Co-Evolution of Language and the Brain

In the AI Story

The empirical foundation is comparative neuroanatomy. Human brains differ from chimpanzee brains not proportionally but selectively: specific regions enlarged far beyond what overall brain size would predict. The prefrontal cortex—supporting the suppression of immediate responses in favor of rule-governed behavior—is disproportionately large, enabling the inhibition required to use arbitrary symbolic conventions rather than indexical pointing. Broca's area, Wernicke's area, and the arcuate fasciculus connecting them are elaborated beyond expectations, supporting the production, comprehension, and integration of symbolic sequences. Regions governing tongue, lip, and larynx movements show refinements enabling the phonemic precision that symbolic language requires.

The Baldwin Effect provides the mechanism by which learned behaviors become progressively innate over evolutionary time. Not through Lamarckian inheritance of acquired characteristics, but through natural selection favoring individuals who can learn the behavior more easily, with less effort, less dependence on environmental scaffolding. Applied to language: early symbolic communicators expended enormous cognitive effort maintaining conventions and computing syntax; individuals whose brains were slightly better suited to these demands had reproductive advantages; across generations, the neural substrate reorganized to make symbolic processing progressively easier. What once required every available cognitive resource became, for modern humans, so automatic that children acquire language without conscious effort.

The co-evolutionary dynamic explains not only the presence of language-specific neural structures but their absence in other primates. Chimpanzees possess the vocal apparatus and the general intelligence to support symbolic communication in principle, yet they do not develop it in the wild. The explanation: without the initial spiral—the proto-linguistic communication that would create selection pressure for the neural reorganizations—the architecture never forms. The co-evolution requires both elements present simultaneously; remove either, and the spiral does not start.

The implications for the AI moment: if a cognitive technology can reshape the brain that uses it, then the question of what AI will do to human cognition is not speculative but structural. The reshaping may not occur at the genetic level—timescales far too short—but at the cultural level, where cognitive habits and skills are transmitted across generations through education and institutional practice. The Berkeley study documenting cognitive reorganization after months of AI use provides the first empirical glimpse of a co-evolutionary process whose full trajectory will take decades to become visible.

Origin

The co-evolutionary framework emerged from Deacon's dissatisfaction with every existing account of language origins. Chomsky's nativism posited a language organ whose evolution was unexplained; generalist theories attributed language to overall intelligence without accounting for the specific neural reorganizations. Deacon's synthesis: language is neither a discrete module nor a general byproduct, but a co-evolutionary partner that built the very brain structures it requires.

The concept of co-evolution itself was well-established in evolutionary biology—the mutual shaping of flowers and pollinators, predators and prey, hosts and parasites. Deacon's innovation was recognizing that the same dynamic could operate between a cultural practice (symbolic communication) and a biological organ (the brain), producing biological changes on evolutionary timescales in response to cultural selection pressures.

Key Ideas

Reciprocal selection pressure. Language created advantages for certain brain features; those features enabled more complex language; the spiral was self-reinforcing across thousands of generations.

No clean temporal priority. Neither language nor the language-capable brain came first—they co-emerged, each constituting the environment in which the other evolved.

Neuroanatomical footprint. The specific, disproportionate enlargements in the human brain—prefrontal, perisylvian, vocal-motor—are the visible traces of language's selection pressure.

Cultural-biological feedback. Cultural practices (symbolic communication) shaped biological evolution (brain structure), which shaped cultural capacity (more complex language), in a feedback loop operating across evolutionary time.

Precedent for AI co-evolution. If symbolic language reshaped the brain across millennia, then AI—operating at the same level of symbolic processing but on cultural rather than biological timescales—carries the potential to reshape cognitive habits, skills, and the developmental environment of the next generation.