The Apparatus · a banned question, replaced with checkable ones

The Question They Banned

On 8 March 1866 the Linguistic Society of Paris put a sentence in its bylaws: it would hear no paper on the origin of language. The rule was real. The legend that grew around it — a century of enforced silence — is not. Here is what the ban's descendants did instead: they stopped trying to answer the unanswerable and started measuring the parts that leave a trace.

You cannot rewind to the first word. Language does not fossilise; there is no layer of rock with a sentence in it. So the honest thing is not to narrate the first sentence — it is to ask what still runs today that must have run then, and to check it. Three of those checks are on this page. You operate them. Before any of them, one distinction has to be nailed down, because the whole field turns on it.

Every card and every instrument below wears one of these two tags. Watch which one it is. The grand narratives — language began as song, as gesture, as grooming, as a mother's lullaby — are almost all in the amber register. That is not an insult. It is the reason the question was banned, and the reason it came back.

1 · The sentence in the bylaws

The Société de Linguistique de Paris, founded in 1864 and given official sanction in 1866, wrote a fence directly into its constitution. Article 2, in full:

Read it twice, because two things are banned, not one. The origin of language — and the construction of a universal language, the Esperanto-shaped dream of a rational tongue. Both were barred for the same reason: they filled the meeting hall with confident papers that nobody could ever check. A society that wanted to be a science drew a line around the parts of its subject that could only be argued, never settled.

That is the honest reading of the ban, and we will come back at the end to correct the version you have probably heard — the one where a cabal of linguists silenced the deepest question about ourselves for a hundred years. (They did not. Five years later Darwin published a whole theory of language origin, and nobody arrested him.) But first: why was the question uncheckable, and what, exactly, did the field keep once it threw the uncheckable part away?

2 · The song spine — six stories, all in amber

The most beautiful family of origin stories says language began as music: that before we had words we had melody, and meaning rode in on the tune. It is a lovely idea with a long pedigree. It is also, every version of it, a hypothesis — a claim about an event that left no evidence, so no observation could ever prove it wrong. Here is the lineage. Read each as a candle, not a measurement.

Six candles. The skeptic who says this is all just-so storytelling is, about these grand arcs, largely right — and that skeptic is the intellectual descendant of the 1866 committee. If the page stopped here, the ban would have won and it would deserve to.

It does not stop here. Because there is a second move, and it is the whole point: the banned question was never answered. It was replaced. You cannot observe the first word — but you can observe, right now, whether the pieces the song stories assume are even real. Does melody carry meaning before words do? Is sound-to-meaning ever non-arbitrary? Does shared rhythm actually bind bodies? Those are not stories. They are experiments, and you can run three of them.

3 · Does a sound have a shape? Experiment

In 1929 the Gestalt psychologist Wolfgang Köhler showed people two shapes — one rounded, one jagged — and two invented words, takete and baluma (later maluma), and asked which was which. Almost everyone agreed. Ramachandran & Hubbard revived it in 2001 with bouba and kiki and reported that 95–98% of English and Tamil speakers give the same answer. If a sound can carry a hint of a shape, then meaning is not purely arbitrary — which is a direct, checkable dent in Saussure's doctrine that the sign is arbitrary.

Do not take our word or theirs. Take the test on your reflex first. Six trials. There is no right answer shown until you finish.

Whatever you got, the cross-cultural picture is now solid. Ćwiek and colleagues (2022) ran the test on 917 participants speaking 25 languages from nine families across ten writing systems. Overall, about 72% chose the "congruent" pairing (a 95% credible interval of roughly 56–82%). Roman-script speakers landed near 75%, non-Roman near 63% — and crucially, both sit far above the 50% a coin would give:

So this leg is genuinely replicated — the strongest of the three. But keep it honest: 72% is not universal. The effect is weaker in some writing systems, differs for people blind from birth, and most of any language's vocabulary really is arbitrary, exactly as Saussure said. What bouba/kiki shows is narrow and real: sound-meaning mapping is not zero. A crack in the arbitrariness of the sign — not the engine of every word.

4 · A cry with an accent Experiment

Here is the eeriest checkable fact in the whole field. Mampe and colleagues (2009) recorded the cries of 60 newborns — 30 from French-speaking homes, 30 from German — all just a few days old, all long before a first word or even a babble. The cries had an accent. French babies cried with a rising melody; German babies with a falling one, mirroring the prosody of the language they had been hearing through the womb wall. The average German cry peaked in pitch and loudness early, around 0.45 s into the cry; the French cry peaked later, around 0.6 s.

Below is a synthesized illustration of those two melodic shapes — not the study's audio. The contour is a model, pinned to the reported peak-timing marks; the playhead and the tone follow it exactly. Play each and watch where the peak lands.

What this does show: prosody — the melody of speech — is learned before words, in the womb. That is a real, if modest, foothold for the song stories: the musical channel of language is up and running before the lexical one exists. What it does not show: that "language began as song." It is one study, n=60, and it drew methodological pushback (Language Log among others). Read it as pre-lexical prosodic learning is real — and stop there. The leap from "babies learn melody first" to "melody came first in evolution" is the amber move again.

5 · When rhythm gets into the body Experiment

The last assumption the song stories lean on is that shared rhythm bonds people. That one is measurable too — and it is where the press most loves to overclaim. Vickhoff and colleagues (2013) wired up a choir of 18-year-old singers (11 in the group analysis, of 15 recorded; 5 more in case studies) and found their heart-rate variability tracked the music. The headline became "choir members' hearts beat as one." That is not what happened.

What happened has a precise name: respiratory sinus arrhythmia (RSA). Your heart speeds up when you breathe in and slows when you breathe out. Singing is guided breathing — the phrase length dictates the breath. So a hymn with 2-, 4- and 8-bar phrases forces breathing cycles of 0.2, 0.1 and 0.05 Hz, and the hearts' variability tracks those frequencies. The hearts are not mystically synchronising. They are following the same breath, which is following the same song. Below, set a breathing pace and watch a simulated heart couple to it.

So the bonding claim, made precise, survives: singing together really does couple people's physiology — through a shared, entrained breath, at frequencies the music sets. The 2020 follow-up found the same when non-experts simply vocalise together, not only trained choirs. Real coupling; no magic. And notice what carried it: not the words, but the rhythm and breath — exactly the pre-verbal layer the song stories care about, here made an experiment instead of a myth.

6 · The myth the ban became

Now the promised correction — the one that keeps this page from repeating a myth while posing as a myth-corrector. You have almost certainly heard the 1866 ban told as a century of enforced silence: linguists so spooked by speculation that they outlawed the deepest question about humanity, and nobody dared touch it until the 1990s. That telling is overstated, and the dates give it away.

The ban was a house rule of one society in Paris, not a law, not a global taboo. It was written in a specific political moment — the young Société de Linguistique defining itself as a rigorous science against the more speculative Société d'Anthropologie of Paul Broca next door — and it was aimed squarely at the flood of unfalsifiable origin papers and rational-universal-language schemes, the two things Article 2 names. And inquiry simply continued. Darwin's Descent of Man, with its full musical-origin theory, appeared in 1871, five years later. Jespersen was writing on origins into the 1920s. The question was out of fashion in some quarters, yes; silenced for a century, no. Historians of linguistics have made exactly this correction — the ban is real, its legend is inflated.

Which leaves the cleanest possible summary of the whole affair. The 1866 committee was right about the unfalsifiable part: the grand origin narratives could not then, and cannot now, be tested against the vanished event. And the field was right to come back — not by defying the ban and answering the old question, but by subtracting it and keeping what a bylaw could never forbid, because it leaves a trace you can measure this afternoon. The banned question was replaced. What replaced it is on this page, and you just ran it.