1 00:00:00,000 --> 00:00:02,740 Welcome to the Deep Dive. Today we are tackling 2 00:00:02,740 --> 00:00:05,480 something that, well, seems utterly simple on 3 00:00:05,480 --> 00:00:07,259 the surface but is actually incredibly profound. 4 00:00:07,799 --> 00:00:11,359 How does a child learn what a word means? You 5 00:00:11,359 --> 00:00:13,380 know, that moment a toddler points and says car 6 00:00:13,380 --> 00:00:17,100 correctly. Looks easy. Almost automatic. It really 7 00:00:17,100 --> 00:00:21,559 does. But that apparent simplicity is deceptive. 8 00:00:21,679 --> 00:00:23,600 We're diving into Paul Bloom's work today and 9 00:00:23,600 --> 00:00:25,640 he argues pretty convincingly that learning a 10 00:00:25,640 --> 00:00:29,370 word is, well, it's basically a miracle. Kind 11 00:00:29,370 --> 00:00:32,020 of a miracle. That sounds strong. It is, but 12 00:00:32,020 --> 00:00:33,840 I think it's justified. For cognitive science, 13 00:00:34,119 --> 00:00:35,880 this isn't just about kids learning vocabulary 14 00:00:35,880 --> 00:00:38,619 lists. It hits right at the core of how language 15 00:00:38,619 --> 00:00:40,939 connects to thought, you know, and what human 16 00:00:40,939 --> 00:00:43,780 learning is even like. Okay. So our mission for 17 00:00:43,780 --> 00:00:46,210 this deep dive is to unpack that miracle. We're 18 00:00:46,210 --> 00:00:48,149 following Bloom's analysis chapter by chapter, 19 00:00:48,270 --> 00:00:50,270 looking at the whole journey. It sounds like 20 00:00:50,270 --> 00:00:51,929 it involves way more than just hearing a word 21 00:00:51,929 --> 00:00:54,350 and seeing a thing. We're talking concepts, social 22 00:00:54,350 --> 00:00:57,210 smarts, understanding language structure. Exactly. 23 00:00:57,289 --> 00:01:01,229 Yeah. Just to map a simple sound like E -O -G 24 00:01:01,229 --> 00:01:07,420 onto the incredibly rich idea of dog. It's not 25 00:01:07,420 --> 00:01:09,920 just association. We need to explore the foundations 26 00:01:09,920 --> 00:01:13,359 that make it possible for you, for me, for everyone. 27 00:01:13,540 --> 00:01:15,879 All right. Let's jump in. Where do we start with 28 00:01:15,879 --> 00:01:18,299 the big philosophical problem children to solve? 29 00:01:18,400 --> 00:01:20,560 We have to start there. Yeah. Yeah. Because it 30 00:01:20,560 --> 00:01:22,659 highlights why just looking and listening isn't 31 00:01:22,659 --> 00:01:25,040 enough. It's the problem of reference. Quine's 32 00:01:25,040 --> 00:01:28,019 gava guy thought experiment is the classic illustration. 33 00:01:28,319 --> 00:01:30,739 Ah, yes, gava guy. So remind us the setup is 34 00:01:30,739 --> 00:01:32,760 a linguist in a completely foreign land, right? 35 00:01:32,799 --> 00:01:34,420 Exactly. You're with a native speaker. You've 36 00:01:34,420 --> 00:01:37,140 got no dictionary, nothing. A rabbit hops by 37 00:01:37,140 --> 00:01:39,700 the speaker points and says gava guy. And naturally, 38 00:01:39,799 --> 00:01:42,459 we assume gava guy means rabbit, the whole rabbit. 39 00:01:42,579 --> 00:01:45,519 We do. But how do we know? Or how does a child 40 00:01:45,519 --> 00:01:49,260 know? Logically, gava guy could mean, well, almost 41 00:01:49,260 --> 00:01:51,319 anything related to that scene. Like what? It 42 00:01:51,319 --> 00:01:53,120 could mean the whole rabbit, sure. Or maybe just 43 00:01:53,120 --> 00:01:55,379 that specific rabbit at that specific moment 44 00:01:55,379 --> 00:01:58,019 in time, a rabbit stage. Or it could mean its 45 00:01:58,019 --> 00:02:01,680 color, whiteness, or its action, hopping. Or 46 00:02:01,680 --> 00:02:06,200 Quine's famous slightly absurd example. All and 47 00:02:06,200 --> 00:02:09,280 sundry, undetached parts of rabbits. Undetached 48 00:02:09,280 --> 00:02:11,840 rabbit parts. Right, so the physical event itself, 49 00:02:12,060 --> 00:02:14,979 seeing the rabbit. Hearing the sound doesn't 50 00:02:14,979 --> 00:02:17,780 actually narrow it down. Not logically, no. Yeah. 51 00:02:17,900 --> 00:02:20,240 There's just this huge, potentially infinite 52 00:02:20,240 --> 00:02:23,539 space of possible meanings. The world doesn't 53 00:02:23,539 --> 00:02:25,960 tell you which one is right. So the child somehow 54 00:02:25,960 --> 00:02:28,500 has to filter out all these crazy possibilities 55 00:02:28,500 --> 00:02:31,319 even before they start learning? Precisely. And 56 00:02:31,319 --> 00:02:32,939 that's just the first hurdle. Let's say, okay, 57 00:02:32,979 --> 00:02:35,080 somehow the child does land on a whole rabbit. 58 00:02:35,199 --> 00:02:37,840 Now they face Goodman's problem. Generalization. 59 00:02:38,120 --> 00:02:40,639 How do they know to apply Gavagai to the next 60 00:02:40,639 --> 00:02:43,270 rabbit they see? Or a picture of a rabbit. Exactly. 61 00:02:43,270 --> 00:02:45,949 Why not extend it to mean white things or fluffy 62 00:02:45,949 --> 00:02:47,990 things? Right. Or Goodman's even weirder example, 63 00:02:48,210 --> 00:02:51,090 rabbits. But only up until the year 2000, after 64 00:02:51,090 --> 00:02:53,229 which it means carrots. Okay, that's just bizarre, 65 00:02:53,490 --> 00:02:56,310 but it makes the point. It does. There's no logical 66 00:02:56,310 --> 00:02:58,810 reason, based only on the evidence, to prefer 67 00:02:58,810 --> 00:03:01,310 the category rabbit over these other stranger 68 00:03:01,310 --> 00:03:03,629 categories. The solution can't be found in logic 69 00:03:03,629 --> 00:03:07,389 applied to the input. So, it must be inside the 70 00:03:07,389 --> 00:03:10,409 child's head, the structure of their mind. That's 71 00:03:10,409 --> 00:03:13,650 the conclusion. The human mind must come equipped 72 00:03:13,650 --> 00:03:17,069 with certain biases or constraints that favor 73 00:03:17,069 --> 00:03:20,189 certain kinds of concepts, like whole object 74 00:03:20,189 --> 00:03:23,110 over others. It's the mind structure, not the 75 00:03:23,110 --> 00:03:26,210 world structure, that solves this. And this internal 76 00:03:26,210 --> 00:03:30,050 structure explains why word learning is so robust, 77 00:03:30,430 --> 00:03:32,990 even when the learning conditions aren't perfect. 78 00:03:33,229 --> 00:03:35,490 Absolutely. We tend to picture this ideal scene, 79 00:03:35,710 --> 00:03:38,469 parent points, child looks, parent says the word, 80 00:03:38,569 --> 00:03:41,349 perfect alignment. The classic naming game. Right. 81 00:03:41,599 --> 00:03:44,120 But research shows that even in, say, typical 82 00:03:44,120 --> 00:03:46,080 Western households where parents do interact 83 00:03:46,080 --> 00:03:48,719 a lot, that perfect alignment only happens maybe 84 00:03:48,719 --> 00:03:51,819 70 % of the time. So 30 % of the time, the word 85 00:03:51,819 --> 00:03:53,759 the child hears doesn't match what they're looking 86 00:03:53,759 --> 00:03:55,840 at. Yeah. The parent might be talking about something 87 00:03:55,840 --> 00:03:58,180 out of sight, or the child might be looking at, 88 00:03:58,199 --> 00:04:00,319 I don't know, a fork, while the parent says, 89 00:04:00,520 --> 00:04:02,539 time for your milk. If it were just association, 90 00:04:02,860 --> 00:04:04,680 wouldn't the child sometimes think milk means 91 00:04:04,680 --> 00:04:07,060 fork? You'd expect those errors all the time. 92 00:04:07,379 --> 00:04:10,969 But they're incredibly rare. Children seem remarkably 93 00:04:10,969 --> 00:04:13,729 good at tuning out that noise and figuring out 94 00:04:13,729 --> 00:04:16,129 the intended meaning anyway. And they don't need 95 00:04:16,129 --> 00:04:18,490 constant correction or feedback either. That's 96 00:04:18,490 --> 00:04:21,910 another big myth. The idea that parents meticulously 97 00:04:21,910 --> 00:04:25,470 correct every misuse. First, evidence suggests 98 00:04:25,470 --> 00:04:28,290 they don't. Not really for meaning anyway. Maybe 99 00:04:28,290 --> 00:04:31,709 for grammar later. But more importantly, learning 100 00:04:31,709 --> 00:04:35,139 happens even without feedback loops. Think about 101 00:04:35,139 --> 00:04:37,439 cultures where adults don't really talk to young 102 00:04:37,439 --> 00:04:39,920 children until the children start talking themselves. 103 00:04:40,339 --> 00:04:42,339 They still learn language just fine. They do. 104 00:04:42,660 --> 00:04:45,439 And consider cases of children who are nonverbal, 105 00:04:45,459 --> 00:04:47,740 maybe due to a physical issue, but have normal 106 00:04:47,740 --> 00:04:50,839 hearing. They can develop huge vocabularies just 107 00:04:50,839 --> 00:04:52,779 by listening. How can they be getting feedback 108 00:04:52,779 --> 00:04:55,519 if they can't speak? They can't. It proves that 109 00:04:55,519 --> 00:04:57,660 reinforcement, in the behavioral sense, isn't 110 00:04:57,660 --> 00:05:00,139 the driver for learning word meanings. They build 111 00:05:00,139 --> 00:05:02,180 the knowledge internally. Okay, that's compelling. 112 00:05:02,399 --> 00:05:04,519 But what about cases where the input itself is 113 00:05:04,519 --> 00:05:07,420 limited? The example of the blind child, Kelly, 114 00:05:07,600 --> 00:05:10,040 seems crucial here. Kelly's case is fascinating. 115 00:05:10,279 --> 00:05:13,040 She was blind from birth, yet she developed a 116 00:05:13,040 --> 00:05:15,000 basically normal understanding of color words. 117 00:05:15,220 --> 00:05:17,720 How is that possible if she couldn't see colors? 118 00:05:18,290 --> 00:05:20,029 Well, she obviously couldn't link the word red 119 00:05:20,029 --> 00:05:22,810 to the visual experience of red. But she learned 120 00:05:22,810 --> 00:05:25,430 about color words. She knew they formed a category. 121 00:05:25,449 --> 00:05:27,990 She knew they applied to objects. She knew red 122 00:05:27,990 --> 00:05:31,149 and blue were colors, but, say, loud wasn't. 123 00:05:31,209 --> 00:05:34,170 So she learned the conceptual domain even without 124 00:05:34,170 --> 00:05:36,839 the sensory data. Exactly. She must have picked 125 00:05:36,839 --> 00:05:39,279 it up from how people talked about colors, the 126 00:05:39,279 --> 00:05:41,959 linguistic context, the social function. It shows 127 00:05:41,959 --> 00:05:44,019 how powerful these other learning routes are, 128 00:05:44,120 --> 00:05:47,139 strong enough to overcome a complete lack of 129 00:05:47,139 --> 00:05:49,879 direct sensory evidence for that specific domain. 130 00:05:50,180 --> 00:05:52,980 So if it's not just sensing and it's robust against 131 00:05:52,980 --> 00:05:56,139 noise and lack of feedback, what is being learned? 132 00:05:56,240 --> 00:05:58,240 What's the structure of a word's meaning in the 133 00:05:58,240 --> 00:06:01,420 mind? This brings us back to basics, to Saussurean 134 00:06:01,420 --> 00:06:04,370 signs. Essentially, learning a word means forming 135 00:06:04,370 --> 00:06:07,389 a mental concept, the signified, and linking 136 00:06:07,389 --> 00:06:10,050 it to a sound pattern, the signifier. Like linking 137 00:06:10,050 --> 00:06:13,110 the idea of a dog to the sound dog. Precisely. 138 00:06:13,329 --> 00:06:17,930 And crucially, that link is arbitrary. There's 139 00:06:17,930 --> 00:06:20,790 nothing dog -like about the sound itself. It's 140 00:06:20,790 --> 00:06:22,970 a convention we learn. And these concepts aren't 141 00:06:22,970 --> 00:06:24,730 just simple labels. You mentioned the word book 142 00:06:24,730 --> 00:06:27,410 having complex semantics. Yeah, even seemingly 143 00:06:27,410 --> 00:06:30,430 simple nouns are incredibly rich. Chomsky pointed 144 00:06:30,430 --> 00:06:32,740 this out. Book. can mean the physical object. 145 00:06:33,019 --> 00:06:37,680 I have five books. Countable, tangible. Right. 146 00:06:37,879 --> 00:06:40,399 Or it can mean the abstract content. John wrote 147 00:06:40,399 --> 00:06:43,540 a book. That's the information, the story. It's 148 00:06:43,540 --> 00:06:45,560 not countable in the same way. It can exist in 149 00:06:45,560 --> 00:06:48,180 many copies or just as an idea. And kids just 150 00:06:48,180 --> 00:06:50,879 absorb this distinction. They seem to, yeah. 151 00:06:51,259 --> 00:06:53,379 Without anyone explicitly teaching them about 152 00:06:53,379 --> 00:06:56,069 abstract versus concrete entities, They grasp 153 00:06:56,069 --> 00:06:58,230 this polysemy, the multiple related meanings, 154 00:06:58,509 --> 00:07:00,649 which shows they're acquiring deep concepts, 155 00:07:00,829 --> 00:07:03,410 not just linking sounds to pictures. Okay, so 156 00:07:03,410 --> 00:07:05,610 wrapping up this first part. Word learning isn't 157 00:07:05,610 --> 00:07:07,970 some simple add -on. It relies on really fundamental 158 00:07:07,970 --> 00:07:10,230 cognitive abilities, how we form concepts, how 159 00:07:10,230 --> 00:07:11,949 we remember, how we figure out what others mean. 160 00:07:12,410 --> 00:07:14,790 Exactly. And building on that complexity, the 161 00:07:14,790 --> 00:07:17,189 next chapter tackles the sheer speed of it all. 162 00:07:17,430 --> 00:07:19,949 This leads us to fast mapping. Fast mapping. 163 00:07:20,089 --> 00:07:22,879 That's the idea that... Kids can learn a word 164 00:07:22,879 --> 00:07:25,120 from just one or two exposures. Oh, it's just 165 00:07:25,120 --> 00:07:27,620 one, yeah. And sometimes in really casual, non 166 00:07:27,620 --> 00:07:30,600 -obvious contexts, it's incredibly efficient. 167 00:07:30,980 --> 00:07:33,920 The classic study here is Carrie and Bartlett's 168 00:07:33,920 --> 00:07:37,060 chromium tray experiment. That's the one. It 169 00:07:37,060 --> 00:07:39,300 really shifted the conversation. They had these 170 00:07:39,300 --> 00:07:41,160 three - and four -year -olds, and they needed 171 00:07:41,160 --> 00:07:44,259 an olive green tray. The experimenter casually 172 00:07:44,259 --> 00:07:46,920 said, bring me the chromium tray, not the blue 173 00:07:46,920 --> 00:07:49,360 one, the chromium one. So just contrasting it 174 00:07:49,360 --> 00:07:52,290 with a known color, blue. Right. And the kids, 175 00:07:52,370 --> 00:07:55,470 just from that single exposure, inferred that 176 00:07:55,470 --> 00:07:57,930 chromium must be the name of this new color, 177 00:07:58,069 --> 00:08:00,329 olive green. Okay, that's quick learning. But 178 00:08:00,329 --> 00:08:02,970 did it stick? That was the amazing part. Weeks 179 00:08:02,970 --> 00:08:05,290 later, like one week and even six weeks later, 180 00:08:05,449 --> 00:08:07,509 many of the children still remembered something 181 00:08:07,509 --> 00:08:10,250 about chromium. Maybe not the exact shade, but 182 00:08:10,250 --> 00:08:13,100 they knew it was a color word. Wow. From one 183 00:08:13,100 --> 00:08:15,360 casual mention, that does seem like a specialized 184 00:08:15,360 --> 00:08:18,120 skill. It seemed that way initially, like maybe 185 00:08:18,120 --> 00:08:20,259 kids have a dedicated word learning device that 186 00:08:20,259 --> 00:08:22,980 does this fast mapping. But then Bloom and Markson 187 00:08:22,980 --> 00:08:25,399 did some crucial follow -up work. What did they 188 00:08:25,399 --> 00:08:27,939 find? They basically showed that this fast mapping 189 00:08:27,939 --> 00:08:31,600 ability is not unique to words. Kids, and adults 190 00:08:31,600 --> 00:08:34,419 too, are just as good at fast mapping other kinds 191 00:08:34,419 --> 00:08:37,419 of arbitrary facts. Such as? Well, in their studies, 192 00:08:37,460 --> 00:08:40,000 people were shown a novel object. They were told 193 00:08:40,000 --> 00:08:42,429 its name, say, Koba. And they were also told 194 00:08:42,429 --> 00:08:44,429 an arbitrary fact about it, like it came from 195 00:08:44,429 --> 00:08:47,350 a place called Coba. Okay, name versus a non 196 00:08:47,350 --> 00:08:50,429 -linguistic fact. Right. And a month later, people 197 00:08:50,429 --> 00:08:52,490 were just as likely to remember the arbitrary 198 00:08:52,490 --> 00:08:54,929 fact about its origin as they were to remember 199 00:08:54,929 --> 00:08:58,110 its name. Fast mapping wasn't word -specific. 200 00:08:58,830 --> 00:09:01,879 So... If it's not just for words, what is it 201 00:09:01,879 --> 00:09:04,840 for? Why are we so good at instantly learning 202 00:09:04,840 --> 00:09:07,360 certain kinds of arbitrary information, but maybe 203 00:09:07,360 --> 00:09:09,759 not others, like where I left my keys? That's 204 00:09:09,759 --> 00:09:12,039 the million dollar question they tackled. Why 205 00:09:12,039 --> 00:09:15,080 names and facts, but not, say, the arbitrary 206 00:09:15,080 --> 00:09:16,960 location of a sticker you just placed? Yeah. 207 00:09:17,059 --> 00:09:20,600 The hypothesis is that fast mapping is specifically 208 00:09:20,600 --> 00:09:23,460 tuned for socially transmitted information that 209 00:09:23,460 --> 00:09:25,600 you can't just figure out by looking. Things 210 00:09:25,600 --> 00:09:28,539 like names, origins. Maybe hidden properties. 211 00:09:28,559 --> 00:09:30,980 It means you need someone else to tell you. Exactly. 212 00:09:31,000 --> 00:09:33,379 Think about it functionally. An object's name 213 00:09:33,379 --> 00:09:37,429 is a social convention. Its history, or where 214 00:09:37,429 --> 00:09:40,389 it came from, is a fact stored in other people's 215 00:09:40,389 --> 00:09:42,990 heads. If you forget it, you need an informant. 216 00:09:43,029 --> 00:09:44,889 Whereas if I forget where I put the sticker... 217 00:09:44,889 --> 00:09:46,269 You can just look around. The information is 218 00:09:46,269 --> 00:09:48,860 stored in the world. The idea is that the brain 219 00:09:48,860 --> 00:09:51,360 has this rapid uptake mechanism specifically 220 00:09:51,360 --> 00:09:54,059 for information that's fragile, that relies on 221 00:09:54,059 --> 00:09:55,919 social transmission, because if you don't grab 222 00:09:55,919 --> 00:09:58,519 it now, you might lose access to it. Is there 223 00:09:58,519 --> 00:10:00,820 direct evidence for this distinction between 224 00:10:00,820 --> 00:10:03,720 social hidden facts and observable facts? There 225 00:10:03,720 --> 00:10:06,600 is. They did another clever study, showed people 226 00:10:06,600 --> 00:10:09,799 a novel object, told them its name, told them 227 00:10:09,799 --> 00:10:11,940 its external visible color, and told them its 228 00:10:11,940 --> 00:10:14,580 internal hidden color like it's white on the 229 00:10:14,580 --> 00:10:17,840 inside. Okay. Name, hidden fact, observable fact. 230 00:10:18,139 --> 00:10:20,019 And what they found was that people remembered 231 00:10:20,019 --> 00:10:24,100 the name and the internal color, the hidden non 232 00:10:24,100 --> 00:10:27,000 -observable fact, significantly better than they 233 00:10:27,000 --> 00:10:29,659 remember the external observable color. Wow. 234 00:10:29,899 --> 00:10:32,159 So we prioritize encoding this stuff. We can't 235 00:10:32,159 --> 00:10:35,100 just check by looking again. It seems so. It 236 00:10:35,100 --> 00:10:37,539 reframes fast mapping away from being a purely 237 00:10:37,539 --> 00:10:40,000 linguistic thing. It looks more like a general 238 00:10:40,000 --> 00:10:43,259 super fast learning mechanism for crucial social 239 00:10:43,259 --> 00:10:46,200 or hidden information. Okay, that makes sense. 240 00:10:46,700 --> 00:10:49,379 Now, speaking of the timing of learning, we have 241 00:10:49,379 --> 00:10:52,000 to talk about the word spurt. This idea that 242 00:10:52,000 --> 00:10:54,840 vocabulary suddenly takes off around 16 or 18 243 00:10:54,840 --> 00:10:57,179 months, is that real? It's a very persistent 244 00:10:57,179 --> 00:10:59,740 idea, but the evidence suggests it's largely 245 00:10:59,740 --> 00:11:02,620 a myth or at least an oversimplification. Why 246 00:11:02,620 --> 00:11:04,879 do people think it happened? Well, if you just 247 00:11:04,879 --> 00:11:06,960 plot the total number of words a child knows, 248 00:11:07,159 --> 00:11:09,519 it looks like the curve suddenly gets steeper 249 00:11:09,519 --> 00:11:12,139 around that age. It's suggested maybe the child 250 00:11:12,139 --> 00:11:14,590 suddenly cracked the code of naming. But that's 251 00:11:14,590 --> 00:11:16,590 not what's really going on. When you look closely 252 00:11:16,590 --> 00:11:18,549 at the rate of learning how many new words per 253 00:11:18,549 --> 00:11:21,230 day, the increase is much more gradual, much 254 00:11:21,230 --> 00:11:23,690 more linear. It starts slow and just steadily 255 00:11:23,690 --> 00:11:27,129 increases. So there's no sudden aha moment where 256 00:11:27,129 --> 00:11:29,850 the rate jumps dramatically? Not really, no. 257 00:11:30,009 --> 00:11:33,610 The rate might be, say, 0 .3 words a day between 258 00:11:33,610 --> 00:11:36,350 12, 16 months, and it just keeps climbing steadily. 259 00:11:36,610 --> 00:11:38,710 And when does the rate actually peak? Is it in 260 00:11:38,710 --> 00:11:41,940 toddlerhood? No. That's the surprising part. 261 00:11:42,059 --> 00:11:44,980 The peak rate of vocabulary acquisition actually 262 00:11:44,980 --> 00:11:48,100 happens much later, often in late childhood or 263 00:11:48,100 --> 00:11:51,360 even early adolescence. We're talking maybe 8 264 00:11:51,360 --> 00:11:53,820 to 10 years old, sometimes hitting rates of over 265 00:11:53,820 --> 00:11:57,240 10, maybe even 12 words a day. 12 words a day. 266 00:11:57,340 --> 00:11:59,899 That's incredible. It is. And it often keeps 267 00:11:59,899 --> 00:12:01,960 accelerating slightly right up until the late 268 00:12:01,960 --> 00:12:05,120 teens. So the really massive vocabulary growth 269 00:12:05,120 --> 00:12:07,960 happens when kids are in school, not when they're 270 00:12:07,960 --> 00:12:11,070 toddlers. The spurt idea was probably an artifact 271 00:12:11,070 --> 00:12:13,549 of how the cumulative data looks over that short 272 00:12:13,549 --> 00:12:15,809 toddler period. Okay, so the timing is gradual. 273 00:12:16,149 --> 00:12:18,669 What about the nature of those very first words 274 00:12:18,669 --> 00:12:20,950 kids use? Are they somehow different, less symbolic 275 00:12:20,950 --> 00:12:23,789 than adult words? Bloom argues strongly against 276 00:12:23,789 --> 00:12:26,570 that idea. He suggests children's first words 277 00:12:26,570 --> 00:12:29,570 are fundamentally the same kind of symbolic representations 278 00:12:29,570 --> 00:12:32,629 as adult words. But what about things like overextensions? 279 00:12:32,629 --> 00:12:34,509 You know, a kid calling every four -legged animal 280 00:12:34,509 --> 00:12:37,070 doggy or calling the cookie jar a cookie. Doesn't 281 00:12:37,070 --> 00:12:39,669 that show confusion? It might look like confusion, 282 00:12:39,830 --> 00:12:42,070 but it's often interpreted as a smart strategy 283 00:12:42,070 --> 00:12:44,450 given their limitations. They have maybe 50 words, 284 00:12:44,590 --> 00:12:46,789 right? Yeah. And very little grammar. So they 285 00:12:46,789 --> 00:12:49,090 use the words they do have flexibly. Exactly. 286 00:12:49,970 --> 00:12:51,649 Huttenlocker and Smiley's research suggested 287 00:12:51,649 --> 00:12:54,649 these overextensions are often context -dependent 288 00:12:54,649 --> 00:12:57,230 communication attempts. The child probably knows 289 00:12:57,230 --> 00:12:59,610 the difference between a dog and a cat conceptually. 290 00:12:59,710 --> 00:13:01,789 They just don't have the word cat yet, so they 291 00:13:01,789 --> 00:13:04,190 use the closest thing they have, doggy, to refer 292 00:13:04,190 --> 00:13:06,350 to it. So it's a vocabulary problem, not a concept 293 00:13:06,350 --> 00:13:09,220 problem. Largely, yes. And even when they use 294 00:13:09,220 --> 00:13:11,639 cookie for the jar, they also use it for actual 295 00:13:11,639 --> 00:13:14,340 cookies, for pictures of cookies. Suggesting 296 00:13:14,340 --> 00:13:17,340 the underlying concept is already quite abstract 297 00:13:17,340 --> 00:13:20,080 and not tied just to one specific instance. They're 298 00:13:20,080 --> 00:13:22,519 using symbols from the get -go. Okay, so we've 299 00:13:22,519 --> 00:13:25,820 got this incredibly fast, robust learning system. 300 00:13:25,879 --> 00:13:28,159 And it's not just simple association. It works 301 00:13:28,159 --> 00:13:32,120 even with messy input. If it's not raw statistics, 302 00:13:32,480 --> 00:13:34,639 what is the driving force? This seems to point 303 00:13:34,639 --> 00:13:36,980 directly towards the social aspect. Theory of 304 00:13:36,980 --> 00:13:40,620 mind. Absolutely. The failure of simple associationism. 305 00:13:40,639 --> 00:13:42,840 Yeah. The idea, going back to Locke, that you 306 00:13:42,840 --> 00:13:45,500 just pair sounds with experiences is pretty clear. 307 00:13:45,639 --> 00:13:47,879 We've mentioned the noisy input, right? 30, 50 308 00:13:47,879 --> 00:13:51,120 percent misalignment. Associationism also relies 309 00:13:51,120 --> 00:13:55,320 heavily on this idea of explicit, ostensive naming 310 00:13:55,320 --> 00:13:58,659 the parent pointing and labeling clearly. Which, 311 00:13:58,759 --> 00:14:00,500 as you said, doesn't happen everywhere. It's 312 00:14:00,500 --> 00:14:03,570 culturally specific. Right. In many cultures, 313 00:14:03,750 --> 00:14:07,029 kids learn primarily by listening in, without 314 00:14:07,029 --> 00:14:09,970 adults directly labeling things for them. If 315 00:14:09,970 --> 00:14:12,470 association was the main game, kids in those 316 00:14:12,470 --> 00:14:14,909 cultures should really struggle, but they learn 317 00:14:14,909 --> 00:14:17,389 language just fine. So the only thing that seems 318 00:14:17,389 --> 00:14:19,889 to work across all these variations, noisy input, 319 00:14:20,049 --> 00:14:22,529 different cultural practices, is the child's 320 00:14:22,529 --> 00:14:24,230 ability to figure out what the speaker means 321 00:14:24,230 --> 00:14:27,590 to say. Exactly. It has to be based on inferring 322 00:14:27,590 --> 00:14:30,250 the speaker's intention. And that requires theory 323 00:14:30,250 --> 00:14:33,429 of mind. Tom, understanding that other people 324 00:14:33,429 --> 00:14:36,470 have minds with beliefs, desires, goals, and 325 00:14:36,470 --> 00:14:39,049 crucially for word learning, referential intentions. 326 00:14:39,230 --> 00:14:41,370 They intend to talk about something. And kids 327 00:14:41,370 --> 00:14:43,590 have this ability really early. Surprisingly 328 00:14:43,590 --> 00:14:46,330 early. We know by nine months or so, infants 329 00:14:46,330 --> 00:14:48,610 are following their caregiver's gaze. That's 330 00:14:48,610 --> 00:14:50,470 huge. They're looking where you're looking. Trying 331 00:14:50,470 --> 00:14:52,049 to figure out what you're paying attention to. 332 00:14:52,309 --> 00:14:56,049 Yes. And by about a year, they're doing more 333 00:14:56,049 --> 00:14:58,389 active things like gaze alternation. They'll 334 00:14:58,389 --> 00:15:00,309 look at an interesting toy. then look at the 335 00:15:00,309 --> 00:15:02,710 adult, then back at the toy. Like saying, hey, 336 00:15:02,889 --> 00:15:05,950 look at this thing with me. Pretty much. They're 337 00:15:05,950 --> 00:15:07,710 actively trying to direct the adult's attention 338 00:15:07,710 --> 00:15:11,269 and establish joint focus. This social machinery 339 00:15:11,269 --> 00:15:13,570 is already running before significant word learning 340 00:15:13,570 --> 00:15:16,190 takes off. Which sets the stage for that Baldwin 341 00:15:16,190 --> 00:15:18,070 -Mody experiment you mentioned earlier. That 342 00:15:18,070 --> 00:15:20,389 seems like a direct test. A fantastic demonstration. 343 00:15:20,909 --> 00:15:24,230 So just to recap, an 18 -month -old is playing 344 00:15:24,230 --> 00:15:27,210 with object A. The experimenter looks over at 345 00:15:27,210 --> 00:15:29,450 object B in a bucket and says enthusiastically, 346 00:15:29,629 --> 00:15:33,169 it's a modi. Okay. The sound modi occurs while 347 00:15:33,169 --> 00:15:35,389 the child is physically focused on object A. 348 00:15:35,549 --> 00:15:39,570 Right. So pure association. Lockean pairing predicts 349 00:15:39,570 --> 00:15:42,289 the child should link Modi to object A. But they 350 00:15:42,289 --> 00:15:44,789 don't. They don't. They link Modi to object B, 351 00:15:44,870 --> 00:15:46,470 the thing the experimenter was looking at and 352 00:15:46,470 --> 00:15:48,889 showing clear intention towards. They use the 353 00:15:48,889 --> 00:15:51,289 speaker's gaze and interest as the primary cue 354 00:15:51,289 --> 00:15:53,570 to reference, overriding their own immediate 355 00:15:53,570 --> 00:15:56,110 sensory experience. That's powerful evidence 356 00:15:56,110 --> 00:15:59,129 for Tom in action. Does this apply beyond just 357 00:15:59,129 --> 00:16:02,370 object names, like for actions or verbs? It seems 358 00:16:02,370 --> 00:16:04,870 to. There's related work on action imitation. 359 00:16:06,090 --> 00:16:08,990 If a researcher demonstrates an action, say, 360 00:16:09,110 --> 00:16:12,029 trying to pull apart a toy and acts like they 361 00:16:12,029 --> 00:16:14,990 succeeded there, toddlers are more likely to 362 00:16:14,990 --> 00:16:17,970 imitate that goal -directed action. But if the 363 00:16:17,970 --> 00:16:20,250 researcher performs the exact same physical movements 364 00:16:20,250 --> 00:16:23,549 but acts like it was an accident, the toddlers 365 00:16:23,549 --> 00:16:25,629 are much less likely to copy it. So they're not 366 00:16:25,629 --> 00:16:27,590 just copying movements, they're copying perceived 367 00:16:27,590 --> 00:16:30,750 intentions or goals. Exactly. They're sensitive 368 00:16:30,750 --> 00:16:33,370 to the emotional and intentional cues surrounding 369 00:16:33,370 --> 00:16:35,769 the action, not just the physical kinematics. 370 00:16:36,049 --> 00:16:38,190 This feeds into understanding verb meanings, 371 00:16:38,370 --> 00:16:40,610 too, which are often about intended actions. 372 00:16:40,929 --> 00:16:43,529 This focus on intention also helps explain something 373 00:16:43,529 --> 00:16:45,870 called the principle of contrast, right? Sometimes 374 00:16:45,870 --> 00:16:48,909 called mutual exclusivity. Yes, that's the idea 375 00:16:48,909 --> 00:16:52,750 that kids assume new words refer to new things. 376 00:16:52,850 --> 00:16:54,850 If they already have a word for something, they 377 00:16:54,850 --> 00:16:56,610 expect a different word to refer to something 378 00:16:56,610 --> 00:16:59,620 else. The classic example. Show a kid a banana, 379 00:16:59,820 --> 00:17:02,419 they know, and a weird new kitchen gadget, like 380 00:17:02,419 --> 00:17:04,940 a whisk. Ask for the fendle. They'll hand you 381 00:17:04,940 --> 00:17:07,839 the whisk. Every time. Because they already have 382 00:17:07,839 --> 00:17:11,380 a word for banana. Banana. So this new word, 383 00:17:11,460 --> 00:17:13,980 fendle, must refer to the other thing. Now, is 384 00:17:13,980 --> 00:17:16,500 this like a built -in rule specifically for language? 385 00:17:17,220 --> 00:17:20,019 Or is it also based on this social reasoning? 386 00:17:20,500 --> 00:17:23,099 Bloom argues strongly it's a pragmatic inference 387 00:17:23,099 --> 00:17:26,079 based on Tom, not a rigid linguistic constraint. 388 00:17:26,589 --> 00:17:28,470 How can we tell the difference? Well, think about 389 00:17:28,470 --> 00:17:31,150 the reasoning process. The child isn't just applying 390 00:17:31,150 --> 00:17:34,869 a rule like one object, one name. They seem to 391 00:17:34,869 --> 00:17:37,569 be thinking something like, okay, this adult 392 00:17:37,569 --> 00:17:39,730 is trying to communicate effectively. That's 393 00:17:39,730 --> 00:17:42,750 Tom. They know the word banana. If they wanted 394 00:17:42,750 --> 00:17:44,670 the banana, they would have said banana. Since 395 00:17:44,670 --> 00:17:47,710 they used a weird new word, fendel, they must 396 00:17:47,710 --> 00:17:49,390 be talking about the other thing here, the whisk. 397 00:17:49,819 --> 00:17:51,720 So it's reasoning about the speaker's likely 398 00:17:51,720 --> 00:17:53,799 communicative strategy. Exactly. It's like the 399 00:17:53,799 --> 00:17:55,740 example with colored cards. If you learn red 400 00:17:55,740 --> 00:17:58,000 card means pick the cube, and then I show you 401 00:17:58,000 --> 00:18:00,480 an orange card next to a sphere, you'll instantly 402 00:18:00,480 --> 00:18:03,000 infer orange means pick the sphere. Right. Yeah, 403 00:18:03,099 --> 00:18:04,859 because you would have used red if you meant 404 00:18:04,859 --> 00:18:07,880 the cube. Precisely. It's a general principle 405 00:18:07,880 --> 00:18:10,380 of informative communication applied to word 406 00:18:10,380 --> 00:18:13,220 learning, not purely lexical. And this social 407 00:18:13,220 --> 00:18:15,299 understanding connects back to the Saussurean 408 00:18:15,299 --> 00:18:18,380 sign idea, that words work both ways. It does. 409 00:18:18,930 --> 00:18:21,910 The child sees an adult use dog with the intention 410 00:18:21,910 --> 00:18:25,420 of referring to a dog. Through Tom, the child 411 00:18:25,420 --> 00:18:28,400 grasps that intention. Then they make the crucial 412 00:18:28,400 --> 00:18:31,000 leap. Okay, if they can use that sound with that 413 00:18:31,000 --> 00:18:32,880 intention, then I can use that sound with the 414 00:18:32,880 --> 00:18:35,000 same intention. It's like learning through imitation, 415 00:18:35,240 --> 00:18:37,579 but imitating the intent behind the word use, 416 00:18:37,759 --> 00:18:40,319 not just the sound. very much like Miltzoff's 417 00:18:40,319 --> 00:18:42,259 findings on imitating goal -directed actions. 418 00:18:42,380 --> 00:18:44,880 You grasp the goal, and then you can reproduce 419 00:18:44,880 --> 00:18:47,640 the action to achieve the same goal. Here, the 420 00:18:47,640 --> 00:18:50,259 goal is reference. Okay, if Tom and intentional 421 00:18:50,259 --> 00:18:52,460 inference are so central, what happens if that 422 00:18:52,460 --> 00:18:55,200 ability is impaired? This is where the evidence 423 00:18:55,200 --> 00:18:57,700 from autism comes in, right? Yes, and it's really 424 00:18:57,700 --> 00:19:00,500 strong support. Children with autism often show 425 00:19:00,500 --> 00:19:02,720 marked difficulties with theory of mind tasks, 426 00:19:03,079 --> 00:19:06,059 understanding others' beliefs, intentions, perspectives. 427 00:19:06,700 --> 00:19:08,900 So how do they do on word learning tasks, like 428 00:19:08,900 --> 00:19:12,079 the MODY experiment? They typically fail the 429 00:19:12,079 --> 00:19:14,960 discrepant -looking task. Unlike typically developing 430 00:19:14,960 --> 00:19:17,720 children, or even mentally handicapped children, 431 00:19:17,900 --> 00:19:21,160 matched for overall cognitive ability, but without 432 00:19:21,160 --> 00:19:24,079 TOM deficits, children with autism tend to make 433 00:19:24,079 --> 00:19:27,259 the associative error. They link Modi to the 434 00:19:27,259 --> 00:19:30,059 object they were holding. Yes. They seem to rely 435 00:19:30,059 --> 00:19:33,319 more on the simple temporal co -occurrence, the 436 00:19:33,319 --> 00:19:35,779 Lockean pairing, because they aren't effectively 437 00:19:35,779 --> 00:19:38,740 using the speaker's intentional cues, like gaze, 438 00:19:38,940 --> 00:19:41,579 to override it. So the contrast between autistic 439 00:19:41,579 --> 00:19:44,119 children and other groups really highlights the 440 00:19:44,119 --> 00:19:46,700 necessity of Tom for typical word learning. It 441 00:19:46,700 --> 00:19:49,140 provides very compelling, almost clinical evidence 442 00:19:49,140 --> 00:19:51,660 that this social inferential capacity is not 443 00:19:51,660 --> 00:19:53,779 just helpful, it's a core component of the mechanism. 444 00:19:54,230 --> 00:19:55,910 All right. We've established the social machinery 445 00:19:55,910 --> 00:19:59,089 inferring intent is key, but you can't map a 446 00:19:59,089 --> 00:20:01,049 word onto a meaning if you don't have the meaning 447 00:20:01,049 --> 00:20:03,309 structured in your head first. So the next part 448 00:20:03,309 --> 00:20:05,170 is about the conceptual prerequisites, right? 449 00:20:05,269 --> 00:20:07,450 How kids carve up the world before they start 450 00:20:07,450 --> 00:20:10,910 naming it. Hashtag, hashtag A, objects and individuals, 451 00:20:11,210 --> 00:20:13,869 chapter four. Exactly. We need to look at the 452 00:20:13,869 --> 00:20:15,950 conceptual landscape the child brings to the 453 00:20:15,950 --> 00:20:18,390 task. And one of the most fundamental aspects 454 00:20:18,390 --> 00:20:21,529 is their understanding of objects. This explains 455 00:20:21,529 --> 00:20:24,289 why kids' early words are so often object names, 456 00:20:24,490 --> 00:20:28,450 ball, cup, dog. Not because they're easier to 457 00:20:28,450 --> 00:20:31,369 say, but because object is a really solid concept 458 00:20:31,369 --> 00:20:34,069 for them early on. That's the idea. They come 459 00:20:34,069 --> 00:20:36,029 into the world, or at least develop very early, 460 00:20:36,130 --> 00:20:38,829 a way of parsing the visual scene into distinct 461 00:20:38,829 --> 00:20:41,680 bounded things. This draws heavily on the work 462 00:20:41,680 --> 00:20:44,140 of people like Elizabeth Spelk. The Spelk object 463 00:20:44,140 --> 00:20:46,200 principles? What are those again? They're these 464 00:20:46,200 --> 00:20:48,579 core principles that infants seem to use to decide 465 00:20:48,579 --> 00:20:51,339 what counts as a single coherent object. The 466 00:20:51,339 --> 00:20:53,839 main ones are cohesion, objects move as connected 467 00:20:53,839 --> 00:20:56,660 holes, continuity, objects move on connected 468 00:20:56,660 --> 00:20:59,799 paths, solidity, objects can't pass through each 469 00:20:59,799 --> 00:21:02,619 other, and contact, objects usually affect each 470 00:21:02,619 --> 00:21:05,329 other only when they touch. Let's focus on cohesion. 471 00:21:05,470 --> 00:21:07,289 How does that help a baby know if they're seeing 472 00:21:07,289 --> 00:21:09,950 one thing or two things? Imagine an infant looking 473 00:21:09,950 --> 00:21:11,509 at something that looks like a single pyramid. 474 00:21:11,910 --> 00:21:14,789 If you then lift the top part and the base stays 475 00:21:14,789 --> 00:21:17,269 sitting on the table. That violates cohesion. 476 00:21:17,430 --> 00:21:19,549 It didn't move as one connected thing. Right. 477 00:21:19,950 --> 00:21:22,329 And studies show infants are surprised by this. 478 00:21:22,789 --> 00:21:25,130 That surprise indicates their internal system 479 00:21:25,130 --> 00:21:28,450 expected it to move as a whole. When it didn't, 480 00:21:28,450 --> 00:21:30,670 they infer, ah, that must actually be two separate 481 00:21:30,670 --> 00:21:33,029 objects. So this isn't learned. It's like a built 482 00:21:33,029 --> 00:21:35,430 -in way of parsing the world into things that 483 00:21:35,430 --> 00:21:38,190 can then receive names. It seems to be a fundamental 484 00:21:38,190 --> 00:21:40,970 pre -linguistic way of individuating the world. 485 00:21:41,529 --> 00:21:43,990 You need to know what the potential things are 486 00:21:43,990 --> 00:21:46,349 before you can start mapping labels onto them. 487 00:21:46,849 --> 00:21:49,049 Hashtag, hashtag, hashtag BB, naming individuals 488 00:21:49,049 --> 00:21:52,440 versus kinds. Chapter 5. Okay, so the baby can 489 00:21:52,440 --> 00:21:55,180 parse the world into individual objects, but 490 00:21:55,180 --> 00:21:57,279 then language throws different kinds of names 491 00:21:57,279 --> 00:22:00,539 at them. Not just dog for the category, but fido 492 00:22:00,539 --> 00:22:03,779 for a specific dog or him, which could refer 493 00:22:03,779 --> 00:22:06,519 to any male dog depending on context. How do 494 00:22:06,519 --> 00:22:08,200 they sort out these different types of names, 495 00:22:08,380 --> 00:22:10,779 particularly proper names versus common nouns? 496 00:22:11,099 --> 00:22:13,039 That's another layer of conceptual understanding. 497 00:22:13,599 --> 00:22:15,960 They need to grasp the difference between naming 498 00:22:15,960 --> 00:22:18,779 a kind, dog, and naming a unique individual, 499 00:22:19,079 --> 00:22:22,190 fido. And research, like Gail Hall's, suggests 500 00:22:22,190 --> 00:22:24,869 they use specific cues. What kind of cues? Two 501 00:22:24,869 --> 00:22:27,690 really stand out. First, the type of entity being 502 00:22:27,690 --> 00:22:30,630 named. And second, the consistency or fidelity 503 00:22:30,630 --> 00:22:33,109 of how the name is used. Let's start with entity 504 00:22:33,109 --> 00:22:35,210 type. Are some things more likely to get proper 505 00:22:35,210 --> 00:22:38,269 names than others? Definitely. Kids seem to expect 506 00:22:38,269 --> 00:22:41,009 proper names primarily for entities that are 507 00:22:41,009 --> 00:22:43,890 special, unique, and often animate or treated 508 00:22:43,890 --> 00:22:47,869 as intentional agents. So, people, pets. Maybe 509 00:22:47,869 --> 00:22:49,930 special dolls? But probably not a random brick 510 00:22:49,930 --> 00:22:52,910 or spoon? Generally not, unless you specifically 511 00:22:52,910 --> 00:22:56,589 animate it or treat it as special. Generic interchangeable 512 00:22:56,589 --> 00:22:59,410 items usually get common nouns. Brick. Not proper 513 00:22:59,410 --> 00:23:02,269 names. Bobby the Brick. This suggests the child's 514 00:23:02,269 --> 00:23:04,650 understanding of animacy and individuality influences 515 00:23:04,650 --> 00:23:07,130 their expectations about naming. Okay, and the 516 00:23:07,130 --> 00:23:10,109 second cue, fidelity of reference. This connects 517 00:23:10,109 --> 00:23:12,190 to those anecdotes about kids getting upset if 518 00:23:12,190 --> 00:23:14,809 a name is reused. Exactly. A common noun like 519 00:23:14,809 --> 00:23:17,599 dog applies to all members of the kind. But a 520 00:23:17,599 --> 00:23:20,720 proper name, like Kieran or Rebecca in those 521 00:23:20,720 --> 00:23:23,299 anecdotes, is expected to refer consistently 522 00:23:23,299 --> 00:23:26,140 and uniquely to one specific individual. So when 523 00:23:26,140 --> 00:23:28,420 McNamara's son Kieran heard another baby called 524 00:23:28,420 --> 00:23:31,359 Lisa, he objected because he already knew a Lisa. 525 00:23:31,660 --> 00:23:34,680 Yes, he seemed to operate under a strict one 526 00:23:34,680 --> 00:23:37,720 -name -one -person rule initially. And the other 527 00:23:37,720 --> 00:23:40,400 child, Matthew, refusing to call a new baby Rebecca 528 00:23:40,400 --> 00:23:43,799 because his sitter's friend was Becca. It shows 529 00:23:43,799 --> 00:23:46,180 they... grasped that proper names are meant to 530 00:23:46,180 --> 00:23:48,140 pick out unique individuals. And they understand 531 00:23:48,140 --> 00:23:50,480 this concept of unique individuals before they 532 00:23:50,480 --> 00:23:53,039 fully master the naming conventions. It appears 533 00:23:53,039 --> 00:23:56,059 so. Evidence suggests that even pre -linguistic 534 00:23:56,059 --> 00:23:58,400 infants can track individual objects over time, 535 00:23:58,579 --> 00:24:01,079 even identical looking ones. They seem to have 536 00:24:01,079 --> 00:24:03,839 concepts for both kinds, grouping similar things, 537 00:24:04,000 --> 00:24:06,680 and specific individuals, tracking unique entities. 538 00:24:06,960 --> 00:24:09,539 This dual system is crucial for things like basic 539 00:24:09,539 --> 00:24:11,339 number sense and keeping track of social agents 540 00:24:11,339 --> 00:24:13,960 like a metal rolodex. The language then maps 541 00:24:13,960 --> 00:24:15,920 onto this pre -existing conceptual structure. 542 00:24:16,119 --> 00:24:18,420 Hashtag, hashtag, hashtag, CC. Categories and 543 00:24:18,420 --> 00:24:21,240 essences. Chapter 6. Okay, the conceptual structure 544 00:24:21,240 --> 00:24:23,380 goes deeper than just objects and individuals. 545 00:24:23,500 --> 00:24:26,059 How do kids form categories? Is it just based 546 00:24:26,059 --> 00:24:27,880 on what things look like? No, and this is a really 547 00:24:27,880 --> 00:24:30,259 critical point. Human categorization, especially 548 00:24:30,259 --> 00:24:33,000 for natural kinds like animals and plants, seems 549 00:24:33,000 --> 00:24:35,400 to go beyond surface appearance. It relies on 550 00:24:35,400 --> 00:24:38,359 a belief in essentialism. Essentialism. The idea 551 00:24:38,359 --> 00:24:41,640 that things have a hidden underlying nature or 552 00:24:41,640 --> 00:24:45,059 essence that makes them what they are. Precisely. 553 00:24:45,099 --> 00:24:48,339 For biological kinds, this essence is often thought 554 00:24:48,339 --> 00:24:51,700 of implicitly, like genetics or internal structure. 555 00:24:52,019 --> 00:24:54,539 It determines the category, even if the appearance 556 00:24:54,539 --> 00:24:57,599 changes. So, the example of putting a tiger in 557 00:24:57,599 --> 00:24:59,980 a lion suit. Kids will still insist it's a tiger. 558 00:25:00,490 --> 00:25:02,609 The costume doesn't change its underlying tigerness. 559 00:25:03,369 --> 00:25:05,970 Jolman and Markman's studies were key here. The 560 00:25:05,970 --> 00:25:08,029 dinosaur example, too. Yeah, that one's striking. 561 00:25:08,529 --> 00:25:11,210 Kids learn a brontosaurus, a dinosaur, has cold 562 00:25:11,210 --> 00:25:14,490 blood, and a rhino, looks similar to a triceratops, 563 00:25:14,569 --> 00:25:16,950 has warm blood. Then you show them a triceratops. 564 00:25:16,990 --> 00:25:19,519 Which looks more like the rhino. Right. But kids 565 00:25:19,519 --> 00:25:21,640 don't generalize based on looks. They're told 566 00:25:21,640 --> 00:25:24,420 that Triceratops is also a dinosaur. So the inferred 567 00:25:24,420 --> 00:25:26,740 also has cold blood like the Brontosaurus because 568 00:25:26,740 --> 00:25:29,039 they share that deeper category, that dinosaur 569 00:25:29,039 --> 00:25:31,619 essence. So the category, sometimes signaled 570 00:25:31,619 --> 00:25:34,319 by the word itself, points to this hidden essence 571 00:25:34,319 --> 00:25:38,160 overriding appearance. Exactly. Essentialism 572 00:25:38,160 --> 00:25:41,599 seems to be a powerful, perhaps innate bias in 573 00:25:41,599 --> 00:25:44,200 how we think about natural kinds. It's adaptive 574 00:25:44,200 --> 00:25:46,420 because it also reflects biological reality. 575 00:25:46,779 --> 00:25:49,569 What about things humans make? Artifacts like 576 00:25:49,569 --> 00:25:52,109 chairs or tools, they don't have a biological 577 00:25:52,109 --> 00:25:55,710 essence. True. For artifacts, the essence seems 578 00:25:55,710 --> 00:25:57,890 to be more about function and creator's intention. 579 00:25:58,230 --> 00:26:01,269 Yeah. What was it made for? So how something 580 00:26:01,269 --> 00:26:03,849 looks is less important than what it does or 581 00:26:03,849 --> 00:26:06,509 what someone intended it to do. That's what the 582 00:26:06,509 --> 00:26:08,690 research suggests. Kendler Nelson, for instance, 583 00:26:08,769 --> 00:26:11,950 showed kids novel objects. If we're told an object 584 00:26:11,950 --> 00:26:14,859 was a stennet. and shown it was used for, say, 585 00:26:14,980 --> 00:26:17,619 making a specific pattern, they would generalize 586 00:26:17,619 --> 00:26:19,880 the name stinnet to other objects that perform 587 00:26:19,880 --> 00:26:21,700 the same function. Even if they looked quite 588 00:26:21,700 --> 00:26:23,640 different, they prioritized function over shape. 589 00:26:23,779 --> 00:26:27,160 So artifacts are categorized based on their intended 590 00:26:27,160 --> 00:26:29,220 purpose, their sort of psychological essence. 591 00:26:29,619 --> 00:26:32,359 Hashtag tag tag D, naming representations, chapter 592 00:26:32,359 --> 00:26:35,140 seven. This reliance on intention seems to extend 593 00:26:35,140 --> 00:26:37,920 even further, right? To naming things like pictures 594 00:26:37,920 --> 00:26:40,890 or drawings. Yes. When we call a picture of a 595 00:26:40,890 --> 00:26:43,730 dog a dog, we're not actually naming the animal 596 00:26:43,730 --> 00:26:46,509 itself. We're naming the representation. Which 597 00:26:46,509 --> 00:26:48,750 is really an artifact paper with marks on it. 598 00:26:48,910 --> 00:26:51,309 Exactly. It's an artifact created with a specific 599 00:26:51,309 --> 00:26:54,349 representational intention. The child needs to 600 00:26:54,349 --> 00:26:56,089 understand that the marks were meant to stand 601 00:26:56,089 --> 00:26:58,730 for something else. And this understanding of 602 00:26:58,730 --> 00:27:01,710 intent can even override how good the picture 603 00:27:01,710 --> 00:27:04,230 is. The anecdote about the child, Simon, and 604 00:27:04,230 --> 00:27:06,009 his scribble. Oh, that's a perfect illustration. 605 00:27:06,410 --> 00:27:09,769 Simon, at three years old, makes this wild scribble 606 00:27:09,769 --> 00:27:12,430 that looks like nothing on earth. But he confidently 607 00:27:12,430 --> 00:27:15,670 declares, it's mommy and daddy. Because that's 608 00:27:15,670 --> 00:27:18,910 who he intended to draw. Precisely. His representational 609 00:27:18,910 --> 00:27:21,329 intent was the sole basis for the name he gave 610 00:27:21,329 --> 00:27:24,349 it. The actual visual resemblance was irrelevant 611 00:27:24,349 --> 00:27:27,329 to him in that moment. It shows that this intentional 612 00:27:27,329 --> 00:27:30,890 stance, understanding that names relate to intended 613 00:27:30,890 --> 00:27:33,829 reference, applies even to the child's own creative 614 00:27:33,829 --> 00:27:36,769 and sometimes failed attempts at representation. 615 00:27:37,410 --> 00:27:39,690 So we've got the social inference engine theory 616 00:27:39,690 --> 00:27:41,970 of mind, and we've got the foundational conceptual 617 00:27:41,970 --> 00:27:44,930 structures, objects, kinds, essences, intentions. 618 00:27:45,390 --> 00:27:48,190 These are powerful, maybe even sufficient for 619 00:27:48,190 --> 00:27:51,029 learning basic concrete nouns. But for other 620 00:27:51,029 --> 00:27:53,529 types of words, especially more abstract ones, 621 00:27:53,690 --> 00:27:56,410 or words that aren't nouns, children seem to 622 00:27:56,410 --> 00:27:59,359 need another source of information, the language 623 00:27:59,359 --> 00:28:03,380 itself, the syntax. So grammar helps you learn 624 00:28:03,380 --> 00:28:05,299 word meanings. It's not just about combining 625 00:28:05,299 --> 00:28:08,410 words you already know. Right. syntax acts like 626 00:28:08,410 --> 00:28:10,950 a set of clues to narrow down the possible meanings 627 00:28:10,950 --> 00:28:14,069 of a new word. It's not strictly necessary for 628 00:28:14,069 --> 00:28:16,789 learning dog or mommy, but it becomes really 629 00:28:16,789 --> 00:28:19,369 important for other things. Like what? Well, 630 00:28:19,430 --> 00:28:21,549 take the basic distinction in nominal syntax 631 00:28:21,549 --> 00:28:24,210 between count nouns and mass nouns. Count nouns 632 00:28:24,210 --> 00:28:26,930 are things you can count, like one dog, two dogs. 633 00:28:27,150 --> 00:28:29,630 Mass nouns are substances you can't easily count, 634 00:28:29,710 --> 00:28:32,970 like water or sand. You say some water, not a 635 00:28:32,970 --> 00:28:35,690 water. Exactly. And the grammar reflects this. 636 00:28:36,029 --> 00:28:38,190 Count nouns typically appear with articles like 637 00:28:38,190 --> 00:28:41,589 A or N, and they pluralize dogs. Mass nouns often 638 00:28:41,589 --> 00:28:44,029 appear with some, and usually don't pluralize 639 00:28:44,029 --> 00:28:46,470 in the same way. And kids pick up on these grammatical 640 00:28:46,470 --> 00:28:49,269 cues to figure out if a new word refers to an 641 00:28:49,269 --> 00:28:52,430 object or a substance. Research suggests they 642 00:28:52,430 --> 00:28:55,269 do. If you introduce a new word in count syntax, 643 00:28:55,509 --> 00:28:58,849 look, a fendel, children are more likely to assume 644 00:28:58,849 --> 00:29:02,250 fendel refers to a whole object. And if you use 645 00:29:02,250 --> 00:29:05,619 mass syntax, look, some blicket. They're more 646 00:29:05,619 --> 00:29:08,039 likely to interpret blicket as referring to a 647 00:29:08,039 --> 00:29:11,559 substance or material. Soja, Carey, and Spelk 648 00:29:11,559 --> 00:29:14,440 did key work showing this sensitivity. But wasn't 649 00:29:14,440 --> 00:29:17,039 there an asymmetry, something about solids? Ah, 650 00:29:17,119 --> 00:29:21,079 yes. The object bias is very strong. Remember 651 00:29:21,079 --> 00:29:23,380 how infants parse the world into cohesive objects? 652 00:29:23,680 --> 00:29:26,859 That bias persists. So even if you use mass syntax 653 00:29:26,859 --> 00:29:29,019 for a solid object, like pointing to a wooden 654 00:29:29,019 --> 00:29:31,140 block and saying, this is some blicket. kids 655 00:29:31,140 --> 00:29:33,420 might resist interpreting blicket as the substance 656 00:29:33,420 --> 00:29:36,079 would. They often do. They're so predisposed 657 00:29:36,079 --> 00:29:38,720 to see a bounded solid as an individual object 658 00:29:38,720 --> 00:29:41,420 that the mass syntax, some blicket, struggles 659 00:29:41,420 --> 00:29:44,160 to override that conceptual default. Learning 660 00:29:44,160 --> 00:29:46,579 names for solid substances like wood or metal 661 00:29:46,579 --> 00:29:48,960 is actually harder for kids than learning object 662 00:29:48,960 --> 00:29:51,200 names, precisely because it fights this object 663 00:29:51,200 --> 00:29:53,759 bias. Syntax helps, but it doesn't always win 664 00:29:53,759 --> 00:29:55,920 against a strong conceptual predisposition. Okay. 665 00:29:55,980 --> 00:29:59,119 What about verbs? Verbs seem really hard to learn 666 00:29:59,119 --> 00:30:01,740 just by observation. They are notoriously difficult. 667 00:30:02,480 --> 00:30:05,460 Think about it. Parents rarely point to an action 668 00:30:05,460 --> 00:30:08,029 and say, running, the way they point to a dog 669 00:30:08,029 --> 00:30:11,670 and say, dog. And many verbs refer to things 670 00:30:11,670 --> 00:30:14,730 you can't even see, like think, believe, want, 671 00:30:14,950 --> 00:30:17,990 dream. You can't point to thinking. Exactly. 672 00:30:18,250 --> 00:30:20,690 So how do kids learn them? This is where Leila 673 00:30:20,690 --> 00:30:22,809 Gleitman's idea of syntactic bootstrapping comes 674 00:30:22,809 --> 00:30:25,210 in. Yeah, bootstrapping. Like pulling yourself 675 00:30:25,210 --> 00:30:27,829 up by your own bootstraps, using syntax to learn 676 00:30:27,829 --> 00:30:30,430 syntax. Sort of. It's using the syntactic frame 677 00:30:30,430 --> 00:30:32,990 a verb appears in to figure out the verb's meaning. 678 00:30:33,190 --> 00:30:35,269 How does that work? Remember the silent video 679 00:30:35,269 --> 00:30:38,369 problem. If adults just watch someone doing something, 680 00:30:38,609 --> 00:30:40,650 they're often terrible at guessing the specific 681 00:30:40,650 --> 00:30:44,170 verb. Is she giving or getting? Chasing or fleeing? 682 00:30:44,329 --> 00:30:47,089 The visual scene is ambiguous. Right. But this 683 00:30:47,089 --> 00:30:49,470 sentence structure provides crucial clues. If 684 00:30:49,470 --> 00:30:53,009 a child hears, Mary is gawping the dog, the structure... 685 00:30:53,339 --> 00:30:56,440 A subject -verb object tells them gorping must 686 00:30:56,440 --> 00:30:58,559 be a transitive verb, an action that involves 687 00:30:58,559 --> 00:31:01,859 an agent, Mary, acting on a patient, the dog. 688 00:31:02,059 --> 00:31:05,000 As opposed to, say, Mary is gorping. Right, which 689 00:31:05,000 --> 00:31:07,519 suggests an intransitive verb, something Mary 690 00:31:07,519 --> 00:31:10,799 does by herself, like running or sleeping. Or, 691 00:31:10,900 --> 00:31:14,200 Mary is gorping to the dog, suggests a verb of 692 00:31:14,200 --> 00:31:16,700 transfer or communication. So the grammar limits 693 00:31:16,700 --> 00:31:19,019 the possible meanings of the unknown verb. It 694 00:31:19,019 --> 00:31:21,480 dramatically narrows the search space. It tells 695 00:31:21,480 --> 00:31:23,920 the child what kind of event or relation to look 696 00:31:23,920 --> 00:31:26,980 for in the world. Gleitman argued this syntactic 697 00:31:26,980 --> 00:31:29,539 information is essential, especially for learning 698 00:31:29,539 --> 00:31:31,619 verbs whose meanings aren't easily observable. 699 00:31:32,079 --> 00:31:33,960 Hashtag, hashtag, the puzzle of number words, 700 00:31:34,079 --> 00:31:35,940 chapter nine. Okay, syntax helps with verbs. 701 00:31:36,279 --> 00:31:39,180 What about really abstract concepts like numbers? 702 00:31:39,599 --> 00:31:42,000 How does a child learn that three isn't just 703 00:31:42,000 --> 00:31:45,039 an adjective like big or red, but refers to a 704 00:31:45,039 --> 00:31:48,140 specific quantity? Number words are a fascinating 705 00:31:48,140 --> 00:31:50,920 case study in using linguistic context. They 706 00:31:50,920 --> 00:31:52,819 don't refer to objects. They don't refer to substances. 707 00:31:53,099 --> 00:31:55,140 They don't refer to simple properties. They refer 708 00:31:55,140 --> 00:31:57,680 to the cardinality of a set. Which is super abstract. 709 00:31:57,960 --> 00:32:01,160 Totally. So how does the child crack this? Bloom 710 00:32:01,160 --> 00:32:03,539 and Winn argued that children use a cluster of 711 00:32:03,539 --> 00:32:06,500 grammatical cues present in the input to figure 712 00:32:06,500 --> 00:32:08,599 out the semantic category of number words long 713 00:32:08,599 --> 00:32:10,920 before they know exactly what three means. They 714 00:32:10,920 --> 00:32:13,259 figure out what kind of word it is first. Yes. 715 00:32:13,359 --> 00:32:15,940 They use the linguistic evidence to classify 716 00:32:15,940 --> 00:32:19,339 number words as quantifiers that apply to sets 717 00:32:19,339 --> 00:32:22,319 of individuals. What are the clues? There are 718 00:32:22,319 --> 00:32:24,539 several consistent patterns in how adults use 719 00:32:24,539 --> 00:32:26,920 number words when talking to children. First, 720 00:32:27,079 --> 00:32:29,359 number words are used almost exclusively with 721 00:32:29,359 --> 00:32:32,880 count nouns, never mass nouns. You say three 722 00:32:32,880 --> 00:32:35,359 dogs, not three waters. Okay, that separates 723 00:32:35,359 --> 00:32:38,359 them from words like much. Second, they are never 724 00:32:38,359 --> 00:32:41,359 modified by degree, words like very. You can 725 00:32:41,359 --> 00:32:44,220 be very tall, but you can't be very three. That 726 00:32:44,220 --> 00:32:46,700 suggests they represent something discrete, not 727 00:32:46,700 --> 00:32:49,559 continuous. Third, their position in the noun 728 00:32:49,559 --> 00:32:52,440 phrase is fixed. They always come before adjectives. 729 00:32:52,480 --> 00:32:55,920 The three brown dogs, never the brown three dogs. 730 00:32:56,329 --> 00:32:58,470 What does that tell the child? It suggests the 731 00:32:58,470 --> 00:33:00,690 number word modifies the entire set, the three 732 00:33:00,690 --> 00:33:03,089 dogs, rather than describing a property of the 733 00:33:03,089 --> 00:33:06,089 individual members, like brownness. It's quantifying 734 00:33:06,089 --> 00:33:09,049 the set. Okay. Any other cues? A fourth clue 735 00:33:09,049 --> 00:33:10,789 is that they frequently appear in the partitive 736 00:33:10,789 --> 00:33:13,789 construction, three of the dogs. This structure 737 00:33:13,789 --> 00:33:15,710 is a very strong indicator of quantification 738 00:33:15,710 --> 00:33:18,309 over sets. So putting all these grammatical clues 739 00:33:18,309 --> 00:33:21,130 together. The child can deduce that words like 740 00:33:21,130 --> 00:33:25,180 two, three, four belong to a unique class. They're 741 00:33:25,180 --> 00:33:28,079 not adjectives, not nouns, not simple modifiers. 742 00:33:28,240 --> 00:33:30,799 They are terms that specify the quantity of a 743 00:33:30,799 --> 00:33:33,660 set of individuals. And this happens before they 744 00:33:33,660 --> 00:33:36,480 know that three means exactly three things. Yes, 745 00:33:36,539 --> 00:33:39,500 that's the key finding. Analysis shows these 746 00:33:39,500 --> 00:33:42,099 grammatical cues are reliably present in speech 747 00:33:42,099 --> 00:33:45,299 to one - and two -year -olds, often a full year 748 00:33:45,299 --> 00:33:47,759 before those children reliably understand the 749 00:33:47,759 --> 00:33:50,609 precise cardinal meanings. One. then two, then 750 00:33:50,609 --> 00:33:52,650 three, which they seem to learn sequentially. 751 00:33:52,809 --> 00:33:55,390 So it's a two -stage process. First, use syntax 752 00:33:55,390 --> 00:33:57,809 to figure out it's a number word, a set quantifier. 753 00:33:57,950 --> 00:34:00,789 Then map that category onto your innate understanding 754 00:34:00,789 --> 00:34:03,809 of quantity to get the exact meaning. That seems 755 00:34:03,809 --> 00:34:06,150 to be exactly right. Syntax provides the conceptual 756 00:34:06,150 --> 00:34:08,130 placeholder, identifying the type of meaning, 757 00:34:08,269 --> 00:34:10,530 and then the child uses their non -linguistic 758 00:34:10,530 --> 00:34:12,690 number sense, combined with observing how the 759 00:34:12,690 --> 00:34:14,889 words are used, often sequentially one, then 760 00:34:14,889 --> 00:34:17,929 two, then three, to nail down the precise cardinality. 761 00:34:18,250 --> 00:34:21,190 Okay, we've covered a lot of ground. The philosophical 762 00:34:21,190 --> 00:34:24,289 problems, the social inferences, the conceptual 763 00:34:24,289 --> 00:34:27,730 underpinnings, the syntactic aids. This brings 764 00:34:27,730 --> 00:34:30,530 us to the really big question, the one that philosophers 765 00:34:30,530 --> 00:34:33,369 and linguists have debated forever. What's the 766 00:34:33,369 --> 00:34:35,829 relationship between language and thought? Does 767 00:34:35,829 --> 00:34:38,150 language actually create our ability to think 768 00:34:38,150 --> 00:34:41,429 complex thoughts? We have to talk about the Horfian 769 00:34:41,429 --> 00:34:44,070 hypothesis. Right. The idea, often associated 770 00:34:44,070 --> 00:34:49,889 with Benjamin doesn't just express thought, it 771 00:34:49,889 --> 00:34:52,769 fundamentally shapes or even determines it. The 772 00:34:52,769 --> 00:34:55,230 categories and structures of your specific language 773 00:34:55,230 --> 00:34:57,289 create the categories and structures of your 774 00:34:57,289 --> 00:35:00,010 mind. So without language, thought would be... 775 00:35:00,010 --> 00:35:03,789 What? Just chaos. Worf used phrases like an unstructured 776 00:35:03,789 --> 00:35:06,550 mess or a kaleidoscopic flux of impressions. 777 00:35:06,929 --> 00:35:09,429 The radical version is that language is the medium 778 00:35:09,429 --> 00:35:11,650 of thought, and without it, complex structured 779 00:35:11,650 --> 00:35:14,090 thought is impossible. But based on everything 780 00:35:14,090 --> 00:35:16,050 we've just discussed... That seems unlikely. 781 00:35:16,210 --> 00:35:17,730 We've spent this whole time talking about the 782 00:35:17,730 --> 00:35:19,510 rich conceptual world of the pre -linguistic 783 00:35:19,510 --> 00:35:22,989 infant. Exactly. The entire premise of Bloom's 784 00:35:22,989 --> 00:35:25,369 argument and much of modern developmental psychology 785 00:35:25,369 --> 00:35:29,769 contradicts that strong Worfian view. We've seen 786 00:35:29,769 --> 00:35:32,070 compelling evidence that infants, before they 787 00:35:32,070 --> 00:35:35,170 learn language, already have sophisticated concepts 788 00:35:35,170 --> 00:35:39,869 of objects, number, causality, agency, intention. 789 00:35:40,500 --> 00:35:43,219 Their minds are anything but an unstructured 790 00:35:43,219 --> 00:35:45,650 mess. So the basic building blocks of thought 791 00:35:45,650 --> 00:35:48,570 seem to be there independent of language. Worf's 792 00:35:48,570 --> 00:35:51,090 starting point was wrong. The strong version, 793 00:35:51,250 --> 00:35:53,889 yes. The foundational premise that the mind starts 794 00:35:53,889 --> 00:35:56,030 with a blank slate structured only by language 795 00:35:56,030 --> 00:35:58,590 just doesn't hold up to the evidence from infant 796 00:35:58,590 --> 00:36:00,690 cognition. Does that mean language has no effect 797 00:36:00,690 --> 00:36:02,449 on thought? That seems too strong the other way. 798 00:36:02,550 --> 00:36:04,789 No, that's too strong as well. The modern perspective, 799 00:36:04,989 --> 00:36:08,110 often called neo -Horfinism, explores more subtle 800 00:36:08,110 --> 00:36:10,329 influences. Language might not create thought 801 00:36:10,329 --> 00:36:12,469 categories, but it might influence how we use 802 00:36:12,469 --> 00:36:14,809 them, or which ones become more salient or habitual. 803 00:36:15,190 --> 00:36:17,489 Like making certain ways of thinking easier or 804 00:36:17,489 --> 00:36:20,090 more automatic. Precisely. One line of research, 805 00:36:20,230 --> 00:36:22,530 associated with Steven Levinson, looks at spatial 806 00:36:22,530 --> 00:36:24,889 frames of reference. How different languages 807 00:36:24,889 --> 00:36:28,210 talk about space, like left -right versus north 808 00:36:28,210 --> 00:36:31,010 -south. Exactly. Some languages, like Tseltol 809 00:36:31,010 --> 00:36:34,579 Mayan, predominantly use an absolute frame. The 810 00:36:34,579 --> 00:36:36,980 cup is north of the plate or uphill from the 811 00:36:36,980 --> 00:36:39,219 plate. English speakers tend to use a relative 812 00:36:39,219 --> 00:36:42,449 frame based on their own body. The cup is to 813 00:36:42,449 --> 00:36:44,849 the left of the plate. Okay, so the languages 814 00:36:44,849 --> 00:36:47,489 differ. Does that change how people think about 815 00:36:47,489 --> 00:36:50,250 space even when not talking? Levinson's group 816 00:36:50,250 --> 00:36:52,869 found evidence that it does. They gave speakers 817 00:36:52,869 --> 00:36:56,030 non -linguistic spatial memory tasks, like remembering 818 00:36:56,030 --> 00:36:58,369 an arrangement of objects on a table after being 819 00:36:58,369 --> 00:37:01,360 turned around. Celtal speakers tended to reconstruct 820 00:37:01,360 --> 00:37:03,599 the array according to the absolute coordinates, 821 00:37:03,860 --> 00:37:06,260 while Dutch speakers, who use relative terms, 822 00:37:06,500 --> 00:37:08,739 reconstructed it relative to their new viewpoint. 823 00:37:09,019 --> 00:37:11,280 So language didn't give them the ability to think 824 00:37:11,280 --> 00:37:13,360 in absolute terms, but it made that way of thinking 825 00:37:13,360 --> 00:37:16,519 their default. That's the interpretation. Both 826 00:37:16,519 --> 00:37:18,880 absolute and relative spatial reasoning systems 827 00:37:18,880 --> 00:37:21,880 seem available to the human brain, but the language 828 00:37:21,880 --> 00:37:24,500 you speak constantly pushes you to use one system 829 00:37:24,500 --> 00:37:26,719 more than the other, making it your habitual 830 00:37:26,719 --> 00:37:29,670 mode of thought for that domain. Language acts 831 00:37:29,670 --> 00:37:32,670 as a kind of cognitive tuning mechanism. Okay, 832 00:37:32,730 --> 00:37:35,469 so language can bias which cognitive tools we 833 00:37:35,469 --> 00:37:37,650 use. What else can it do? You mentioned it acting 834 00:37:37,650 --> 00:37:40,449 as a cue for categorization. Yes, this is another 835 00:37:40,449 --> 00:37:42,949 powerful effect. Even if the underlying concepts 836 00:37:42,949 --> 00:37:45,969 exist, language provides a very clear signal 837 00:37:45,969 --> 00:37:48,769 about how to group things, especially when appearance 838 00:37:48,769 --> 00:37:50,809 might be misleading. Like the dinosaur example 839 00:37:50,809 --> 00:37:53,610 again, saying Triceratops and Brontosaurus are 840 00:37:53,610 --> 00:37:56,289 both dinosaurs prompts the child to generalize 841 00:37:56,289 --> 00:37:58,429 the hidden property. cold blood between them. 842 00:37:58,590 --> 00:38:01,510 Right. The shared label acts as an invitation 843 00:38:01,510 --> 00:38:04,389 or even an instruction to treat these visually 844 00:38:04,389 --> 00:38:06,530 different things as members of the same deeper 845 00:38:06,530 --> 00:38:09,130 kind and to generalize properties accordingly. 846 00:38:09,989 --> 00:38:12,230 Davidson and Gelman's work showed how powerful 847 00:38:12,230 --> 00:38:14,869 labels are in overriding superficial similarity 848 00:38:14,869 --> 00:38:17,389 when kids make inferences about hidden properties. 849 00:38:17,710 --> 00:38:20,210 The word becomes a catalyst for accessing that 850 00:38:20,210 --> 00:38:23,030 deeper, essentialist mode of thought. So pulling 851 00:38:23,030 --> 00:38:25,969 it all together, what's the final synthesis on 852 00:38:25,969 --> 00:38:28,570 language and thought? The synthesis following 853 00:38:28,570 --> 00:38:31,230 Bloom's line of argument is that the core conceptual 854 00:38:31,230 --> 00:38:34,030 capacities we've discussed, understanding objects, 855 00:38:34,389 --> 00:38:37,309 intentions, time, kinds, number, are fundamentally 856 00:38:37,309 --> 00:38:40,610 independent of and prior to language. They are 857 00:38:40,610 --> 00:38:42,449 necessary for language learning to even get off 858 00:38:42,449 --> 00:38:44,469 the ground. Language doesn't create these core 859 00:38:44,469 --> 00:38:47,969 ideas. It doesn't seem to. Language is instead 860 00:38:47,969 --> 00:38:50,150 an incredibly powerful tool that builds upon 861 00:38:50,150 --> 00:38:53,079 this pre -existing cognitive foundation. It's 862 00:38:53,079 --> 00:38:55,519 a tool for communication, obviously. It's a tool 863 00:38:55,519 --> 00:38:57,400 for transmitting culture and complex knowledge 864 00:38:57,400 --> 00:39:00,260 efficiently. And it allows us to construct and 865 00:39:00,260 --> 00:39:03,300 manipulate incredibly complex ideas like higher 866 00:39:03,300 --> 00:39:06,280 mathematics or... intricate logical arguments 867 00:39:06,280 --> 00:39:08,519 that might be very difficult or impossible to 868 00:39:08,519 --> 00:39:11,400 manage without the symbolic system language provides. 869 00:39:11,739 --> 00:39:14,760 So essential for complex expression and cultural 870 00:39:14,760 --> 00:39:17,460 learning, but not the source of the basic capacity 871 00:39:17,460 --> 00:39:19,780 to have ideas. That seems to be the most accurate 872 00:39:19,780 --> 00:39:21,719 picture based on the evidence. The foundations 873 00:39:21,719 --> 00:39:24,260 are conceptual. Language is the remarkable structure 874 00:39:24,260 --> 00:39:27,900 we build upon them. Hashtag tag outro. Wow. So 875 00:39:27,900 --> 00:39:30,000 this deep dive really changes how you think about 876 00:39:30,000 --> 00:39:32,039 something as common as a child learning a word. 877 00:39:32,119 --> 00:39:34,639 It's not just mimicry or association. It's this 878 00:39:34,639 --> 00:39:37,340 incredible cognitive feat drawing on philosophy, 879 00:39:37,599 --> 00:39:40,539 social intelligence, deep conceptual knowledge. 880 00:39:40,780 --> 00:39:43,280 It really is. Every time a child learns a word, 881 00:39:43,400 --> 00:39:45,500 they're implicitly solving problems that have 882 00:39:45,500 --> 00:39:47,940 stumped philosophers for centuries. They're acting 883 00:39:47,940 --> 00:39:50,920 like tiny anthropologists figuring out alien 884 00:39:50,920 --> 00:39:54,610 intentions. like nascent scientists carving the 885 00:39:54,610 --> 00:39:56,809 world at its concessional joints. It's quite 886 00:39:56,809 --> 00:39:58,909 amazing. It makes you appreciate the complexity 887 00:39:58,909 --> 00:40:01,829 simmering beneath that seemingly simple surface. 888 00:40:02,150 --> 00:40:04,469 Definitely. And it leaves us with a really interesting, 889 00:40:04,590 --> 00:40:06,949 maybe provocative final thought for you to chew 890 00:40:06,949 --> 00:40:10,260 on. If these core conceptual abilities, tracking 891 00:40:10,260 --> 00:40:13,460 objects, understanding intentions, basic categorization, 892 00:40:13,559 --> 00:40:16,039 are indeed present before language, what does 893 00:40:16,039 --> 00:40:18,480 that tell us about the minds of people who, for 894 00:40:18,480 --> 00:40:20,619 tragic reasons, perhaps never acquire language, 895 00:40:20,820 --> 00:40:24,199 like profoundly deaf individuals raised in isolation 896 00:40:24,199 --> 00:40:26,519 without sign language? What does the evidence 897 00:40:26,519 --> 00:40:28,980 suggest about their thinking? Well, it's limited 898 00:40:28,980 --> 00:40:31,739 and complex, but the evidence we have suggests 899 00:40:31,739 --> 00:40:34,380 that While they obviously miss out on the vast 900 00:40:34,380 --> 00:40:36,539 universe of knowledge transmitted through language 901 00:40:36,539 --> 00:40:39,639 and culture, their fundamental capacity for structured, 902 00:40:39,800 --> 00:40:43,380 abstract thought about space, number, even potentially 903 00:40:43,380 --> 00:40:46,380 others' minds, seems to remain largely intact. 904 00:40:47,000 --> 00:40:49,380 It's perhaps the most powerful argument for the 905 00:40:49,380 --> 00:40:52,059 independence of core thought from language. So 906 00:40:52,059 --> 00:40:55,460 the human mind has this robust core that language 907 00:40:55,460 --> 00:40:58,300 builds upon rather than creates. That seems to 908 00:40:58,300 --> 00:41:00,659 be the picture. The ongoing challenge, really, 909 00:41:00,760 --> 00:41:02,619 isn't whether things like theory of mind are 910 00:41:02,619 --> 00:41:04,920 needed for word learning, but getting much more 911 00:41:04,920 --> 00:41:07,219 precise about which specific aspects of social 912 00:41:07,219 --> 00:41:09,440 cognition are crucial for learning different 913 00:41:09,440 --> 00:41:11,840 kinds of words, nouns versus verbs versus abstract 914 00:41:11,840 --> 00:41:14,059 terms. That's where the intricate research continues.