1 00:00:00,000 --> 00:00:04,219 Welcome back to the Deep Dive. Today we're tackling 2 00:00:04,219 --> 00:00:07,500 a really huge question, maybe the biggest one, 3 00:00:07,540 --> 00:00:10,320 about what makes us human. Yeah. How did language 4 00:00:10,320 --> 00:00:12,419 evolve? It's definitely up there. I mean, it 5 00:00:12,419 --> 00:00:15,339 sets us apart so dramatically from, well, everything 6 00:00:15,339 --> 00:00:17,300 else. And the clues are scattered everywhere, 7 00:00:17,379 --> 00:00:19,879 aren't they? Fossils, genes, animal behavior, 8 00:00:20,280 --> 00:00:22,739 linguistics. Exactly. And that's why today we're 9 00:00:22,739 --> 00:00:25,500 really emphasizing this interdisciplinary approach. 10 00:00:26,239 --> 00:00:28,359 Our sources stress you just can't crack this 11 00:00:28,359 --> 00:00:30,519 looking through one single lens. No way. You 12 00:00:30,519 --> 00:00:33,500 try just genetics, you hit a wall, just archaeology, 13 00:00:33,539 --> 00:00:35,280 same thing. You really have to bring it all together, 14 00:00:35,399 --> 00:00:37,539 look for the overlaps, and importantly, accept 15 00:00:37,539 --> 00:00:40,659 that there isn't going to be one neat, simple 16 00:00:40,659 --> 00:00:44,079 answer. Right. And the core puzzle, the thing 17 00:00:44,079 --> 00:00:46,299 that starts this whole quest, is actually pretty 18 00:00:46,299 --> 00:00:48,859 simple on the surface, but deep down it's profound. 19 00:00:49,539 --> 00:00:52,939 Which is? Why can basically any human child... 20 00:00:53,229 --> 00:00:55,810 Learn language, this incredibly complex system, 21 00:00:55,929 --> 00:00:57,929 from what seems like pretty limited examples, 22 00:00:58,270 --> 00:01:00,929 linguists call it sparse data. Right. While even 23 00:01:00,929 --> 00:01:02,609 the smartest chimp, you know, give it all the 24 00:01:02,609 --> 00:01:04,629 human interaction in the world, it just can't 25 00:01:04,629 --> 00:01:07,010 get there. Yeah, that difference is stark. It 26 00:01:07,010 --> 00:01:09,629 really points towards some kind of unique, built 27 00:01:09,629 --> 00:01:12,109 -in biological capacity we have. Understanding 28 00:01:12,109 --> 00:01:15,030 how that capacity arose, that's the goal. Okay, 29 00:01:15,069 --> 00:01:17,950 so before we really dive in, there's a framework 30 00:01:17,950 --> 00:01:19,390 we need. You mentioned it before we started, 31 00:01:19,430 --> 00:01:22,260 the multi -component approach. Ah, yes. Absolutely 32 00:01:22,260 --> 00:01:24,359 crucial. We have to ditch the idea of language 33 00:01:24,359 --> 00:01:28,280 as this single solid block. You know, like it's 34 00:01:28,280 --> 00:01:29,959 one thing you either have or you don't. That 35 00:01:29,959 --> 00:01:32,819 old view makes it impossible to study evolutionarily, 36 00:01:32,879 --> 00:01:36,299 doesn't it? Totally. Because how does half a 37 00:01:36,299 --> 00:01:39,939 language evolve? But if you break it down, think 38 00:01:39,939 --> 00:01:42,659 of it more like a complex machine. It's got different 39 00:01:42,659 --> 00:01:44,939 parts, different subsystems doing different jobs. 40 00:01:45,099 --> 00:01:47,560 Like what kind of parts? Well, you've got semantics. 41 00:01:47,560 --> 00:01:49,560 That's the meaning side, the concepts. You've 42 00:01:49,560 --> 00:01:53,000 got vocal control. The physical ability to make 43 00:01:53,000 --> 00:01:55,260 and control speech sounds. And then there's syntax, 44 00:01:55,579 --> 00:01:57,959 the rules, the grammar, how we put things together. 45 00:01:58,239 --> 00:02:00,560 And the key insight is that these different bits 46 00:02:00,560 --> 00:02:03,019 probably have different evolutionary stories, 47 00:02:03,140 --> 00:02:05,239 different origins in the brain, maybe different 48 00:02:05,239 --> 00:02:08,080 genes. Precisely. They didn't all have to pop 49 00:02:08,080 --> 00:02:10,659 into existence at the same time. Some parts might 50 00:02:10,659 --> 00:02:12,979 be ancient, shared with other animals. Others 51 00:02:12,979 --> 00:02:15,400 might be uniquely human and relatively recent. 52 00:02:15,620 --> 00:02:18,340 Which leads to that. FLB -FLN distinction, right? 53 00:02:18,460 --> 00:02:21,819 Exactly. FLB is the faculty of language in the 54 00:02:21,819 --> 00:02:23,860 broad sense. That's kind of the big umbrella, 55 00:02:23,939 --> 00:02:26,060 includes everything involved, even things we 56 00:02:26,060 --> 00:02:28,659 share like memory or basic conceptual abilities. 57 00:02:29,039 --> 00:02:32,460 We cast that net wide first. To see what's not 58 00:02:32,460 --> 00:02:34,199 unique. Yeah, to see the shared foundations. 59 00:02:34,580 --> 00:02:38,360 Then we narrow down to FLN, the faculty of language 60 00:02:38,360 --> 00:02:40,819 in the narrow sense. That's usually where people 61 00:02:40,819 --> 00:02:44,610 put the really specific. human -only stuff, often 62 00:02:44,610 --> 00:02:47,509 focusing on that computational power, like recursion 63 00:02:47,509 --> 00:02:50,289 and syntax. Okay, got it. So starting broad with 64 00:02:50,289 --> 00:02:53,449 FLB lets us do comparative work. Yeah. Which 65 00:02:53,449 --> 00:02:56,129 brings us neatly, I think, to part one, syntax, 66 00:02:56,449 --> 00:02:59,210 semantics, and this big debate about innateness. 67 00:02:59,330 --> 00:03:01,789 Right. So that multi -component view, it really 68 00:03:01,789 --> 00:03:03,789 opens the door for comparing ourselves to other 69 00:03:03,789 --> 00:03:06,009 animals in a meaningful way. Because even if 70 00:03:06,009 --> 00:03:08,310 no other species has the whole package, lots 71 00:03:08,310 --> 00:03:10,030 of species have some of the parts. Give me an 72 00:03:10,030 --> 00:03:11,930 example. Okay, take semantics, the conceptual 73 00:03:11,930 --> 00:03:15,199 stuff. A lot of that basic ability to form concepts, 74 00:03:15,560 --> 00:03:18,639 understand relationships, maybe even basic event 75 00:03:18,639 --> 00:03:21,740 structures. Well, primates clearly have sophisticated 76 00:03:21,740 --> 00:03:24,860 versions of that. Dogs, birds, too, to some extent. 77 00:03:24,919 --> 00:03:27,979 That's part of FLB, but it's not unique. So the 78 00:03:27,979 --> 00:03:30,060 thinking part, the concepts themselves, weren't 79 00:03:30,060 --> 00:03:33,020 the big evolutionary leap for language per se. 80 00:03:33,199 --> 00:03:35,599 Not necessarily the concepts themselves, no. 81 00:03:35,740 --> 00:03:38,280 But then look at vocal learning, the ability 82 00:03:38,280 --> 00:03:41,280 to imitate sounds. Pymates, surprisingly, are 83 00:03:41,280 --> 00:03:43,479 pretty terrible at it. Our closest relatives 84 00:03:43,479 --> 00:03:45,840 can't really learn new calls. But other animals 85 00:03:45,840 --> 00:03:49,120 can. Oh, yeah. Songbirds are masters. Seals, 86 00:03:49,120 --> 00:03:51,699 dolphins, whales, even elephants show complex 87 00:03:51,699 --> 00:03:54,759 vocal imitation. So that piece, the ability to 88 00:03:54,759 --> 00:03:57,599 learn and produce complex sounds, evolved independently 89 00:03:57,599 --> 00:04:00,400 in other lineages. It's also part of FLB, but 90 00:04:00,400 --> 00:04:02,979 again, not uniquely human in its basic form. 91 00:04:03,379 --> 00:04:05,120 Evolution's trick was maybe putting these shared 92 00:04:05,120 --> 00:04:07,080 components together in a new way. Okay, so we 93 00:04:07,080 --> 00:04:08,919 have shared concepts, shared vocal learning ability 94 00:04:08,919 --> 00:04:11,580 in some species, but then we hit syntax. That 95 00:04:11,580 --> 00:04:13,340 seems to be the real sticking point, the core 96 00:04:13,340 --> 00:04:15,620 of FLN, maybe. That's often the argument, yes. 97 00:04:16,790 --> 00:04:19,230 Syntax, especially this idea of discrete infinity, 98 00:04:19,550 --> 00:04:22,310 is where things get really complex and potentially 99 00:04:22,310 --> 00:04:27,089 uniquely human. Discrete infinity. Sounds daunting. 100 00:04:27,230 --> 00:04:29,649 Break that down. It basically means we can generate 101 00:04:29,649 --> 00:04:31,870 a limitless number of unique sentences, unique 102 00:04:31,870 --> 00:04:35,230 meanings, from a finite set of elements, words, 103 00:04:35,389 --> 00:04:38,629 and a finite set of rules, grammar. Like Lego 104 00:04:38,629 --> 00:04:41,560 bricks. limited brick types infinite structures 105 00:04:41,560 --> 00:04:44,579 exactly like lego bricks and crucially it's not 106 00:04:44,579 --> 00:04:46,660 just stringing them together it's hierarchical 107 00:04:46,660 --> 00:04:50,319 we embed structures within structures think about 108 00:04:50,319 --> 00:04:53,220 sentences like the dog that chased the cat that 109 00:04:53,220 --> 00:04:56,480 stole the fish ran away your nesting ideas that's 110 00:04:56,480 --> 00:04:58,949 recursion right Embedding a phrase or clause 111 00:04:58,949 --> 00:05:01,670 inside another of the same type. Yes. Recursion 112 00:05:01,670 --> 00:05:04,370 is key to this unbounded generation. And the 113 00:05:04,370 --> 00:05:06,310 rules for how we combine things are structure 114 00:05:06,310 --> 00:05:08,230 dependent. They care about grammatical roles, 115 00:05:08,550 --> 00:05:11,009 noun, phrase, verb, phrase, not just the linear 116 00:05:11,009 --> 00:05:13,769 order of words. Which is complex stuff. Yet kids, 117 00:05:13,949 --> 00:05:16,310 they just seem to pick it up. Effortlessly, it 118 00:05:16,310 --> 00:05:18,509 seems. And that's what fuels the classic argument 119 00:05:18,509 --> 00:05:20,649 for innate language constraints the poverty of 120 00:05:20,649 --> 00:05:22,920 the stimulus's argument. Right, the idea that 121 00:05:22,920 --> 00:05:25,439 the input kids get is too messy, too limited, 122 00:05:25,500 --> 00:05:28,279 to logically deduce these incredibly abstract 123 00:05:28,279 --> 00:05:31,560 and subtle rules just from listening. Precisely. 124 00:05:31,560 --> 00:05:33,699 Take those reflexive pronouns again, himself, 125 00:05:34,019 --> 00:05:36,639 herself. The rules governing where they can point 126 00:05:36,639 --> 00:05:39,040 back to in a sentence are really quite intricate. 127 00:05:39,279 --> 00:05:40,839 Let's use that example because it makes it clear. 128 00:05:41,040 --> 00:05:44,519 The sentence, Akbar asked Maria to bathe himself. 129 00:05:44,959 --> 00:05:47,600 Why does that sound wrong to... basically any 130 00:05:47,600 --> 00:05:50,000 English speaker. Because himself needs to refer 131 00:05:50,000 --> 00:05:54,439 back to the subject of the bathing action. Structurally, 132 00:05:54,540 --> 00:05:56,379 in that sentence, the one doing the bathing is 133 00:05:56,379 --> 00:05:59,259 Maria. But Maria is female, so herself would 134 00:05:59,259 --> 00:06:01,860 be needed. Himself tries to jump over Maria to 135 00:06:01,860 --> 00:06:04,060 get back to Akbar, but the grammar says no, it 136 00:06:04,060 --> 00:06:06,439 has to be bound more locally. But you don't teach 137 00:06:06,439 --> 00:06:09,279 a kid that rule explicitly. They don't get examples 138 00:06:09,279 --> 00:06:11,579 of sentences like that being wrong. Not usually, 139 00:06:11,620 --> 00:06:14,519 no. And yet by four or five, they implicitly 140 00:06:14,519 --> 00:06:18,100 know it. Now tweak it slightly. Akbar asked Maria 141 00:06:18,100 --> 00:06:20,220 to help him bathe himself. And now it works perfectly 142 00:06:20,220 --> 00:06:22,639 fine. Himself refers to him, which refers to 143 00:06:22,639 --> 00:06:25,920 Akbar. Right. The child understands without instruction 144 00:06:25,920 --> 00:06:28,740 that the binding depends on the hierarchical 145 00:06:28,740 --> 00:06:31,279 grammatical structure, not just who's closest 146 00:06:31,279 --> 00:06:33,560 in the linear string of words. So the argument 147 00:06:33,560 --> 00:06:35,959 is the blueprint for those kinds of abstract 148 00:06:35,959 --> 00:06:38,839 rules must be built in somehow, part of our biology. 149 00:06:39,139 --> 00:06:41,060 That's the core of the innateness hypothesis. 150 00:06:41,620 --> 00:06:45,420 But, and this is a huge but, that innate capacity 151 00:06:45,420 --> 00:06:49,160 needs... Input. It's not enough on its own. This 152 00:06:49,160 --> 00:06:51,160 is where stories like Kaspar Hauser come in. 153 00:06:51,240 --> 00:06:53,779 The idea that if you don't get linguistic input 154 00:06:53,779 --> 00:06:56,540 during critical periods, the system doesn't develop 155 00:06:56,540 --> 00:06:59,600 properly. Exactly. Kaspar Hauser was tragically 156 00:06:59,600 --> 00:07:02,699 deprived of language input for years. When found, 157 00:07:03,019 --> 00:07:05,459 despite seeming cognitively normal otherwise, 158 00:07:05,680 --> 00:07:08,019 he struggled massively to learn language. He 159 00:07:08,019 --> 00:07:10,759 developed some communication, but it was rudimentary, 160 00:07:10,759 --> 00:07:13,980 lacking complex grammar. So the biological capacities 161 00:07:13,980 --> 00:07:16,850 there, it's necessary. but it needs the environment, 162 00:07:16,949 --> 00:07:19,290 the linguistic food, to actually grow and structure 163 00:07:19,290 --> 00:07:21,889 itself. Necessary but not sufficient. It's a 164 00:07:21,889 --> 00:07:24,129 biological potential that requires environmental 165 00:07:24,129 --> 00:07:27,069 triggering and shaping. Okay, so that's syntax 166 00:07:27,069 --> 00:07:29,910 and innateness. What about the other big piece, 167 00:07:30,069 --> 00:07:32,389 semantics? You mentioned concepts might be older. 168 00:07:32,779 --> 00:07:34,899 Yeah, this is a really foundational point in 169 00:07:34,899 --> 00:07:37,319 Fitch's work, especially. The evidence suggests 170 00:07:37,319 --> 00:07:39,779 that complex conceptual structures, thinking, 171 00:07:40,000 --> 00:07:42,399 remembering, planning, categorizing the world 172 00:07:42,399 --> 00:07:45,660 evolved well before language as we know it. And 173 00:07:45,660 --> 00:07:47,620 they seem to rely on different brain systems. 174 00:07:47,740 --> 00:07:50,560 So our ancestors, even before anything like modern 175 00:07:50,560 --> 00:07:54,079 language, already had rich mental lives. Absolutely. 176 00:07:54,459 --> 00:07:58,040 Think about the last common ancestor. LCA we 177 00:07:58,040 --> 00:08:00,779 share with chimpanzees. They likely had sophisticated 178 00:08:00,779 --> 00:08:03,459 memory, goal -directed planning, understanding 179 00:08:03,459 --> 00:08:06,860 of social hierarchies, maybe even basic causal 180 00:08:06,860 --> 00:08:10,740 reasoning, transitive inference. Complex tool 181 00:08:10,740 --> 00:08:13,139 use points to this too. And crucially. Crucially, 182 00:08:13,139 --> 00:08:15,579 they likely already had the ability to link a 183 00:08:15,579 --> 00:08:18,060 mental concept like danger or food source to 184 00:08:18,060 --> 00:08:20,300 an arbitrary sign, maybe a specific grunt, maybe 185 00:08:20,300 --> 00:08:23,040 a gesture. The basic symbolic capacity seems 186 00:08:23,040 --> 00:08:25,300 to be there. So evolution didn't need to invent 187 00:08:25,300 --> 00:08:27,470 thought. It needed to invent a communication 188 00:08:27,470 --> 00:08:30,230 system powerful enough to express the thoughts 189 00:08:30,230 --> 00:08:32,309 that were already possible. That's a really nice 190 00:08:32,309 --> 00:08:35,009 way to put it. The internal world was rich. The 191 00:08:35,009 --> 00:08:37,110 challenge was building the interface to share 192 00:08:37,110 --> 00:08:40,149 it effectively. But that raises a question. If 193 00:08:40,149 --> 00:08:42,710 the concepts were there and the basic symbolic 194 00:08:42,710 --> 00:08:45,809 link was possible, why didn't complex language 195 00:08:45,809 --> 00:08:47,909 just happen much earlier? Why the long delay? 196 00:08:48,350 --> 00:08:50,950 Ah, that's a great question. It forces us to 197 00:08:50,950 --> 00:08:53,710 maybe question linguistic uniformitarianism a 198 00:08:53,710 --> 00:08:56,789 bit. The idea that all languages, past and present, 199 00:08:56,909 --> 00:08:59,149 are equally complex. No such thing as a primitive 200 00:08:59,149 --> 00:09:01,809 language. Right. Linguists generally hold that 201 00:09:01,809 --> 00:09:04,710 view for modern languages. But when thinking 202 00:09:04,710 --> 00:09:07,470 about evolution, some researchers, like Neumeier, 203 00:09:07,590 --> 00:09:10,110 suggest we might need a more nuanced view. Maybe 204 00:09:10,110 --> 00:09:12,529 language or proto -language didn't even start 205 00:09:12,529 --> 00:09:14,669 primarily for communication. What else would 206 00:09:14,669 --> 00:09:18,470 it be for? Thinking. Possibly. Using linguistic 207 00:09:18,470 --> 00:09:21,029 -like structures internally to organize thoughts, 208 00:09:21,309 --> 00:09:24,389 plan complex sequences, represent knowledge better, 209 00:09:24,529 --> 00:09:27,009 maybe that internal structuring happened first. 210 00:09:27,230 --> 00:09:29,509 And only later got co -opted for talking to each 211 00:09:29,509 --> 00:09:32,370 other. It's a hypothesis. It suggests the initial 212 00:09:32,370 --> 00:09:34,730 evolutionary driver wasn't sharing information, 213 00:09:34,990 --> 00:09:38,629 but refining the internal mental toolkit. Communication 214 00:09:38,629 --> 00:09:42,019 was a later, secondary benefit. It means accepting 215 00:09:42,019 --> 00:09:44,279 that early language might have looked very different 216 00:09:44,279 --> 00:09:46,840 and served a different primary purpose than today's 217 00:09:46,840 --> 00:09:49,960 languages. That idea of stages leads us right 218 00:09:49,960 --> 00:09:52,159 into part two, these proto -language models. 219 00:09:52,519 --> 00:09:54,799 What did those intermediate steps look like? 220 00:09:54,919 --> 00:09:56,639 Let's start with the lexical proto -language 221 00:09:56,639 --> 00:09:59,220 idea. Okay, this is closely associated with Derek 222 00:09:59,220 --> 00:10:01,919 Bickerton. It's essentially a syntax final model. 223 00:10:02,139 --> 00:10:04,720 The idea is that the first major step towards 224 00:10:04,720 --> 00:10:07,019 language was building up a large vocabulary, 225 00:10:07,419 --> 00:10:10,980 a lexicon. So lots of words for things. Objects, 226 00:10:10,980 --> 00:10:13,399 actions, maybe basic properties. Exactly. Hundreds, 227 00:10:13,500 --> 00:10:15,759 maybe thousands of learned words with specific 228 00:10:15,759 --> 00:10:18,379 meanings. But critically, without the complex 229 00:10:18,379 --> 00:10:21,299 grammatical machinery, the syntax, that glues 230 00:10:21,299 --> 00:10:23,600 modern language together. So communication would 231 00:10:23,600 --> 00:10:26,799 be kind of like pigeon language. Yeah. Or Tarzan 232 00:10:26,799 --> 00:10:29,460 speak. Yeah. Me hungry. Danger water. Find mammoth 233 00:10:29,460 --> 00:10:32,159 west. Pretty much. Simple concatenation of words 234 00:10:32,159 --> 00:10:34,679 relying heavily on context or shared knowledge 235 00:10:34,679 --> 00:10:36,840 to figure out the full meaning. The structure 236 00:10:36,840 --> 00:10:39,570 is minimal. What would drive the need for so 237 00:10:39,570 --> 00:10:42,690 many distinct words, though? Why not just a few 238 00:10:42,690 --> 00:10:45,610 dozen basic calls like other primates? Bakerton's 239 00:10:45,610 --> 00:10:48,940 argument focuses on foraging. He argues that 240 00:10:48,940 --> 00:10:51,820 early hominids, as omnivorous scavengers and 241 00:10:51,820 --> 00:10:54,580 hunters, dealt with resources that were scattered, 242 00:10:54,740 --> 00:10:57,820 unpredictable, but potentially very high value, 243 00:10:57,860 --> 00:11:00,159 like a large animal carcass. Finding one would 244 00:11:00,159 --> 00:11:02,679 be a big deal for the group. A huge deal. And 245 00:11:02,679 --> 00:11:04,639 being able to communicate specifics about that 246 00:11:04,639 --> 00:11:07,220 find would be incredibly advantageous. Not just 247 00:11:07,220 --> 00:11:11,399 food, but big animal killed, two days walk west, 248 00:11:11,559 --> 00:11:14,590 many hyenas. The more detail, the better the 249 00:11:14,590 --> 00:11:16,809 group can coordinate and benefit. So the pressure 250 00:11:16,809 --> 00:11:18,990 is for more content, more specific information 251 00:11:18,990 --> 00:11:22,059 transfer. Yes, driven by these high -spakes foraging 252 00:11:22,059 --> 00:11:25,220 scenarios. That constant need for better, more 253 00:11:25,220 --> 00:11:27,960 precise information pushes the lexicon to expand. 254 00:11:28,279 --> 00:11:30,340 Okay, that's a very practical, survival -focused 255 00:11:30,340 --> 00:11:32,620 driver. But there's another major idea about 256 00:11:32,620 --> 00:11:34,659 the driving force, focusing more on social life, 257 00:11:34,820 --> 00:11:37,759 Robin Dunbar's hypothesis. Right, Dunbar comes 258 00:11:37,759 --> 00:11:40,019 at it from a different angle. He looks at primate 259 00:11:40,019 --> 00:11:42,940 social groups and the role of grooming in maintaining 260 00:11:42,940 --> 00:11:45,519 social bonds. Physical grooming, like picking 261 00:11:45,519 --> 00:11:48,580 fleas. Exactly. It's crucial for building trust, 262 00:11:48,799 --> 00:11:51,620 alliances, managing conflicts in primate groups. 263 00:11:51,879 --> 00:11:54,240 But it's very time consuming. You can only groom 264 00:11:54,240 --> 00:11:56,779 one individual at a time. And Dunbar noticed 265 00:11:56,779 --> 00:11:59,740 that as primate brain size, and presumably group 266 00:11:59,740 --> 00:12:02,620 size, increased. They'd eventually hit a time 267 00:12:02,620 --> 00:12:05,379 crunch. Hominid groups were likely getting larger, 268 00:12:05,519 --> 00:12:07,659 maybe for better defense against predators on 269 00:12:07,659 --> 00:12:10,659 the savanna. Dunbar calculated that for a typical 270 00:12:10,659 --> 00:12:13,620 estimated early human group size of around 150, 271 00:12:14,240 --> 00:12:16,600 they'd need to spend something like 40 % of their 272 00:12:16,600 --> 00:12:19,940 day just grooming to keep the social fabric intact. 273 00:12:20,340 --> 00:12:23,100 Which is unsustainable. Pretty much. So Dunbar 274 00:12:23,100 --> 00:12:25,279 proposes language evolved as a form of vocal 275 00:12:25,279 --> 00:12:28,179 grooming. talking, gossiping, sharing social 276 00:12:28,179 --> 00:12:30,259 information. You can do that with multiple people 277 00:12:30,259 --> 00:12:32,559 at once. Exactly. It's far more efficient. You 278 00:12:32,559 --> 00:12:34,500 can groom several individuals simultaneously, 279 00:12:35,059 --> 00:12:38,139 reinforcing bonds, sharing crucial social intel, 280 00:12:38,399 --> 00:12:40,559 who's trustworthy, who's cheating, who's allied 281 00:12:40,559 --> 00:12:42,580 with whom. And this ties into the free writer 282 00:12:42,580 --> 00:12:44,840 problem. How does language help there? Well, 283 00:12:44,919 --> 00:12:47,080 shared language, especially the development of 284 00:12:47,080 --> 00:12:49,940 dialects or accents, can act like a social badge. 285 00:12:50,429 --> 00:12:52,429 It helps you quickly identify who belongs to 286 00:12:52,429 --> 00:12:54,710 your cooperative group, who shares your norms, 287 00:12:54,850 --> 00:12:58,230 who you can likely trust. It helps maintain cooperation 288 00:12:58,230 --> 00:13:01,049 in these larger, potentially more anonymous groups. 289 00:13:01,230 --> 00:13:03,610 So for Bickerton, it's about the information 290 00:13:03,610 --> 00:13:07,370 needed for survival tasks like foraging. For 291 00:13:07,370 --> 00:13:09,980 Dunbar... It's about the social bonding needed 292 00:13:09,980 --> 00:13:12,860 for group cohesion and cooperation. Yeah, they 293 00:13:12,860 --> 00:13:15,080 focus on different primary benefits, though both 294 00:13:15,080 --> 00:13:17,320 ultimately tie back to the advantages of cooperation 295 00:13:17,320 --> 00:13:20,299 in larger groups. Was the key innovation the 296 00:13:20,299 --> 00:13:22,639 ability to share detailed environmental info 297 00:13:22,639 --> 00:13:25,559 or the ability to manage complex social relationships 298 00:13:25,559 --> 00:13:28,720 efficiently? Or maybe both played a role. Fascinating 299 00:13:28,720 --> 00:13:31,659 contrast. Now let's completely shift gears away 300 00:13:31,659 --> 00:13:34,480 from what was said or why to how. The gestural 301 00:13:34,480 --> 00:13:37,220 proto -language idea. Right. This set of theories, 302 00:13:37,419 --> 00:13:39,820 pushed by people like Michael Corbalis, suggests 303 00:13:39,820 --> 00:13:42,100 that the first complex symbolic communication 304 00:13:42,100 --> 00:13:45,720 system wasn't vocal at all. It was manual gestural. 305 00:13:45,820 --> 00:13:48,480 Using hands and bodies, like sign language precursors. 306 00:13:48,679 --> 00:13:51,960 Essentially, yes. The argument is that the pathway 307 00:13:51,960 --> 00:13:55,360 through gesture is actually much smoother, evolutionarily 308 00:13:55,360 --> 00:13:58,279 speaking, than jumping straight to complex vocal 309 00:13:58,279 --> 00:14:01,350 control. Why smoother? Several reasons. First, 310 00:14:01,570 --> 00:14:04,090 primates already have pretty sophisticated voluntary 311 00:14:04,090 --> 00:14:06,570 control over their hands and arms, much more 312 00:14:06,570 --> 00:14:09,250 so than their vocalizations. Think about tool 313 00:14:09,250 --> 00:14:12,509 use. Okay, makes sense. Second, gestures are 314 00:14:12,509 --> 00:14:15,389 inherently visible. This makes learning easier. 315 00:14:15,629 --> 00:14:17,690 You can see exactly what someone else is doing 316 00:14:17,690 --> 00:14:21,169 and imitate it. And third, gestures can be iconic. 317 00:14:21,509 --> 00:14:23,929 Meaning they can look like the thing they represent, 318 00:14:24,149 --> 00:14:26,690 like... Mimicking, drinking, or eating. Exactly. 319 00:14:26,909 --> 00:14:29,129 That provides a natural bridge from action to 320 00:14:29,129 --> 00:14:31,690 symbol. You can pantomime actions, trace shapes, 321 00:14:31,889 --> 00:14:34,830 point. It offers a more direct route to establishing 322 00:14:34,830 --> 00:14:37,210 shared meanings, perhaps, than arbitrary sounds 323 00:14:37,210 --> 00:14:39,590 initially. And wasn't there already a sophisticated 324 00:14:39,590 --> 00:14:43,409 gestural system in our ape ancestors? Yes. Chimpanzees 325 00:14:43,409 --> 00:14:45,889 and other apes have complex gestural repertoires 326 00:14:45,889 --> 00:14:49,379 used intentionally in social communication. The 327 00:14:49,379 --> 00:14:51,279 foundation was likely already there in the last 328 00:14:51,279 --> 00:14:53,700 common ancestor. We didn't have to invent gesture 329 00:14:53,700 --> 00:14:56,279 from scratch. And this connects to the mirror 330 00:14:56,279 --> 00:14:58,840 neuron discovery, doesn't it? It does, very strongly. 331 00:14:58,960 --> 00:15:00,980 The discovery of mirror neurons, neurons that 332 00:15:00,980 --> 00:15:03,320 fire both when an individual performs an action 333 00:15:03,320 --> 00:15:05,519 and when they see someone else perform that same 334 00:15:05,519 --> 00:15:08,220 action, provides a potential neural basis for 335 00:15:08,220 --> 00:15:11,559 understanding and imitation. The brain mechanism 336 00:15:11,559 --> 00:15:14,580 for achieving parity, recognizing the equivalence 337 00:15:14,580 --> 00:15:18,379 between seeing and doing. Precisely. Arbib and 338 00:15:18,379 --> 00:15:21,279 others have argued this mirror system, likely 339 00:15:21,279 --> 00:15:25,019 present in the LCA, could be a crucial pre -adaptation 340 00:15:25,019 --> 00:15:27,940 for language, especially a gestural one. It helps 341 00:15:27,940 --> 00:15:30,299 solve the problem of how meanings become shared 342 00:15:30,299 --> 00:15:33,320 through imitation. If my brain simulates your 343 00:15:33,320 --> 00:15:35,620 action when I see your gesture, we have a shared 344 00:15:35,620 --> 00:15:37,879 understanding almost immediately. It sounds like 345 00:15:37,879 --> 00:15:39,980 a really plausible starting point. Hands are 346 00:15:39,980 --> 00:15:42,539 controllable. Gestures can be iconic. The mirror 347 00:15:42,539 --> 00:15:45,019 system helps. But then you hit the big problem. 348 00:15:45,240 --> 00:15:49,440 Ah, yes. The gestural problem. If gestural proto 349 00:15:49,440 --> 00:15:51,740 -language was the start, and we know that modern 350 00:15:51,740 --> 00:15:54,299 sign languages are fully expressive, complex 351 00:15:54,299 --> 00:15:57,940 languages on par with spoken ones, why did we 352 00:15:57,940 --> 00:16:00,720 ever switch? Why did speech take over almost 353 00:16:00,720 --> 00:16:02,980 universally? That's the million -dollar question 354 00:16:02,980 --> 00:16:06,299 for gestural theories. The move to vocal language 355 00:16:06,299 --> 00:16:08,919 must have offered some massive advantage to outweigh 356 00:16:08,919 --> 00:16:12,080 the benefits of gesture and, crucially, to overcome 357 00:16:12,080 --> 00:16:15,059 the significant downsides of speech. Like? Well. 358 00:16:15,289 --> 00:16:18,570 As we discussed, speech requires incredibly fine 359 00:16:18,570 --> 00:16:21,149 motor control of apparatus not initially designed 360 00:16:21,149 --> 00:16:24,250 for it. It takes over the mouth needed for breathing 361 00:16:24,250 --> 00:16:26,330 and eating. Remember those tough Pleistocene 362 00:16:26,330 --> 00:16:29,870 diets? And perhaps most significantly, the anatomical 363 00:16:29,870 --> 00:16:32,830 changes needed, like the lowered larynx. Increase 364 00:16:32,830 --> 00:16:35,169 the risk of choking, a potentially lethal tradeoff. 365 00:16:35,190 --> 00:16:37,509 A huge risk. So what advantage could be worth 366 00:16:37,509 --> 00:16:40,070 that? Gestural proponents struggle to provide 367 00:16:40,070 --> 00:16:42,590 a killer argument. Maybe communicating at night. 368 00:16:42,970 --> 00:16:45,169 or when your hands are full carrying tools or 369 00:16:45,169 --> 00:16:48,429 babies, or over longer distances. But are those 370 00:16:48,429 --> 00:16:50,669 advantages big enough to drive such a fundamental 371 00:16:50,669 --> 00:16:54,049 risky shift across the entire species? It remains 372 00:16:54,049 --> 00:16:56,210 a point of debate. The pressure for switching 373 00:16:56,210 --> 00:16:58,649 must have been immense, which leads us nicely 374 00:16:58,649 --> 00:17:00,889 into part three, thinking about sound itself 375 00:17:00,889 --> 00:17:03,750 and the role music might have played. The musical 376 00:17:03,750 --> 00:17:06,069 proto -language hypothesis. This feels different 377 00:17:06,069 --> 00:17:09,230 again. It is. These models suggest the intermediate 378 00:17:09,230 --> 00:17:12,269 stage was, well, song -like. Not necessarily 379 00:17:12,269 --> 00:17:16,869 words set to music, but complex, structured vocalizations 380 00:17:16,869 --> 00:17:19,910 that had rhythm, melody, intricate phonology, 381 00:17:20,009 --> 00:17:23,230 but maybe lacked precise compositional meaning 382 00:17:23,230 --> 00:17:25,950 initially. More about emotion, social bonding, 383 00:17:26,109 --> 00:17:29,250 display. Exactly. Think elaborate bird song rather 384 00:17:29,250 --> 00:17:31,410 than conversation. And this idea actually goes 385 00:17:31,410 --> 00:17:33,730 way back to Darwin himself. Darwin had a theory 386 00:17:33,730 --> 00:17:36,769 on this. He did. Quite a detailed one. He proposed 387 00:17:36,769 --> 00:17:40,009 three stages. First, general cognitive development 388 00:17:40,009 --> 00:17:42,609 in our ancestors, driven by things like tool 389 00:17:42,609 --> 00:17:44,789 use and social complexity. That's the foundation. 390 00:17:45,170 --> 00:17:47,450 Okay. Second, the evolution of vocal imitation, 391 00:17:47,630 --> 00:17:50,450 but used musically. He explicitly talked about 392 00:17:50,450 --> 00:17:53,130 true musical cadences, essentially singing, and 393 00:17:53,130 --> 00:17:54,950 he thought the main driver was sexual selection. 394 00:17:55,210 --> 00:17:57,369 Like birds singing to attract mates or defend 395 00:17:57,369 --> 00:18:00,769 territory. Precisely. Early hominids using complex 396 00:18:00,769 --> 00:18:03,630 songs in courtship displays or to challenge rivals. 397 00:18:03,950 --> 00:18:06,509 It provides a powerful context for honing vocal 398 00:18:06,509 --> 00:18:08,910 control and imitation skills. Think of it as 399 00:18:08,910 --> 00:18:11,250 practice for the complex motor sequences needed 400 00:18:11,250 --> 00:18:13,609 later for speech. And only after that musical 401 00:18:13,609 --> 00:18:16,329 vocal capacity was established. Only in Stage 402 00:18:16,329 --> 00:18:19,710 3, according to Darwin, did specific propositional 403 00:18:19,710 --> 00:18:22,890 meanings start to infiltrate the system. Words 404 00:18:22,890 --> 00:18:24,849 attached themselves to the pre -existing musical 405 00:18:24,849 --> 00:18:27,630 structures. Meaning came later. So the ability 406 00:18:27,630 --> 00:18:30,329 to make complex sounds came first, driven by 407 00:18:30,329 --> 00:18:33,269 social display. And then we figured out how to 408 00:18:33,269 --> 00:18:36,329 attach specific meanings to those sounds. That's 409 00:18:36,329 --> 00:18:38,650 the core idea. And it highlights the power of 410 00:18:38,650 --> 00:18:41,470 phonology, the sound system itself. Which brings 411 00:18:41,470 --> 00:18:44,519 us to duality of patterning. Explain that. This 412 00:18:44,519 --> 00:18:47,500 is a key design feature of spoken language. We 413 00:18:47,500 --> 00:18:50,259 have a small set of meaningless basic sounds, 414 00:18:50,319 --> 00:18:53,119 foams. Think K -A -T. Individually, they mean 415 00:18:53,119 --> 00:18:55,619 nothing. Right. But we can recombine this small 416 00:18:55,619 --> 00:18:57,960 set in different orders to create a vast number 417 00:18:57,960 --> 00:19:01,880 of meaningful units, words. Cat, act, tack. This 418 00:19:01,880 --> 00:19:04,700 is the first patterning. Then we take these meaningful 419 00:19:04,700 --> 00:19:06,740 words and combine them according to grammatical 420 00:19:06,740 --> 00:19:09,059 rules to make limitless sentences. That's the 421 00:19:09,059 --> 00:19:10,960 second patterning. Our levels of combination. 422 00:19:11,559 --> 00:19:14,109 Meaningless sounds make words. Words make sentences. 423 00:19:14,450 --> 00:19:17,170 Exactly. And this system is incredibly efficient. 424 00:19:17,509 --> 00:19:19,910 A small inventory sounds gives you potentially 425 00:19:19,910 --> 00:19:23,329 tens of thousands of distinct words. This generative 426 00:19:23,329 --> 00:19:26,890 power is a huge advantage of vocal systems. Arbitrariness, 427 00:19:26,930 --> 00:19:29,269 the fact that the word cat doesn't sound like 428 00:19:29,269 --> 00:19:32,109 a cat, becomes easy, almost automatic, once you're 429 00:19:32,109 --> 00:19:35,190 combining meaningless sounds. And this. intricate, 430 00:19:35,289 --> 00:19:38,009 efficient design. This is where Pinker argues 431 00:19:38,009 --> 00:19:40,009 natural selection must have been involved. Yes. 432 00:19:40,509 --> 00:19:43,250 Pinker's point is that such a complex, well -engineered 433 00:19:43,250 --> 00:19:46,589 system is unlikely to arise just by chance or 434 00:19:46,589 --> 00:19:49,210 general learning mechanisms. Its efficiency, 435 00:19:49,569 --> 00:19:52,349 its rapid acquisition by children, its universal 436 00:19:52,349 --> 00:19:55,170 structure, it looks like a biological adaptation 437 00:19:55,170 --> 00:19:58,299 shaped by natural selection over time. Selection 438 00:19:58,299 --> 00:20:00,299 favored individuals whose brains were better 439 00:20:00,299 --> 00:20:02,980 wired for this kind of structured vocal communication. 440 00:20:03,339 --> 00:20:05,559 Okay, so selection shapes the brain for sound 441 00:20:05,559 --> 00:20:07,880 processing and production. What about the physical 442 00:20:07,880 --> 00:20:10,559 apparatus, the vocal tract itself? This brings 443 00:20:10,559 --> 00:20:13,140 us to the lowered larynx debate. Ah, the famous 444 00:20:13,140 --> 00:20:15,460 descended larynx. It's one of the most obvious 445 00:20:15,460 --> 00:20:17,779 anatomical differences between us and apes related 446 00:20:17,779 --> 00:20:21,019 to speech potential. Our larynx sits much lower 447 00:20:21,019 --> 00:20:23,400 in the throat. Which, as we said, is dangerous. 448 00:20:23,849 --> 00:20:25,869 It creates a shared pathway for food and air, 449 00:20:25,990 --> 00:20:28,509 low down the pharynx, increasing choking risk. 450 00:20:28,710 --> 00:20:31,289 Why take that risk? The classic explanation came 451 00:20:31,289 --> 00:20:34,750 from Philip Lieberman. He argued forcefully that 452 00:20:34,750 --> 00:20:36,829 the lowered larynx was a specific adaptation 453 00:20:36,829 --> 00:20:39,569 for speech production. How so? By lowering the 454 00:20:39,569 --> 00:20:41,589 larynx, you create a larger pharyngeal cavity 455 00:20:41,589 --> 00:20:44,680 above it. This two -tube shape of the human vocal 456 00:20:44,680 --> 00:20:47,900 tract, oral cavity and pharyngeal cavity, allows 457 00:20:47,900 --> 00:20:50,859 us to produce a wider range of distinct vowel 458 00:20:50,859 --> 00:20:54,400 sounds, particularly the point vowels A, I, and 459 00:20:54,400 --> 00:20:56,920 U, which are acoustically very stable and distinct. 460 00:20:57,230 --> 00:20:59,690 And Lieberman really emphasized the vowel I, 461 00:20:59,890 --> 00:21:02,569 the E sound. He did. He argued it is particularly 462 00:21:02,569 --> 00:21:04,930 important because it represents an acoustic extreme. 463 00:21:05,329 --> 00:21:08,190 Having these clear, distinct vowel sounds, especially 464 00:21:08,190 --> 00:21:10,930 the extremes, makes speech much more robust and 465 00:21:10,930 --> 00:21:13,029 easier to understand, especially when dealing 466 00:21:13,029 --> 00:21:15,430 with variation between speakers, different vocal 467 00:21:15,430 --> 00:21:18,210 tract sizes. It helps listeners normalize the 468 00:21:18,210 --> 00:21:20,849 signal. So the benefit for reliable speech perception 469 00:21:20,849 --> 00:21:23,049 outweighed the choking risk. It sounds like a 470 00:21:23,049 --> 00:21:25,660 neat story. But comparative evidence complicated 471 00:21:25,660 --> 00:21:28,940 things. Hugely. Lieberman's initial work was 472 00:21:28,940 --> 00:21:32,680 based mostly on humans and chimps. But then researchers 473 00:21:32,680 --> 00:21:35,500 started looking more broadly. The killer finding 474 00:21:35,500 --> 00:21:38,740 came from studies of red deer. Red deer? What 475 00:21:38,740 --> 00:21:41,220 do they do? Male red deer temporarily retract 476 00:21:41,220 --> 00:21:44,039 their larynx way down into their chest when they 477 00:21:44,039 --> 00:21:46,279 roar during the mating season. Why on earth would 478 00:21:46,279 --> 00:21:47,859 they do that? To make themselves sound bigger. 479 00:21:48,500 --> 00:21:51,059 Lowering the larynx effectively lengthens the 480 00:21:51,059 --> 00:21:54,029 vocal tract tube. which lowers the resonant frequencies, 481 00:21:54,269 --> 00:21:57,190 formants of their calls. Lower frequencies typically 482 00:21:57,190 --> 00:21:59,869 correlate with larger body size, so they're essentially 483 00:21:59,869 --> 00:22:02,609 using laryngeal descent as a form of acoustic 484 00:22:02,609 --> 00:22:05,569 bluffing, exaggerating their size to intimidate 485 00:22:05,569 --> 00:22:08,509 rivals. Wow, so the anatomical ability to lower 486 00:22:08,509 --> 00:22:10,690 the larynx wasn't originally for speech vowels 487 00:22:10,690 --> 00:22:13,160 at all. It seems not. It looks like it was an 488 00:22:13,160 --> 00:22:15,640 adaptation for size exaggeration, likely driven 489 00:22:15,640 --> 00:22:18,460 by sexual selection or competition, perhaps present 490 00:22:18,460 --> 00:22:21,640 much earlier in mammals. In humans, this pre 491 00:22:21,640 --> 00:22:24,559 -existing ability was then co -opted for speech. 492 00:22:24,960 --> 00:22:27,339 We repurposed the size exaggeration machinery 493 00:22:27,339 --> 00:22:30,559 to make a wider range of speech sounds. Exactly. 494 00:22:30,660 --> 00:22:33,539 The anatomy came first, likely for display, and 495 00:22:33,539 --> 00:22:35,920 speech took advantage of it later. This fits 496 00:22:35,920 --> 00:22:37,660 better with the musical proto -language idea, 497 00:22:37,819 --> 00:22:41,359 too. Display first. referential meaning later 498 00:22:41,359 --> 00:22:43,980 this also connects back to the idea of holistic 499 00:22:43,980 --> 00:22:46,539 meaning doesn't it if early vocalizations were 500 00:22:46,539 --> 00:22:49,920 more like roars or songs there were likely holistic 501 00:22:49,920 --> 00:22:54,099 signals allison ray champions this idea a complex 502 00:22:54,099 --> 00:22:56,460 vocalization pattern wouldn't be broken down 503 00:22:56,460 --> 00:22:59,240 into words the whole chunk of sound would carry 504 00:22:59,240 --> 00:23:02,920 a meaning like danger or come here or this is 505 00:23:02,920 --> 00:23:05,720 my territory so the big transition then was moving 506 00:23:05,720 --> 00:23:08,640 from those holistic chunks to an analytic compositional 507 00:23:08,640 --> 00:23:11,279 system where meaning resides in smaller parts, 508 00:23:11,400 --> 00:23:13,900 words that get combined. That's the crucial shift 509 00:23:13,900 --> 00:23:15,859 needed to get to modeled language. How do you 510 00:23:15,859 --> 00:23:17,960 break down those holistic signals into meaningful 511 00:23:17,960 --> 00:23:20,339 segments? How does analysis emerge? And this 512 00:23:20,339 --> 00:23:22,759 might not be purely biological evolution. This 513 00:23:22,759 --> 00:23:25,059 is where cultural evolution likely plays a massive 514 00:23:25,059 --> 00:23:28,299 role. The process of language change itself over 515 00:23:28,299 --> 00:23:30,559 generations of speakers learning and using the 516 00:23:30,559 --> 00:23:33,920 language can drive this analytic structure. How 517 00:23:33,920 --> 00:23:36,140 does that work? Can you model it? Yes, researchers 518 00:23:36,140 --> 00:23:38,259 like Bart DeBoer have used computer simulations. 519 00:23:38,680 --> 00:23:41,980 You can create artificial agents with basic biological 520 00:23:41,980 --> 00:23:45,220 predispositions like the ability to imitate sounds 521 00:23:45,220 --> 00:23:48,220 and perceptual motor constraints similar to humans 522 00:23:48,220 --> 00:23:51,200 and then have them learn from each other over 523 00:23:51,200 --> 00:23:53,859 generations. And what happens? What often happens 524 00:23:53,859 --> 00:23:56,980 is that structured systems like vowel systems 525 00:23:56,980 --> 00:23:59,859 that resemble human languages emerge spontaneously. 526 00:24:00,640 --> 00:24:02,960 The pressures of learnability and usability, 527 00:24:03,440 --> 00:24:06,700 acting culturally over time, push the system 528 00:24:06,700 --> 00:24:09,220 towards becoming more segmented, more structured, 529 00:24:09,359 --> 00:24:12,160 more analytic. So the specific structures, like 530 00:24:12,160 --> 00:24:14,619 the exact vowel inventory of English or Japanese, 531 00:24:14,920 --> 00:24:17,759 might not be directly coded in our genes. Probably 532 00:24:17,759 --> 00:24:20,240 not. They might be culturally discovered solutions, 533 00:24:20,559 --> 00:24:22,839 stable patterns that emerge from the interaction 534 00:24:22,839 --> 00:24:25,279 of our general biological capacities with the 535 00:24:25,279 --> 00:24:27,200 demands of communication and learning passed 536 00:24:27,200 --> 00:24:30,019 down culturally. The biology provides the canvas 537 00:24:30,019 --> 00:24:32,180 and maybe some basic tools, but the specific 538 00:24:32,180 --> 00:24:34,980 painting is a result of cultural history. That 539 00:24:34,980 --> 00:24:37,079 integration of biology and culture seems key. 540 00:24:37,299 --> 00:24:39,880 Okay, let's move to part four. What are the actual 541 00:24:39,880 --> 00:24:42,500 physical traces, the neural and genetic footprints? 542 00:24:42,819 --> 00:24:46,559 Right. First, let's revisit motor control. We 543 00:24:46,559 --> 00:24:50,049 said primates lack complex vocal learning. Speech 544 00:24:50,049 --> 00:24:52,910 requires incredibly fast, precise, coordinated 545 00:24:52,910 --> 00:24:56,730 movements of the tongue, lips, jaw, larynx. Much 546 00:24:56,730 --> 00:24:59,029 more complex than eating or breathing requires. 547 00:24:59,250 --> 00:25:02,150 Far more. And neuroanatomically, humans seem 548 00:25:02,150 --> 00:25:04,829 to have something unique. Direct neural connections 549 00:25:04,829 --> 00:25:07,329 from the motor cortex straight down to the motor 550 00:25:07,329 --> 00:25:09,529 neurons controlling the larynx. Bypassing some 551 00:25:09,529 --> 00:25:11,670 of the intermediate pathways seen in other primates. 552 00:25:11,809 --> 00:25:15,170 Exactly. This direct wiring allows for much finer, 553 00:25:15,250 --> 00:25:17,490 more voluntary moment -to -moment control over 554 00:25:17,490 --> 00:25:20,230 vocalization. It's thought to be essential for 555 00:25:20,230 --> 00:25:22,390 the dexterity needed for speech articulation. 556 00:25:22,750 --> 00:25:25,509 Any evolutionary story has to explain when and 557 00:25:25,509 --> 00:25:27,829 why these novel circuits evolved. And this brings 558 00:25:27,829 --> 00:25:31,339 us to the famous FOXP2 gene. It's often called 559 00:25:31,339 --> 00:25:33,039 the language gene, but that's too simple, right? 560 00:25:33,099 --> 00:25:35,680 Way too simple. FOXP2 is not the gene for language. 561 00:25:35,779 --> 00:25:38,200 It's a regulatory gene, a transcription factor, 562 00:25:38,359 --> 00:25:40,059 meaning it affects the expression of many other 563 00:25:40,059 --> 00:25:42,660 genes. But it's clearly important for language, 564 00:25:42,799 --> 00:25:44,759 specifically the motor aspect. How do we know? 565 00:25:44,900 --> 00:25:47,759 From studies of a family known as the KD family, 566 00:25:48,079 --> 00:25:50,319 where about half the members had a mutation in 567 00:25:50,319 --> 00:25:54,140 FOXP2. They suffered from a severe speech disorder, 568 00:25:54,480 --> 00:25:57,660 developmental verbal dyspraxia. They had trouble 569 00:25:57,660 --> 00:26:00,000 coordinating the complex sequences of muscle 570 00:26:00,000 --> 00:26:02,799 movements for clear speech. They also had some 571 00:26:02,799 --> 00:26:05,059 grammatical difficulties, possibly linked to 572 00:26:05,059 --> 00:26:07,519 the sequencing problems. So it's crucial for 573 00:26:07,519 --> 00:26:10,380 the motor skills underlying fluent speech production. 574 00:26:10,579 --> 00:26:12,759 Yeah, that seems to be its core role in humans. 575 00:26:13,039 --> 00:26:16,500 And interestingly, FOXT2 shows deep homology. 576 00:26:16,759 --> 00:26:18,680 Meaning the gene itself is ancient and found 577 00:26:18,680 --> 00:26:21,640 in many animals. Yes, and it often plays analogous 578 00:26:21,640 --> 00:26:23,920 roles related to motor control or learning in 579 00:26:23,920 --> 00:26:26,440 other species. In songbirds, it's involved in 580 00:26:26,440 --> 00:26:28,400 the brain circuits, essential for learning and 581 00:26:28,400 --> 00:26:31,119 producing their complex songs. In mice, it's... 582 00:26:31,119 --> 00:26:32,819 It's involved in motor learning, like learning 583 00:26:32,819 --> 00:26:35,940 complex sequences of movements. Studying FOXP2 584 00:26:35,940 --> 00:26:38,200 in birds or mice can actually tell us something 585 00:26:38,200 --> 00:26:40,700 about the fundamental mechanisms relevant to 586 00:26:40,700 --> 00:26:43,440 human speech. Absolutely. It justifies that broad 587 00:26:43,440 --> 00:26:46,019 comparative approach. The basic molecular tools 588 00:26:46,019 --> 00:26:48,400 might be conserved, even if the final output 589 00:26:48,400 --> 00:26:51,000 birdsong versus human speech is different. Okay, 590 00:26:51,059 --> 00:26:55,000 timing. When did the human version of FOXP2 arise 591 00:26:55,000 --> 00:26:58,279 and become fixed? Genetic analyses looking at 592 00:26:58,279 --> 00:27:00,380 the differences between the human and chimpanzee 593 00:27:00,380 --> 00:27:03,799 FOXP2 genes and the patterns of variation in 594 00:27:03,799 --> 00:27:06,599 modern humans suggest the human -specific version 595 00:27:06,599 --> 00:27:09,460 underwent strong positive selection. It was highly 596 00:27:09,460 --> 00:27:11,900 advantageous. The estimates for when it became 597 00:27:11,900 --> 00:27:14,119 fixed in the human lineage cluster somewhere 598 00:27:14,119 --> 00:27:16,819 between, say, 200 ,000 years ago and perhaps 599 00:27:16,819 --> 00:27:19,259 as recently as 100 ,000 years ago, though there's 600 00:27:19,259 --> 00:27:21,700 still debate. That time frame, roughly coinciding 601 00:27:21,700 --> 00:27:23,920 with the appearance of anatomically modern humans, 602 00:27:24,140 --> 00:27:26,859 seems significant. It does. It suggests that 603 00:27:26,859 --> 00:27:28,740 the fine -tuning of the motor control system 604 00:27:28,740 --> 00:27:31,759 for articulate speech might be a relatively recent 605 00:27:31,759 --> 00:27:34,680 event in human evolution specific to our lineage. 606 00:27:34,980 --> 00:27:37,599 How does that square with the fossil evidence? 607 00:27:38,119 --> 00:27:40,599 Because that seems to point earlier. Ah, the 608 00:27:40,599 --> 00:27:43,059 eternal tension between genes and bones. Yes, 609 00:27:43,099 --> 00:27:45,019 the fossil evidence is tricky because language 610 00:27:45,019 --> 00:27:48,240 doesn't fossilize. We have to rely on proxies. 611 00:27:48,400 --> 00:27:50,740 Like the ipoglossal canal. Right. That's the 612 00:27:50,740 --> 00:27:52,960 hole at the base of the skull where the nerve 613 00:27:52,960 --> 00:27:55,410 controlling the tongue. the hypoglossal nerve 614 00:27:55,410 --> 00:27:58,150 passes through. In humans, it's significantly 615 00:27:58,150 --> 00:28:01,369 larger relative to body size than in apes. Suggesting 616 00:28:01,369 --> 00:28:03,849 more nerve fibers, finer muscle control of the 617 00:28:03,849 --> 00:28:06,490 tongue? That's the inference. More wiring for 618 00:28:06,490 --> 00:28:08,769 more dexterous tongue, presumably needed for 619 00:28:08,769 --> 00:28:11,269 complex speech articulation. And when does that 620 00:28:11,269 --> 00:28:14,069 canal reach modern human size in the fossil record? 621 00:28:14,539 --> 00:28:16,759 Well, studies on fossils like those from Sima 622 00:28:16,759 --> 00:28:19,519 de los Huesos in Spain, which are ancestors of 623 00:28:19,519 --> 00:28:21,359 Neanderthals, and on Neanderthals themselves, 624 00:28:21,559 --> 00:28:24,220 suggest the canal size was already in the modern 625 00:28:24,220 --> 00:28:27,539 human range by at least 300 ,000 years ago, maybe 626 00:28:27,539 --> 00:28:29,740 even earlier. Okay, so that's potentially 100 627 00:28:29,740 --> 00:28:33,220 ,000 years or more before the estimated FOXP2 628 00:28:33,220 --> 00:28:35,420 sweep. The tongue control might have been in 629 00:28:35,420 --> 00:28:38,200 place earlier? It creates a puzzle. Maybe the 630 00:28:38,200 --> 00:28:40,359 basic anatomy for tongue dexterity was there, 631 00:28:40,480 --> 00:28:43,059 but the specific neural programming facilitated 632 00:28:43,059 --> 00:28:45,980 by the modern FOXP2 variant came later, allowing 633 00:28:45,980 --> 00:28:48,259 for faster learning or more complex sequencing. 634 00:28:48,539 --> 00:28:50,539 And we can go back even further, can't we? Brain 635 00:28:50,539 --> 00:28:54,000 shape changes. Yes. Looking at endocasts, impressions 636 00:28:54,000 --> 00:28:56,339 of the brain's surface on the inside of the skull, 637 00:28:56,500 --> 00:28:58,799 there's evidence suggesting that the region around 638 00:28:58,799 --> 00:29:01,160 Broca's area, typically associated with language 639 00:29:01,160 --> 00:29:03,259 processing and motor control in modern humans, 640 00:29:03,440 --> 00:29:05,779 might have started reorganizing or expanding 641 00:29:05,779 --> 00:29:08,319 as far back as Homo habilis around two million 642 00:29:08,319 --> 00:29:11,339 years ago. Two million years. That's vastly earlier. 643 00:29:11,619 --> 00:29:13,960 What could have been driving that change so early? 644 00:29:14,319 --> 00:29:17,099 One influential idea links it to toolmaking. 645 00:29:17,299 --> 00:29:20,319 The appearance of the first stone tools. Oldowan 646 00:29:20,319 --> 00:29:23,660 tradition and later, the more complex Acheulean 647 00:29:23,660 --> 00:29:26,619 hand axes require sophisticated planning, sequential 648 00:29:26,619 --> 00:29:29,519 action, and fine motor control. often biased 649 00:29:29,519 --> 00:29:31,660 towards the right hand. Which is controlled by 650 00:29:31,660 --> 00:29:33,839 the left hemisphere, where Broca's area is typically 651 00:29:33,839 --> 00:29:36,579 located. Exactly. So the hypothesis is that the 652 00:29:36,579 --> 00:29:38,700 neural circuits for complex sequential motor 653 00:29:38,700 --> 00:29:41,180 planning, perhaps initially evolved for tool 654 00:29:41,180 --> 00:29:43,839 manufacture, were later co -opted or provided 655 00:29:43,839 --> 00:29:46,099 a foundation for the sequential processing required 656 00:29:46,099 --> 00:29:49,579 in language, both for production, syntax, articulation, 657 00:29:49,799 --> 00:29:52,529 and maybe even comprehension. The brain circuits 658 00:29:52,529 --> 00:29:54,789 got good at sequencing actions for making tools, 659 00:29:54,990 --> 00:29:57,210 and that ability turned out to be useful for 660 00:29:57,210 --> 00:29:59,589 sequencing sounds. It's a compelling example 661 00:29:59,589 --> 00:30:03,529 of exactation. The ability evolved for one purpose, 662 00:30:03,809 --> 00:30:06,250 or maybe multiple purposes related to motor control, 663 00:30:06,390 --> 00:30:09,529 and then became available for this new, incredibly 664 00:30:09,529 --> 00:30:11,910 powerful function of language. It really paints 665 00:30:11,910 --> 00:30:14,069 a picture of language being assembled piece by 666 00:30:14,069 --> 00:30:16,769 piece over a vast timescale, doesn't it? Absolutely. 667 00:30:16,829 --> 00:30:19,690 It's a mosaic, not one single language organ 668 00:30:19,690 --> 00:30:22,339 appearing overnight. So wrapping this up, the 669 00:30:22,339 --> 00:30:25,559 overall picture is complex integration. Definitely. 670 00:30:25,680 --> 00:30:28,700 It's an integrated mosaic. You have ancient components 671 00:30:28,700 --> 00:30:31,380 related to concepts and social intelligence shared 672 00:30:31,380 --> 00:30:33,799 widely with other animals. You have intermediate 673 00:30:33,799 --> 00:30:36,319 steps like changes in motor control, possibly 674 00:30:36,319 --> 00:30:39,200 linked to tool use, maybe... basic vocal control 675 00:30:39,200 --> 00:30:41,980 for display and then you have the more recent 676 00:30:41,980 --> 00:30:45,339 perhaps human specific innovations like the fine 677 00:30:45,339 --> 00:30:48,539 -tuned vocal learning circuits involving foxp2 678 00:30:48,539 --> 00:30:51,779 and potentially the full recursive power of syntax 679 00:30:51,779 --> 00:30:53,920 all layered on top of each other over millions 680 00:30:53,920 --> 00:30:57,200 of years Exactly. And molecular genetics is becoming 681 00:30:57,200 --> 00:31:01,519 incredibly powerful here. Genes like FOXP2 act 682 00:31:01,519 --> 00:31:04,500 like genetic fossils, as Fix calls them. They 683 00:31:04,500 --> 00:31:06,539 let us start putting dates and sequences on these 684 00:31:06,539 --> 00:31:08,740 events. We can see which changes likely came 685 00:31:08,740 --> 00:31:10,380 first, which selective pressures were acting 686 00:31:10,380 --> 00:31:12,960 when. It helps unravel the timeline in a way 687 00:31:12,960 --> 00:31:15,299 fossils alone can't. Right. For example, the 688 00:31:15,299 --> 00:31:18,240 evidence seems to lean towards anatomical potential 689 00:31:18,240 --> 00:31:20,500 for speech sounds, like the hypoglossal canal, 690 00:31:20,700 --> 00:31:23,299 maybe preceding the full refinement scene with 691 00:31:23,299 --> 00:31:26,420 FOXP2, which in turn might precede some aspects 692 00:31:26,420 --> 00:31:29,059 of complex grammar emerging culturally. Which 693 00:31:29,059 --> 00:31:31,200 leads us to that final, really thought -provoking 694 00:31:31,200 --> 00:31:33,779 tension you mentioned earlier. If the core biological 695 00:31:33,779 --> 00:31:36,460 capacity for language, the ability to acquire 696 00:31:36,460 --> 00:31:39,259 a structured system, is innate, how much of the 697 00:31:39,259 --> 00:31:42,140 actual language you or I speak is just... cultural 698 00:31:42,140 --> 00:31:44,279 happenstance. That's the big question, isn't 699 00:31:44,279 --> 00:31:46,339 it? How much is biology? How much is history 700 00:31:46,339 --> 00:31:48,940 and culture? Many specific features of languages, 701 00:31:49,059 --> 00:31:51,539 why English uses subject -verb -object order, 702 00:31:51,740 --> 00:31:54,680 why German has complex case markings, why French 703 00:31:54,680 --> 00:31:57,759 has grammatical gender. These might not be direct 704 00:31:57,759 --> 00:32:00,180 biological adaptations. They could be. What was 705 00:32:00,180 --> 00:32:03,460 the term? Grammaticalization. Yes, grammaticalization. 706 00:32:04,160 --> 00:32:07,599 They could be solutions that cultures have stumbled 707 00:32:07,599 --> 00:32:10,920 upon and standardized over time to solve the 708 00:32:10,920 --> 00:32:13,640 incredibly difficult problem of mapping complex 709 00:32:13,640 --> 00:32:16,480 multidimensional thoughts onto a simple linear 710 00:32:16,480 --> 00:32:19,039 string of sounds. They're culturally evolved 711 00:32:19,039 --> 00:32:21,940 conventions shaped by efficiency, learnability 712 00:32:21,940 --> 00:32:25,799 and historical accidents. Our brains are biologically 713 00:32:25,799 --> 00:32:28,819 prepared to find and use structure -dependent 714 00:32:28,819 --> 00:32:30,500 rules, maybe recursive rules. That's the innate 715 00:32:30,500 --> 00:32:33,019 part, the FLN. Arguably, yes. But the specific 716 00:32:33,019 --> 00:32:35,200 form those rules take in English versus Japanese 717 00:32:35,200 --> 00:32:38,359 versus Swahili, that's largely shaped by cultural 718 00:32:38,359 --> 00:32:40,720 transmission. That's a very prominent view. The 719 00:32:40,720 --> 00:32:42,819 biology provides the constraints, the potential, 720 00:32:43,059 --> 00:32:45,900 the learning biases, but the actual linguistic 721 00:32:45,900 --> 00:32:48,180 system is fleshed out and transmitted culturally. 722 00:32:48,750 --> 00:32:51,009 Chomsky himself talked about many surface features 723 00:32:51,009 --> 00:32:53,670 of language being down to accidental historical 724 00:32:53,670 --> 00:32:56,589 and cultural events. Which leaves us with a fascinating 725 00:32:56,589 --> 00:32:59,089 final thought for you, the listener, to mull 726 00:32:59,089 --> 00:33:01,349 over. Think about the language you speak every 727 00:33:01,349 --> 00:33:04,549 day, its specific word order, its quirky idioms, 728 00:33:04,569 --> 00:33:07,490 its grammatical rules. How much of that structure 729 00:33:07,490 --> 00:33:10,029 reflects a deep biological blueprint for human 730 00:33:10,029 --> 00:33:12,950 language, and how much is just the echo of countless 731 00:33:12,950 --> 00:33:15,569 historical accidents, innovations, and conventions 732 00:33:15,569 --> 00:33:17,630 passed down through generations of speakers? 733 00:33:18,029 --> 00:33:20,190 Where does biology end and cultural history begin? 734 00:33:20,410 --> 00:33:22,589 It's a blurry line, and exploring it is what 735 00:33:22,589 --> 00:33:25,269 makes this field so endlessly fascinating. Thanks 736 00:33:25,269 --> 00:33:27,190 for joining us on this deep dive into language 737 00:33:27,190 --> 00:33:27,450 evolution.