The Nature of Language — Abduction, and How Side Effects Color the World

The Nature of Language (Mutsumi Imai & Kimi Akita, 2023) is a book that tries to answer the big question “how was language born and how did it evolve?” from both cognitive science and linguistics¹. It won the grand prize of the 2024 Shinsho Awards and was widely read. At its center sit two key concepts — onomatopoeia, which bridges body and sign, and the abduction (inference to a hypothesis) that crosses that bridge to raise up a linguistic system.

Here I read the book through the framework this site has used again and again — thinking in signs, différance, and Verification and Validation (V&V). To state the conclusion first: the book can be read as an empirical study showing that the universal form of intellectual activity can be observed, as is, at the very site of language acquisition.

The symbol grounding problem, and onomatopoeia as mediation

An abstract sign does not, on its own, ground in reality. The string of sounds “apple” has no necessary relation to the apple itself — Saussure’s arbitrariness. So how does a child tie the first sign to reality? This is the symbol grounding problem³.

The book’s answer is onomatopoeia. Signs like “korokoro” (rolling) or “fuwafuwa” (fluffy), where sound and meaning are joined through bodily sensation — these become the first bridge spanning between the body (mono, things) and the abstract sign-system (koto, relations/meaning). Rather than diving straight into the sea of fully arbitrary signs, you start crossing from a foothold grounded in the body.

This can be read as the language version of what this site calls mediation. We can grasp the world only through something. Onomatopoeia is the case where that “something we pass through” appears in its form closest to the body.

Abduction — the leap that cannot be grounded

What the book extracts as the engine of the bootstrapping cycle is abduction. C. S. Peirce divided inference into three — deduction (rule + cause → necessary result), induction (cases → rule), and abduction (result → best explanatory hypothesis)². The first two are operations inside a known framework and generate no new concept. Only abduction can introduce a new distinction.

A child leaps from limited cues to the meaning of a word. That leap is not logically justified — inference from result to cause is not necessary. The book argues that this very “unjustified leap” is the substance of language acquisition.

This is the same structure this site discussed in Why V&V Never Ends (Part II · Application). Verification can confirm a necessary condition, but “what is correct in the first place” does not close formally. Somewhere, you must leap without grounding and decide. Abduction is the name given to that leap as an inference.

Only an ungrounded leap generates a new distinction.

A misunderstanding is a side effect — and side effects are exactly what is useful

The leap often misses. A child over-applies a freshly learned rule and utters a conjugation no adult uses. This is an error. Yet what the book shows is that such errors evolve language — misuse settles in, rules recombine, and language changes across generations.

This is the point I resonated with most. The color of the world arises from errors like misunderstandings. And it has the same structure as a side effect in software. A side effect is an unplanned consequence. It is disliked as a bug — but a pure function with no side effects at all is safe and, in exchange, makes nothing happen in the outside world. Without side effects, nothing is useful.

Penicillin (the side effect of a mold), the Post-it (a repurposed failed adhesive) — novelty often arises from the residue of a plan. The error of abduction and the side effect of software are the same “incomplete inference / unplanned consequence,” and that is exactly the generative surplus that gives the world its color⁴.

Without side effects, nothing is useful.

In this site’s framework, this is the part where the iteration “does not close” — the residue. Not closing fully is not a defect but the seed of the next new différance. So we actively choose not to close. We design a system that turns errors into the next nourishment, rather than erasing them.

Between grounding and arbitrariness — moderation as design

Another achievement of the book is explaining why language, while starting from “body-grounded onomatopoeia,” could take off toward an “arbitrary, abstract sign-system.” Ground it too much and you cannot generalize; make it too arbitrary and you lose the foothold for learning. Language stands up at the moderate point between grounding and arbitrariness.

This is the same structure as the Whiteheadian “moderate abstraction” discussed in Scale as a Degree of Freedom (Part III · Scale). Too short or too long, and no order is visible; only at the appropriate magnification does a meaningful structure stand up — language too, one can read, crystallizes at a moderate magnification between body and abstraction.

The Nature of Language can be read as a theory of the origin of words and, at the same time, as a self-portrait of intellectual activity itself. Take a difference (articulate the world with onomatopoeia), leap to a hypothesis (abduction), check, err, and turn that error (the side effect) into the next nourishment, and repeat — this is exactly the universal form of a being that thinks in signs, the form this site stated in Relative, Not Deterministic.

Without absolute ground, words have kept updating themselves through the leap of abduction. And it is the errors of that leap that have given language — and thus the world — its color. The book, one could say, is a volume that demonstrated this from the mouths of children.

References