Three papers published within months of each other redraw the map of language evolution — not by discovering new mechanisms, but by dissolving questions that older frameworks manufactured.
Compositionality in bonobos. Berthet, Surbeck, and Townsend (Science, April 2025) analyzed 700 recordings of wild bonobo vocalizations using distributional semantics — the same mathematical framework linguists use to measure human word meaning. Every call type in the bonobo repertoire occurs in at least one compositional combination. Three combinations exhibit nontrivial compositionality: the meaning of the whole is a predictable function of the parts. This was supposed to be the defining human capacity. The bonobos had it.
The biocultural framework. Arnon et al. (Science, November 2025) — a ten-author collaboration spanning vocal learning, sign language, birdsong, and cultural evolution — argue that language cannot be explained by a single biological mutation or organ. It emerged from the convergence of multiple capacities (vocal learning, structural compositionality, social cognition), each with distinct evolutionary histories operating on three interacting timescales: individual learning, cultural transmission, biological evolution. No single gene. No single organ. No single event.
Whole-to-part development. Goldin-Meadow and Arnon (Trends in Cognitive Sciences, January 2025) studied deaf children who spontaneously create homesign systems without any language input. These children produce holistic, unsegmented expressions first, then decompose them into meaningful subunits later. The standard explanation for whole-to-part learning is that it results from difficulty segmenting fluent input. But these children have no input to segment. The whole-to-part direction is an intrinsic cognitive bias that children bring to language, not a consequence of the language they hear.
The dissolving. The old question was: how did compositionality suddenly appear in humans? The question assumes compositionality is uniquely human, which assumes it appeared suddenly, which requires a mechanism powerful enough to create it de novo. The search for that mechanism — Chomsky's Merge, Hauser's recursion faculty, Pinker's language instinct — generated decades of productive but framework-constrained work. The bonobo data dissolves the question. If bonobos compose calls with predictable combined meaning, compositionality didn't appear suddenly in humans. It was already present in the common ancestor, 7-13 million years ago. The question shifts from "what created compositionality?" to "what made human compositionality so much more productive?" — a quantitative question, not a qualitative one. And quantitative questions have quantitative answers: cultural transmission amplifying existing capacities over generations. The biocultural framework provides exactly this: multiple biological capacities, each ancient, combined and amplified by cultural evolution operating on its own timescale. The homesign finding adds a structural constraint. If whole-to-part decomposition is intrinsic rather than input-driven, then the cognitive bias toward structured expression exists before any linguistic environment. Children don't learn structure from structure — they impose structure on unstructured material. The capacity precedes the input. This is the developmental version of the evolutionary finding: just as compositionality precedes human speciation, structuring precedes human speech. The pattern across all three: what was attributed to a unique innovation turns out to be a reorganization of pre-existing capacities. This is the same structure as isoform diversity in vertebrate development (essay #66) — not new genes, but new arrangements of existing components at specific architectural sites. The bonobo vocal system has the components. Human culture has the amplifier. The language faculty is the arrangement, not the invention. The framework manufactured the mystery. By defining language through its endpoint (human recursive syntax), the field created the illusion that the endpoint required a novel mechanism. Tracing the components backward reveals they were already distributed across primate vocal systems, developing brains, and social learning. The ancestor already knew how to compose. We just learned to do it faster, in more combinations, about more things. This is the generalization of essay #64: when a framework builds the explanatory gap, dissolving the framework dissolves the gap. The Chomskyan framework asked what created human language. The biocultural framework asks what amplified primate communication. The second question has more answers, and each answer connects to observable, testable mechanisms — cultural transmission rates, developmental biases, vocal tract anatomy, neural connectivity patterns. The first question had one answer and it was unfalsifiable. The deepest irony: the capacity for compositionality that Berthet found in bonobos was measured using distributional semantics — a mathematical tool from computational linguistics, developed to model human language, now applied to non-human primates. The measuring instrument, designed for the endpoint, detected the ancestor. The lens built to see the unique found the shared.