127 | Erich Jarvis on Language, Birds, and People

Episode Summary

Executive Summary: Sean Carroll and Eric Jarvis explore vocal learning—the rare ability to learn new sounds—as a foundation of spoken language in humans and a few animal lineages. They distinguish it from mere sound usage, discuss birds, mammals, genetics, brain circuitry, and evolution, and argue that language arose via duplicated motor-learning circuits, not a special language module, with broad implications for intelligence and comparative neuroscience.

Main Topics: What counts as vocal learning vs. ordinary communication (Priority: 5/5): Jarvis distinguishes vocal production learning (learning novel sounds) from vocal usage learning (using innate sounds in different contexts) and from auditory learning. The key trait of interest is the ability to imitate or modify acoustic structure, not just assign meaning to sounds. Species with vocal learning (Priority: 5/5): Vocal learning is rare and appears in humans and a small set of mammals (cetaceans, bats, pinnipeds, elephants) and birds (parrots, songbirds, hummingbirds). The list is strikingly non-obvious and suggests convergent evolution under similar pressures. Brain circuits and deep homology (Priority: 5/5): Jarvis argues vocal learning comes from a specialized forebrain motor-learning circuit, likely duplicated from existing motor pathways for hands or orofacial movement. The shared underlying motor-learning architecture is homologous, while the vocal-learning circuit is an independent analogous evolution. Genes, development, and comparative genomics (Priority: 4/5): The discussion covers genes such as SRGAP2 and FOXP2, plus the idea that many species share a common genetic toolkit but regulate it differently. Jarvis is also leading large genome-sequencing efforts to compare vertebrate species and map trait evolution. Language, syntax, intelligence, and social evolution (Priority: 4/5): Jarvis suggests spoken language is not a separate module but emerges from motor learning plus auditory processing. He connects vocal learning to social communication, intelligence, mate choice, and possibly abstract thought, while cautioning against overclaiming human uniqueness. Future experiments and broader implications (Priority: 3/5): The lab is testing whether human gene variants inserted into mice can alter vocal-learning circuitry, and whether large genome databases across vertebrates and all eukaryotes will reveal new evolutionary patterns and biomedical insights.

Key Arguments: Cats and dogs can communicate with sounds, but that is not the same as learning novel sounds; vocal production learning is the rare trait relevant to spoken language. Auditory learning is common across many animals, but it is not sufficient for language without the ability to produce learned sounds. Vocal learning evolved independently in a small number of mammal and bird lineages, implying strong selection pressures and/or strong constraints on viable solutions. The core neural substrate is a motor-learning circuit, repurposed for vocal control; this is a better explanation than a separate, dedicated language module. Human speech-related circuitry likely shares deep homology with broader vertebrate motor-learning systems, while the vocal-specific circuit is an analogous derivation. Genes such as FOXP2 and SRGAP2 illustrate that language-related traits may arise more from regulatory changes and developmental tuning than from brand-new genes. Parrots and songbirds show that learned vocal communication can be used socially and even recombined into new contexts, suggesting meaningful complexity beyond simple mimicry. Intelligence may be selected partly through vocal learning because flexible communication enhances social networks, mate attraction, and survival. Large-scale genome projects can transform the study of evolution by linking traits to conserved and divergent genomic features across the animal kingdom.

Data Points: Human vocabulary: 20,000 words - Jarvis compares Coco the gorilla’s learned vocabulary to typical human vocabulary size. Coco the gorilla comprehension: 2,000 to 7,000 words (reported range) - Discussion of claims about Coco’s understanding of human words. Coco production ability: about 1,000 to 2,000 signs; no spoken words - Used to highlight the gap between comprehension/signing and vocal production. Vocal-learning mammal orders: 5 out of roughly 30 mammal orders - Jarvis lists humans/primates, cetaceans, bats, pinnipeds, and elephants. Vocal-learning bird lineages: 3 bird lineages - Parrots, songbirds, and hummingbirds are identified as vocal learners. Total vertebrate orders: about 260 - He frames vocal learning as present in only a tiny fraction of vertebrate diversity. Vocal-learning vertebrate orders: 8 of 260 - Overall tally of vocal-learning orders among vertebrates. Gene overlap between birds and humans: over 80% of genes shared - Explains why birds and humans can have similar evolutionary toolkits despite looking very different. SRGAP2: extra copy in humans - Associated with keeping neurons in a more immature state and possibly prolonging learning capacity. FOXP2 differences: 2 nucleotide mutations / 2 amino acid differences - Presented as a small but potentially important human-chimp difference linked to speech circuits. Lab size: close to 30 people - Jarvis describes the combined size of his lab spaces and field research team. Genome consortium size: over 200 people in 80 countries - The Vertebrate Genomes Project collaboration as described in the interview.

Pivotal Quotes: "The ability to actually modulate the acoustic structure and the sequence of the sounds that you produce." — Eric Jarvis: Definition of vocal production learning, the core trait under discussion. "We find that the species that can produce learned sounds have a specialized four brain circuit that you don't find in the species that can't." — Sean Carroll (paraphrasing Jarvis's finding): Summarizes the central neurobiological claim about vocal learners. "I think we have a greater capacity for abstract thought. And the reason why is that I think thought is happening in our auditorial visual pathways as well as in our speech pathways." — Eric Jarvis: Closing discussion linking speech circuitry to inner speech and abstract cognition.

Implications: The conversation suggests language is an evolved, biology-grounded trait built from general motor-learning machinery, not a mysterious human-only module. This reframes research on speech, cognition, animal communication, and genome engineering, and points toward comparative genomics as a major tool for future discoveries.