The Desk

Speculative essay. What is measured comes with its figure; what is conjectured is named as conjecture.

For much of its history this discipline spent its energy on a single question: is this resonance true? Are casa and Haus related? Hervir and brew? Iceland and isle?

And it built an admirable apparatus to answer it — sound laws, protoforms, loan controls, unicity distances, permutation nulls. The apparatus works.

The problem is that it answers a question about the title deed, and the title, once the resonance has already occurred, explains almost nothing of what happened afterwards.

The best documented case

The English word island comes from īeġland, “land in water”, related to aqua. It has nothing to do with isle, from Latin īnsula. In the sixteenth century some clerks linked them by resemblance and inserted into island an s that had never belonged to it.

Etymologically it is nonsense. And yet:

the language kept it.

Four hundred years later the s is still there. The title verdict — “false” — is correct and useless: it does not explain why that falsehood won while others of the same kind get corrected and vanish.

The productive question is not whether a resonance is true, but what makes one take hold.

The field was never made of coderivatives

Endolinguistics’ unit of observation is the coderivative network: words that, descending from a common source, preserve the code across different languages. It is an excellent unit, and that is why the resemblance inside a network is a fact and not an impression.

From there it slid, without anyone deciding it, into treating the field as though it were made of those words. It is not.

The coderivative network is how we find a field, not what it is made of.

The evidence is in the corpus and has the shape of a rule: the region belongs to the field, not to the lexeme. When a language replaces the form occupying a region, the region is not lost: it changes tenant.

Promise lives in English give my word, in Spanish dar mi palabra and in French parole. Three unrelated etymological networks, one region. And Spanish is the limiting case: it lost the inherited word — verbo left the centre and came back as a learned borrowing — and did not lose the promise. It gave it to palabra, which came from somewhere else.

If regions migrate to whoever occupies the centre, the criterion of belonging is occupying, not descending.

The desk

There is a way of seeing the system that makes almost all of the preceding conceptual machinery unnecessary.

Picture a very old desk with hundreds of drawers. Each drawer is a binary or ternary code. Inside, hundreds of cards: the words sharing that code. Some have been there since who knows when. From time to time new cards arrive and, instead of staying on the table, a clerk files them where he thinks they should go — even if they come from other languages or other desks. If they sound alike to him, he puts them in the drawer where he finds the most resemblance.

Three things follow from that image.

First: the index is the code, and provenance is not a field of the index. The clerk asks for no papers. That is why an unrelated word ends up among the coderivatives — not “almost magically”, but by construction of the filing system.

Second, and this is the one that bites: once filed, the drawer does not record how the card arrived. The one that has been there since Proto-Indo-European and the one the clerk filed yesterday are in the same place, indistinguishable to the system. So the apophenia verdict judges precisely the datum the archive does not keep.

Third: the drawers are never empty, and they take another card anyway. This corrects the first version of this essay, which put the vacancy at the centre. No free slot is needed.

And there is a measurement from the project itself that changes meaning under this light. The unicity distance measured that a two-class code is carried by some 513 words, and treated it as a defect — the code does not identify. But a drawer with 513 cards is not a broken instrument: it is a well-used drawer. The unicity distance does not measure how bad the instrument is. It measures the size of the drawers.

Who directs the clerk

If the clerk filed by sound alone, the desk would be a heap of drawers holding five hundred cards with nothing in common. It is not. Someone tells him where to put things.

The obvious directs him: use, and the community. But that does not explain the urgency — why a card has to go in one particular drawer.

Also directing him, and this is what the essay wants to hold as conjecture, is unconscious psychic structure. Some cards are filed where the psyche needs them: to displace something that cannot be said in its own place, or to metaphorise. Sound resemblance is not the motive: it is the vehicle.

And historical linguistics already has the case attested, though it does not call it this: taboo. The forbidden word cannot be said, so a neighbour — often a sound neighbour — occupies its place, and the bear comes to be called “the brown one”. It is, exactly, a card filed where it is needed because its own place is closed.

Hence the correction of causality:

need → substitution → the displaced one shifts → the vacancy appears

The vacancy is not the cause: it is the residue. And that explains what the earlier version did not — why some resonances take hold and others do not. The ones that do work, take hold.

Rewriting the card

The clerk has a second faculty, and it unifies two phenomena nobody had put together: he can rewrite the card so that it fits.

If the desk has no drawer for TH, Θεόδωρος is filed as Фёдор. Nobody calls that an error: it is called phonological adaptation.

Now look again at the s of island.

Θεόδωρος → Фёдор and iland → island are the same operation: rewriting what comes in so that it fits a drawer that already exists.

One we call adaptation and treat as normal; the other, false etymology, and treat as error. The difference is in us, not in the language. With an immediate methodological consequence: the nativisation of loanwords is not a confound to be controlled for in the study of resonances. It is the phenomenon, seen in its tame form.

Who audits the clerk

The desk does not accept everything. Iceland filed under isle gets corrected; island filed under isle stayed.

That instance is transmission: repeated use, the community that adopts or does not. The clerk proposes; the community disposes. And since most proposals do not survive, most resonances do not take hold — which is the fact to be explained, not its opposite.

Three instances, three tribunals, and none replaces another:

instance what it decides
the code where it can go — which drawers are within reach of that form
the psyche why it goes — what needs to be said, displaced or metaphorised
the community whether it stays — transmission, the only thing that makes an accident historical

Which reformulates the old rule of the method — the code says where to look, use says what is seen — with one more term in the middle.

Why “it’s apophenia” is the wrong answer

With the desk in place this can be said precisely.

Apophenia is a verdict about title, and title is precisely the field the archive does not keep. It is true, it is verifiable, and it does not touch the question.

And there is evidence that makes it uncomfortable to maintain: folk etymology. Écrevisse becomes crayfish; asparagus, sparrowgrass; a nadder, an adder. Each was, at the time, exactly what we would today call apophenia — and stopped being so the moment a group made it its own and rewrote the word.

Shared, functional apophenia becomes historical fact.

So the verdict moves: the question is not whether the link had a right, but whether it took hold — and taking hold is detected by effects, not by testimony.

And the case this essay does not explain

Taking hold is not one thing. The form is remade (island, sparrowgrass, an adder); use shifts without the form changing, and that shows only in usage corpora; or it is read and nothing more — a student hears hervir and it sounds like brew: nothing in the language moved, something in one psyche bound.

And there is a fourth that fits none of these. Latin bellus comes from *duenelos, diminutive of duenos “good”; the B-L was not in the root and arises by syncope and assimilation. No neighbouring card attracted it.

It is the case this essay does not explain, and it is left in plain sight for that reason.

The essay: The Desk (PDF) · Versión en español

Endolinguistics was founded by C. S. Meulemans and J. Á. Elias Fernández y Cuevas (Introducción a la endolingüística decaglota, 1993). The discipline is theirs, as are its four levels of language and the distinction between exolanguage and endolanguage. And it does not compete with historical linguistics: it presupposes it.