WEBVTT - Ep47 "Wheels rotate backwards on TV, but do they ever in real life?" 0:00:05.040 --> 0:00:07.840 When you watch a car commercial, have you ever noticed 0:00:07.880 --> 0:00:12.000 that sometimes the wheels seem to turn backwards? Or sometimes 0:00:12.080 --> 0:00:14.760 you see a video of a helicopter and it seems 0:00:14.760 --> 0:00:18.360 like the blades are barely turning, But you never see 0:00:18.360 --> 0:00:21.400 that in real life. So what's going on there? And 0:00:21.480 --> 0:00:24.479 what does this have to do with yellow street lamps? 0:00:24.960 --> 0:00:27.880 Or whether all four legs of a horse come off 0:00:27.920 --> 0:00:31.960 the ground when it runs, or the very surprising thing 0:00:32.000 --> 0:00:36.200 that happens if you stare and stare at your ceiling 0:00:36.280 --> 0:00:42.519 fan while it turns. Welcome to Inner Cosmos with me 0:00:42.600 --> 0:00:46.080 and David Eagleman. I'm a neuroscientist and author at Stanford, 0:00:46.440 --> 0:00:49.760 and in these episodes we sail deeply into our three 0:00:49.840 --> 0:00:54.240 pound universe to understand why and how our lives look 0:00:54.280 --> 0:00:57.400 the way they do, and in this case, why the 0:00:57.480 --> 0:01:05.839 world looks the way it does. Today's episode is about 0:01:06.319 --> 0:01:12.040 visual perception and a series of really strange surprises about 0:01:12.120 --> 0:01:17.920 whether our visual systems analyze the world continuously or instead 0:01:18.360 --> 0:01:23.000 whether we see in frames like a movie camera. So 0:01:23.760 --> 0:01:26.679 have you ever noticed what happens when you film a 0:01:26.720 --> 0:01:29.400 car going by on your cell phone camera and then 0:01:29.440 --> 0:01:32.480 you watch the video When you look at the hubcaps 0:01:32.520 --> 0:01:35.600 on the car, say with some spokes on it or 0:01:35.600 --> 0:01:38.640 some pattern. When you look at the hubcaps, it looks 0:01:38.680 --> 0:01:42.680 like they're spinning the wrong way, or maybe they occasionally 0:01:42.720 --> 0:01:45.440 look like they're not spinning at all, even though the 0:01:45.480 --> 0:01:49.600 car is moving. So the first question is why do 0:01:49.680 --> 0:01:52.840 we so rarely notice this, Like, why doesn't it blow 0:01:52.880 --> 0:01:55.600 our minds and we say, oh my god, that's not 0:01:55.960 --> 0:01:59.120 consistent with what I just saw with my own eyes 0:01:59.160 --> 0:02:01.720 and what I filmed. Why is it that we're so 0:02:01.880 --> 0:02:06.200 nonchalant about that. I'll be addressing that in some future episodes, 0:02:06.240 --> 0:02:09.120 but today the main question I want to ask is 0:02:09.600 --> 0:02:12.800 why does it happen? Why does the wheel look on 0:02:12.840 --> 0:02:17.320 your video like it's not spinning correctly even though that's 0:02:17.360 --> 0:02:20.480 not what you just witnessed in real life. So to 0:02:20.600 --> 0:02:25.639 understand that, let's step back to the eighteen seventies here 0:02:25.840 --> 0:02:30.799 in Palo Alto, California, where I am so Leland Stanford, 0:02:31.200 --> 0:02:35.080 who was a wealthy industrialist and the governor of California 0:02:35.440 --> 0:02:40.240 and started Stanford University. Leland Stanford had some horses that 0:02:40.320 --> 0:02:43.359 he loved, and so he hired a guy named Edward 0:02:43.400 --> 0:02:49.239 Moybridge to take pictures of his horses running. Now, Moybridge 0:02:49.320 --> 0:02:53.239 was this really talented guy with a collection of cameras 0:02:53.720 --> 0:02:57.440 that could achieve shutter speeds of about one to one 0:02:57.480 --> 0:03:01.320 thousandth of a second, and this was a really big 0:03:01.360 --> 0:03:05.680 deal in eighteen seventy eight. So Stanford and Moydbridge wanted 0:03:05.680 --> 0:03:09.240 to take these really fast photographs of the running horses 0:03:09.680 --> 0:03:12.359 because no one had ever done that. And it turns 0:03:12.360 --> 0:03:17.160 out there was a debate about how horses arranged their 0:03:17.360 --> 0:03:21.400 legs when they galloped, and the question was whether all 0:03:21.639 --> 0:03:25.160 four legs ever come off the ground at the same time. 0:03:25.800 --> 0:03:29.480 And Stanford realized he could put these things together, his 0:03:29.800 --> 0:03:34.880 horses and this new photographic technology to finally answer this question, 0:03:35.120 --> 0:03:38.400 do all the legs come off the ground? Now, if 0:03:38.400 --> 0:03:42.280 you've ever watched a galloping horse, you know that everything 0:03:42.360 --> 0:03:46.119 is moving just slightly too fast for you to confidently 0:03:46.200 --> 0:03:49.880 be able to answer this question, and so they needed 0:03:50.040 --> 0:03:53.640 a new way to address this. As an interesting side note, 0:03:53.640 --> 0:03:57.640 there were earlier paintings that showed a horse with all 0:03:57.760 --> 0:04:00.360 four of its legs off the ground when it was 0:04:00.360 --> 0:04:03.160 in the middle of a gallop. And in these paintings, 0:04:03.480 --> 0:04:05.960 the two front legs were extended in the air, and 0:04:06.000 --> 0:04:08.320 the two back legs were kicked out behind the horse. 0:04:08.600 --> 0:04:13.960 But nobody really knew if this was possible, and right 0:04:14.120 --> 0:04:18.719 in eighteen seventy eight, this was the birth of chronophotography, 0:04:18.839 --> 0:04:23.520 which meant taking these fast pictures of complicated movements to 0:04:23.640 --> 0:04:28.360 really get what was happening there. So Moybridge set up 0:04:28.400 --> 0:04:33.320 a series of cameras on Leland Stanford's track in Palo Alto, 0:04:33.760 --> 0:04:37.640 and he took these photographs as the horse went by, 0:04:38.240 --> 0:04:42.840 and these photographs became very famous because of their clarity 0:04:43.160 --> 0:04:45.840 and because they answered the question. It turns out that 0:04:46.240 --> 0:04:49.200 all four of a horse's legs do come off the ground, 0:04:49.600 --> 0:04:52.880 but this happens when the legs are gathered underneath the 0:04:52.920 --> 0:04:57.279 horse rather than extended front and back. But what happened 0:04:57.320 --> 0:05:00.760 next is the important part. Weybridge. He took his very 0:05:00.760 --> 0:05:05.240 clear photographs and he put them in what's called a zootrope, 0:05:05.560 --> 0:05:09.360 which is like an upright cylinder, like a big jar 0:05:10.080 --> 0:05:13.880 with vertical slits in it, and you put the photographs 0:05:14.000 --> 0:05:17.400 inside the cylinder, and then you spin the cylinder and 0:05:17.480 --> 0:05:21.680 as it rotates, you see one of the photos through 0:05:21.760 --> 0:05:25.000 one of the slits, and then when the next slit 0:05:25.160 --> 0:05:29.400 rotates around, you see the second photo in that same spot, 0:05:29.600 --> 0:05:32.160 and then as the cylinder continues to spin, you then 0:05:32.240 --> 0:05:36.080 see the third photo through the third slit, and so on. 0:05:36.640 --> 0:05:39.960 And Moidbridge was able to make the first prototype of 0:05:40.000 --> 0:05:45.279 a motion picture this way. And there's very rich history 0:05:45.320 --> 0:05:49.120 to all the pieces that came together for movie technology 0:05:49.120 --> 0:05:52.120 over the next decades. But this is the main idea. 0:05:52.560 --> 0:05:56.000 Your brain sees picture one, and then picture two, and 0:05:56.000 --> 0:06:01.440 then picture three, and it interprets that as smooth motion. Now, 0:06:01.440 --> 0:06:05.520 this is exactly how modern movies work. On a video 0:06:05.600 --> 0:06:07.920 you watch on your cell phone, you look at a 0:06:08.440 --> 0:06:12.400 snapshot of Tom Cruise with his left foot in the air, 0:06:12.880 --> 0:06:16.559 and then another still shot of him with his foot 0:06:16.600 --> 0:06:19.280 a little lower, and another with his foot now touching 0:06:19.320 --> 0:06:22.119 the ground. And in the next photograph his right foot 0:06:22.200 --> 0:06:24.679 is a few inches off the floor. And as long 0:06:24.720 --> 0:06:28.720 as you flash these snapshots quickly, then it looks like 0:06:28.760 --> 0:06:31.880 he's racing down the sidewalk after the bad guy. And 0:06:31.920 --> 0:06:34.600 you're totally caught up in the emotion of the scene 0:06:35.080 --> 0:06:38.960 and not even considering that your visual cortex is being 0:06:39.120 --> 0:06:46.120 fooled into believing something motion that is not actually there. Now, 0:06:46.160 --> 0:06:48.679 what strikes me. Is interesting is that we are so 0:06:48.880 --> 0:06:53.680 used to this that it's difficult to recreate for ourselves 0:06:53.960 --> 0:06:57.400 the absolute shock that people must have had when they 0:06:57.800 --> 0:07:00.600 saw this phenomenon for the first time. I mean, how 0:07:00.680 --> 0:07:04.080 stunning would that be to witness a series of still 0:07:04.160 --> 0:07:07.800 shots looking like they were moving. Keep in mind that 0:07:07.839 --> 0:07:10.560 in the entire history of the world before this moment, 0:07:11.000 --> 0:07:14.880 no one had ever had a chance to capture photons 0:07:14.920 --> 0:07:19.160 from the scene, make a photograph, and then swap those 0:07:19.200 --> 0:07:24.160 out so rapidly that it looks like smooth motion. It wasn't, 0:07:24.200 --> 0:07:27.160 in fact, even until eighteen sixty eight that somebody made 0:07:27.240 --> 0:07:30.480 a flip book. You remember those little books where you 0:07:30.560 --> 0:07:33.680 hold your thumb to zip through all the pages rapidly, 0:07:34.000 --> 0:07:38.080 and it looks like smooth motion of some drawing. And 0:07:38.280 --> 0:07:42.920 although the zootrope was around with little hand drawings since 0:07:42.960 --> 0:07:47.320 eighteen sixty six, no one had done this with photographs 0:07:47.560 --> 0:07:51.679 until Moybridge did. Before that, no one had ever seen 0:07:52.240 --> 0:07:56.200 motion pictures. I mean, just imagine the next time you're 0:07:56.200 --> 0:08:00.440 watching a TikTok video that a smart person like Benjamin 0:08:00.480 --> 0:08:05.200 Franklin went his whole life without ever seeing a moving picture. 0:08:05.640 --> 0:08:08.560 And just imagine how gobsmacked he would be to see that. 0:08:09.640 --> 0:08:12.840 And remember that this technology was so stunning that it 0:08:12.880 --> 0:08:16.440 came to be called a move e, as in something 0:08:16.720 --> 0:08:21.320 that seems to move, and we still call it that. Okay, now, 0:08:21.640 --> 0:08:24.840 how does this work? Exactly? So, whether you're talking about 0:08:24.880 --> 0:08:28.320 a zootrope or a flip book, or a movie projector 0:08:28.920 --> 0:08:31.280 or what happens on your cell phone screen when you're 0:08:31.320 --> 0:08:33.760 watching a YouTube video, it's all the same thing. It 0:08:33.840 --> 0:08:38.960 relies on the phenomenon of apparent motion. So what is 0:08:38.960 --> 0:08:43.600 a parent motion. It's a visual illusion that occurs when 0:08:43.720 --> 0:08:47.840 a series of still images are shown in rapid succession, 0:08:48.320 --> 0:08:52.800 and you get this perception of continuous, smooth movement. So 0:08:52.880 --> 0:08:55.800 that is the basis for the zootropes and the flip 0:08:55.800 --> 0:08:59.040 books and modern films and videos. Now, the reason this 0:08:59.160 --> 0:09:03.480 works is because the brain retains the impression of the 0:09:03.520 --> 0:09:07.040 previous image for just a brief moment after it's disappeared, 0:09:07.640 --> 0:09:11.400 and that creates a smooth transition to the next image. 0:09:12.040 --> 0:09:15.800 And this phenomenon of apparent motion is closely related to 0:09:15.920 --> 0:09:20.520 a different phenomenon called the persistence of vision, which is 0:09:20.760 --> 0:09:23.920 that your brain holds onto an image for a split 0:09:24.000 --> 0:09:27.640 second after it disappears, and that allows you to perceive 0:09:27.920 --> 0:09:32.920 smooth motion even when there are gaps between the individual images, 0:09:33.440 --> 0:09:36.960 like when you're spinning the cylinder the zootrope. You see 0:09:37.000 --> 0:09:40.160 a flash of image one, and then you don't see 0:09:40.200 --> 0:09:42.480 anything while the cylinder spins a bit more, and then 0:09:42.480 --> 0:09:44.880 you see a flash of image two, and then you 0:09:44.920 --> 0:09:48.360 don't see anything, and so on. But your brain clocks 0:09:48.400 --> 0:09:52.160 the image and retains it long enough that it bridges 0:09:52.200 --> 0:10:12.520 the gap between So it's because of apparent motion and 0:10:12.640 --> 0:10:16.720 persistence of vision that we can make and enjoy motion 0:10:16.880 --> 0:10:20.920 pictures or TV, or video games or virtual reality. Okay, 0:10:21.160 --> 0:10:24.120 now there's one more concept that's important here. Imagine that 0:10:24.280 --> 0:10:27.960 I held up the first photograph of Moybridge's horse, and 0:10:28.000 --> 0:10:30.480 then I picked up the second picture, and I held 0:10:30.520 --> 0:10:32.880 that in front of it, and then I took the 0:10:32.960 --> 0:10:35.440 third picture and held that up and so on. I'd 0:10:35.480 --> 0:10:39.560 be going way too slow to fool you into thinking 0:10:39.600 --> 0:10:43.560 there's motion. You might cognitively get it that there's a 0:10:43.760 --> 0:10:47.680 sequence being shown, but you wouldn't have the direct perceptual 0:10:47.720 --> 0:10:52.120 experience of motion. But if I flashed the photos quickly, 0:10:52.200 --> 0:10:56.360 then it works. So how quickly do I have to 0:10:56.440 --> 0:10:59.760 flash these. So to get at this answer, you can 0:10:59.760 --> 0:11:03.240 do a simple experiment where you just flash an LED 0:11:03.520 --> 0:11:06.000 light on and off and on and off, and if 0:11:06.000 --> 0:11:10.440 you're doing it slowly, like flash flash, then you see 0:11:10.480 --> 0:11:13.520 a flashing light. If you do it more quickly like 0:11:13.559 --> 0:11:17.040 flash flash, flash flash, then you still see that it's flashing. 0:11:17.440 --> 0:11:20.160 But of course it's a bit faster now. But now 0:11:20.440 --> 0:11:23.360 you would just a little bit faster than that, and 0:11:23.520 --> 0:11:28.520 suddenly the flashing light looks solid to you. You can't 0:11:28.559 --> 0:11:33.400 distinguish this from a light that's just on. So the 0:11:33.480 --> 0:11:37.000 speed of flashing at which the light suddenly looks like 0:11:37.040 --> 0:11:41.600 a continuous light is called the flicker fusion threshold. Because 0:11:41.640 --> 0:11:47.079 suddenly the flicker fuses to look like a solid your 0:11:47.120 --> 0:11:49.200 brain just can't see the fact that the light is 0:11:49.200 --> 0:11:53.000 turning on and off. So movies only work when the 0:11:53.080 --> 0:11:58.240 successive images are flashed faster than the flicker fusion threshold, 0:11:58.280 --> 0:12:00.880 so that you don't see one photo and then the 0:12:00.920 --> 0:12:05.240 next and the next, but instead you see a smooth transition. Now, 0:12:05.320 --> 0:12:07.600 how fast do you need to flash it? Well, the 0:12:07.640 --> 0:12:10.880 exact threshold that you can measure in people that changes 0:12:10.920 --> 0:12:13.320 a bit. When you're talking about the size of the 0:12:13.360 --> 0:12:16.400 thing flashing and the intensity, and whether it's black or 0:12:16.400 --> 0:12:19.959 white or color. But generally you need to flash something 0:12:20.120 --> 0:12:25.040 like forty times per second and then it looks smooth. Now, 0:12:25.040 --> 0:12:26.960 as a quick side note, you may know that old 0:12:27.040 --> 0:12:30.920 movies were shot at twenty four frames per second, so 0:12:31.400 --> 0:12:35.200 why didn't people see a flickering between the frames. The 0:12:35.320 --> 0:12:40.240 answer is that movie projectors would flash each frame two 0:12:40.520 --> 0:12:43.520 or sometimes three times before going to the next frame, 0:12:43.880 --> 0:12:47.640 so you had a flicker of forty eight or sometimes 0:12:47.640 --> 0:12:51.920 seventy two flashes per second, so the whole thing looked smooth. 0:12:52.320 --> 0:12:57.760 And similarly, television is traditionally shot at thirty frames per second, 0:12:58.200 --> 0:13:02.400 and in the same way, all the frames are doubled. Phones, 0:13:02.440 --> 0:13:06.520 by the way, typically use sixty refreshes per second, although 0:13:06.559 --> 0:13:08.680 a lot of high end phones have an even higher 0:13:08.679 --> 0:13:11.640 refresh rate. So the point is, when you look at 0:13:11.679 --> 0:13:14.960 any of these technologies, movies or television or phone, they're 0:13:15.000 --> 0:13:20.120 all flickering above the flicker fusion threshold, and so everything 0:13:20.200 --> 0:13:25.679 looks wonderfully smooth. So when you're walking around in the world, 0:13:25.720 --> 0:13:29.080 a lot of lighting can look like it's continuous, but 0:13:29.120 --> 0:13:32.239 you can demonstrate to yourself that it must be flickering 0:13:32.720 --> 0:13:36.280 because of strange effects that you can get. For example, 0:13:36.640 --> 0:13:41.600 many cars have moved to led headlights or blinkers that flicker, 0:13:42.040 --> 0:13:44.520 and you can't tell that when you're looking right at 0:13:44.559 --> 0:13:47.000 the light, But if you move your eyes to the side, 0:13:47.040 --> 0:13:50.400 you'll see a series of images of the headlight or 0:13:50.400 --> 0:13:53.920 the blinker because it hits your eye in different spots 0:13:54.000 --> 0:13:58.600 when it comes on. Or take those yellow street lights 0:13:58.640 --> 0:14:01.920 that you see. They're called sodium vapor lamps, and even 0:14:01.960 --> 0:14:06.040 though it looks like a solid light, they're actually flickering 0:14:06.160 --> 0:14:09.920 on and off with the alternating current, which is sixty 0:14:09.920 --> 0:14:12.679 times a second in America and fifty times a second 0:14:12.760 --> 0:14:17.040 in Europe. Now, the light looks solid because it's flickering 0:14:17.640 --> 0:14:22.520 faster than the critical flicker fusion frequency. But if you 0:14:23.040 --> 0:14:27.240 swing your hand around, you'll notice a strobing effect. You'll 0:14:27.240 --> 0:14:31.480 see multiple locations of your hand that are spaced apart, 0:14:31.920 --> 0:14:34.080 and the amount there spaced has to do with how 0:14:34.160 --> 0:14:37.520 fast you're moving your hand. So when you see these 0:14:37.560 --> 0:14:40.680 sorts of effects, that's how you know something is actually 0:14:40.680 --> 0:14:45.440 flickering even though it looks solid to you. Okay, so 0:14:45.480 --> 0:14:48.200 we see how movies work by flashing a bunch of images, 0:14:48.600 --> 0:14:51.840 and even though it looks like continuous motion, it's actually 0:14:52.320 --> 0:14:56.880 discrete frames. And this trick has revolutionized the way that 0:14:56.920 --> 0:15:01.080 we communicate information. Compared to two hundred years ago, we 0:15:01.200 --> 0:15:04.560 have YouTube and TikTok and television and video conferencing and 0:15:04.600 --> 0:15:10.280 so on. Cool. But this particular trick of stringing together 0:15:10.520 --> 0:15:16.320 discrete frames can yield strange illusions, and people started noticing 0:15:16.360 --> 0:15:21.240 these things during the first days of movies. So originally 0:15:21.280 --> 0:15:25.320 the movies were westerns with wagons, and the wagon would 0:15:25.320 --> 0:15:29.240 be rolling forward along the dirt road, but it often 0:15:29.280 --> 0:15:32.280 looked in the movie like the spokes were turning the 0:15:32.440 --> 0:15:35.360 other way, or sometimes the spokes would look like they 0:15:35.400 --> 0:15:38.200 were turning the right way but not at the right speed, 0:15:38.320 --> 0:15:43.920 or sometimes not turning at all. So why does that happen? Well, 0:15:44.200 --> 0:15:49.480 imagine the wagon wheel turning clockwise. The camera captures a 0:15:49.520 --> 0:15:52.240 frame where the wheel is in this position, and then 0:15:52.280 --> 0:15:55.800 when the next frame is captured some tens of milliseconds later, 0:15:56.240 --> 0:15:59.120 the wheel has turned a bit and the spokes are 0:15:59.160 --> 0:16:03.200 in a slightly position. But here's the tricky part. All 0:16:03.320 --> 0:16:06.920 the spokes look alike, and so your brain's job is 0:16:06.960 --> 0:16:09.680 to figure out how the spokes in the first frame 0:16:10.160 --> 0:16:13.200 match up to the spokes in the second frame. And 0:16:13.320 --> 0:16:16.200 generally it can only do this by looking for the 0:16:16.280 --> 0:16:20.840 shortest distance that things must have changed. So, let's say 0:16:20.880 --> 0:16:24.600 the spoke has rotated ninety percent of the way to 0:16:24.680 --> 0:16:29.240 the position of the next spoke. The brain will erroneously 0:16:29.320 --> 0:16:33.280 think it has rotated the other way, because from frame 0:16:33.360 --> 0:16:36.440 one to frame two, your brain sees that the shortest 0:16:36.480 --> 0:16:39.760 distance is not a rotation this way, but a rotation 0:16:39.840 --> 0:16:43.560 the other way. And so this effect came to be 0:16:43.640 --> 0:16:47.800 known as the wagon wheel illusion. And now that you're 0:16:47.800 --> 0:16:49.320 going to keep an eye out for it, you're gonna 0:16:49.320 --> 0:16:52.880 see this everywhere. In car commercials, the wheel doesn't appear 0:16:52.920 --> 0:16:56.440 to be turning correctly, but instead the turning of the 0:16:56.520 --> 0:16:59.920 hubcap seems to run the wrong way sometimes or slow 0:17:00.120 --> 0:17:03.120 down in its rotation to a halt, even though the 0:17:03.160 --> 0:17:05.880 car is zooming down the highway. And you can see 0:17:05.880 --> 0:17:09.359 this with helicopters or drones on your television or on 0:17:09.400 --> 0:17:12.159 your phone. It always looks like something is wrong. The 0:17:12.280 --> 0:17:15.680 rotor blades are hardly turning it all, or maybe they 0:17:15.720 --> 0:17:18.680 turn the other way and the helicopter lifts up even 0:17:18.680 --> 0:17:21.840 though it doesn't make any sense. There's no adequate spinning 0:17:21.960 --> 0:17:28.400 to make that happen. This is all the wagon wheel effect. Okay, 0:17:28.480 --> 0:17:32.160 so now we are all set up for an observation 0:17:32.280 --> 0:17:35.160 that came as a big surprise in the neuroscience world. 0:17:35.359 --> 0:17:38.359 Some colleagues of Mind published a paper in nineteen ninety 0:17:38.400 --> 0:17:41.719 six pointing out that you could get these kind of 0:17:41.800 --> 0:17:46.199 motion illusions not just in movies, but in real life 0:17:46.400 --> 0:17:49.920 under a continuous light. Now this was a big claim 0:17:50.440 --> 0:17:54.720 because it suggested the possibility that our brains are actually 0:17:54.760 --> 0:17:58.520 seeing in snapshots like the frames of a video camera. 0:17:59.280 --> 0:18:01.959 Is that true or not true? Well, I'm gonna come 0:18:02.000 --> 0:18:03.960 to that in a second, But first I want to 0:18:04.000 --> 0:18:08.359 tell you how to experience the illusion. So, if you 0:18:08.440 --> 0:18:12.320 have a ceiling fan that's turning, lie back on your 0:18:12.320 --> 0:18:15.479 bed and stare at the fan. Do this in the 0:18:15.520 --> 0:18:18.520 middle of the day with no lights on, so there's 0:18:18.560 --> 0:18:22.000 no flickering lights. There's only sunlight. Now, let's say your 0:18:22.080 --> 0:18:25.960 fan is turning clockwise and you stare at it. You'll 0:18:26.000 --> 0:18:28.639 see that it's turning clockwise. But if you stare and 0:18:28.760 --> 0:18:34.080 stare long enough, occasionally you will see it turn the 0:18:34.119 --> 0:18:36.840 other way. For just a second or two, the fan 0:18:37.080 --> 0:18:41.600 seems to reverse direction. So please try this, although it 0:18:41.640 --> 0:18:44.399 might take a few minutes of just staring and staring 0:18:44.440 --> 0:18:46.400 at it before you see it, but you'll know when 0:18:46.440 --> 0:18:49.439 you see it the fan suddenly runs backward for just 0:18:49.480 --> 0:18:52.119 a moment. Now, I just want to be clear that 0:18:52.600 --> 0:18:55.920 you can see this illusion in the day under sunlight, 0:18:56.040 --> 0:18:59.520 so this is not explained by something like a subtle 0:18:59.840 --> 0:19:19.880 flickering of the lights because of the electrical current. Now, 0:19:19.920 --> 0:19:23.040 this illusion with the fan seems like a subtle thing 0:19:23.080 --> 0:19:25.560 that no one would care about, but it caused a 0:19:25.560 --> 0:19:29.360 lot of discussion in the neuroscience community when this was noticed. 0:19:29.800 --> 0:19:33.240 Even one of my mentors, Francis Crick, the co discoverer 0:19:33.320 --> 0:19:36.120 of the structure of DNA, he wrote about this too, 0:19:36.480 --> 0:19:38.080 And I'll tell you why it was such a big 0:19:38.160 --> 0:19:42.640 deal to neuroscientists because it wasn't clear why this would 0:19:42.680 --> 0:19:45.920 happen in the visual system, which seems to be taking 0:19:45.960 --> 0:19:51.439 in information continuously. So the hypothesis that took hold was 0:19:51.440 --> 0:19:54.359 that maybe the brain is taking in the world in 0:19:54.440 --> 0:19:58.160 frames like a movie camera. And this is what the 0:19:58.200 --> 0:20:03.120 researchers who found this suggested. They suggested that vision might 0:20:03.200 --> 0:20:05.919 be like the filming of a spoked wheel with a 0:20:06.040 --> 0:20:10.040 video camera, and that perhaps this illusion with the fan 0:20:10.800 --> 0:20:15.080 was evidence of discrete perception, in other words, that we 0:20:15.240 --> 0:20:19.040 see in frames like a video camera. Now, if the 0:20:19.080 --> 0:20:24.080 claim were true, that's a big deal. But I started 0:20:24.119 --> 0:20:28.080 to suspect that something wasn't right here, because, first of all, 0:20:28.400 --> 0:20:31.760 there are some important differences between watching the fan in 0:20:31.800 --> 0:20:36.520 the daylight and watching the wagon wheel effect in movies. First, 0:20:37.119 --> 0:20:39.800 in the movie, if you have the wagon rolling at 0:20:39.800 --> 0:20:43.639 a particular speed, the wheel seems to be going backwards 0:20:43.680 --> 0:20:46.560 at a fixed speed the whole time. But that doesn't 0:20:46.600 --> 0:20:49.439 happen with the fan. It only happens for just a 0:20:49.480 --> 0:20:53.400 moment after you've been staring at it for a while. Second, 0:20:53.440 --> 0:20:56.200 when you see the fan reverse, it seems to happen 0:20:56.359 --> 0:20:59.760 at a faster than normal speed, even though in the 0:20:59.800 --> 0:21:03.120 movies the reverse spinning of the wheel is always slower. 0:21:03.880 --> 0:21:07.600 And Third, in the movies, sometimes the wagon wheel effect 0:21:07.680 --> 0:21:10.919 can look like the wheel is stopped because the camera 0:21:11.040 --> 0:21:13.159 keeps catching the wheel when the spokes are in the 0:21:13.200 --> 0:21:16.200 same position, so it looks like nothing's turning. But that 0:21:16.320 --> 0:21:19.280 never happens with the fan in daylight. You never see 0:21:19.320 --> 0:21:22.960 it look like it stopped, And little problems like that 0:21:23.119 --> 0:21:27.439 started making me suspicious that maybe the wagon wheel affect 0:21:27.480 --> 0:21:31.080 in movies and this issue about the fan reversing under 0:21:31.200 --> 0:21:36.240 daylight had only a superficial similarity, and the two effects 0:21:36.880 --> 0:21:41.240 had totally different reasons for actually happening. And I really 0:21:41.320 --> 0:21:44.800 wanted to understand this very fundamental point about the visual system. 0:21:45.080 --> 0:21:47.000 So what I did is I drove to the pawn 0:21:47.040 --> 0:21:50.880 shop and I bought an old record player for eight dollars, 0:21:51.359 --> 0:21:53.760 and then I spent two dollars at the art store 0:21:53.800 --> 0:21:56.960 to get a circle of styrofoam picture this like a 0:21:56.960 --> 0:22:00.800 big hockey puck. And then with my student Keith Klein 0:22:00.880 --> 0:22:05.199 and my colleague Alex Holcomb, we put evenly spaced black 0:22:05.320 --> 0:22:08.679 dots all around the outside of the styrofoam puck, and 0:22:08.720 --> 0:22:11.960 we put that on the record player. So when you 0:22:12.000 --> 0:22:14.360 flick on the record player and look at it from 0:22:14.359 --> 0:22:17.960 the side, you see these little black dots moving from 0:22:18.080 --> 0:22:21.920 left to right. Now, why did I use a record 0:22:21.920 --> 0:22:25.439 player instead of a computer screen, Because the computer screen 0:22:25.960 --> 0:22:28.640 inherently has a frame rate, and I wanted to make 0:22:28.680 --> 0:22:33.160 sure we were really doing smooth motion here, continuous motion, 0:22:34.119 --> 0:22:37.159 and the way this experiment goes when you watch this 0:22:37.320 --> 0:22:39.760 record player in the daylight next to a big window 0:22:39.800 --> 0:22:42.440 with no lights on. That's how we made sure there 0:22:42.480 --> 0:22:45.680 was no flicker that influenced anything. When you stare at 0:22:45.680 --> 0:22:49.199 these smoothly moving dots going from left to right, and 0:22:49.240 --> 0:22:51.320 you keep your eyes in one place and you stare 0:22:51.359 --> 0:22:54.240 and stare for a few minutes, it eventually works. You 0:22:54.280 --> 0:22:58.679 see the stream of dots suddenly reverse direction, just for 0:22:58.720 --> 0:23:02.040 a few seconds. It looks like they're zipping from right 0:23:02.080 --> 0:23:04.520 to left, and then you see things going back to 0:23:04.560 --> 0:23:09.040 normal again. So we were able to reproduce the illusion. 0:23:09.560 --> 0:23:14.080 But now here was the important trick. We now placed 0:23:14.119 --> 0:23:17.480 a mirror right next to the record player. So now 0:23:17.520 --> 0:23:21.879 what you see are two pucks rotating side by side 0:23:21.920 --> 0:23:25.919 in opposite directions. So you're seeing one stream of dots 0:23:25.960 --> 0:23:29.320 moving left to right and the other reflected stream of 0:23:29.359 --> 0:23:33.439 dots moving right to left. Now, the question is do 0:23:33.480 --> 0:23:37.960 you see both streams of dots reverse direction at the 0:23:38.000 --> 0:23:41.760 same time, or do you see one puck reverse and 0:23:41.800 --> 0:23:44.520 then maybe later the other puck reverses, and then the 0:23:44.560 --> 0:23:50.320 first puck reverses again. If your visual system is snapping frames, 0:23:51.000 --> 0:23:53.320 and here we have to assume that the length of 0:23:53.359 --> 0:23:56.800 the frames are changing for some reason. If your visual 0:23:56.800 --> 0:24:00.479 system is snapping frames, then both pucks should reverse at 0:24:00.520 --> 0:24:04.040 the same time, because that's what would happen if you 0:24:04.080 --> 0:24:07.719 were filming with a video camera whose frame rate was changing. 0:24:08.359 --> 0:24:13.359 But if the two pucks reverse independently, that suggests something 0:24:13.440 --> 0:24:16.679 very different is going on. So you keep doing gets 0:24:16.720 --> 0:24:20.040 fixed right in the middle between these two streams of dots, 0:24:20.680 --> 0:24:24.840 and what happens. What happens is you see one puck reverse, 0:24:25.160 --> 0:24:27.399 and then the other, and then the first one again. 0:24:27.920 --> 0:24:32.760 And this result seems to rule out snapshots, because again, 0:24:32.800 --> 0:24:36.560 if our brains processed visual information like a camera in 0:24:36.680 --> 0:24:40.600 discrete frames, then you would expect both sets of dots 0:24:40.680 --> 0:24:43.800 to always switch directions at the same time. But that's 0:24:43.840 --> 0:24:47.240 not what happens. Now. You might argue a point here, 0:24:47.280 --> 0:24:49.879 which is that because I had people stare right in 0:24:50.000 --> 0:24:53.200 between the dots, they were seeing one puck in the 0:24:53.320 --> 0:24:56.680 left hemisphere and the reflected puck in the right hemisphere. 0:24:57.040 --> 0:25:00.000 And what if perhaps the two hemispheres of the brain 0:25:00.600 --> 0:25:04.719 both do snapshots, but with independent frame rates. So I 0:25:04.800 --> 0:25:08.320 address that frame in fifteen seconds by turning the whole 0:25:08.359 --> 0:25:11.240 contraption on its side, so that you could see both 0:25:11.320 --> 0:25:14.439 pucks in the same hemisphere and you get the same result. 0:25:14.520 --> 0:25:19.720 The pucks reverse direction independently at different times from one another. 0:25:20.440 --> 0:25:24.280 So it appears that the visual system is not taking snapshots. 0:25:24.280 --> 0:25:28.000 But what is the explanation here, Well, you have some 0:25:28.400 --> 0:25:32.120 populations of neurons in your visual cortex that detect right 0:25:32.160 --> 0:25:36.160 word motion, and when those are active, you say, ah, 0:25:36.200 --> 0:25:39.240 there's clearly right word motion in the world. But you 0:25:39.320 --> 0:25:42.880 also have populations of cells that pick up on leftward motion. 0:25:43.320 --> 0:25:47.240 And these populations are always balanced in a competition. But 0:25:47.400 --> 0:25:50.920 here's the key, for technical reasons that you can read 0:25:50.920 --> 0:25:53.440 about in my papers on this. It turns out that 0:25:53.480 --> 0:25:58.640 those leftward populations can sometimes be fooled by a lot 0:25:58.680 --> 0:26:01.360