03:39.687 mean many different things and it's probably best if individuals just say what 00:03:39.687 --> 00:03:43.147 they mean by it when they start talking about it i think that's a really good 00:03:43.147 --> 00:03:52.027 idea so in your group is it fair to characterize what you're doing as studying perception in general, 00:03:52.087 --> 00:03:54.707 but through the route of looking at active touch. 00:03:55.327 --> 00:04:01.867 I think that's a fair characterization. I'm very interested in the question of perception, 00:04:02.147 --> 00:04:12.607 that is how an animal or a human can take energy in, in the form of a sound wave or a light wave, 00:04:12.887 --> 00:04:18.547 and somehow turn that energy signal into a perception of the world. 00:04:18.727 --> 00:04:21.667 That's something we have very little understanding of. 00:04:22.747 --> 00:04:26.847 But also, we don't know that animals do form perceptions. 00:04:27.127 --> 00:04:32.607 This is true. You're absolutely right. We do not know that animals form perceptions. 00:04:32.687 --> 00:04:37.307 We infer that they form perceptions based on their behavior. 00:04:38.147 --> 00:04:43.447 And so, what would be an example of a behaviour of, say, an animal, 00:04:43.567 --> 00:04:46.927 from which you might infer that they've got a perception of an environment, 00:04:47.027 --> 00:04:51.967 rather than just directly using sensing to control movement? Yeah. 00:04:54.665 --> 00:05:00.925 Uh, some, some sort of tasks that you perform in the lab or where you ask them 00:05:00.925 --> 00:05:03.165 to discriminate would be an example or. 00:05:03.425 --> 00:05:09.865 Certainly in the laboratory, we can, uh, measure what we like to think of as perception. 00:05:10.365 --> 00:05:19.385 Um, so for example, you can train a rat to, uh, go to the left if there's a 00:05:19.385 --> 00:05:23.465 square object in its path and go to the right if there's a round object in its path. 00:05:23.465 --> 00:05:28.785 And if the objects are in every other way identical and your task is appropriately 00:05:28.785 --> 00:05:34.745 controlled, you might argue that the rat has had a perception of a cube or a sphere. 00:05:35.445 --> 00:05:41.105 But in this case and in the studies you're talking about, the rat might well be using its whiskers. 00:05:41.325 --> 00:05:48.045 The rat is very likely to be using its whiskers to perform such a tactile discrimination. 00:05:49.085 --> 00:05:54.005 Rats have four paws, but they don't use them very much to explore the world. 00:05:54.125 --> 00:05:58.625 Instead, they use their whiskers to explore the world. And people don't have whiskers. 00:05:58.745 --> 00:06:04.285 So it's going to be quite different experience of the world or perceptions that 00:06:04.285 --> 00:06:05.885 rats are going to have compared to people. 00:06:06.345 --> 00:06:10.445 It does seem likely that rats will have very different perceptions of the world 00:06:10.445 --> 00:06:11.645 than a person would have. 00:06:12.225 --> 00:06:18.965 Humans are very visual animals. Rats are very tactually and olfactory-based creatures. 00:06:19.645 --> 00:06:24.045 So you have some hypotheses about how a rat might use its whiskers to build 00:06:24.045 --> 00:06:25.185 up a tactile perception. 00:06:26.794 --> 00:06:32.094 Our laboratory has several different ideas about how the rat might use its whiskers 00:06:32.094 --> 00:06:34.114 to build up perceptions of the world. 00:06:34.774 --> 00:06:38.254 We are particularly interested in the mechanics of sensing. 00:06:38.874 --> 00:06:45.674 Rats actively move their whiskers back and forth rhythmically as they explore objects. 00:06:46.654 --> 00:06:50.034 Now, you may be familiar with the whiskers on a cat or a dog. 00:06:50.694 --> 00:06:55.054 Cats and dogs do use their whiskers to explore different things, 00:06:55.134 --> 00:06:58.414 but they can't actively move them in the same way that a rat does. 00:06:58.694 --> 00:07:02.414 So a rat brushes its whiskers against objects rhythmically. 00:07:03.394 --> 00:07:12.674 And we think that the forces that the whiskers exert on different objects tell 00:07:12.674 --> 00:07:15.634 the rat something about the properties of those objects. 00:07:16.094 --> 00:07:19.634 So the whisker is just a hair, isn't it? A particular kind of hair. 00:07:19.914 --> 00:07:25.374 And so it's not sensing anything on the whisker itself, but at the base. That's right. 00:07:25.474 --> 00:07:29.714 The whiskers are just like the hairs on your head in the sense that there are 00:07:29.714 --> 00:07:31.094 no sensors along their length. 00:07:31.294 --> 00:07:34.614 All of the sensors are at the base of the whisker in a follicle. 00:07:35.014 --> 00:07:39.734 So we could, for example, trim the whiskers of a rat and it would not hurt the 00:07:39.734 --> 00:07:41.974 rat at all, same way as you can get a haircut. 00:07:42.954 --> 00:07:47.314 And so it has a whole bunch of these whiskers on each side of its face and it's 00:07:47.314 --> 00:07:53.054 moving them quickly against objects and sensing that at the base of the whisker. That's correct. 00:07:53.134 --> 00:07:58.874 There are about 60 whiskers in total on the rat, 30 approximately on each side, 00:07:59.054 --> 00:08:07.074 and it moves the whiskers very rapidly, up to, say, 1,000 degrees per second. 00:08:07.494 --> 00:08:10.134 And why do you think it moves its whiskers? 00:08:12.534 --> 00:08:15.414 It's not entirely clear to the 00:08:15.414 --> 00:08:22.774 field what the advantages are of active movements of the whiskers why for example 00:08:22.774 --> 00:08:28.614 does the rat actively move its whiskers while the cat and dog don't one clear 00:08:28.614 --> 00:08:34.074 advantage is that it's able to cover more volume but then you might easily ask well well, 00:08:34.094 --> 00:08:37.674 why doesn't it just grow more whiskers to cover that same volume? 00:08:38.094 --> 00:08:43.574 So I think it's a large open question that's certainly worth investigating about 00:08:43.574 --> 00:08:48.614 why some animals perform this whisking behavior, as we call it, 00:08:48.634 --> 00:08:49.814 and other animals don't. 00:08:50.677 --> 00:08:56.137 So the whisking that the animal does against surfaces causes deflections of 00:08:56.137 --> 00:09:02.677 the whiskers and that bending is picked up in the follicle, the base of the whisker. 00:09:03.477 --> 00:09:08.917 What kind of things do you think the animal is able to extract from that signal? 00:09:09.037 --> 00:09:12.097 What kind of properties of the world do you think it can detect? 00:09:12.097 --> 00:09:20.937 So, a number of laboratories use the rat whisker system as a model to study the sense of touch. 00:09:21.157 --> 00:09:29.477 There are maybe 20, 30 laboratories interested in the rat vibrissal or whisker system. 00:09:29.697 --> 00:09:35.577 So, what I'm about to say comes from many different laboratories all working on this question. 00:09:36.497 --> 00:09:40.697 There are a number of laboratories that have shown that rats can discriminate 00:09:40.697 --> 00:09:42.997 different textures. with their whiskers. 00:09:43.637 --> 00:09:48.137 There are some studies that show that rats can do as well with their whiskers 00:09:48.137 --> 00:09:52.397 as we might be able to do with our fingertips in judging texture. 00:09:53.277 --> 00:09:58.877 Rats can certainly use their whiskers to navigate along the contours of walls. 00:09:59.057 --> 00:10:01.257 They're very good at scurrying along walls. 00:10:01.637 --> 00:10:06.817 And from this, we infer that they can tell something about the way that the 00:10:06.817 --> 00:10:10.377 wall is bending or curving back and forth. 00:10:10.557 --> 00:10:15.657 So we infer that they're able to obtain information about the local curvature 00:10:15.657 --> 00:10:19.897 of an object, at least far enough in advance to tell themselves how to locomote. 00:10:20.897 --> 00:10:24.937 So certainly texture and the shape of a wall. 00:10:25.997 --> 00:10:30.377 There's work by Ron Frostig showing that rats can determine the orientation 00:10:30.377 --> 00:10:33.877 of a bar, whether it's vertical or horizontal. 00:10:35.017 --> 00:10:41.497 And there's, I believe, some recent evidence showing they can do compliance, 00:10:41.937 --> 00:10:43.957 but I'm not confident about that. 00:10:45.114 --> 00:10:51.894 And is studying this system going to help us understand James Gibson's question 00:10:51.894 --> 00:10:54.534 of active touch in people? Are there similarities? 00:10:55.674 --> 00:11:00.314 One of the nice advantages of the rat vibrissal system, 00:11:00.394 --> 00:11:04.754 and the reason that I've chosen it as a model for use in my laboratory, 00:11:04.754 --> 00:11:11.674 is that the neural pathways that carry information from the vibrissae through 00:11:11.674 --> 00:11:17.654 the rat's brain are in many ways analogous to the neural pathways that carry 00:11:17.654 --> 00:11:21.754 information from the human hand through the human brain. 00:11:22.234 --> 00:11:27.434 They're not entirely similar, but they are in many ways analogous, 00:11:27.434 --> 00:11:32.654 and so we think that by better understanding the neural processing at different 00:11:32.654 --> 00:11:34.734 stages of the rat whisker system, 00:11:34.994 --> 00:11:40.334 we may gain insight into how sensory data from the hand is processed. 00:11:41.074 --> 00:11:46.194 So what's your strategy for understanding this system? Where do you think is 00:11:46.194 --> 00:11:48.254 the best place to focus the effort? 00:11:49.314 --> 00:11:56.714 As an engineer, I argue that one of the most essential things we can do is to 00:11:56.714 --> 00:11:58.134 characterize the input. 00:11:58.474 --> 00:12:04.194 And what I mean by that is that we must know what each whisker is doing, 00:12:04.294 --> 00:12:08.714 and the sensory data obtained by each of those whiskers, that is the mechanics 00:12:08.714 --> 00:12:13.334 of those whiskers and the mechanical signals that each of those whiskers is receiving. 00:12:14.354 --> 00:12:20.314 And then we need to know how those mechanical signals correspond to spikes, 00:12:20.454 --> 00:12:27.194 or electrical activity, that is, in the very first stage of neurons in the system. 00:12:27.194 --> 00:12:33.914 And if we understand that input, then we can go further on up the different 00:12:33.914 --> 00:12:36.374 brain structures and look at how that input is transformed. 00:12:37.966 --> 00:12:43.486 But we can't begin to understand the processor until we know what's going into the system. 00:12:43.926 --> 00:12:47.946 And for that, you think we need to look at the material properties of the whiskers themselves? 00:12:48.946 --> 00:12:54.846 We need to look at the material properties, and our laboratory has done some of that. 00:12:54.966 --> 00:12:59.966 We need to look at the basic mechanics of single vibrissae, and we need to look 00:12:59.966 --> 00:13:06.546 at how rats use their whiskers naturally during their normal, regular behavior. 00:13:06.546 --> 00:13:14.026 How is the brain used as the rat is exploring its objects as it would in its natural environment? 00:13:15.066 --> 00:13:21.046 So as well as being interesting for understanding the rat brain and potentially the human brain, 00:13:21.846 --> 00:13:26.906 is there uses of vibrational sensing that you could imagine, 00:13:27.106 --> 00:13:31.446 perhaps in technology, that we could develop by understanding the system better? 00:13:32.866 --> 00:13:38.326 I think there are many technologies that we might develop that are inspired 00:13:38.326 --> 00:13:45.566 by the rat whisker system, even if they're not 100% faithful to the way that 00:13:45.566 --> 00:13:46.926 the rat uses its whiskers. 00:13:47.826 --> 00:13:55.106 One example I can think of is what you might think of as a whisker paintbrush. 00:13:55.386 --> 00:14:00.666 So imagine a paintbrush with lots of bristles coming out, and you can paint 00:14:00.666 --> 00:14:04.066 this brush over objects, move this brush over objects. 00:14:04.226 --> 00:14:09.926 And the idea is that each of its bristles picks up mechanical data that allows 00:14:09.926 --> 00:14:14.246 us to reconstruct the three-dimensional features of that object. 00:14:15.406 --> 00:14:19.906 Essentially a tactile three-dimensional scanner. And so you're planning to build one of these yourself? 00:14:20.326 --> 00:14:26.426 We'd love to be able to build a whisker paintbrush, as we call it, a tactile paintbrush. 00:14:26.806 --> 00:14:31.006 And you previously built a whisking robot with a few whiskers, 00:14:31.006 --> 00:14:31.946 and this would have several. 00:14:32.686 --> 00:14:37.146 We've built a limited whisking robot that can move in. 00:14:37.146 --> 00:14:41.006 We only used it to move in one dimension and 00:14:41.006 --> 00:14:44.426 it was able to uh to determine 00:14:44.426 --> 00:14:51.826 uh an object's spatial features uh but this would be a high have more degrees 00:14:51.826 --> 00:14:55.526 of freedom you would be able to move this in more ways than we could move the 00:14:55.526 --> 00:14:59.586 previous version is the idea that the whiskers would move independently or they 00:14:59.586 --> 00:15:02.266 would just be on the it would it be literally a paintbrush with. 00:15:03.353 --> 00:15:08.193 I think in the initial prototype, it would literally be like a paintbrush where 00:15:08.193 --> 00:15:09.533 the bristles don't move independently. 00:15:09.993 --> 00:15:15.953 But a very interesting question we could answer that you're getting at is, 00:15:16.173 --> 00:15:21.433 could we extract more information if each of the whiskers were allowed to have 00:15:21.433 --> 00:15:27.613 its own degree of freedom and rotate independent of how the paintbrush itself was oriented? 00:15:27.613 --> 00:15:33.333 And so building this paintbrush could help us answer some questions about the 00:15:33.333 --> 00:15:35.673 biology, which is still unresolved, perhaps. 00:15:36.053 --> 00:15:43.813 I could imagine potentially that investigating independent whisker movement 00:15:43.813 --> 00:15:48.753 in the context of a robotic whisker array could help us answer some questions 00:15:48.753 --> 00:15:52.593 about what advantages there are to active whisking movements. 00:15:52.593 --> 00:15:58.353 And even suggest some hypotheses about how information from the viscous systems 00:15:58.353 --> 00:16:01.413 process in the brain? I suppose that that's true, yes. 00:16:02.313 --> 00:16:05.313 Okay, well, thanks very much for talking to us, Mitra. Thank you, Tony. 00:16:08.913 --> 00:16:14.553 The CSN podcast was produced by the Convergent Science Network of Biometrics 00:16:14.553 --> 00:16:21.293 and Biohybrid Systems, a project funded by the European Sevens Research Framework Programme. 00:16:22.593 --> 00:16:27.813 For more interviews, recorded lectures, or upcoming conferences in the field 00:16:27.813 --> 00:16:33.833 of biometrics and biohybrid systems, go to csnnetwork.com. 00:16:34.160 --> 00:16:42.320 Music.