WEBVTT 00:00:01.640 --> 00:00:04.040 The following content is provided under a Creative 00:00:04.040 --> 00:00:05.580 Commons license. 00:00:05.580 --> 00:00:07.880 Your support will help MIT OpenCourseWare 00:00:07.880 --> 00:00:12.270 continue to offer high-quality educational resources for free. 00:00:12.270 --> 00:00:14.870 To make a donation or view additional materials 00:00:14.870 --> 00:00:18.830 from hundreds of MIT courses, visit MIT OpenCourseWare 00:00:18.830 --> 00:00:22.049 at ocw.mit.edu. 00:00:22.049 --> 00:00:23.465 LIZ SPELKE: I want to talk about-- 00:00:23.465 --> 00:00:27.506 you asked about development of knowledge within infancy-- 00:00:27.506 --> 00:00:28.880 and I want to talk about one case 00:00:28.880 --> 00:00:31.088 where we've seen an interesting developmental change. 00:00:31.088 --> 00:00:33.687 One that I don't think we really fully understand. 00:00:33.687 --> 00:00:35.270 I'm sure we don't fully understand it. 00:00:35.270 --> 00:00:37.620 I'm not sure we understand it at all. 00:00:37.620 --> 00:00:39.620 But it seems to be there. 00:00:39.620 --> 00:00:43.520 And it has to do with effects of both, maybe, 00:00:43.520 --> 00:00:45.140 inertial properties of object motion 00:00:45.140 --> 00:00:48.200 and also effects of gravity on object motion. 00:00:48.200 --> 00:00:50.090 Let me just tell you the result first. 00:00:50.090 --> 00:00:52.850 These are studies that were done by In-Kyeong Kim 00:00:52.850 --> 00:00:54.770 a long time ago. 00:00:54.770 --> 00:00:57.020 But recently enough, that video was 00:00:57.020 --> 00:01:00.650 part of our toolkit, which wasn't always the case, 00:01:00.650 --> 00:01:04.670 and involved showing babies videotapes of events 00:01:04.670 --> 00:01:06.980 in which an object held by a hand 00:01:06.980 --> 00:01:12.130 was placed on an inclined plane, released, and started to move. 00:01:12.130 --> 00:01:14.870 And in one study, babies either saw 00:01:14.870 --> 00:01:18.110 a plane that was inclined downward, the hand released it, 00:01:18.110 --> 00:01:22.250 and it rolled downward with the natural acceleration, 00:01:22.250 --> 00:01:26.660 or the plane was inclined upward, the hand released it, 00:01:26.660 --> 00:01:29.660 and it rolled upward, decelerating. 00:01:29.660 --> 00:01:30.470 OK? 00:01:30.470 --> 00:01:32.090 We studied this in adults as well, 00:01:32.090 --> 00:01:34.490 and for adults, adults reported that it 00:01:34.490 --> 00:01:37.107 looked like this hand set the ball in motion. 00:01:37.107 --> 00:01:38.690 In fact, to control the motion better, 00:01:38.690 --> 00:01:42.980 there was a sling shot apparatus that actually set it in motion. 00:01:42.980 --> 00:01:46.250 But it underwent what looked like a natural deceleration 00:01:46.250 --> 00:01:51.959 from the point at which the hand released it and as it rose up. 00:01:51.959 --> 00:01:53.750 Half the babies were bored with this event, 00:01:53.750 --> 00:01:55.490 and half were bored with that. 00:01:55.490 --> 00:01:59.270 And then, at test, we switched the direction 00:01:59.270 --> 00:02:03.140 of the orientation of the ramp so 00:02:03.140 --> 00:02:06.140 that if babies had been seeing something move downward, now 00:02:06.140 --> 00:02:09.110 they were seeing events in which it moved upward. 00:02:09.110 --> 00:02:10.889 And in one of those two events, it 00:02:10.889 --> 00:02:13.010 underwent the natural acceleration 00:02:13.010 --> 00:02:16.620 that adults would expect, which is to say, 00:02:16.620 --> 00:02:19.670 a change in the motion that the infants had previously seen. 00:02:19.670 --> 00:02:22.100 So these guys saw this thing rolling downward 00:02:22.100 --> 00:02:23.360 and accelerating. 00:02:23.360 --> 00:02:25.100 And here in the natural event, they're 00:02:25.100 --> 00:02:27.200 seeing something that's being propelled upward, 00:02:27.200 --> 00:02:28.610 and it's decelerating. 00:02:28.610 --> 00:02:29.840 OK? 00:02:29.840 --> 00:02:33.080 In this event, they saw the same motion pattern 00:02:33.080 --> 00:02:37.160 that they saw before, so if they saw speeding up here, 00:02:37.160 --> 00:02:41.720 they saw speeding up here, contrary to effects of gravity 00:02:41.720 --> 00:02:42.620 on object motion. 00:02:42.620 --> 00:02:43.160 OK. 00:02:43.160 --> 00:02:44.720 So the question was, if the babies 00:02:44.720 --> 00:02:46.550 were sensitive to effects of gravity, 00:02:46.550 --> 00:02:49.280 they should look longer when the thing accelerates, 00:02:49.280 --> 00:02:51.610 you know, speeds up as it's moving upward 00:02:51.610 --> 00:02:54.369 than they do when it slows down as it's moving upward. 00:02:54.369 --> 00:02:56.660 On the other hand, if they're not sensitive to gravity, 00:02:56.660 --> 00:02:59.060 and they're just representing these events 00:02:59.060 --> 00:03:00.560 as speeding up or slowing down, they 00:03:00.560 --> 00:03:01.893 might show the opposite pattern. 00:03:01.893 --> 00:03:03.080 OK? 00:03:03.080 --> 00:03:05.960 And that's actually what we found in both conditions 00:03:05.960 --> 00:03:06.710 of that study. 00:03:06.710 --> 00:03:11.690 At five months of age, babies showed the pattern opposite 00:03:11.690 --> 00:03:15.440 to what adults would expect, as if they thought 00:03:15.440 --> 00:03:17.480 an object that speeds up in the beginning 00:03:17.480 --> 00:03:20.810 is going to continue speeding up irrespective of the orientation 00:03:20.810 --> 00:03:23.180 of that plane, with respect to gravity, OK? 00:03:23.180 --> 00:03:28.090 So failure at five months, success at seven months. 00:03:28.090 --> 00:03:29.870 At seven months, we get a reversal, 00:03:29.870 --> 00:03:33.930 and they flip in their looking times here. 00:03:33.930 --> 00:03:36.470 Well, we thought, maybe the 5-month-olds 00:03:36.470 --> 00:03:38.720 will succeed at a simpler task. 00:03:38.720 --> 00:03:40.550 Suppose we start out just showing 00:03:40.550 --> 00:03:42.500 them a flat surface and constant speed, 00:03:42.500 --> 00:03:46.850 so we're not engendering any expectation 00:03:46.850 --> 00:03:48.860 that this object is going to speed up over time 00:03:48.860 --> 00:03:50.090 or slow down over time. 00:03:50.090 --> 00:03:51.680 They're just seeing a constant motion, 00:03:51.680 --> 00:03:55.040 and then they get tested with events 00:03:55.040 --> 00:03:57.260 in which the object is placed either at the top 00:03:57.260 --> 00:03:58.970 or at the bottom of this ramp. 00:03:58.970 --> 00:04:01.490 It's released, and it speeds up in both cases. 00:04:01.490 --> 00:04:06.080 In that case, will they view this motion as more natural 00:04:06.080 --> 00:04:08.330 and look longer at that one? 00:04:08.330 --> 00:04:11.060 Or, you might ask, would they show the opposite pattern 00:04:11.060 --> 00:04:14.060 [AUDIO OUT] if these events are difficult for them, 00:04:14.060 --> 00:04:15.320 show the opposite pattern? 00:04:15.320 --> 00:04:17.089 What they actually showed was absolutely 00:04:17.089 --> 00:04:20.180 no expectation, the 5-month-olds, in that case. 00:04:20.180 --> 00:04:21.170 OK? 00:04:21.170 --> 00:04:24.770 They seemed utterly uncommitted as to whether it 00:04:24.770 --> 00:04:27.020 was more natural for the thing to be moving downward 00:04:27.020 --> 00:04:29.270 than to be moving upward after seeing 00:04:29.270 --> 00:04:32.270 it move on this flat surface. 00:04:32.270 --> 00:04:37.850 Then, at seven months, kids responded as adults would. 00:04:37.850 --> 00:04:40.280 Then we thought, maybe there's just a problem with video. 00:04:40.280 --> 00:04:43.700 Maybe kids don't understand that vertical in a video 00:04:43.700 --> 00:04:46.082 corresponds to vertical in the real world. 00:04:46.082 --> 00:04:47.540 So we just ran a control experiment 00:04:47.540 --> 00:04:50.000 to see if that was true, by familiarizing kids 00:04:50.000 --> 00:04:51.770 with downward motion in the real world 00:04:51.770 --> 00:04:54.070 and then presenting those same two test events. 00:04:54.070 --> 00:04:57.237 And now they showed the pattern that adults would show. 00:04:57.237 --> 00:04:59.570 That's not showing any knowledge here, it's just saying, 00:04:59.570 --> 00:05:01.570 yeah, they can distinguish downward from upward, 00:05:01.570 --> 00:05:03.650 and they can relate real events to video. 00:05:03.650 --> 00:05:07.280 But they don't seem to have any prior expectation that objects 00:05:07.280 --> 00:05:11.090 will move downward when they're rolling on an inclined plane. 00:05:11.090 --> 00:05:13.790 And all these things seem to develop between five and seven 00:05:13.790 --> 00:05:15.200 months, which I think provides us 00:05:15.200 --> 00:05:18.625 with an interesting window for asking what's developing here. 00:05:18.625 --> 00:05:20.330 One possibility is what's developing 00:05:20.330 --> 00:05:21.800 is something very local. 00:05:21.800 --> 00:05:24.020 But I don't think so, because there's 00:05:24.020 --> 00:05:28.160 another change also occurring between five and seven 00:05:28.160 --> 00:05:30.967 months in what looks like a very similar situation. 00:05:30.967 --> 00:05:33.050 These are experiments that were conducted in Renée 00:05:33.050 --> 00:05:36.620 Baillargeon's lab, where she did these very simple studies where 00:05:36.620 --> 00:05:38.600 you'd have a single object in an array, 00:05:38.600 --> 00:05:41.060 and then a hand would come in holding another object 00:05:41.060 --> 00:05:42.740 and place it on that array. 00:05:42.740 --> 00:05:45.890 And in that top study, she compares infants' reactions 00:05:45.890 --> 00:05:49.820 when the hand places the object on top of the box, releases it, 00:05:49.820 --> 00:05:51.890 and it stays there, versus places 00:05:51.890 --> 00:05:54.710 the object on the side of the box, releases it, 00:05:54.710 --> 00:05:55.970 and it stays there. 00:05:55.970 --> 00:05:57.290 OK? 00:05:57.290 --> 00:05:58.880 Now, if you're a 3-month-old infant, 00:05:58.880 --> 00:06:00.800 you are equally happy with those two events. 00:06:00.800 --> 00:06:02.690 You do not look differentially at them. 00:06:02.690 --> 00:06:05.180 But by five months, infants do. 00:06:05.180 --> 00:06:09.710 They look longer when the object seems to be stably supported 00:06:09.710 --> 00:06:12.500 by the side of the box, than when it's stably supported 00:06:12.500 --> 00:06:14.160 by the top of the box. 00:06:14.160 --> 00:06:18.620 But then she goes on and asks, do infants 00:06:18.620 --> 00:06:20.950 make distinctions among objects that are stably 00:06:20.950 --> 00:06:22.100 supported from below? 00:06:22.100 --> 00:06:24.725 As, by the way, they were in all those studies with the rolling 00:06:24.725 --> 00:06:26.319 on the ramps, right? 00:06:26.319 --> 00:06:27.860 Do they make the kind of distinctions 00:06:27.860 --> 00:06:29.540 that we would make based on the mass 00:06:29.540 --> 00:06:31.610 of the object, the physical force on the object, 00:06:31.610 --> 00:06:32.180 and so forth? 00:06:32.180 --> 00:06:34.382 And to get at that, she did a second study 00:06:34.382 --> 00:06:35.840 where she has a box, and she either 00:06:35.840 --> 00:06:38.540 places it in the middle of the top box or way 00:06:38.540 --> 00:06:41.270 off to the side where we would expect it to fall, OK, 00:06:41.270 --> 00:06:42.920 not to be stably supported. 00:06:42.920 --> 00:06:45.650 7-month-olds get that right, 5-month-olds do not. 00:06:45.650 --> 00:06:47.622 So across these different situations, 00:06:47.622 --> 00:06:49.580 it looks like we have this developmental change 00:06:49.580 --> 00:06:52.230 in what infants know. 00:06:52.230 --> 00:06:55.320 I mean, I think these studies raise a lot of questions 00:06:55.320 --> 00:06:57.690 that they don't answer about what infants are 00:06:57.690 --> 00:06:59.887 learning over this time period. 00:06:59.887 --> 00:07:01.470 I'm excited about that because I think 00:07:01.470 --> 00:07:06.080 they also give us a method for addressing those questions. 00:07:06.080 --> 00:07:06.990 OK. 00:07:06.990 --> 00:07:11.475 And Tomer has been working on chasing that [AUDIO OUT],, 00:07:11.475 --> 00:07:12.560 which would be great. 00:07:12.560 --> 00:07:13.170 OK. 00:07:13.170 --> 00:07:15.697 So I've been asked already by a bunch of you, 00:07:15.697 --> 00:07:18.030 what happens at the very beginning of visual experience? 00:07:18.030 --> 00:07:19.304 I do have some slides on that. 00:07:19.304 --> 00:07:20.970 I do want us to take a break, but let me 00:07:20.970 --> 00:07:24.460 go through them very quickly. 00:07:24.460 --> 00:07:27.420 There's been a little bit done with newborn infants, 00:07:27.420 --> 00:07:30.870 and where they've been studied under conditions where 00:07:30.870 --> 00:07:34.620 we are confident that they're able to see what they're 00:07:34.620 --> 00:07:36.000 being presented with. 00:07:36.000 --> 00:07:37.840 It looks as if some of these-- 00:07:37.840 --> 00:07:40.476 there's evidence for these abilities in newborns, 00:07:40.476 --> 00:07:41.850 but most of the things I told you 00:07:41.850 --> 00:07:44.220 about have not been tested in newborn infants 00:07:44.220 --> 00:07:46.350 and would be really hard to test in them. 00:07:46.350 --> 00:07:49.410 But fortunately, humans aren't the only animals 00:07:49.410 --> 00:07:53.190 that have to be able to find the objects in a scene 00:07:53.190 --> 00:07:54.690 and track them through time. 00:07:54.690 --> 00:07:57.030 Other animals do that as well. 00:07:57.030 --> 00:08:01.260 And many animals seem to succeed at representing objects 00:08:01.260 --> 00:08:03.840 under the conditions where infants succeed. 00:08:03.840 --> 00:08:05.970 The big problem we have with animal models 00:08:05.970 --> 00:08:07.590 is that in many cases, the animals 00:08:07.590 --> 00:08:09.510 are way more competent than the infants. 00:08:09.510 --> 00:08:11.940 And they succeed in cases where infants would fail. 00:08:11.940 --> 00:08:12.540 OK? 00:08:12.540 --> 00:08:14.081 But at least where they succeed where 00:08:14.081 --> 00:08:15.750 infants succeed, we can ask, what would 00:08:15.750 --> 00:08:17.400 happen with controlled rearing? 00:08:17.400 --> 00:08:20.110 And can we at least get an existence proof 00:08:20.110 --> 00:08:22.680 that in a mechanism of the sort, that the abilities 00:08:22.680 --> 00:08:24.930 that we see in infants can be performed 00:08:24.930 --> 00:08:29.040 by some nervous system on first encounters 00:08:29.040 --> 00:08:32.100 with visible objects and types of events 00:08:32.100 --> 00:08:34.620 that it's now being asked to reason about? 00:08:34.620 --> 00:08:38.510 So this has been done with controlled reared experiments. 00:08:38.510 --> 00:08:40.590 I'm going to talk about just a few, 00:08:40.590 --> 00:08:42.659 show you the results of just a few experiments 00:08:42.659 --> 00:08:44.250 that were done on chicks. 00:08:44.250 --> 00:08:46.140 Chicks are being used a lot lately 00:08:46.140 --> 00:08:50.580 because they grow up pretty quickly, they're easy to raise, 00:08:50.580 --> 00:08:54.930 and they show this innate behavior toward objects. 00:08:54.930 --> 00:08:56.610 That gives us a nice indicator, which 00:08:56.610 --> 00:08:59.010 is in some ways a little bit like preferential looking, 00:08:59.010 --> 00:09:01.470 though opposite in sign. 00:09:01.470 --> 00:09:04.532 They imprint if you show them an object repeatedly, 00:09:04.532 --> 00:09:06.240 and a chick has been isolated, so there's 00:09:06.240 --> 00:09:09.930 no other chicks or hens around in its environment. 00:09:09.930 --> 00:09:13.170 The object is the only moving object they see. 00:09:13.170 --> 00:09:18.720 They will imprint to it, treat it like another-- 00:09:18.720 --> 00:09:22.680 behave as they would with their mom, were she there. 00:09:22.680 --> 00:09:25.080 And in particular, if you then take the chick 00:09:25.080 --> 00:09:27.180 and put them in a novel environment 00:09:27.180 --> 00:09:28.890 where they're a little bit stressed, 00:09:28.890 --> 00:09:31.100 they will tend to approach that object. 00:09:31.100 --> 00:09:35.130 So this has the same logic as looking longer 00:09:35.130 --> 00:09:35.910 at a novel thing. 00:09:35.910 --> 00:09:38.130 You have a selective approach to a familiar thing, 00:09:38.130 --> 00:09:41.580 and you can now run, on chicks, the kinds of experiments that 00:09:41.580 --> 00:09:44.190 have been run on human infants. 00:09:44.190 --> 00:09:46.724 So here is one imprinted chick to a-- 00:09:46.724 --> 00:09:48.390 these are experiments that have actually 00:09:48.390 --> 00:09:52.020 been done a while ago, in some case they're much more 00:09:52.020 --> 00:09:55.230 recent, imprint a chick to a triangle, 00:09:55.230 --> 00:09:58.230 and then present them with a triangle whose center is 00:09:58.230 --> 00:10:00.210 missing, either because it's occluded 00:10:00.210 --> 00:10:02.430 or because there's a gap there. 00:10:02.430 --> 00:10:03.946 Who's mom in this case? 00:10:03.946 --> 00:10:05.820 The one with the occluded center, not the one 00:10:05.820 --> 00:10:07.320 with the gap at the center. 00:10:07.320 --> 00:10:07.980 OK? 00:10:07.980 --> 00:10:09.990 Consistent with findings with infants 00:10:09.990 --> 00:10:14.250 in one way, these chicks are perceiving occlusion, 00:10:14.250 --> 00:10:16.350 but they're doing better than the infants. 00:10:16.350 --> 00:10:18.944 This works when the objects move behind the occluder. 00:10:18.944 --> 00:10:20.610 If you move objects behind the occluder, 00:10:20.610 --> 00:10:22.260 you get all the same kind of motion effects 00:10:22.260 --> 00:10:23.880 that I was talking about with infants. 00:10:23.880 --> 00:10:25.990 But it also works if the object is stationary. 00:10:25.990 --> 00:10:30.160 So the chicks are better than the infants in that case. 00:10:30.160 --> 00:10:31.800 What about object permanence? 00:10:31.800 --> 00:10:33.810 Here's a task that babies can't solve 00:10:33.810 --> 00:10:36.390 until they're 12 months old-- 00:10:36.390 --> 00:10:40.950 well, eight months in this case, 12 months in this case-- 00:10:40.950 --> 00:10:43.420 that chicks solved the first time 00:10:43.420 --> 00:10:47.560 they're presented with an imprinted object that moves out 00:10:47.560 --> 00:10:48.060 of you. 00:10:48.060 --> 00:10:49.976 I should have said that on the previous slide. 00:10:49.976 --> 00:10:53.730 The chicks never saw occlusion until the imprinting test. 00:10:53.730 --> 00:10:57.750 Here, they're imprinted to mom on the first two days of life, 00:10:57.750 --> 00:10:59.970 but there's no other surfaces in the environment, 00:10:59.970 --> 00:11:02.610 so they never see her occluded by another surface. 00:11:02.610 --> 00:11:05.410 Maybe they see her occluded by their own wing or something, 00:11:05.410 --> 00:11:08.070 but not by other objects in the environment. 00:11:08.070 --> 00:11:08.940 OK. 00:11:08.940 --> 00:11:10.050 There's two screens there. 00:11:10.050 --> 00:11:12.210 A chick is restrained in a Plexiglas box 00:11:12.210 --> 00:11:15.060 and sees mom disappear behind one of the screens. 00:11:15.060 --> 00:11:19.350 They will go and search behind that screen for her. 00:11:19.350 --> 00:11:21.570 Only at 12 months do human infants 00:11:21.570 --> 00:11:23.130 solve the following problem. 00:11:23.130 --> 00:11:25.380 Mom has disappeared behind and been 00:11:25.380 --> 00:11:28.440 found behind this screen five times in a row. 00:11:28.440 --> 00:11:31.590 Now she goes there, where will chicks search? 00:11:31.590 --> 00:11:34.260 A baby, until 12 months, will search behind the first place 00:11:34.260 --> 00:11:37.254 where she was hidden, where they found her in the past. 00:11:37.254 --> 00:11:40.770 A chick is more like a 12-month-old baby, 00:11:40.770 --> 00:11:42.101 shows the more mature pattern. 00:11:42.101 --> 00:11:43.350 He goes where mom actually is. 00:11:43.350 --> 00:11:44.970 AUDIENCE: Are these images or actual moms? 00:11:44.970 --> 00:11:46.386 LIZ SPELKE: These are actual moms. 00:11:46.386 --> 00:11:48.480 This is imprinting to either a ping pong 00:11:48.480 --> 00:11:51.620 ball in some of these studies or a cylinder in other studies, 00:11:51.620 --> 00:11:53.600 dangles on a wire during the imprinting period 00:11:53.600 --> 00:11:55.190 so you can see that it moves. 00:11:55.190 --> 00:11:57.740 It moves here, and it kind of dangles 00:11:57.740 --> 00:12:00.061 and moves behind one screen or another screen. 00:12:00.061 --> 00:12:02.060 And then the chick is released, and the question 00:12:02.060 --> 00:12:03.325 is where will the chick go? 00:12:03.325 --> 00:12:03.825 Yeah. 00:12:03.825 --> 00:12:06.249 AUDIENCE: Are these other chicks? 00:12:06.249 --> 00:12:07.790 LIZ SPELKE: I think they imprint them 00:12:07.790 --> 00:12:09.470 for two days post hatching. 00:12:09.470 --> 00:12:11.970 So they're hatched in an incubator. 00:12:11.970 --> 00:12:14.600 In the killer study I'm going to tell you about, 00:12:14.600 --> 00:12:16.280 the incubator's in the dark. 00:12:16.280 --> 00:12:18.647 They spend two days in-- 00:12:18.647 --> 00:12:20.480 I think they spend a day in the dark getting 00:12:20.480 --> 00:12:22.040 used to just walking around. 00:12:22.040 --> 00:12:24.080 And then starts the visual experience, 00:12:24.080 --> 00:12:26.150 and it's controlled, so no occlusion. 00:12:26.150 --> 00:12:30.080 But you do see this one object that they get to imprint you, 00:12:30.080 --> 00:12:32.080 and then they later go toward the object. 00:12:32.080 --> 00:12:32.630 OK? 00:12:32.630 --> 00:12:39.080 So an existence proof that such a capacity could exist 00:12:39.080 --> 00:12:41.900 doesn't tell us that it does exist in a young infant, 00:12:41.900 --> 00:12:43.310 but it could. 00:12:43.310 --> 00:12:45.830 Here's the one that was done recently, 00:12:45.830 --> 00:12:49.850 that's my favorite study in this series on solidity. 00:12:49.850 --> 00:12:55.130 OK, here's a study where the chick is hatched in the dark 00:12:55.130 --> 00:12:57.740 and spends most of the day for the first three days 00:12:57.740 --> 00:12:59.240 of its life in the dark. 00:12:59.240 --> 00:13:01.370 But during a certain period each day, 00:13:01.370 --> 00:13:04.415 it's put in a Plexiglas box with black walls, 00:13:04.415 --> 00:13:07.610 a black floor, black ceiling. 00:13:07.610 --> 00:13:10.430 And through the Plexiglas, there is a single object that 00:13:10.430 --> 00:13:12.550 dangles that they imprint to. 00:13:12.550 --> 00:13:13.160 OK? 00:13:13.160 --> 00:13:15.740 Now, they can't touch the object, 00:13:15.740 --> 00:13:18.230 so they never get evidence whether this object is solid 00:13:18.230 --> 00:13:18.730 or not. 00:13:18.730 --> 00:13:20.450 They can't peck at it, right? 00:13:20.450 --> 00:13:23.270 They can peck, but they can only peck 00:13:23.270 --> 00:13:25.670 at this black surface that they can't see 00:13:25.670 --> 00:13:29.070 or at this transparent surface that they also can't see. 00:13:29.070 --> 00:13:31.857 So they might learn there are objects in the environment, 00:13:31.857 --> 00:13:34.190 but they're not going to have any visual characteristics 00:13:34.190 --> 00:13:36.200 like this guy does. 00:13:36.200 --> 00:13:37.670 OK? 00:13:37.670 --> 00:13:38.210 All right. 00:13:38.210 --> 00:13:44.330 So then after this experience, the chick stays in this box, 00:13:44.330 --> 00:13:47.090 and in that box, sees a series of events 00:13:47.090 --> 00:13:49.820 involving this object. 00:13:49.820 --> 00:13:52.530 In the first set of experiments, the object 00:13:52.530 --> 00:13:56.490 moves behind one screen and then is revealed there, 00:13:56.490 --> 00:14:00.600 or moves behind the other screen and then is revealed there. 00:14:00.600 --> 00:14:02.940 In this particular study, the chick 00:14:02.940 --> 00:14:05.130 does not get to go out to get to that object. 00:14:05.130 --> 00:14:07.280 They did a series of studies before this one where they did, 00:14:07.280 --> 00:14:08.738 but in this case study, they don't. 00:14:13.170 --> 00:14:16.620 Then they see a second set of events where all they see 00:14:16.620 --> 00:14:20.940 is mom starts moving toward the space between the two screens. 00:14:20.940 --> 00:14:24.120 Then there, a screen comes down, so they 00:14:24.120 --> 00:14:25.770 can't see what happens next. 00:14:25.770 --> 00:14:27.550 And then when the screen comes up, 00:14:27.550 --> 00:14:29.850 she is no longer visible, but she then 00:14:29.850 --> 00:14:32.310 emerges behind either one screen or the other, 00:14:32.310 --> 00:14:34.290 basically teaching the chicks, in effect, 00:14:34.290 --> 00:14:37.290 mom can go behind each of these two screens, 00:14:37.290 --> 00:14:38.970 OK, either of these two screens. 00:14:38.970 --> 00:14:40.590 And then comes the critical test. 00:14:40.590 --> 00:14:43.530 In the critical test, mom starts moving toward the midpoint 00:14:43.530 --> 00:14:44.550 of the two screens. 00:14:44.550 --> 00:14:47.980 The screen comes down, and when the screen-- sorry, 00:14:47.980 --> 00:14:51.630 the big occluder-- vision is blocked, 00:14:51.630 --> 00:14:55.537 and then when that curtain raises again 00:14:55.537 --> 00:14:57.120 and the two screens are again visible, 00:14:57.120 --> 00:14:58.860 they've been rotated backward. 00:14:58.860 --> 00:15:02.550 And across a series of studies, they vary the size of mom 00:15:02.550 --> 00:15:05.250 and the degree of rotation of the screen. 00:15:05.250 --> 00:15:07.860 And the question is, will the chick go to the side 00:15:07.860 --> 00:15:12.000 where the screen's rotation is consistent with a solid object? 00:15:12.000 --> 00:15:13.470 And they do. 00:15:13.470 --> 00:15:17.430 I want to argue from that that knowledge of solidity-- well, 00:15:17.430 --> 00:15:19.620 knowledge in some sense-- representations 00:15:19.620 --> 00:15:22.620 that accord with solidity is innate in chicks. 00:15:22.620 --> 00:15:24.780 And what I mean by innate is simply 00:15:24.780 --> 00:15:27.330 that it's present and functional on first encounters 00:15:27.330 --> 00:15:29.497 with objects that exhibit that property. 00:15:29.497 --> 00:15:31.080 This chick has not had the opportunity 00:15:31.080 --> 00:15:32.610 to peck at these objects, or observe 00:15:32.610 --> 00:15:35.350 them bumping into each other, or anything like that. 00:15:35.350 --> 00:15:37.890 And the first time they see this degree of rotation, 00:15:37.890 --> 00:15:40.870 they make inferences that are consistent with that principle. 00:15:40.870 --> 00:15:44.490 This doesn't tell us that human infants do it. 00:15:44.490 --> 00:15:46.180 To be convinced of that, I'd want 00:15:46.180 --> 00:15:49.200 to see evidence for this ability in an animal that's 00:15:49.200 --> 00:15:55.210 a much better model for human object cognition. 00:15:55.210 --> 00:15:57.519 Or I'd want to see more evidence from the chicks 00:15:57.519 --> 00:15:59.310 to convince me that, actually, the chick is 00:15:59.310 --> 00:16:01.374 a really good model of human object cognition. 00:16:01.374 --> 00:16:02.790 But I do think it should encourage 00:16:02.790 --> 00:16:06.990 us to start thinking about how nervous systems could be wired 00:16:06.990 --> 00:16:10.610 to exhibit this property in the absence of specific learning 00:16:10.610 --> 00:16:12.830 and experiences with it. 00:16:12.830 --> 00:16:18.440 Here are some questions that we could talk about in the Q&A 00:16:18.440 --> 00:16:20.365 later. 00:16:20.365 --> 00:16:23.640 I can give you the one-liner on it. 00:16:23.640 --> 00:16:25.290 On the issue of compositionality, 00:16:25.290 --> 00:16:29.640 the question is, do babies have a laundry list of rules 00:16:29.640 --> 00:16:32.040 about how objects are going to behave, 00:16:32.040 --> 00:16:34.830 or are they building some kind of unitary model 00:16:34.830 --> 00:16:38.790 of the world that accords with certain general properties? 00:16:38.790 --> 00:16:41.100 I think the evidence, starting at least 00:16:41.100 --> 00:16:42.330 at eight months of age-- 00:16:42.330 --> 00:16:43.860 lower than that, we don't know-- 00:16:43.860 --> 00:16:45.810 but starting at that age, I think the evidence 00:16:45.810 --> 00:16:47.820 favors the second possibility. 00:16:47.820 --> 00:16:49.830 It comes primarily from these beautiful studies 00:16:49.830 --> 00:16:53.580 that were conducted by Susan Carey and her students 00:16:53.580 --> 00:16:58.470 and collaborators where they presented infants 00:16:58.470 --> 00:17:01.740 with an event that violated cohesion. 00:17:01.740 --> 00:17:05.400 So this could be a sand pile that's poured onto a stage, 00:17:05.400 --> 00:17:08.790 or a cookie that's broken into two pieces 00:17:08.790 --> 00:17:13.440 before those pieces are put in boxes, 00:17:13.440 --> 00:17:21.089 or a block that falls apart when it's hit by another block. 00:17:21.089 --> 00:17:25.740 And then, she asks, having seen that this object violated 00:17:25.740 --> 00:17:28.410 this property, do they expect it to accord 00:17:28.410 --> 00:17:30.660 with the other properties? 00:17:30.660 --> 00:17:33.120 And the answer is no, they don't. 00:17:33.120 --> 00:17:33.690 OK? 00:17:33.690 --> 00:17:37.620 So for example, if you've seen, there's-- 00:17:37.620 --> 00:17:40.500 in that causality study where an object moves behind a screen, 00:17:40.500 --> 00:17:42.330 and then another thing starts to move-- 00:17:42.330 --> 00:17:45.270 if instead of starting to move, it falls apart-- 00:17:45.270 --> 00:17:48.660 it violates cohesion, falls apart-- 00:17:48.660 --> 00:17:51.890 infants no longer expect the first object to hit it. 00:17:51.890 --> 00:17:53.250 OK? 00:17:53.250 --> 00:17:55.086 Maybe it fell apart on its own, OK? 00:17:55.086 --> 00:17:56.460 They don't assume that it's going 00:17:56.460 --> 00:17:58.650 to behave like an object in other respects. 00:17:58.650 --> 00:18:01.680 They also don't assume that sand piles will move continuously, 00:18:01.680 --> 00:18:04.650 that they won't pass through surfaces, 00:18:04.650 --> 00:18:08.280 that a cookie that is broken in two-- if you have 00:18:08.280 --> 00:18:10.980 two cookies in one box and one cookie in the other, the babies 00:18:10.980 --> 00:18:11.965 crawl to the two. 00:18:11.965 --> 00:18:14.340 But if you have one large cookie, and you break it in two 00:18:14.340 --> 00:18:16.840 and put it in one box, they're neutral between those two 00:18:16.840 --> 00:18:17.340 options. 00:18:17.340 --> 00:18:17.871 OK? 00:18:17.871 --> 00:18:20.370 So it looks like the system is acting like an interconnected 00:18:20.370 --> 00:18:22.310 whole. 00:18:22.310 --> 00:18:25.320 What the babies learn about one aspect of an object's behavior 00:18:25.320 --> 00:18:28.410 bears on the inferences they make about other aspects 00:18:28.410 --> 00:18:30.330 of its behavior. 00:18:30.330 --> 00:18:33.060 The final thing I want to end with that I think 00:18:33.060 --> 00:18:35.730 makes this point is very recent work 00:18:35.730 --> 00:18:38.370 that was conducted by Lisa Feigenson-- 00:18:38.370 --> 00:18:41.070 that I think Laura Schulz may talk about as well. 00:18:41.070 --> 00:18:45.510 Josh mentioned it last Friday, but didn't really describe it-- 00:18:45.510 --> 00:18:48.660 that I think suggests that at least at the end of infancy-- 00:18:48.660 --> 00:18:52.666 and again, we don't know what's happening earlier-- 00:18:52.666 --> 00:18:54.040 infants' understanding of objects 00:18:54.040 --> 00:18:57.850 isn't only forming an interconnected whole, 00:18:57.850 --> 00:19:01.060 but it centers on abstract notions about the causes 00:19:01.060 --> 00:19:04.060 of object motions. 00:19:04.060 --> 00:19:08.560 So these are studies that picked up on some old findings 00:19:08.560 --> 00:19:10.390 from a variety of labs. 00:19:10.390 --> 00:19:12.680 Here's one that Baillargeon had worked on, 00:19:12.680 --> 00:19:14.920 and I did studies on it as well, where infants 00:19:14.920 --> 00:19:21.080 see an object that's moving in an array that has two barriers, 00:19:21.080 --> 00:19:24.486 but the barriers are mostly hidden by a screen, 00:19:24.486 --> 00:19:26.110 and the object moves so that it's fully 00:19:26.110 --> 00:19:27.160 hidden by the screen. 00:19:27.160 --> 00:19:29.590 The question is, where will it come to rest? 00:19:29.590 --> 00:19:31.365 Looking time studies say babies expect 00:19:31.365 --> 00:19:34.240 it to come to rest at the first barrier in its path. 00:19:34.240 --> 00:19:37.120 They look longer if they find it when the screen is raised 00:19:37.120 --> 00:19:40.440 behind the second barrier, as if it passed through that surface 00:19:40.440 --> 00:19:42.770 in its path, a solidity violation. 00:19:42.770 --> 00:19:44.770 Here's another study that Baillargeon and others 00:19:44.770 --> 00:19:47.650 have studied showing that by three months of age, 00:19:47.650 --> 00:19:50.890 infants expect objects to fall in the absence of support. 00:19:50.890 --> 00:19:54.340 If you push a truck along a block 00:19:54.340 --> 00:19:57.730 so that it stops, but under conditions where it stops, when 00:19:57.730 --> 00:20:01.000 it gets to the end of the block or continues off the block 00:20:01.000 --> 00:20:04.750 and doesn't fall, infants look longer in the latter case. 00:20:04.750 --> 00:20:07.324 And what Feigenson asked is, suppose 00:20:07.324 --> 00:20:09.490 we don't let look infants look as long as they want. 00:20:09.490 --> 00:20:11.170 Suppose we limit their looking time 00:20:11.170 --> 00:20:14.020 to just a couple of seconds, so they've looked equally 00:20:14.020 --> 00:20:15.680 at these two outcomes. 00:20:15.680 --> 00:20:17.950 What is their internal state? 00:20:17.950 --> 00:20:21.430 Are they just bewildered in the case where the object did 00:20:21.430 --> 00:20:23.380 something impossible? 00:20:23.380 --> 00:20:27.310 Or are they seeking to understand 00:20:27.310 --> 00:20:28.750 what's happening in the world? 00:20:28.750 --> 00:20:30.610 Is this a learning signal for them? 00:20:30.610 --> 00:20:32.719 So she did two experiments-- 00:20:32.719 --> 00:20:33.760 actually, more than two-- 00:20:33.760 --> 00:20:36.460 but asked two questions with experiments. 00:20:36.460 --> 00:20:39.820 One question was, suppose after showing 00:20:39.820 --> 00:20:43.390 this event, or this event, or this, or this one, 00:20:43.390 --> 00:20:45.790 after showing an event where an object behaves naturally 00:20:45.790 --> 00:20:48.880 versus apparently unnaturally, you now 00:20:48.880 --> 00:20:51.520 teach infants in effect, try to teach infants in effect 00:20:51.520 --> 00:20:53.210 some new property of the object. 00:20:53.210 --> 00:20:54.940 So you pick the object up and squeak it. 00:20:54.940 --> 00:20:56.356 And the question is, do the babies 00:20:56.356 --> 00:20:58.900 learn that the object makes that sound? 00:20:58.900 --> 00:21:04.150 And what she finds is that the infants learn much more 00:21:04.150 --> 00:21:07.750 consistently about the object whose previous behavior was 00:21:07.750 --> 00:21:10.600 unpredicted than about the object whose previous behavior 00:21:10.600 --> 00:21:11.550 was predictable. 00:21:11.550 --> 00:21:12.275 OK. 00:21:12.275 --> 00:21:14.650 I think this is both good news and bad news about all the 00:21:14.650 --> 00:21:16.811 looking time methods I've been telling you about. 00:21:16.811 --> 00:21:18.310 The good news is looking time really 00:21:18.310 --> 00:21:21.040 does seem, when you let babies look for as long as they want, 00:21:21.040 --> 00:21:23.620 it really does look like that's tracking what they're 00:21:23.620 --> 00:21:27.850 seeking to learn about their exploration and learning. 00:21:27.850 --> 00:21:30.580 The bad news is that when you restrict looking time 00:21:30.580 --> 00:21:32.396 to just a very short amount of time, 00:21:32.396 --> 00:21:34.520 there are many, many things that could be going on. 00:21:34.520 --> 00:21:35.710 They could be attending a lot, or they 00:21:35.710 --> 00:21:36.876 could be attending a little. 00:21:36.876 --> 00:21:39.680 It's a very crude measure that's not telling us that. 00:21:39.680 --> 00:21:42.340 But here's a richer measure suggesting differential 00:21:42.340 --> 00:21:44.200 learning in these two cases. 00:21:44.200 --> 00:21:46.570 The other study, I think, is even cooler. 00:21:46.570 --> 00:21:49.410 In the other studies, after showing infants a violation 00:21:49.410 --> 00:21:53.110 event, she handed infants the object, 00:21:53.110 --> 00:21:55.220 and allowed them to explore it, and looked 00:21:55.220 --> 00:21:57.730 to see what they would do. 00:21:57.730 --> 00:22:00.430 And what she found is that they do different things depending 00:22:00.430 --> 00:22:02.050 on the nature of violation. 00:22:02.050 --> 00:22:04.660 So when there was a solidity violation, 00:22:04.660 --> 00:22:07.780 they take the object, and they bang it on the tray 00:22:07.780 --> 00:22:08.500 in front of them. 00:22:08.500 --> 00:22:09.220 OK? 00:22:09.220 --> 00:22:12.070 In the case of a support violation, 00:22:12.070 --> 00:22:13.990 they take the object, and they release it. 00:22:13.990 --> 00:22:14.550 OK? 00:22:14.550 --> 00:22:16.854 So if they're specifically oriented at 11 months, 00:22:16.854 --> 00:22:18.520 and we don't know what happened earlier, 00:22:18.520 --> 00:22:21.590 they're specifically oriented to expected properties 00:22:21.590 --> 00:22:25.030 of the object that could be relevant to understanding 00:22:25.030 --> 00:22:27.940 the apparent violation that they saw. 00:22:27.940 --> 00:22:29.530 To summarize, it looks like there 00:22:29.530 --> 00:22:33.280 is a system that is growing over the course of infancy, 00:22:33.280 --> 00:22:35.380 as it's clear from all the questions 00:22:35.380 --> 00:22:37.630 that we can continue to discuss. 00:22:37.630 --> 00:22:40.930 There's a lot we don't know about this system. 00:22:40.930 --> 00:22:43.210 But at no point in development do 00:22:43.210 --> 00:22:45.610 infants seem to be perceiving just 00:22:45.610 --> 00:22:50.920 a two-dimensional, disorganized world of sensory experiences. 00:22:50.920 --> 00:22:54.220 At no point do they seem oriented to events going on 00:22:54.220 --> 00:22:56.830 in the visual field as opposed to the real world 00:22:56.830 --> 00:22:58.414 when you test them in these situations 00:22:58.414 --> 00:22:59.830 where you're asking them questions 00:22:59.830 --> 00:23:01.210 about properties of the world. 00:23:01.210 --> 00:23:04.390 They seem to be oriented to properties of the world itself. 00:23:04.390 --> 00:23:05.920 And to start out already, as early 00:23:05.920 --> 00:23:07.960 as we can test them, in a few cases, 00:23:07.960 --> 00:23:10.750 like the center-occluded objects, that means newborns. 00:23:10.750 --> 00:23:13.420 To start out already with a system which, 00:23:13.420 --> 00:23:15.340 although it's radically different from ours, 00:23:15.340 --> 00:23:18.370 we see that most of the things that we know about, 00:23:18.370 --> 00:23:21.580 most of the information we can get about objects from a scene, 00:23:21.580 --> 00:23:23.050 they do not get. 00:23:23.050 --> 00:23:27.246 Nevertheless, we're seeing core skeletal abilities 00:23:27.246 --> 00:23:28.870 that we continue to use throughout life 00:23:28.870 --> 00:23:33.250 and that seem to be present to serve as a basis for learning. 00:23:33.250 --> 00:23:35.800 There have been lots of studies in which infants 00:23:35.800 --> 00:23:40.000 have been presented with people or other self-propelled objects 00:23:40.000 --> 00:23:42.610 that have used the same logic as the studies for looking 00:23:42.610 --> 00:23:44.800 at naive physics to get at something like naive 00:23:44.800 --> 00:23:49.150 psychology in infancy, what they know about object motion. 00:23:49.150 --> 00:23:50.920 Most of them are on babies that are 00:23:50.920 --> 00:23:53.505 older than the babies from the object studies. 00:23:53.505 --> 00:23:57.220 Those studies ran mostly up to about four months of age. 00:23:57.220 --> 00:24:00.730 These studies are starting later than that, 00:24:00.730 --> 00:24:02.755 but they give us a starting point. 00:24:02.755 --> 00:24:06.106 And here's one slide that tries to capture just 00:24:06.106 --> 00:24:08.230 about everything that I think we know about what 6- 00:24:08.230 --> 00:24:12.910 to 12-month-old infants, how they represent agents. 00:24:12.910 --> 00:24:15.970 First of all, they represent people's actions on objects 00:24:15.970 --> 00:24:17.840 as directed to goals. 00:24:17.840 --> 00:24:20.110 So this is a basic study that was conducted-- 00:24:20.110 --> 00:24:22.180 a whole series of studies, one of many-- 00:24:22.180 --> 00:24:24.280 conducted by Amanda Woodward, who 00:24:24.280 --> 00:24:26.170 was the advisor for Jessica Sommerville, who 00:24:26.170 --> 00:24:29.717 will be here giving a talk on Thursday afternoon. 00:24:29.717 --> 00:24:32.050 They're very simple studies in which she presents babies 00:24:32.050 --> 00:24:34.600 with two objects side by side, and then they 00:24:34.600 --> 00:24:38.680 see a hand reach out and grasp one of those two objects. 00:24:38.680 --> 00:24:42.070 And the question is, how do babies represent that action? 00:24:42.070 --> 00:24:45.610 Do they represent it as a motion to a position on the left, 00:24:45.610 --> 00:24:48.550 or do they represent it as an action 00:24:48.550 --> 00:24:51.130 with the goal of obtaining the ball? 00:24:51.130 --> 00:24:52.930 And to distinguish those two possibilities, 00:24:52.930 --> 00:24:55.810 she then reverses the positions of the two objects, 00:24:55.810 --> 00:24:57.640 presents the hand in alternation, 00:24:57.640 --> 00:25:00.190 taking a new trajectory to the old object 00:25:00.190 --> 00:25:02.410 versus the old trajectory to a new object. 00:25:02.410 --> 00:25:04.720 And the babies look longer when the hand goes 00:25:04.720 --> 00:25:06.940 to the new object, suggesting not 00:25:06.940 --> 00:25:08.590 that they can't represent trajectories, 00:25:08.590 --> 00:25:11.380 but that they care more about-- 00:25:11.380 --> 00:25:13.990 bigger news is the agent's goal. 00:25:13.990 --> 00:25:16.180 When the goal changes, that's a bigger change 00:25:16.180 --> 00:25:18.460 for the infant than when the motion simply 00:25:18.460 --> 00:25:21.870 changes to a new path and a new endpoint. 00:25:21.870 --> 00:25:25.150 That's at as young as five months of age. 00:25:25.150 --> 00:25:27.880 Then there's been studies showing that infants' goal 00:25:27.880 --> 00:25:32.380 attribution depends on what is visible from the perspective 00:25:32.380 --> 00:25:33.950 of the agent. 00:25:33.950 --> 00:25:37.330 So if two objects are present, and the agent consistently 00:25:37.330 --> 00:25:40.990 reaches for one of them, babies in some sense, 00:25:40.990 --> 00:25:44.500 in some poorly-understood sense, represent that agent 00:25:44.500 --> 00:25:46.150 as having a preference for the object 00:25:46.150 --> 00:25:48.920 that they've chosen to go for over the other object. 00:25:48.920 --> 00:25:51.910 But if the object they didn't go for is occluded, 00:25:51.910 --> 00:25:54.430 and there's no evidence that that agent ever saw it, 00:25:54.430 --> 00:25:56.710 they don't make that preference, inference. 00:25:56.710 --> 00:25:58.120 OK? 00:25:58.120 --> 00:26:02.030 Third, infants represent agents as acting efficiently. 00:26:02.030 --> 00:26:04.030 And this is true whenever they're given evidence 00:26:04.030 --> 00:26:06.640 for self-propulsion, whether the object that's 00:26:06.640 --> 00:26:08.080 moving in a self-propelled manner 00:26:08.080 --> 00:26:10.160 has the features of an agent or not. 00:26:10.160 --> 00:26:12.460 So with these classic studies by Gergely 00:26:12.460 --> 00:26:15.850 and his collaborators that continue to the present day, 00:26:15.850 --> 00:26:19.180 but started, now, I guess 20 years ago, 00:26:19.180 --> 00:26:21.220 they present infants with two balls 00:26:21.220 --> 00:26:23.650 that engage in self-propelled motion and even 00:26:23.650 --> 00:26:25.420 a little kind of interaction. 00:26:25.420 --> 00:26:27.910 And then one of the balls jumps over a barrier 00:26:27.910 --> 00:26:29.200 to get to the other ball. 00:26:29.200 --> 00:26:32.230 And across a series of familiarization trials, 00:26:32.230 --> 00:26:33.850 the barrier varies in height. 00:26:33.850 --> 00:26:36.100 The jump is appropriately adapted to it. 00:26:36.100 --> 00:26:37.960 And the question is, what do infants 00:26:37.960 --> 00:26:41.980 infer this agent will do when the barrier is taken away? 00:26:41.980 --> 00:26:45.400 Will it engage in one of its familiar patterns of motion, 00:26:45.400 --> 00:26:47.530 or will it engage in a new pattern of motion 00:26:47.530 --> 00:26:49.682 that's more efficient to get to the object? 00:26:49.682 --> 00:26:51.640 And the finding is, they expect that new motion 00:26:51.640 --> 00:26:54.070 and look longer at the more superficially 00:26:54.070 --> 00:26:59.950 familiar but inefficient indirect action. 00:26:59.950 --> 00:27:01.720 Here are my favorite studies. 00:27:01.720 --> 00:27:06.640 This is one that Shimon talked about yesterday afternoon. 00:27:06.640 --> 00:27:08.290 He talked about one version of this. 00:27:08.290 --> 00:27:12.130 The original study was conducted by Sabina Pauen, 00:27:12.130 --> 00:27:15.921 and Rebecca Saxe and Susan Carey did interesting extensions 00:27:15.921 --> 00:27:16.420 on it. 00:27:16.420 --> 00:27:19.500 It uses the simplest imaginable method. 00:27:19.500 --> 00:27:22.600 They present babies first with two objects, stationary, 00:27:22.600 --> 00:27:25.390 side by side, one with animate features, 00:27:25.390 --> 00:27:28.940 a face and sort of a fuzzy body, tail-like body, 00:27:28.940 --> 00:27:31.880 the other, an inanimate ball. 00:27:31.880 --> 00:27:33.997 They're both stationary, and the objects 00:27:33.997 --> 00:27:35.830 were chosen to be about equally interesting. 00:27:35.830 --> 00:27:38.320 The babies look about half the time at each one. 00:27:38.320 --> 00:27:42.070 Then, they see a series of events in which these two 00:27:42.070 --> 00:27:44.620 objects are stuck together and undergo this very 00:27:44.620 --> 00:27:46.460 irregular pattern of motion. 00:27:46.460 --> 00:27:48.340 This was actually some kind of parlor trick 00:27:48.340 --> 00:27:49.810 toy that was sold for a while. 00:27:49.810 --> 00:27:51.610 They got them started on this study. 00:27:51.610 --> 00:27:54.250 There's a mechanism inside the ball that's actually propelling 00:27:54.250 --> 00:27:55.810 the two objects around. 00:27:55.810 --> 00:27:59.170 But the question is, which of these two objects 00:27:59.170 --> 00:28:02.290 does the baby attribute the motion to? 00:28:02.290 --> 00:28:05.290 And to find that out, they subsequently separated the two 00:28:05.290 --> 00:28:08.410 objects again, and ask again, which one will the infant look 00:28:08.410 --> 00:28:09.040 at more? 00:28:09.040 --> 00:28:11.020 And after seeing this motion, the infant 00:28:11.020 --> 00:28:13.960 looks more at the one that has the animate features. 00:28:13.960 --> 00:28:16.720 Although both objects underwent the same motion, 00:28:16.720 --> 00:28:20.170 they attribute that motion to this guy, not the other guy. 00:28:20.170 --> 00:28:22.480 It follows then that they're perceiving 00:28:22.480 --> 00:28:25.630 this-- they're representing this guy as causing the other guy's 00:28:25.630 --> 00:28:26.200 motion. 00:28:26.200 --> 00:28:26.440 Right? 00:28:26.440 --> 00:28:28.780 The other guy is not seen as causing its own motion. 00:28:28.780 --> 00:28:30.610 The motion is being caused by this guy. 00:28:30.610 --> 00:28:33.400 This is at seven months of age. 00:28:33.400 --> 00:28:36.766 Now, even stronger evidence that infants infer causal agents 00:28:36.766 --> 00:28:38.140 come from these beautiful studies 00:28:38.140 --> 00:28:40.600 that Rebecca Saxe, and Susan Carey, and Josh Tenenbaum 00:28:40.600 --> 00:28:44.050 conducted 10 years ago now, that went back 00:28:44.050 --> 00:28:45.910 to this efficient situation where 00:28:45.910 --> 00:28:48.370 you see an object efficiently going over 00:28:48.370 --> 00:28:50.650 barriers of variable heights to get 00:28:50.650 --> 00:28:53.590 to a new position on the stage. 00:28:53.590 --> 00:28:55.570 The one thing they added is that the object 00:28:55.570 --> 00:28:57.100 is manifestly inanimate. 00:28:57.100 --> 00:28:58.064 It's a beanbag. 00:28:58.064 --> 00:29:00.730 And the kid had a chance to play with it before the study began. 00:29:00.730 --> 00:29:03.530 They can feel that this is not an animate object. 00:29:03.530 --> 00:29:06.610 So if this isn't an animate object, 00:29:06.610 --> 00:29:09.520 and infants are actively trying to explain 00:29:09.520 --> 00:29:11.560 the motions of objects that they see, 00:29:11.560 --> 00:29:13.750 they're going to need another kind of explanation. 00:29:13.750 --> 00:29:16.210 And what Saxe, Tenenbaum, and Carey showed 00:29:16.210 --> 00:29:18.100 is that infants infer that there is 00:29:18.100 --> 00:29:20.920 an agent, off-screen, on this side of the screen, that 00:29:20.920 --> 00:29:22.690 set that object in motion. 00:29:22.690 --> 00:29:26.020 And they show that by their relative looking time 00:29:26.020 --> 00:29:30.460 to an event where a hand comes in on that side of the screen, 00:29:30.460 --> 00:29:33.880 which is consistent with that causal attribution, versus 00:29:33.880 --> 00:29:36.880 it comes in on the other side, which is inconsistent with it. 00:29:36.880 --> 00:29:39.460 And really, a pretty manipulation. 00:29:39.460 --> 00:29:42.790 If in fact babies are making causal attributions here, 00:29:42.790 --> 00:29:44.860 then you ought to be able to screen them off. 00:29:44.860 --> 00:29:47.770 If you make a simple change to the method, 00:29:47.770 --> 00:29:51.880 you show evidence that actually, this object is animate. 00:29:51.880 --> 00:29:54.642 If it is animate, then you don't need to infer another cause. 00:29:54.642 --> 00:29:56.350 So they ran that study as well and showed 00:29:56.350 --> 00:29:58.224 that when you change from an inanimate object 00:29:58.224 --> 00:30:00.740 to an animate one, you no longer get this effect. 00:30:00.740 --> 00:30:01.240 OK? 00:30:01.240 --> 00:30:03.865 So it really looks like they're seeking to explain these events 00:30:03.865 --> 00:30:06.010 and doing so in accord with the principle 00:30:06.010 --> 00:30:08.110 that agents can cause not only their own motion, 00:30:08.110 --> 00:30:10.510 but also can make changes in the world, cause motions 00:30:10.510 --> 00:30:12.130 and other changes in objects. 00:30:12.130 --> 00:30:13.780 And a final study that shows this, 00:30:13.780 --> 00:30:17.500 I told you that if a truck goes behind a screen toward a box, 00:30:17.500 --> 00:30:20.020 and then the box subsequently collapses, 00:30:20.020 --> 00:30:23.690 babies do not infer that the truck hit the box. 00:30:23.690 --> 00:30:24.190 OK? 00:30:24.190 --> 00:30:26.636 Collapsing takes this outside the domain 00:30:26.636 --> 00:30:28.510 of physical reasoning for those young babies. 00:30:28.510 --> 00:30:29.110 OK? 00:30:29.110 --> 00:30:31.510 But suppose instead of a truck going behind that-- this 00:30:31.510 --> 00:30:32.950 is work by Susan Carey as well. 00:30:32.950 --> 00:30:36.280 Sorry, I should have put it on here, Muentener and Carey. 00:30:36.280 --> 00:30:39.700 If instead of a truck, a hand goes behind that screen, 00:30:39.700 --> 00:30:43.000 now they infer that the hand did contact the object. 00:30:43.000 --> 00:30:43.570 OK? 00:30:43.570 --> 00:30:46.192 So hands can make things move, and they not only 00:30:46.192 --> 00:30:47.650 can make things move, they can make 00:30:47.650 --> 00:30:49.816 things do all sorts of things that objects otherwise 00:30:49.816 --> 00:30:52.270 won't do, like fall apart, OK? 00:30:52.270 --> 00:30:52.910 All right. 00:30:52.910 --> 00:30:56.320 So one problem, as I said, with all these studies 00:30:56.320 --> 00:30:58.810 is that they're all older infants, many of them 00:30:58.810 --> 00:31:03.520 much older infants like 8- and 10-month-olds, just about all 00:31:03.520 --> 00:31:06.850 of them 6 months old or older. 00:31:06.850 --> 00:31:08.480 And when you test younger infants, 00:31:08.480 --> 00:31:10.550 you get failures on some of these tasks. 00:31:10.550 --> 00:31:13.840 So these goal attribution tasks work with 5-month-old infants, 00:31:13.840 --> 00:31:16.350 but they fail with 3-month-old infants. 00:31:16.350 --> 00:31:18.940 3-month-old infants seem uncommitted as 00:31:18.940 --> 00:31:21.460 to where this hand is going to go once the two 00:31:21.460 --> 00:31:26.080 objects exchange locations. 00:31:26.080 --> 00:31:28.060 And that raises the question-- 00:31:28.060 --> 00:31:29.860 to me, the most interesting question-- 00:31:29.860 --> 00:31:32.810 what kinds of representations of agents, if any, 00:31:32.810 --> 00:31:36.190 does a younger infant have before they're 00:31:36.190 --> 00:31:39.940 able to act on the world by manipulating 00:31:39.940 --> 00:31:43.360 things, which starts around five months of age? 00:31:43.360 --> 00:31:47.150 And tell you just quickly, two kinds of studies that I think 00:31:47.150 --> 00:31:48.400 get at this-- 00:31:48.400 --> 00:31:51.250 imperfectly, but they're getting there. 00:31:51.250 --> 00:31:53.620 One is, again, going back to studies 00:31:53.620 --> 00:31:55.540 of controlled reared animals. 00:31:55.540 --> 00:31:57.814 Chicks, again, in particular. 00:31:57.814 --> 00:31:59.230 There have been imprinting studies 00:31:59.230 --> 00:32:01.510 done where chicks are raised in the dark, 00:32:01.510 --> 00:32:05.140 and all they get to see are video screens 00:32:05.140 --> 00:32:09.250 in which you get two objects engaging in this simple causal 00:32:09.250 --> 00:32:11.410 billiard ball type event. 00:32:11.410 --> 00:32:14.570 One object starts to move, contacts a second object, 00:32:14.570 --> 00:32:17.710 and at the point of contact, it stops moving, 00:32:17.710 --> 00:32:19.890 and the other object starts to move. 00:32:19.890 --> 00:32:20.620 OK? 00:32:20.620 --> 00:32:22.400 And they see that repeatedly. 00:32:22.400 --> 00:32:26.230 And now, OK, I told you that if chicks are raised in isolation, 00:32:26.230 --> 00:32:28.960 but there's a moving object there, they'll imprint to it. 00:32:28.960 --> 00:32:30.790 I was treating the imprinting object 00:32:30.790 --> 00:32:32.650 as just an example of an object. 00:32:32.650 --> 00:32:34.800 But of course, the imprinting object is mom, 00:32:34.800 --> 00:32:35.770 and she's an agent. 00:32:35.770 --> 00:32:38.500 So it should be a self-propelled object, really, right? 00:32:38.500 --> 00:32:43.030 So given a choice between these two, 00:32:43.030 --> 00:32:45.850 do they selectively imprint to one over the other? 00:32:45.850 --> 00:32:49.550 And the finding is that they do. 00:32:49.550 --> 00:32:51.234 The two objects have different features, 00:32:51.234 --> 00:32:53.650 I think, different colors, maybe different shapes as well. 00:32:53.650 --> 00:32:56.140 If you now present one of them at one end of their cage 00:32:56.140 --> 00:32:57.790 and the other at the other end, they'll 00:32:57.790 --> 00:33:01.660 go to object A over object B as if they saw that 00:33:01.660 --> 00:33:04.480 as the object that set the other object in motion. 00:33:04.480 --> 00:33:07.900 But now you could ask, is that because they see A as causal, 00:33:07.900 --> 00:33:09.670 as causing its own motion? 00:33:09.670 --> 00:33:12.740 Or is it for other superficial reasons like A moved first, 00:33:12.740 --> 00:33:15.460 or A was moving at the time of the collision, 00:33:15.460 --> 00:33:17.140 and B was stationary? 00:33:17.140 --> 00:33:19.600 So they tested for all of these things one by one, 00:33:19.600 --> 00:33:21.520 but this is their killer experiment, the test 00:33:21.520 --> 00:33:23.110 for all of them at once. 00:33:23.110 --> 00:33:24.880 They present exactly the same events. 00:33:24.880 --> 00:33:29.440 The only difference is you don't see A initially at rest 00:33:29.440 --> 00:33:31.041 and starting to move. 00:33:31.041 --> 00:33:31.540 OK? 00:33:31.540 --> 00:33:32.910 You never get to see that. 00:33:32.910 --> 00:33:35.000 There are screens on the two ends of the display, 00:33:35.000 --> 00:33:38.860 so all you see is that A enters the display already in motion 00:33:38.860 --> 00:33:42.170 and contacts B. So you have no evidence as to whether A 00:33:42.170 --> 00:33:44.470 is self-propelled or not, but you 00:33:44.470 --> 00:33:47.680 see the two objects interacting otherwise in the same way. 00:33:47.680 --> 00:33:49.760 That makes the effect completely go away. 00:33:49.760 --> 00:33:51.790 I think this is logically a little 00:33:51.790 --> 00:33:53.530 like the studies with infants where 00:33:53.530 --> 00:33:55.630 you show that the thing is animate or not, 00:33:55.630 --> 00:33:57.940 and it effects whether they expect that there's 00:33:57.940 --> 00:33:59.950 an agent there or not. 00:33:59.950 --> 00:34:02.680 It suggests that infants are representing A 00:34:02.680 --> 00:34:04.600 as causing its own motion. 00:34:04.600 --> 00:34:06.850 They're representing it as causing B's motions, 00:34:06.850 --> 00:34:10.420 so A is a better object of imprinting than B is. 00:34:10.420 --> 00:34:15.824 And all of this can be abolished if A 00:34:15.824 --> 00:34:17.699 isn't seen to have the first causal property, 00:34:17.699 --> 00:34:19.960 they're not inferring the second. 00:34:19.960 --> 00:34:22.780 So it's, again, an existence proof kind of argument saying, 00:34:22.780 --> 00:34:25.540 you can get this kind of system working in an animal that 00:34:25.540 --> 00:34:29.469 hasn't had prior visual experiences in which they've 00:34:29.469 --> 00:34:32.854 seen objects in motion or in which they themselves have 00:34:32.854 --> 00:34:35.270 been-- they haven't had any other encounters with objects, 00:34:35.270 --> 00:34:37.659 so they haven't been able to move things around. 00:34:37.659 --> 00:34:40.360 So that's one way of trying to get at early representations. 00:34:40.360 --> 00:34:43.250 Here's another way that I hope Sommerville 00:34:43.250 --> 00:34:45.850 will talk about on Thursday night, 00:34:45.850 --> 00:34:48.310 because she pioneered this method with Amanda Woodward 00:34:48.310 --> 00:34:51.130 and Amy Needham. 00:34:51.130 --> 00:34:54.130 You can give a 3-month-old infant, who otherwise wouldn't 00:34:54.130 --> 00:34:56.770 be reaching for things for another two months, 00:34:56.770 --> 00:35:00.550 you can give them the ability to pick up objects 00:35:00.550 --> 00:35:03.894 by equipping them with mittens that have Velcro on them 00:35:03.894 --> 00:35:05.560 and presenting them with Velcro objects. 00:35:05.560 --> 00:35:09.020 So now they can make things move. 00:35:09.020 --> 00:35:11.650 And when you do that, you see interesting changes 00:35:11.650 --> 00:35:13.940 in their representations of these events. 00:35:13.940 --> 00:35:16.330 So an infant who has played with an object 00:35:16.330 --> 00:35:20.800 while wearing a sticky mitten, who subsequently looks 00:35:20.800 --> 00:35:23.800 at events in which another person wearing that mitten 00:35:23.800 --> 00:35:27.970 reaches for that object or for another object, 00:35:27.970 --> 00:35:30.910 now they look like a 5-month-old and represent 00:35:30.910 --> 00:35:33.280 that reaching as goal-directed. 00:35:33.280 --> 00:35:35.950 But on the other hand, if they saw those same events 00:35:35.950 --> 00:35:39.080 without the mittens, they failed. 00:35:39.080 --> 00:35:41.500 So that's saying that the infant's own action 00:35:41.500 --> 00:35:46.150 experience can elicit these action representations. 00:35:46.150 --> 00:35:48.820 And it raises, I think, all sorts of interesting questions 00:35:48.820 --> 00:35:52.120 about, do infants have to learn one by one 00:35:52.120 --> 00:35:54.010 what the properties of agents are? 00:35:54.010 --> 00:35:57.100 Or is it possible that once these representations 00:35:57.100 --> 00:35:59.020 of goal-directed actions are elicited, 00:35:59.020 --> 00:36:02.559 we'll see other knowledge of agents already present? 00:36:02.559 --> 00:36:04.600 So let me just give you one finding that suggests 00:36:04.600 --> 00:36:06.100 we might get the second. 00:36:06.100 --> 00:36:12.400 This is a study that Amy Skerry performed rather recently. 00:36:12.400 --> 00:36:14.830 She was interested in this representation 00:36:14.830 --> 00:36:17.230 of efficient action before babies 00:36:17.230 --> 00:36:20.390 are able to reach for and pick up objects themselves. 00:36:20.390 --> 00:36:24.800 Do they already represent the actions of agents 00:36:24.800 --> 00:36:27.380 as directed to goals efficiently? 00:36:27.380 --> 00:36:30.100 Did they already expect that agents will move efficiently 00:36:30.100 --> 00:36:32.270 to achieve their goals? 00:36:32.270 --> 00:36:36.220 So to get at that, she gave infants, 3-month-old infants, 00:36:36.220 --> 00:36:38.410 sticky mittens experience with an object where 00:36:38.410 --> 00:36:44.830 they got to pick up the object and then showed them 00:36:44.830 --> 00:36:48.460 events in one-- split them into two different conditions, 00:36:48.460 --> 00:36:50.380 and in both conditions, they see events 00:36:50.380 --> 00:36:54.520 in which a person moves on an indirect path to an object. 00:36:54.520 --> 00:36:57.670 In one of the series of events, there's a barrier in the way, 00:36:57.670 --> 00:37:01.270 so that motion is the best way to get there, probably the only 00:37:01.270 --> 00:37:02.709 obvious way to get there. 00:37:02.709 --> 00:37:04.750 In the other case, the barrier is in the display, 00:37:04.750 --> 00:37:08.320 but it's out of the way, so this is not an efficient action. 00:37:08.320 --> 00:37:12.240 And then at test, the barrier is gone in both conditions, 00:37:12.240 --> 00:37:15.550 and the person either moves efficiently or moves 00:37:15.550 --> 00:37:16.690 inefficiently. 00:37:16.690 --> 00:37:18.820 Now, in this control condition, the baby 00:37:18.820 --> 00:37:20.760 does not expect efficient action. 00:37:20.760 --> 00:37:22.510 But in the condition where they previously 00:37:22.510 --> 00:37:25.060 had the sticky mittens experience, which presumably 00:37:25.060 --> 00:37:28.000 told them that this is actually the person's goal, 00:37:28.000 --> 00:37:30.340 that they're attempting to get there, once 00:37:30.340 --> 00:37:33.190 they have that, what seems to follow immediately 00:37:33.190 --> 00:37:35.856 from that is this efficiency, this principle 00:37:35.856 --> 00:37:37.480 of efficient action, right, that you'll 00:37:37.480 --> 00:37:39.940 move to the goal on the most direct path possible, 00:37:39.940 --> 00:37:42.070 even though you've never seen that direct path. 00:37:42.070 --> 00:37:43.660 And even though when the infant was 00:37:43.660 --> 00:37:46.420 playing with the object with the mittens 00:37:46.420 --> 00:37:49.360 on and picking up objects for the first time in his life, 00:37:49.360 --> 00:37:50.890 there were no barriers present. 00:37:50.890 --> 00:37:52.630 They never had to do anything indirect. 00:37:52.630 --> 00:37:55.840 They could always get that object directly. 00:37:55.840 --> 00:37:57.310 Yet they're expecting direct action 00:37:57.310 --> 00:38:02.890 only when they've seen efficient action in another agent. 00:38:02.890 --> 00:38:05.180 OK. 00:38:05.180 --> 00:38:08.560 So in summary, it looks like these abilities are all there 00:38:08.560 --> 00:38:11.130 quite early. 00:38:11.130 --> 00:38:12.880 They're unlearned, at least some of them 00:38:12.880 --> 00:38:15.160 are unlearned, in chicks. 00:38:15.160 --> 00:38:17.500 And they can be elicited before-- 00:38:17.500 --> 00:38:20.740 at least some of them can be elicited before reaching-- 00:38:20.740 --> 00:38:22.075 develops in young infants. 00:38:22.075 --> 00:38:23.950 But infants-- and this is important, I think, 00:38:23.950 --> 00:38:25.900 for what Alia's going to be talking about-- 00:38:25.900 --> 00:38:28.180 infants' understanding of agents is limited. 00:38:28.180 --> 00:38:30.550 It's radically limited, just like their understanding 00:38:30.550 --> 00:38:34.120 of inanimate objects is. 00:38:34.120 --> 00:38:37.990 And here are a few, I think, really interesting limits. 00:38:37.990 --> 00:38:40.450 Although infants are sensitive to what's 00:38:40.450 --> 00:38:45.190 perceptually accessible to an agent, 00:38:45.190 --> 00:38:49.120 they don't seem to represent agents as seeing objects. 00:38:49.120 --> 00:38:50.570 What do I mean by that? 00:38:50.570 --> 00:38:53.470 Well, here's a experiment with exactly the structure 00:38:53.470 --> 00:38:55.410 of the Woodward reaching experiment. 00:38:55.410 --> 00:38:59.020 Two objects present, and a person acts with respect 00:38:59.020 --> 00:39:00.010 to one of them. 00:39:00.010 --> 00:39:01.780 But instead of reaching for one object, 00:39:01.780 --> 00:39:03.940 she looks at one of the two objects. 00:39:03.940 --> 00:39:07.060 And now, you ask, after you exchange the two objects' 00:39:07.060 --> 00:39:09.790 positions, do babies find that she's 00:39:09.790 --> 00:39:13.129 doing something newer if she looks at the other object? 00:39:13.129 --> 00:39:14.920 Or do they find she's doing something newer 00:39:14.920 --> 00:39:16.600 if she looks in a different direction 00:39:16.600 --> 00:39:17.850 than she looked in before? 00:39:17.850 --> 00:39:18.580 OK? 00:39:18.580 --> 00:39:22.630 At 12 months, infants view looking as object-directed, 00:39:22.630 --> 00:39:24.190 by this measure. 00:39:24.190 --> 00:39:26.320 Younger than 12 months, they do not. 00:39:26.320 --> 00:39:30.400 They don't seem to view the orientation of the person 00:39:30.400 --> 00:39:34.860 as a look that's directed to one of these two objects, 00:39:34.860 --> 00:39:35.950 by that measure. 00:39:35.950 --> 00:39:37.550 Now, the age at which they succeeded, 00:39:37.550 --> 00:39:39.070 this is also the first age at which 00:39:39.070 --> 00:39:41.560 infants start to show this really interesting pattern 00:39:41.560 --> 00:39:43.390 in their communication with other people, 00:39:43.390 --> 00:39:45.592 that Mike Tomasello has written about a lot. 00:39:45.592 --> 00:39:47.050 It's the first age at which they'll 00:39:47.050 --> 00:39:51.280 start to alternately look at objects, 00:39:51.280 --> 00:39:53.050 and look at another person, and attempt 00:39:53.050 --> 00:39:56.020 to engage their attention to the object, by checking back 00:39:56.020 --> 00:39:59.170 and forth between looks at the person and looks at the object, 00:39:59.170 --> 00:40:02.080 or by pointing to the object, or by following another person's 00:40:02.080 --> 00:40:02.977 point to an object. 00:40:02.977 --> 00:40:05.060 All of that comes in at the end of the first year. 00:40:05.060 --> 00:40:06.820 It's not there earlier. 00:40:06.820 --> 00:40:09.640 Younger infants, younger than about 12 months of age, 00:40:09.640 --> 00:40:12.841 don't even seem to expect that if a person reaches 00:40:12.841 --> 00:40:15.340 for one of two objects, they'll tend to reach for the object 00:40:15.340 --> 00:40:16.670 that they were looking at. 00:40:16.670 --> 00:40:17.170 OK? 00:40:17.170 --> 00:40:19.900 We thought for sure babies would succeed at that task early. 00:40:19.900 --> 00:40:21.400 They don't reliably succeed at it 00:40:21.400 --> 00:40:24.250 until 12 or even 14 months of age. 00:40:24.250 --> 00:40:27.070 And finally, babies attribute first order but not 00:40:27.070 --> 00:40:29.170 second order goals to agents. 00:40:29.170 --> 00:40:32.050 So if they see an agent pull on a rake 00:40:32.050 --> 00:40:34.030 and then reach for an object, they 00:40:34.030 --> 00:40:36.697 perceive the agent's goal as the rake, not the object 00:40:36.697 --> 00:40:37.780 that they're reaching for. 00:40:37.780 --> 00:40:40.570 So very limited representations. 00:40:40.570 --> 00:40:44.580 Also, as far as we know, none of these are unique to humans. 00:40:44.580 --> 00:40:46.330 I think this raised all sorts of questions 00:40:46.330 --> 00:40:49.960 that we might want to talk about in the Q&A later. 00:40:49.960 --> 00:40:53.020 What's the relationship between infants' representations 00:40:53.020 --> 00:40:55.330 of agents and their representations of the things 00:40:55.330 --> 00:40:56.770 that agents act on? 00:40:56.770 --> 00:41:02.540 Clearly, these should be related in some interesting ways. 00:41:02.540 --> 00:41:05.230 Although, agents can do things that objects can't do, 00:41:05.230 --> 00:41:07.420 like move on their own, and formulate goals, 00:41:07.420 --> 00:41:10.870 and act on things that are visually accessible to them. 00:41:10.870 --> 00:41:13.060 Agents also are objects, right? 00:41:13.060 --> 00:41:15.680 And we're subject to all the constraints on objects. 00:41:15.680 --> 00:41:16.911 We can't walk through walls. 00:41:16.911 --> 00:41:19.285 If we want to make something move, we have to contact it. 00:41:19.285 --> 00:41:23.320 And babies are sensitive to those constraints. 00:41:23.320 --> 00:41:25.270 And I think this raises all sorts of questions 00:41:25.270 --> 00:41:28.870 that to date, research on infants hasn't really answered. 00:41:28.870 --> 00:41:31.060 Is there some hierarchy of representations 00:41:31.060 --> 00:41:33.340 where you've got all objects, and then you've 00:41:33.340 --> 00:41:36.820 got these especially talented objects that are agents, right? 00:41:36.820 --> 00:41:39.910 Or I think one way Tomer has put it is maybe we're all agents, 00:41:39.910 --> 00:41:41.980 but objects are just really bad agents, right, 00:41:41.980 --> 00:41:44.270 that can't do very much. 00:41:44.270 --> 00:41:45.917 That's one possibility. 00:41:45.917 --> 00:41:47.500 It's also possible that these are just 00:41:47.500 --> 00:41:48.916 separate systems at the beginning, 00:41:48.916 --> 00:41:52.960 and they have to get linked together over time. 00:41:52.960 --> 00:41:56.630 I think these are all answerable questions. 00:41:56.630 --> 00:41:59.780 I said that babies don't see looking as directed to objects, 00:41:59.780 --> 00:42:06.230 but they do from birth respond in a pro-social way to looking 00:42:06.230 --> 00:42:07.640 that's directed to them. 00:42:07.640 --> 00:42:09.680 So an infant will look longer at someone 00:42:09.680 --> 00:42:12.440 whose eyes are directed to them than someone 00:42:12.440 --> 00:42:14.010 whose eyes are looking away. 00:42:14.010 --> 00:42:15.260 That's true for human infants. 00:42:15.260 --> 00:42:18.020 It's also true for infant monkeys. 00:42:18.020 --> 00:42:20.352 If a monkey is presented with this display, 00:42:20.352 --> 00:42:22.310 and they're over here, they'll look longer than 00:42:22.310 --> 00:42:23.660 if they're over here. 00:42:23.660 --> 00:42:25.190 Right? 00:42:25.190 --> 00:42:28.130 When they're here, it looks like that guy is looking at them. 00:42:28.130 --> 00:42:30.695 They also, human and monkey infants 00:42:30.695 --> 00:42:39.110 engage in eye-to-eye contact with adults from birth. 00:42:39.110 --> 00:42:41.390 They also, human infants and monkey infants 00:42:41.390 --> 00:42:43.490 tend to imitate the gestures of other people, 00:42:43.490 --> 00:42:46.430 interestingly, only when those people are looking at them, 00:42:46.430 --> 00:42:47.960 not when they close their eyes. 00:42:47.960 --> 00:42:52.160 Then they'll imitate them as you saw in those beautiful films 00:42:52.160 --> 00:42:56.150 that Winrich showed from the Ferrari group, 00:42:56.150 --> 00:42:59.000 attentive to the person and then trying to reproduce 00:42:59.000 --> 00:43:00.450 the person's action. 00:43:00.450 --> 00:43:03.020 This has been shown with baby chimpanzees and baby monkeys 00:43:03.020 --> 00:43:06.590 as well as with human newborn infants. 00:43:06.590 --> 00:43:10.970 And finally, I said infants don't follow gaze to objects. 00:43:10.970 --> 00:43:14.330 That's true, but it's also true that if they see a person who's 00:43:14.330 --> 00:43:16.370 looking directly at them, and then 00:43:16.370 --> 00:43:18.650 the person's gaze shifts to the side, 00:43:18.650 --> 00:43:20.870 they'll continue to look at the person, 00:43:20.870 --> 00:43:23.840 but their attention will undergo a momentary shift 00:43:23.840 --> 00:43:27.290 in the direction of the shift of the other person's gaze. 00:43:27.290 --> 00:43:27.860 OK? 00:43:27.860 --> 00:43:30.080 So this has been shown at two months of age 00:43:30.080 --> 00:43:31.460 and also with newborn infants. 00:43:31.460 --> 00:43:33.560 It's been shown with photographs of real faces 00:43:33.560 --> 00:43:35.840 and also with schematic faces. 00:43:35.840 --> 00:43:38.204 The way to show that attention is shifting 00:43:38.204 --> 00:43:40.370 is to present infants with an event that could never 00:43:40.370 --> 00:43:41.780 happen in the real world. 00:43:41.780 --> 00:43:44.450 You have an image of a face, the eyes shift to the side, 00:43:44.450 --> 00:43:47.240 either left or right, and then the face disappears, 00:43:47.240 --> 00:43:49.820 and a probe appears either on the left or on the right. 00:43:49.820 --> 00:43:51.890 They'll get to the probe faster if it 00:43:51.890 --> 00:43:54.980 appears on the side to which the person shifted their gaze. 00:43:54.980 --> 00:43:57.230 And nice control studies show that it's not 00:43:57.230 --> 00:44:00.770 just any kind of low level motion in that direction that 00:44:00.770 --> 00:44:02.390 gets them there. 00:44:02.390 --> 00:44:05.210 So this could be a sign of something 00:44:05.210 --> 00:44:07.160 like infancy direct gaze. 00:44:07.160 --> 00:44:10.100 And it engages something like a state of engagement with 00:44:10.100 --> 00:44:13.880 another person in which evidence about the other person's state 00:44:13.880 --> 00:44:18.110 of attention-- possibly also emotion if the questionable 00:44:18.110 --> 00:44:21.320 literature on empathy emerging early is right-- 00:44:21.320 --> 00:44:24.770 this can be automatically spread from one person 00:44:24.770 --> 00:44:27.110 to a social partner. 00:44:27.110 --> 00:44:29.510 So the hypothesis is that infants 00:44:29.510 --> 00:44:35.090 are finding other potential social partners 00:44:35.090 --> 00:44:38.780 from the beginning of life, by looking 00:44:38.780 --> 00:44:40.820 at things like gaze direction, maybe also 00:44:40.820 --> 00:44:43.610 infant-directed speech, as a signal that someone else is 00:44:43.610 --> 00:44:47.420 engaging with them, by interpreting patterns 00:44:47.420 --> 00:44:49.790 of imitation as a communicative signal 00:44:49.790 --> 00:44:51.890 that somebody is tracking what they do, 00:44:51.890 --> 00:44:56.990 and acting in kind with them, and interpreting shifts 00:44:56.990 --> 00:44:59.390 of attention as-- you're responding to shifts 00:44:59.390 --> 00:45:02.240 of attention by shifting their own mental states 00:45:02.240 --> 00:45:04.510 in the same direction.