The tally stick

Before England had banknotes, or double-entry books, or anything a modern accountant would recognize as a ledger, it had wood. From around the twelfth century until the practice was formally abolished in 1826, the royal Exchequer recorded debts and payments on tally sticks — lengths of seasoned hazel or willow, notched along one edge to mark a sum. The width of the notch encoded the amount: a cut the breadth of a hand for a thousand pounds, a thumb's width for a hundred, a barleycorn for a shilling, each denomination given its own unmistakable size so that no one could later claim a shilling had been a pound. This was a serious instrument of state. Enormous sums, whole rounds of royal taxation, passed through these notched sticks for the better part of six hundred years. And the genius of the thing was not in the notching. It was in what happened next.

Once the amount was cut into the stick, the stick was split. It was cleft lengthwise, straight down through the notches, but deliberately unevenly — one part left longer than the other. The larger piece, called the stock, went to the party who had paid money into the treasury; it is from this that we get the word stockholder. The smaller piece, the foil, stayed with the Exchequer. Each party now held half of a single record. And because the two halves had once been one piece of wood, split along the grain, the notches and the wandering line of the split matched only each other. No second stick could be made to fit. To prove a debt, you brought your half and laid it against the other; if they married — notch to notch, grain to grain — the account was true.

The record split in two

Consider what this arrangement quietly refuses to rely on. There is no strongbox, no seal, no clerk of unimpeachable character whose word settles the matter. The Exchequer holds one half and the subject holds the other, and neither half means anything on its own — a foil without its stock is a splinter of wood. The integrity of the record does not sit inside either piece. It lives in the relationship between them, in whether they still fit. The proof is the match. This is what makes the tally a genuinely different kind of object from a signed parchment locked in a chest: its trustworthiness is not guarded, it is distributed. You cannot corrupt the record by getting to where it is kept, because it is not kept anywhere. It is kept in two places at once, by two parties with opposed interests, and the discrepancy between them is the alarm.

Tamper-evidence, in the tally, is a property of the geometry, not of anyone's vigilance. The historian's word for the mechanism is homely and exact: the stock and the foil are indentures of each other, the same irregular edge read from both sides. Nature made each split unrepeatable — the specific way a green branch tears along its fibers cannot be forged by cutting a fresh stick to order. So the humblest possible material, a stick of hazel, was turned into an instrument that a king and a taxpayer could each trust without trusting each other. That is a remarkable thing to have built out of firewood, and it is worth being precise about why it worked.

The record was not kept safe. It was kept split — and safety was whatever survived the two halves being brought back together.

Forgery that requires two conspirators

Try to cheat a tally and you discover the shape of its defense. Suppose you are the debtor and you want to shrink what you owe. You could shave a notch on your half, narrowing a pound into a shilling. But your half is only the stock; the foil sits in the Exchequer, uncut, and the moment the two are laid together the fraud is not argued about, it is simply seen — the notches no longer line up. To make the lie hold you would need to alter both halves identically, which means reaching the piece in the other party's keeping and reproducing your forgery on it down to the grain. A single party cannot do it. Forgery here is not a matter of skill or nerve; it requires a conspiracy of everyone who holds a half, acting in concert, against their own opposed interests. The one alteration that would go undetected is the one that both sides make together — which is to say, not a forgery at all but an agreement.

This is a different and stronger thing than the protection a notary offers, and the difference is the whole point. A notary — or a seal, or a trusted registrar — defends a record by inserting an authority you are asked to believe: a third party who was present, who attests, whose reputation stands behind the document. The tally needs none of that. There is no trusted third party in a tally, and nothing turns on anyone's honesty, not the Exchequer's and not the subject's. The matching does the work. Integrity rests on the fact that the record was split, not on the fact that someone reputable is holding it. Take away every honest man in the story and the tally still cannot be quietly altered, because the proof was never a person's word. It was two pieces of wood that agree.

A tally stick for the decision

Now turn the lens forward. Nearly every consequential record produced by a modern machine fails the tally's test in the most basic way. A model scores an applicant, flags a transaction, denies a claim, and the account of what happened is held by exactly one party — the party that ran the system and has every interest in how the story reads. There is no second half in anyone else's keeping. The record can be revised, backdated, regenerated, or quietly improved, and no one on the other side holds the matching piece that would reveal the change. Its integrity rests entirely on trusting the holder not to touch it. That is precisely the pre-tally condition: a single strongbox, and a promise. We have built extraordinary machinery for making decisions and reverted, for the record of them, to the security model the Exchequer had already outgrown by the reign of Henry II.

The tally names the fix, and it is not more surveillance of the holder. It is distribution. A record whose tamper-evidence comes from being split and mutually held — where the party affected by a decision holds a piece that must still match the piece held by the party who made it — cannot be silently rewritten by either side, because alteration breaks the match, and the break is the finding. This is the discipline behind what we call a Decision Receipt: not a prettier account handed down by the deciding party, but a record structured so that its integrity is verifiable by comparing independently-held parts, its provenance carried rather than asserted, its chain of custody something an adversary can check instead of something they must take on faith. The point is not that the record is stored more carefully. It is that no one party can hold the whole of it and still be believed. Contestability begins there — with a second half in the hands of the person who might want to fight.

We reach, when we imagine accountability for machines, for the language of oversight and audit — for a better clerk watching the strongbox. The Exchequer knew a deeper answer eight hundred years ago, and it fits in the hand. Do not ask anyone to be trustworthy. Split the record, give a half to the party who could be wronged, and let the truth be whatever survives the two halves being laid together. When the sticks stopped being needed and the accumulated centuries of foils were finally burned in 1834, the fire got out of hand and took the Houses of Parliament with it. It is a fittingly literal image. We had a technology for records no single party could forge, and we set fire to it — and are now paying, the expensive way, to rediscover what it knew.

— Dispatches · Summit Cognitive

---