Chapter 20: Language in the Brain

Core Idea

Broca’s area (language production) and Wernicke’s area (language understanding) look like a language organ, but chimpanzees, bonobos, and even macaques have the same areas with nearly identical wiring. In monkeys these areas have nothing to do with communication. Ape hoots and gestures match human laughing and crying, which come from an older, hardwired emotional-expression system, not from language. What is new in humans is a hardwired learning curriculum: proto-conversations, joint attention, and question asking. This curriculum is set up by changes in older structures such as the amygdala and brainstem. It repurposes the older mentalizing areas of the neocortex for language. There is no language organ, just as there is no flight organ in a bird.

Frameworks Introduced

• Two communication systems (Figure 20.3 shows a chimpanzee brain and a human brain side by side). Both have a dark-shaded emotional-expression system deep in the brainstem and amygdala, labeled hoots, squeals for the chimp and laughing, crying for the human. Only the human brain also has the shaded Broca’s and Wernicke’s areas in the neocortex, labeled Language system.

	Emotional-expression system	Language system
Who has it	All primates, going back to early mammals or earlier	Humans only
Where	Older structures such as the amygdala and brainstem	Neocortex, usually the left side
Origin	Mostly genetically hardwired	Learned, with a critical window in childhood
Raised in isolation	Monkeys still produce their normal gestures and calls	Never develops
Across groups	Chimps and bonobos share almost 90 percent of gestures; humans everywhere (even those born blind and deaf) cry, smile, laugh, and frown alike	Features differ greatly across cultures
Baby born without neocortex	Still expresses emotions	Never speaks

• The language curriculum:
  • When to use: to explain how a complex skill can be universal without being hard-coded.
  • How: evolution pairs a generic learning system (the cortex) with a hardwired curriculum. Baby birds are not born knowing how to fly. Instincts to jump, flap, and glide make nearly 100 percent of them learn it. For language, the curriculum has three parts:
    1. Proto-conversations: by four months, infants take turns with parents in vocalizations, expressions, and gestures, and match their mother’s pause durations. Chimpanzee infants show no such behavior.
    2. Joint attention: by nine months, infants look back and forth between an object and the parent to confirm they share attention. An infant keeps pointing until the parent alternates gaze between the object and the infant, and is often satisfied without getting the object. The more joint attention at one year, the larger the vocabulary twelve months later.
    3. Question asking: children ask “Want this?” and “Hungry?” before they can form grammatical sentences. All languages use the same rising intonation for yes/no questions. Kanzi, Washoe, and the other trained apes never asked even simple questions about another’s inner world.
  • Why it works: a skill like flying or language is too information-dense to fit in a genome, so it is cheaper to encode a curriculum. “The language curriculum requires both a teacher and a student.”
• Curriculum in AI:
  • Jeffrey Elman (UC San Diego, 1990s) trained a neural network to predict the next word. Showing it sentences of every complexity at once failed. Starting with very simple sentences and adding complexity only after it did well made it work.
  • TD-Gammon learned backgammon by playing against itself, so it always had an evenly matched opponent. Against a human expert it would never have won and so never learned. AlphaZero was also trained by self-play.
  • Lesson: “The curriculum used to train a model is as crucial as the model itself.”

Key Concepts

• Broca’s area and Broca’s aphasia: a left frontal region. Damage removes the ability to produce language in speech, writing, and sign alike.
• Wernicke’s area and Wernicke’s aphasia: a left posterior region. Damage removes understanding of spoken and written language. Patients produce fluent nonsense (“You know that smoodle pinkered…”).
• Modality independence: both areas handle language in general, not hearing, reading, or signing in particular. The same areas activate when a hearing person listens and when a deaf person watches signing.
• Larynx control (a red herring): the human motor cortex connects directly to the brainstem area for the larynx, one of few structural differences from apes. It did not enable language, because humans learn nonverbal language just as fluently.
• Dissociation: language ability can separate from general intelligence. Christopher (Neil Smith and Ianthi-Maria Tsimpli, 1995) could not button a shirt or win at tic-tac-toe but could read, write, and speak over fifteen languages.
• Joint attention vs gaze following: gaze following is a reflex to look where others look, which even turtles show. Joint attention deliberately confirms that both minds attend to the same object. Chimpanzees follow gaze but show no interest in joint attention.
• Repurposing: children with the whole left hemisphere removed still learn language using right-side areas. About 10 percent of people use the right side for language anyway. Newer studies place language areas all over the neocortex and even in the basal ganglia.

Mental Models

• Think of language like bird flight: a general learner plus inborn instincts that make the learner practice the right things.
• Change the curriculum, not the learner, when a model will not learn a skill. Evolution repeatedly chose this route for monkey climbing, bird flying, and human language.
• Think of an ape’s hoot as closer to a human laugh than to a human word. Jane Goodall noted that for chimps “the production of a sound in the absence of the appropriate emotional state seems to be an almost impossible task”.
• Asking where language lives in the brain “may be as silly as asking where playing baseball or playing guitar lives in the brain.” Complex skills come from many areas trained to work together.

Anti-patterns

• “Case closed” localization: the author sets up and then rejects the claim that humans evolved two new areas wired into a language subnetwork. These areas emerged in the first primates as part of the mentalizing neocortex.
• Treating ape gestures as the ancestor of words: comparing them is “comparing apples to oranges”. Damage to Broca’s and Wernicke’s areas in a monkey leaves its communication intact.
• Treating language as free with a bigger brain: specific dissociations (aphasia with normal intellect, Christopher’s gift) show that language is “a specific and independent skill that evolution wove into our brains.”
• Assuming apes lack the hardware: the ape neocortex is capable. Apes struggle because “they don’t have the required instincts to learn it.”

Rules of Thumb (author’s wording)

• “To teach a new skill, it is often easier to change the curriculum instead of changing the learning system.”
• “There is no language organ in the human brain, just as there is no flight organ in the bird brain.”
• “It isn’t Broca’s or Wernicke’s areas that are new, it is the underlying learning program that repurposes them for language that is new.”
• What is unique in the human brain “is hidden and subtle, tucked deep in older structures like the amygdala and brain stem.”

Worked Example

The teacher who could laugh but not smile on command (Trepel et al., 1996).

1. In the mid-1990s a teacher in his fifties developed slurred speech and facial paralysis over three days. Asked to smile, he produced a lopsided smirk.
2. When the doctor told a joke, he smiled normally on both sides. Figure 20.2 shows the two photos: Attempt to smile on command (left paralysis) and Automatic smile when told a joke (no paralysis).
3. A brainstem lesion had cut the connection from his neocortex to the muscles on the left side of his face. It spared the connection from his amygdala to the same muscles. He could not raise an eyebrow at will, but he could laugh, frown, and cry.
4. Conclusion: facial expression has two parallel controllers, an old reflexive emotional system and a newer voluntary neocortical one. The same split explains why people with severe aphasia who cannot say a word still laugh and cry, and why monkeys with damaged Broca’s and Wernicke’s areas still hoot.
5. Historical background: Broca found the production area from patient “Tan” (Louis Victor Leborgne, who lost speech in 1830 at age thirty). Broca autopsied twelve brains and published in 1865. Wernicke found the comprehension area a few years later.

Key Takeaways

1. Language uses specific neocortical areas, but those areas are not new. Other primates have them.
2. Human emotional expressions, not human words, correspond to ape calls.
3. The innovation is a hardwired curriculum of turn-taking, joint attention, and questions, built into old subcortical structures.
4. The curriculum repurposes primate mentalizing areas for language and needs a teacher as well as a student.
5. In AI and in evolution, the training curriculum matters as much as the learning system.
6. Apes can learn the basics of language but lack the instincts that make humans learn it easily.

Connects To

• Ch 5–9: TD-Gammon and self-play from breakthrough #2.
• Ch 15–18: Broca’s and Wernicke’s areas belong to the primate mentalizing regions. Joint attention needs theory of mind.
• Ch 19: the ape-language results that this chapter explains.
• Ch 21: the curriculum starts between parents and children, which kin selection can explain.
• Ch 22: Wernicke’s area overlaps the temporoparietal junction, a mentalizing area. Children tether words to their inner simulation, unlike LLMs.