 Okay, good morning. Welcome back. I realised with a start to surprise at the weekend that we were nearly finished. There seems to be no reason really why we shouldn't just keep going forever. But after all, they are finally back squarely confronting the problem of the relation between the name and the object. So, why stop? But we'll do one session in pollution today and then two on Russell and acquaintance on Wednesday and Friday. And did we discuss the format for the exam? Okay, well, here is how we plan to go with the final exam that on Wednesday we'll give out a set of 12 study questions. When you go into the exam, you will see six of those study questions, word for word, and you will have to choose three of them to answer. Does that make sense? Okay, we'll work over it again on Wednesday. Okay, so today is pollution and multiple object tracking. And we ended with Davidson and the question, is Davidson really right to say that you can say whatever you like about reference. You can say that the name Raleigh refers to Raleigh or you can say that the name Raleigh refers to the place one mile west of Raleigh or to the place two miles west of Raleigh or whatever so long as you make adjustments in what you say about the predicates like smokes or fishes or whatever. And I said, doesn't it really matter that you learn the name Raleigh by being introduced to Raleigh, not by encountering the place one mile west of Raleigh? Don't we really have to explain the role of perception in our learning of individual words? It's perception, surely there's some sense in which perception is basic to your understanding of representation. If we're going to think about the relation between the sign and the object, you have to get something about how seeing the object, sensing the objects in some way, perceiving the object is what lets you think about that particular thing. And pollution's work gives a very, in some ways it's a very simple framework for thinking about how perception allows you to think about objects. So I'll first of all just try and state what's going on with pollution's style of work on multiple object tracking and then say what it has to do with acquaintance and the questions about sense and reference and so on that we started with. And really, although we're going on to rustle in acquaintance on Wednesday and Friday, we'll be developing the same line of thought. So the pollution's an experimental psychologist and the basic kind of study here is you're showing a screen containing a number of simple identical objects. You've got four targets, they flash briefly. So four of the things in the screen flash to indicate their yours and then all the objects begin to move in unpredictable ways. So the scenario is something like this. That's your initial screen with display of eight objects on it. Four flash briefly saying they're yours. Then they all start moving and then at the end one flashes and you're asked, was that one of yours? And the thing is people turned out to be pretty good at this, not perfect but well above chance. And if you think about why this would be a real world skill, well, just for example, if you think of a football player, someone on a field with a whole bunch of other people, you've got to keep track of key individuals where they all are. If you think of someone with a family and a dog out shopping, then of course you need to keep clumped where everybody is. So these are very abstract displays, but this is a highly real world skill. I just want to listen to some Shakespeare. There's a battlefield with your key enemies with, as it might be, Ajax and the killies around you. But you need to keep tabs on these guys even though you and they are moving. So although this is the kind of artificial displays, there is a real world skill. It's practically very important to be able to keep tabs on where the things around you are even though they're moving. I want to show you some displays, but I have to say right off, I'm not the best person to be introducing you to these things. First of all because these are tasks, right? You're going to be asked at the end, is this one one of yours? And I myself can never ever do these tasks. I've looked at them just dozens of times and I just can't do it. I mean I have some rare neurological damage, but the results are that most of you guys will be able to do this. The other thing is, on the displays I'm going to show you, there aren't any answers in the back of the book as to what the right answers are. So we will have to decide right or wrong on a kind of Vikashtinyan basis, if you see what I mean. So anyway, with all that said, try this one. So can you see the eight targets? And try and focus, this point in the middle is just to fix it on, just to keep your eye in the centre of the screen. Ready? So four of them will flash to indicate you're yours and you then keep tabs on them through the rest of the experiment. It's only a few seconds. Was this one one of yours? Aha! Was this one one of yours? Was this one one of yours? Was this one one of yours? I'm not really in a position to assess your performance. It's not a bad idea, isn't it? So this is a fairly uncluttered display, all the dots keep out of each other's way. And the question just to have in mind as you're doing this is, how is it that you're doing it? Here's a simple task. You should be able to think about something and be able to do this. So these first couple are from Zen and Politions Lab and this elaborates the first demonstration a bit to ask, does it matter whether the objects that you're keeping tabs on go behind each other? Does it matter if they cross each other's paths? Was that one of yours? Yes. The top one there was one of yours? No. So you still do it fine, even though they're crossing each other's paths? Suppose they go behind barriers? Oops. Suppose they go behind barriers? Okay. Was this one one of yours? No. This one? No. This one? Yes. Okay. So including strips, don't matter. What about this one? This is all for regular solid physical objects. So here's Brian Skoll, a demonstration from Brian Skoll. So here you're going to get your four targets flash again and keep tabs on them. So these are kind of like liquids being poured here and there. Okay, was this one one of yours? No. Was this one one of yours? Yes. Well, you guys are doing much better than the regular population who find this virtually impossible. Right. Can you put your hand up if you feel less confident about this one than the other? Yeah. That's normal. It seems like the visual system is good at solid objects. It doesn't mind about them going behind barriers or crisscrossing each other. But keeping track of liquids like that is very difficult. And if you ask why that is, well, you might say, well, there they were all changing shape. So you could try something where the objects you have to keep track of change shape in the course of the thing, but they stay rigid. Let's try this one. Was that one one of yours? No. That one? Yeah. This is easy, right? This is just as easy as the first ones. Yeah, that's the... No? Look, I can't do any of them. I'd probably be more of a guide if I could. But what people say anyway is that there's a lot of change of shape going on here, but they stay rigid. That really doesn't affect your ability to keep track. It seems more like what's going on in the liquids case. Why the liquids are difficult is not that the liquids are changing shape, but that they're elongating, they're stretching out in such a way that it's very difficult to say what the position of the thing is. And if you say, where is a snake? What do you point to? Do you point to the head? Do you point to the middle bit? Do you see what I mean? It's very hard to look if you want to know exactly what to look if you want to look at the location of the snake because it's so stretched out. The liquids is two things. They're deforming in shape, but they're stretching out, too. And it doesn't seem to be the deforming in shape that matters. It seems more like the stretching out. So here's one more from Brian Skull. Ready? OK, what about that one? No. Ah-ha. By the usual Wiggish city in criteria, that was a very hard one. There's no right or wrong here. OK, I take it that your confidence level was less, though. Yes. Yes. So what was happening there was just, you've got the same kind of movement, but this stretching and the contracting makes it very, very hard. In some sense, you're keeping track of these things using location. And there, there isn't any such thing as location. So that's just to give a sense of there's a whole set of experiments here exploring how the visual system carves up things into objects. So these slinkies don't really count as objects or the liquids don't count as objects for the visual system, whereas rigid, coherently moving things do count as objects, even if they're being occluded. So I said this is a real-world skill, so I really shouldn't be a great surprise that people can do this. But if you think about how are you doing this, the general problem we had right from the start of the class was what's the connection between a sign and an object? So here you've got a special case of this, probably a pretty fundamental case of this. What's the relation between your perceptual representation of the world and that object out there? You were keeping track of four objects out there. How were you keeping track of them? Should you think in terms of a Freguian sense and reference? There's a way I was getting onto the object there, and I reckon specify what the sense was that you were using to lock onto the object. Well, the position's argument is you can't be keeping track of these objects by having descriptions of them. And if you think of a Freguian or Serl kind of description picture of how you lock onto a particular object here, I know these objects are very similar. The only thing that differentiates them is their location. But the argument is you can't be keeping track of these objects by saying the one that's over at the left here, the one that's furthest over at the left, then saying, and now that's the one that's further over here, now that's the one that's further over there, now that's the one that's further over there. Because you're having to keep track of four of them simultaneously, you can only really keep making explicit the locations of any one of those objects at a time. So if you try to get track of four of them, making explicit the locations as you went along, you'd have to be cycling back and forth. And these things are moving just a little bit too fast for you to be able to do that to keep cycling back and forth. And in fact, if at the end of the experiment I say, okay, is that one of yours, then although you guys are all clearly pretty good at doing that, saying which one is yours, if you're asked, what trajectory did that particular one go through in between the start point and the end? People in general can't say. You have no idea. Did that particular one loop round here, loop round there? Does that ring true? If you're probed and said, did it just stay where it was? Did it go right to the far corner? You don't know. Your vision doesn't answer that. The way the thing is being done is not by you having a description of the object that lets you keep track of it. Your visual system is locking on to the object in a more primitive way than that. Polition gives this example of Plastic Man who had very long plastic hands and could sprout extra appendages. If you want to keep track of four of the people in the class, you just go boom. Or if you think about a parent keeping track of a child by just putting a hand on their head, it's not that you're representing the child. You're keeping track of them, but not by representing anything about them. So if you had these plastic fingers, you could go blink and keep tabs on four people without representing anything about them. They could move about and at the end, you could still say which ones were yours. The idea is that vision is doing something like that in these multiple object tracking experiments. There are visual indexes in the visual system that can just lock onto objects in some very primitive way, keep tabs on them, and you can use them to find out about the characteristics of the object. So just one more thing about Polition's picture of what's going on. This is a somewhat comic. This may look forbiddingly technical. The first time you look at it, it's a somewhat comic, to my eye, model of Polition's for what the visual system is doing. The idea is here's early vision. Here's what's happening in early vision. Before you decide to attend to anything in particular, you just gape vacantly out of the room in front of you. You know the kind of thing. And then you decide, okay, I'll keep tabs on those ones, those are the troublemakers or whatever you're doing. So what you do that is these four buttons there are for the visual indexes, the four visual indexes that you keep tabs on particular objects. Now, they don't do that by making explicit any of the characteristics of those objects. In fact, those multiple object tracking tasks is very striking that the objects can change a lot. You can do real world versions of these where an object changes from being an animal of one kind to being an animal of another kind. And people just don't notice that it's changed. They still keep track of it just fine. So this level is more primitive than the level at which you know any of the properties of the object. You just got that lock onto the particular thing. And then what you do when you're finding out about the world is you push a button to say, let's find out more about that one. And then upstream, there are all these different kind of features you might be looking for like, what color is it? What shape is it? Is it tilted or upright? Is it horizontal? Is it square? So there are all these different characteristics you can ask about for any particular object. So in visually interrogating the room, what you do is you push one button, you choose one object, and then you push another button to ask, does it have this characteristic? So if you're just sitting here and the lecture fails to grip and you're looking at the people around you and you're thinking, oh, look at that one, look at that one, then what you do is you ask, okay, what are they? You choose that person and you say, okay, what are they wearing? How tall are they? Well, whatever your questions are. As Polition Wisely says, you need two hands to operate this machine. There are two of these two things, these two kinds of buttons, right? So there are two things that you're doing when you visually look over a scene. One is you visually lock onto the object and the other is you interrogate the scene to ask, you interrogate that object to ask what characteristics does it have? That's the basic structure of you getting knowledge about your surroundings. And what these experiments are doing is starting to probe what notion of object we have here. So the picture is there are four or five visual indexes in your visual system, each of which can be assigned a particular scene object. And the visual index is a very primitive kind of tagging of the object. It's something that's going to be causally responsive to the object, but it doesn't involve any kind of description of the object. His position is a way to name or refer to individual parts of a scene independent of their properties or locations. So there's nothing like a freguin sense here. It's really primitive, it's really simple what the visual system is doing, giving you that simple lock onto the parts of a scene. So I'm not sure how much this diagram adds, but the picture is there are these causal connections from the objects out there through to these central structures in the brain via the visual indexes, X, Y, Z, and so on here. And these in virtual, the causal connections to the indexes, they represent the particular parts of the world outside. This has actually turned out to be quite a fruitful idea. This idea of visual indexes. A simple way of starting to see the broader significance is to say, suppose you've got five pennies thrown in front of you. Can you see at a glance how many pennies there are? Yes. If you've got three, four, five, you can do that at a glance, say I get three, four, five there. If you're throwing 17 pennies and asked how many are there there, can you do that at a glance? Well, you'd be unusual if you could, right? Most people can't do that at all, you just have to count. Why is that? How come you can do up to four or five without having to count? But once you get beyond that, you do have to count. What's going on? How did that? Why is that? Well, the answer suggested here is, if you've got enough visual indexes that you can just throw them at the stuff in front of you, that lets you do a very simple type of counting, a very immediate kind of counting. So up to four or five, the number of your visual indexes, you can give the number without having to count. Beyond that, your visual indexes give out, so you do have to count. It used to be said, I have no idea whether this is true or not, but I've often had it said to me, there are these people who can only count up to three. Does that story still go up? Maybe it was only me that was told that story. But whether or not that's true for any culture is certainly through children learning number of words. There is a point at which they can count up to three, four, five, and then it just stops. After that, the most you get is a lot. So one of the basic phases in learning to count is getting beyond that barrier of one, two, three, four to a lot and getting to be able to count systematically to any arbitrary number. And one diagnosis of what's going on there is that at an early stage, children are just using their visual indexes to count. And then they run into this real brick wall beyond which you can't use the visual indexes and they have to use some completely different strategy to learn the meanings of words like 17. Or if you take learning the object concept, one basic thing that children have to do is get the concept of a physical object. And the traditional answer to how children do this is by learning language. That's how you get the concept of an object. You learn a concept like woman or man or tree or chair and that's what allows you to carve out the tree or the person from the surroundings. You will still hear that said as if it's obviously true in many, many areas that our conceptual scheme carves out the world for us. What these kind of experiments are suggesting is that that's not the right way to think of it. The visual system itself, the thing that you get just as a generic human being is a system that is already carving up the world into objects before you have any concepts. So that what these experiments are doing is they're probing the question, what is your visual system? Count as an object. And by this criterion, Raleigh will count as an object. One of the first things that children get is the ability to track people. I mean, because people are clearly the most important things in a child's early life. So Raleigh will count as an object. The place one mile west of Raleigh will not count as an object. What the visual system has got is this capacity to hook on to Raleigh rather than the place one mile west of Raleigh. And that's more primitive than any use of concepts you might make. Your whole use of language to refer to individual things in your surroundings depends on that primitive perceptual capacity. Okay, so there's a rapid initial introduction to multiple object tracking and visual indexes. Okay. Okay. So remember when Russell was talking about reference? I mean, pollution really makes this connection to Russell, but I know it's a long time ago now. But Russell was arguing there has to be a basic class of names that gets tied up to objects, but not because they're defined in terms of descriptions. Why was that? What was Russell's objection? What was Russell's objection to the description theory of reference? I do think these terms are unreasonably long. Okay. Remember the distinction between how many terms and names? Suppose you get a how many phrase like exactly one person. Does that refer to anyone? No, that is not referred to anyone. So if you say exactly one person wrote Waverly and anyone who wrote Waverly was Scotch, have you referred to anyone? If you take the description, the author of Waverly was Scotch, does that mean the same as exactly one person wrote Waverly and anyone who wrote Waverly was Scotch? Yes, that's the right answer. Very good. Not a very confident answer, but yeah. That was Russell's theory of description. The author of Waverly was Scotch means exactly the same thing as exactly one person wrote Waverly and anyone who wrote Waverly was Scotch. Yes? So Russell said it can't be right to analyse names in terms of descriptions because there has to be a more basic level of reference than one that involves descriptions. So if you've got something like Scotch wrote Waverly and Scotch there is really a genuine proper name, then it can't be making its connection to the object by way of a description. But then if names are more basic than descriptions, what is it that's fixing the reference of a name? What does Russell say? Very good. Acquaintance. Yeah, something about acquaintance. That's a fair summary. Something about acquaintance is what fixes the reference of the name. There's some notion of a really primitive kind of prepropositional connection between the subject and the thing out there that doesn't go by way of a description, doesn't go by way of knowing any truths about the object out there, is just some primordial link between you and it. So does pollution's account suggest any picture of what acquaintance might be? Yes, this kind of thing is a picture of what acquaintance might be. Before you know any of the properties of the object, before you're able to say anything about what characteristics the object has, you've got this level at which you have this simple tagging of the object, this simple locking onto the object by the visual system. And that is kind of intuitive. If you just look around you right now, if you just let your gaze roam lightly over the room, landing on one object now on another, that's what happens is that your gaze lands on one object and then you can interrogate to say, to vision, let me find out a bit more about that thing. So that picture is giving a way of being more explicit about what acquaintance is. And that after all is just what pollution said, that indexing is a way of naming or referring to individual parts of a scene independent of their properties or locations. And just as Russell thought, acquaintance is more primitive than knowledge of truths about the thing, but still acquaintance is the foundation for your knowledge of truths about the thing. And that's what that suggests. The kind of connect that you have here is the foundation, being able to push this button for a particular object is the found, isn't itself a matter of knowing any truths about the thing. Then when you press the further buttons and say now which characteristics does it have, that's where your knowledge of truths about the thing is coming from. So we can think of tagging as being what Russell meant by acquaintance, logically proper names, are just the same as visual indexes. Those visual indexes are logically proper names. I mean, where it all goes so awfully wrong in Russell is when he says that you can only be acquainted with your own sense data and that leaves each of us trapped in a world of their own sense data and makes communication possible. But if you think of visual indexes as logically proper names, then you have a way of making sense of what Russell's saying, but you don't cut us off from the medium-sized world and you don't cut us off from communication because each of us can be visually indexing the same things and we can know that that's what's going on. My list of topics seems to be changing. As you go through this, but there you go, that's the creative process. Is that reasonable sense? Do you remember Russell well enough to see the connect there? That's what's important. I think so engaging about these multiple object tracking experiments that what it feels like when you're doing it is not that you're keeping tabs on the thing by having a description of it. It feels like you're doing something much more basic. Does that ring true? One thing you can ask is when you're thinking of tagging an object, just keeping tabs on an object, then you can make sense of the notion of a mistake. Keeping tabs on an object is not something at the level of knowing a proposition. If you're just following someone, if after the class I'm tracking someone, then you might get muddled up. There might be a substitution of one person for another in the course of the tracking. So another object might cause you to form a representation involving the tag. Even though this isn't a matter of knowing propositions, you can make sense of going right or wrong here, succeeding in keeping track or failing in keeping track. But what does it mean to have made a mistake when you're using a visual index? To have made a mistake when you're using a tag? Well, at this point, I think we could go right back to Dretschke on biological functions. I mean, what you've got here in the visual system is something that's evolved, something that has an adaptive point. The point of you having a visual system with visual indexes in it is to let you keep track of objects. What the experiments seem to be showing is that it's not to let you keep track of liquids, it's not to let you keep track of objects that are extending and contracting the whole time. There's a conception of a rigid, medium-sized object that the thing is allowing you to keep track of. That's what your visual system is for. So you can think of the biological function of a visual index as being to indicate where an assigned object is. Yeah? Unless the object is in front of them. That's right. Render remains only being acquainted in such a manner when it's direct interaction with the object. That's right. It seems like Russell in that way. There is that limit on what each of us can be acquainted with. But for learning a language and for the basis of communication, it's really arguable. I mean, isn't pointing something like the basis of communication? So when you're teaching a child, when a child's learning language, pointing is really so basic. Yeah? But what pointing is a way of drawing attention to things and drawing attention to things on this kind of model. Oops. Drawing attention to things on this kind of model is just a matter of tagging them with one of those visual indexes. So if I say, look at that chair, then what you do is you tag it instantly. So you're now in a position to interrogate for more information and say, what about it? What about it? Yeah? So it's true that this is limited to what's directly in front of us. But it's not like Russell because in Russell, I could only be acquainted with my own sense data and you couldn't know anything about that stuff. Yeah? So here we've got like a limited initial domain that we do have in common. It's the basis in which communication can get going. Yeah? So you write it, it only goes a little way beyond Russell, but it's quite important to have that little way. Or suppose you do lose track of an object and when you're watching it, you confuse it with another object, then that other object is making you form representations involving the tag. So if I'm trying to keep tabs on you and then I've lost you and now following someone else instead, then my causal connection between that other person and me is only there because of the original causal connection between you and me. Does that make sense? So if I'm... I mean something very simple here. If it was me and you and I'm trying to follow you, I'm trying to follow you so you're causally impacting on me and I'm trying to keep you in view so you keep causally impacting on me and then at a certain point I get you muddled up with the other and I start following the other. Yeah? That causal connection with the other is only there because of the original causal connection to you. So you could say that causal relation between the other and me depends on the causal relation between you and me. That's an asymmetric dependence just as Fodor said. So we can think of these visual indexes as just exemplifying the kind of thing that Dretschke and Fodor were talking about. Here you've got a real simple example of how you can have a primitive level of reference of the kind that Russell was describing that illustrates Dretschke and Fodor's points. You've got a system here that was evolved to allow you as biological function was to allow you to keep tabs and objects and when you're making a mistake there will be a causal connection between you and the other object that is asymmetrically dependent on the original causal connection between the object and you. So you think about how this plays out for Frege's sense and reference and remember Frege's picture you've got the sign, the sense and the reference. Oh yes, happy days, happy days. It was so sunny back then. Yeah, so you had the sign and the sense and the reference and Polition's picture is now doing away with the sense and saying at this basic level of visually tagging things around you, we don't have sense we just have the sign and the reference but remember there were two basic problems for that picture. One was informative identities and we'll discuss informative identities further on Wednesday and Friday and how are you going to make sense of informative identities if you don't have descriptions associated with the sign and what about meaning without reference? Can't you have meaning without reference? Well, the thing is that with the visual system we're talking about something very primitive if you think about what you were doing with those multiple object tracking experiments as you watch the things moving around you've exhausted your supply of visual indexes but within vision itself it doesn't make sense there's no way you can even frame the question is this object that object? You see what I mean? You're just going to wear keeping tabs on these four independent objects if you reflect in the situation later and you can frame descriptive questions like was the one that started out in the right hand corner one in the same as the one that started out in the middle could there have been two different ways of looking at the same object but you can't ask that within vision itself you can't even formulate identity questions involving visual indexes so Frig's first problem just doesn't arise there isn't a problem about informativeness at the level of the visual system so here we've got a system in which visual indexes refer to external things where the visual system can't formulate informative identities involving them and what about a visual index that didn't manage to tag anything there was nothing there for it to tag well a visual index that didn't tag anything just doesn't have any meaning doesn't have any significance so you could just bite the bullet in that there and say there is no meaning for a visual index that hasn't been assigned to a particular object it has a potential to have meaning if you now do assign it to a particular object but so long as it's not been assigned to a particular object it doesn't have meaning so visual indexes really seem like they are true, resilient, proper names you don't need a sense you don't need a descriptive mode of identification because you can't frame identities involving them they can't have meaning unless there's something there you can think of them as getting assigned to the object by having a causal connection set up between the object and the visual system so they really seem like a way in which you can think of the ability to represent particular objects around you as being grounded in perception we'll go on to Wittgensteinian resonances of this picture and the role of sensory consciousness of the object next time and we'll also tie up a loose end from our discussion of Twin Earth onwards! okay, thank you