Dr. Erich Jarvis on the Neuroscience of Speech, Language, Song and Dance

High-Level Overview

Andrew Huberman sits down with Dr. Erich Jarvis, a professor at Rockefeller University and Howard Hughes Medical Institute investigator who studies the neurobiology and genetics of vocal learning. Using songbirds, parrots, and hummingbirds as models alongside humans, Jarvis explains why the traditional split between speech and language may not match how the brain actually works, why only a handful of species can imitate sounds, and why those same species are the only ones that can dance to a beat. The conversation ranges from critical periods and learning multiple languages to stuttering, reading and writing, texting, brain-computer interfaces, and an ambitious effort to sequence the genomes of every vertebrate on Earth.

The tone is warm, curious, and often delighted. Huberman is openly enchanted by the material, and Jarvis weaves in his personal story: a family of singers, a youth spent competing in dance contests, training at the Alvin Ailey school, and the choice to pursue science because he believed it could have a bigger positive impact. The conversation flows from basic brain circuitry to evolution, then to practical questions about learning, communication, and keeping the brain healthy, before closing on genomics and conservation.

The Discussion Explores:

• Why Jarvis doubts a separate language module and sees language built into speech production and auditory circuits
• Innate versus learned vocalizations, and how rare true vocal learning is in the animal kingdom
• Striking parallels between songbird and human brains, down to shared genes and mutations like FOXP2
• Critical periods, why children learn languages more easily, and why multilingual people pick up new languages faster
• The motor theory of vocal learning: speech evolved from movement circuits, which is why only vocal learners dance
• How reading and writing recruit speech circuits, and why writing speed should match your inner speech
• Stuttering, turn-taking in conversation, texting, and brain-computer interfaces that decode speech
• Vertebrate genome projects, the Genome Ark, and using genetic data for conservation

Deep Dive into Main Ideas

Speech and Language Are Not Separate Modules

Huberman opens by asking how to distinguish speech from language. Jarvis says he struggled with this for years before realizing the terms come from psychology and don't align neatly with brain function. The common view holds that a separate language module computes the rules and then instructs the speech production and auditory pathways. "And it turns out I don't think there is any good evidence for a separate language module", he says. Instead, the speech production pathway that controls the larynx and jaw has the complex algorithms for spoken language built into it, and the auditory pathway has the algorithms for understanding speech.

A key point is that these two pathways are not equally rare. The speech production pathway is specialized to humans, parrots, and songbirds, while auditory perception is widespread. That is why, as Jarvis puts it, dogs can understand several hundred human words, and great apes can be taught thousands, yet none of them can say a word. He also notes that brain regions for hand gesturing sit right next to those for speech, which is why people gesture while talking even on the phone. Koko the gorilla learned sign language and could understand speech but could not produce it, because non-human primates have learned-gesture circuits but lack the extra pathway for learned sound.

Innate Sounds, Learned Sounds, and the Origins of Language

Huberman wonders whether primitive emotional sounds, like inhaling with pleasure or exhaling in disgust, are the building blocks of language. Jarvis says the idea isn't crazy and points to the central distinction in his field: most vertebrates vocalize, but with innate sounds they are born able to make, like a baby's cry or a dog's bark. Only a few species can learn to imitate sounds, and that is what makes spoken language special. Innate sounds are handled largely by brainstem circuits, while learned behaviors like speaking or playing piano use forebrain circuits. In humans and parrots, the forebrain has taken over the brainstem vocal circuits so they can produce learned sounds.

On when language evolved, Jarvis cautions that humans tend to overrate themselves. Genomic data show our ancestors hybridized with Neanderthals and Denisovans, and he knows of no living vocal learner that can have offspring with a non-vocal-learning species. Those ancient hominids also share the human versions of genes that function in speech circuits. "So I think Neanderthals had spoken language", he says, estimating that the ability has existed for somewhere between 500,000 and a million years and has grown more advanced culturally and possibly genetically.

Songbirds and Humans: Convergent Brains and Genes

Jarvis explains that only three of the roughly 40-something bird orders, songbirds, parrots, and hummingbirds, can imitate sounds. Beginning with researchers like Peter Marler and Fernando Nottebohm, scientists found a package of parallels: these birds have critical periods for learning, and like humans (and unlike chickens or non-human primates), their vocalizations deteriorate if they become deaf. Their brains have specialized song pathways missing in non-learners, and Jarvis's own work showed these pathways have functional parallels to human speech areas such as Broca's area and the laryngeal motor cortex. More recently, his team found that the genes specially expressed in these regions, and even specific mutations, are similar across humans, songbirds, and parrots despite roughly 300 million years of separation. Mutations in genes like FOXP2 that cause speech deficits in humans produce similar deficits in the birds.

The genes fall into three groups. The first controls axon guidance, the wiring of connections. Surprisingly, several of these genes were turned off in speech circuits; they normally repel connections, so silencing them allows a direct connection from the cortex to the motor neurons controlling the larynx or syrinx. "So by turning it off, you got to gain a function for speech", Jarvis explains. The second group involves calcium buffering and neural protection, such as parvalbumin and heat shock proteins. Because the larynx is the fastest-firing muscle in the body, the neurons controlling it fire at high rates and need protection from toxicity. The third group supports neuroplasticity, since learning to speak is a more complex learning task than learning to walk. Along the way, Jarvis shares a delightful detail: some hummingbirds snap their wings together in unison with their song, making a sound that blends in with their voice.

Critical Periods and Learning Multiple Languages

Marler found that young songbirds learn their own species' song best, but can learn another species' song imperfectly, a principle he called the innate predisposition to learn, which Jarvis likens to universal grammar in linguistics. A zebra finch raised with a canary sings a hybrid song, and given a choice between tutors, it prefers its own species, though it will learn a related species well if its own is removed. Social bonding with one's own kind matters. Jarvis applies similar logic to human children who grow up between two languages: they merge phonemes and words during their critical period in a way adults can't.

Huberman, who has struggled to learn Spanish as an adult, asks whether early learning helps and whether multilingual people learn new languages more easily. Jarvis says yes to both. The whole brain goes through critical periods, which is why piano or bike riding are easier to learn as a child, but speech has a stronger one. His explanation is that "the brain can only hold so much information", so after rapid early learning, circuits stabilize for survival and efficiency. Humans remain more juvenile-like than other animals, thanks in part to an extra copy of a gene called SRGAP2. Children are born able to produce a wide set of phonemes and prune it to those their language uses. People who keep more phonemes active across several languages have an easier time with a third or fourth, not because their brains are more plastic but because they retain the ability to produce more sounds.

Emotion, Song, and the Motor Theory of Dance

Jarvis distinguishes semantic communication, which carries meaning, from affective communication, which carries emotion. Although emotional centers such as the hypothalamus and cingulate cortex add tone, he believes both kinds of communication use the same speech or song circuits: "It's the same brain circuits." Speech is more left-hemisphere dominant, while singing and music processing are more balanced across both sides. Because all vocal learners use learned sounds for emotional communication like courtship, but only a few use them for semantic meaning, many researchers think spoken language first evolved for singing and only later for abstract communication. This also helps explain why singing or music can help Parkinson's patients move and speak more easily.

That leads to Jarvis's own story. He came from a family of singers, discovered he could dance as a teenager, attended the High School of Performing Arts, trained at the Alvin Ailey and Joffrey ballet schools, and turned down a path toward the Ailey company to study science at Hunter College and then Rockefeller. Years later, research by Aniruddh Patel and others showed that only vocal learning species can learn how to dance, like Snowball the cockatoo. Jarvis's lab found that vocal learning pathways are embedded within circuits that control learned movement, leading to the motor theory of vocal learning origin: speech circuits evolved through a duplication of surrounding motor circuits. The tight auditory-motor integration needed for speech spilled over to neighboring regions, allowing humans and parrots to synchronize body movement to sound. As Huberman sums it up: "Well, so we're speaking with our bodies when we dance." Jarvis adds that dance is mostly affective, used for emotional bonding, though ballet can tell stories. He mentions a new neurobiology-of-dance field in which EEG studies suggest brain activity may synchronize between dance partners and audiences, while stressing the findings aren't ready for prime time. He also notes his 23andMe results show variants associated with fast-twitch muscle and with difficulty singing on pitch.

Faces, Reading, Writing, and Inner Speech

Non-human primates have diverse facial expressions with strong cortical control of facial muscles, but weak control of the voice. Humans added learned vocalizations on top of an older system of facial communication. Jarvis says facial expressions have both innate and learned components, which is why it's hard to speak without gesturing or to say something politely while your face conveys anger. Facial expressions remove the ambiguity that plagues email.

On reading, Jarvis describes at least three pathways working together: visual signals travel to speech motor areas like Broca's, where you silently speak the words, sometimes with measurable activity in your laryngeal muscles, and then to auditory areas so you hear them in your head. Writing adds a fourth, as hand areas next to speech circuits translate that inner speech onto the page. Huberman shares that he must stop talking to write on a board, and that he prefers writing by hand with Pilot V5 and V7 pens because they match his thinking speed. Jarvis's view is that writing goes smoothly only when your hand or typing keeps up with the speech in your head. Huberman summarizes that the link between thought and writing is speech, and Jarvis agrees: "That bridge is speech." Huberman also mentions warming up for long solo episodes by silently singing song lyrics in his head, which Jarvis says really does exercise speech circuits and even the vocal muscles at low levels.

Stuttering, Turn-Taking, and Texting

Jarvis's lab stumbled onto stuttering in songbirds. When part of the basal ganglia (the striatum) involved in the song pathway was damaged, the birds began to stutter as the region recovered, driven by newly generated neurons integrating imperfectly. Unlike humans, the birds largely recovered within three or four months because bird brains produce new neurons. In humans, damage or disruption to the basal ganglia at a young age can cause stuttering, and developmental stuttering often involves this region too. Jarvis says behavioral therapies such as speaking more slowly or tapping out a rhythm can help, and that they share a common thread of sensory-motor integration.

When Huberman describes people who say the last word of his sentences along with him, Jarvis offers several explanations: the listener's speech circuits may be predicting the words, it may be a social bonding signal, or it may be a bid to take a turn. On texting, Jarvis is optimistic, framing the brain as use it or lose it: "I think texting is not decreasing the speech prowess or the intellectual prowess of speech." It is being converted and used differently, and the thumb circuits are getting a workout. The real casualty, he says, is interpretation, since short messages can be over- or under-interpreted, though quick messages may also reveal what someone truly feels. Huberman raises Eddie Chang's work at UCSF decoding speech from the brain activity of paralyzed people, and Jarvis says it supports his view that there is no separate language module, while noting the ethical questions if inner speech can be read and transmitted.

Movement Keeps the Brain Sharp

Asked for practical advice, Jarvis says that although science separates perception from production, he has found that dancing helps him think and keeps his brain fresh, because controlling the whole body takes a lot of brain tissue near the speech circuits. "And so I argue, if you want to stay cognitively intact until your old age, you better be moving." He recommends consistent movement, whether dancing, walking, or running, plus practicing oratory or singing to engage the circuits that move the face. He is clear this comes from personal experience. Huberman adds that long, slow runs help him prepare for hours-long solo recordings.

Genomes, Conservation, and the Genome Ark

Jarvis explains that nature has run natural experiments: vocal learning has evolved several times, with similar genetic changes. To find which changes relate to vocal learning rather than traits like flight, you need comparative genomics across species and a good family tree, which requires high-quality genomes. That led him to help lead the Vertebrate Genomes Project, aiming to sequence all roughly 70,000 vertebrate species, and to participate in the Earth BioGenome Project for all eukaryotes. He discovered that older genome assemblies were full of errors, including false gene duplications and missing regions. Working with sequencing companies and computer scientists helped produce the first complete telomere-to-telomere human genome, an effort led by Adam Phillippy. In the previously missing dark matter of the genome, his team is finding regulatory regions specialized in vocal learners that may help build speech circuits.

He also feels a responsibility to endangered species: "I feel like it's a moral duty." His team created a database called the Genome Ark to store complete genomes, and conservation groups such as Revive and Restore, which hopes to bring back the passenger pigeon, and Colossal, which aims to revive the woolly mammoth, now come to him for high-quality genomes. He closes with skin color as an example of convergent evolution: darker and lighter skin evolved independently several times in humans, depending on sunlight and vitamin D, often via the same melanin-related genes, and similar patterns appear in animals near and far from the equator.

Conclusion

Dr. Erich Jarvis offers a unifying view of human communication: language isn't a standalone module but lives in the circuits that produce and perceive speech, which themselves evolved from the brain's movement circuits. That single idea ties together why only vocal learners like parrots can dance, why we gesture while talking, why we silently speak as we read, and why singing may have come before speech. Along the way, listeners get practical takeaways: early exposure to multiple languages makes later learning easier, writing flows best when it keeps pace with your inner voice, stuttering has identifiable roots in the basal ganglia, and staying physically active may help keep the mind sharp. Jarvis's work in genomics shows how studying songbirds can illuminate human biology while also preserving the genetic codes of species we risk losing. For anyone curious about language, music, or the brain, the full conversation is a rewarding listen.

This is a summary of "The Neuroscience of Speech, Language & Music | Dr. Erich Jarvis" from Huberman Lab, giving you the GIST so you can decide if you want the REST. Listen to the full episode.