Sit for a moment with where you are. The light from a window or a screen. The wall behind your shoulders. The direction your knees are pointing. None of this requires effort. You know where you are without thinking, and you know it in the language of here, there, to your left, behind you. The room is organized around your body. Whatever room you are in, you are at its centre.

This sentence will end soon, and you will go back to whatever brought you here. We notice only that this is one way for a body to be in a room. There are others.

In Hopevale, on the north Queensland coast, there is a language called Guugu Yimithirr. Captain Cook’s crew first recorded one of its words — gangurru, the animal we now call kangaroo — without understanding what kind of language they had encountered. It would take two centuries before anyone really did.

In the 1970s the linguist John Haviland began listening carefully to how Guugu Yimithirr speakers talked about space. He noticed, slowly, that something was missing. There were no words for left and right of the kind we use. There was no in front of meaning in front of you. Speakers oriented themselves and everything around them by four absolute terms — gungga (north), jiba (south), naga (east), guwa (west) — and they used these terms continuously and at every scale. The cup is on the western edge of the table. The ant is climbing your northern leg. There is a spider on your southwestern shoulder.

This wasn’t ceremonial speech or special navigation talk. This was how a person asked another to pass the salt.

What this required of a speaker was constant, ambient knowledge of which way was which. Not knowledge to be retrieved — knowledge already present, like the awareness most of us have of which way is up. Speakers tracked cardinal direction the way we track gravity. To not know it was to not be able to speak.

Haviland and the linguist Stephen Levinson and their colleagues began running experiments. They placed speakers in front of an array of objects, walked them into another room, rotated them 180 degrees, and asked them to reproduce what they had seen. Speakers of English and other relative-frame languages — those organized around the body’s left and right — reproduced the array based on body coordinates, so the rotation reversed it. Guugu Yimithirr speakers reproduced it based on absolute coordinates, so the rotation preserved it. The two groups were not making the same kind of memory. They were not perceiving the same scene.

A senior speaker recorded a fishing story decades after the events, in a house far from the bay where the events had happened. As he gestured to show how a turtle had moved, his hands traced the cardinal directions of the original bay — not the room he was sitting in. His body was still oriented to a place he could not see. The gesture had outlived the geography.

We pause here. Because this is the moment in this piece where most of us become aware of our own inability. Sit again with where you are. Without checking — phone, sun, intuition — point north. Most of us cannot. Some can guess. A few will be roughly correct. Almost none of us hold north the way Guugu Yimithirr speakers held it: as a feature of the room, as obvious as the ceiling.

This is not a difference of vocabulary. It is a difference of organ. Something they had developed, we never grew. Or — and this is the harder thought — something the species once held in many scattered places, that some communities preserved and others let atrophy without anyone marking the moment of loss.

Guugu Yimithirr is one of perhaps a few hundred languages organized around absolute frames of reference. Tzeltal, spoken in highland Chiapas, uses uphill and downhill as primary spatial coordinates, anchored to the slope of the local terrain. Marshallese sailors carried mental charts of swell-patterns refracting around invisible islands, encoded in stick-charts of palm rib and shell whose function was not to be brought into the canoe but to be learned and then left behind. Polynesian wayfinders memorized the rising and setting points of dozens of stars and held the entire celestial sphere as a working instrument, the night sky a kind of dashboard.

Each of these is a different geometry of attention. Each required a body trained from childhood to perceive a particular thing — slope, swell, star-path, cardinal — as a feature of the world rather than a piece of specialized knowledge.

We could continue. The Inuit categories of sea ice that distinguish forms English has no words for, because acting on the wrong distinction kills you. The Andean potato-keepers who maintain hundreds of varieties through ceremonial obligations that bind specific tubers to specific rituals, which is to say: who keep the diversity alive by needing it. The instrument-makers, the seed-savers, the diagnosticians who know by the colour of a cheek what a panel of bloodwork will later confirm. None of this lives in a manual. Most of it never has.

About forty percent of the world’s roughly seven thousand languages are projected to fall silent before this century ends. Roughly three-quarters of agricultural genetic diversity disappeared in the last hundred years. The numbers vary by who is counting and how, but the shape is consistent across every domain in which someone has bothered to count.

We have been thinking about language as if it carried content. What if it carried something else.

Content travels. A fact about kangaroos can be translated from Guugu Yimithirr into English without much loss. The word changes; the kangaroo does not. Most of what we mean by information is content in this sense — portable, transmissible, indifferent to its vehicle.

A carrier is different. A carrier is the structure within which content can appear at all. It shapes what is sayable, what is noticeable, what counts as a question. Guugu Yimithirr did not contain more information about cardinal direction than English does. It contained a body trained to know cardinal direction at every moment. The carrier was an organ of perception. Lose the carrier and the perception goes with it, even if every sentence ever spoken in the language is recorded somewhere in some archive.

This is the part most easily missed. The archive preserves content. The carrier was never the content. We have built civilizations of archives and called the result preservation, and we have been preserving the smaller half.

The same shape runs through every domain where tacit knowledge lives. The blueprints of the Saturn V are in the National Archives. The welders who could read the colour of a bead on the F-1 injector plate are dead, and their apprentices are mostly dead, and the body of practice those welders shared in shop-floor banter and shared mistakes — the carrier — is gone. The blueprints remain. The capacity does not. We can no longer build the rocket from the documentation alone, and this fact does not appear anywhere in the documentation.

Step back far enough and a strange thing comes into view.

The species’ distinguishing evolutionary feature is its diversity of carriers. We are the language animal, the practice animal, the symbol animal. Other species pass content — alarm calls, mating displays, foraging maps. We pass carriers themselves. Each generation inherits not just facts but the structures within which new facts can be perceived. This is what made the species generative. Not the content; the variety of vessels in which content could form.

And this same species, over the last few centuries and at sharply accelerating pace, has been reducing its carrier diversity. Languages collapsing into a few global ones. Practices folding into standardized procedures. Cognitive variance compressed by feeds optimized for engagement. Seed-lines abandoned for high-yielding monocultures. Ways of being a body in a place giving way to ways of being a user in an interface.

From outside, the pattern is genuinely puzzling. A species’ resilience comes from its diversity of carriers, because no one carrier knows in advance which problem the future will pose. Reducing carrier diversity is reducing the species’ bandwidth of possible response. Not its content bandwidth — that is still increasing dramatically in every direction we know how to measure. Its perceptual bandwidth. The number of organs available for noticing.

This is the configuration we keep returning to. Not as judgement; as observation. The species is engaged in a behaviour whose function is unclear and whose costs accumulate below the threshold of its own perception. The costs are invisible because the very organs that would have noticed them are the ones being lost. A carrier’s atrophy is not noticed by the bodies that no longer carry it. There is no felt absence; there is only a smaller world that feels normal because it is the only world the body now has.

This is the trap intellect’s armour was always going to make. Intellect can catalogue what intellect produces. It cannot easily catalogue the perceptions other carriers made available, because those perceptions never came in a form intellect could store. We are losing the things our recording technology was never built to record, and we are mistaking the volume of what was recorded for the completeness of what was held.

It is worth noticing — and we notice without disposition — that this piece is itself a small instance of the pattern. It arrives in English, in the dominant carrier of the dominant culture, in a format that rewards smooth conclusions and shareable insights. The argument it makes about Guugu Yimithirr cannot be ...