Eight Tenths of a Second

David Kinnebrook was an assistant at the Royal Observatory from 23 May 1794. His job was to catch the exact moment a star crossed the meridian wire.

The method was Bradley's — eye and ear. You watch the star approach the wire while counting the clock's beats aloud, and you judge where between two beats the crossing fell. It is a whole-body skill: eyes, ears, working memory, and the timing of your own attention, all at once. Maskelyne set out what it demanded, and the sentence is more revealing than he can have known:

In the same published account, the Astronomer Royal recorded that his assistant had "began, from the beginning of August last, to set them down half a second of time later than he should do," and that in January of 1796 "he increased his error to 8/10ths of a second."

Kinnebrook was dismissed on 12 February 1796. Maskelyne, who had already tried and failed to correct him, still described him in print as a diligent and useful assistant to me in other respects. He went home to Norfolk, taught school, and computed for the Nautical Almanac from a house in Norwich. He died in 1802, aged about thirty.

Nobody at the time thought anything had been discovered. A young man was simply not accurate — which is to say, not the same as his supervisor.

contested  Greenwich → the personal equation

Two decades later, Friedrich Bessel read Maskelyne's account and did the thing nobody at Greenwich had done. He checked whether it was only Kinnebrook.

It was not. Bessel compared his own transit timings against other practised astronomers and found gaps as large as Kinnebrook's, and larger. More importantly, he found they were not random. Each observer ran consistently early or consistently late against each other observer — a stable, personal offset that stayed put across nights and stars.

That is a genuine discovery about human beings, made with a telescope: perception and reaction differ between people, systematically, and the difference does not go away when you try harder.

And here is what astronomy did with it. It did not screen for fast observers. It did not train the difference out, or grade observers against a standard nervous system. It wrote down, for each person, how they differed — the personal equation — and added that constant to their readings.

Be careful about why. Bessel was not defending anybody. He wanted accurate stellar positions, and a systematic error is simply a correctable one. Nobody was aiming at dignity. That is the shape of every link in this chain, in both directions — people solving their own problem, producing an outcome they never had in view.

documented  the personal equation → experimental psychology

Once you can measure how long a person takes to notice something, you have an instrument that does not care what it is pointed at.

Reaction time came out of astronomy as a nuisance to be corrected, and arrived in the new psychological laboratories as a subject in its own right. Wilhelm Wundt opened his laboratory at Leipzig in 1879, and mental chronometry — timing the components of perceiving, deciding, responding — was among the things it was built to do.

James McKeen Cattell studied with Wundt, and in 1890 published the phrase that names the rest of this chain: mental tests. He then took it to Columbia.

Notice exactly where the substitution happens. Bessel measured a difference and used it to correct a reading. The new discipline measured a difference and used it to rank the person. Those are not the same operation, and nothing in the physics licenses the second.

This joint is well documented as a matter of who taught whom and what got measured. The inference that a speed is a quantity of mind is not documented at all — it is an assumption the apparatus quietly carried across the threshold, and it is the load-bearing error of the next hundred years.

documented  Cattell → Thorndike, at Columbia

Cattell's doctoral student finished his dissertation in a year. It was called Animal Intelligence, and it put cats in boxes.

Edward L. Thorndike built puzzle boxes with a latch inside. A hungry cat goes in; food sits outside; the cat scrabbles until it happens to trip the latch. Put it back in and time it again. And again.

What Thorndike published was not the story of any cat. It was the learning curve — escape time on one axis, trials on the other, dropping away as fumbling gives way to the successful move. From it came the law of effect: acts followed by satisfaction get stamped in, acts followed by discomfort get stamped out.

This is the joint where the instrument that started at a telescope reaches something it can shape rather than merely read. A transit time is a fact you record. A learning curve is a fact you can try to bend — by arranging what follows a behaviour.

Thorndike went to Teachers College, Columbia, and stayed. Everything after this happens in one building or downstream of it.

documented  Thorndike 1898 → Thorndike 1918

Twenty years after the cats, Thorndike wrote the sentence that made a child measurable.

Read as epistemology it is almost banal. Read as a warrant for a profession it is enormous, because it converts a question about what schooling is into a question about what schooling can be scored on. And Thorndike said what would be scored: education is known to us, he wrote, only by the products it produces — things made, words spoken, acts performed, and the like.

Products. Not children — products, of which children are the raw material. That framing was not hidden or subtextual in this period; it was the textbook. Ellwood Cubberley put it in the standard administration manual of the era, in words nobody has had to dig for:

Bessel's constant said this observer differs, here is by how much, keep them. Thorndike's credo says this child differs, here is by how much — and stops. What you do next is left to whoever is holding the ledger.

contested  Thorndike → Kelly

A second track has been running underneath this whole time, and it started in the same place — an astronomer's error curve, turned on people.

We have told that story already, and we will not re-tell it here: the curve for observational error becomes Quetelet's average man, becomes Galton's ranking, becomes eugenics. It is The Lines We Drew, zine No. 9, and the corrections we had to make to it are in the public changelog.

What matters for this chain is where that rail goes next. Emil Kraepelin extended Galton's sorting logic across the whole range of mental functioning, building the classificatory scheme that descends into the DSM — a system whose business is to say what you are.

They are separable in principle and they were built by different people for different reasons. But a classification with no intervention attached is inert, and an intervention with no classification has no one to be applied to. The rails need each other, and in 1987 they meet.

documented  Galton → Kraepelin → DSM  (verified at No. 9)

The multiple-choice test was invented by a man trying to solve an injustice, and he spent his later life trying to undo it.

Frederick J. Kelly finished his doctorate at Teachers College, Columbia, in 1914. The dissertation was Teachers' Marks, Their Variability and Standardization, and it had two complaints, both fair. Teachers' grades varied wildly for the same work — a matter of luck for the child. And marking swallowed the hours a teacher should be teaching.

His answer was the Kansas Silent Reading Test: the first published multiple-choice test. A whole class at once, no writing, scored by running an eye down a column.

Hold on to how reasonable this is. Kelly is not a villain and the problem he named was real — Autistic and otherwise neurodivergent children have always been on the receiving end of a teacher's private, unaccountable judgment. Standardization is a genuine protection against that, and it is also how a child's morning becomes a set of items with one right answer each.

Kelly himself turned against what he had made. As a university president decades later he tried to pull his institution back from standardized testing, and lost his job for it.

leap  Kelly's item → Pressey's machine

Sidney Pressey assembled a machine out of typewriter parts that could present a multiple-choice question, take an answer, and score it.

The order here is the point, and it is the reverse of how people tell it. The machine did not create the standardized item. The item created the machine. Once a child's understanding had been packed into four options with exactly one correct, a device could handle it — because the hard part, deciding what counts as knowing, had already been done and frozen into the answer key.

Pressey believed he was building a liberation: immediate feedback, self-pacing, a teacher freed from drill. Every claim made for personalized learning in the last decade was made for a box of typewriter parts a century ago, and Audrey Watters' history of these devices is the record of how reliably the claim recurs and how reliably the machines fail commercially while the idea survives.

Note what has quietly happened to Bessel's discovery by this point. Self-pacing is almost the personal equation — an acknowledgment that people move at different rates. But the machine paces you toward one fixed endpoint. Difference is allowed as a speed, never as a destination.

documented  Pressey → Skinner

B. F. Skinner took Thorndike's law of effect, sharpened it into operant conditioning, and pointed it at the classroom in "The Science of Learning and the Art of Teaching."

His complaint about schools was, in his own terms, technical. Reinforcement arrives too late, too rarely, and too crudely — a mark on a page days after the work. Behaviour is shaped by what immediately follows it, so schooling is an engineering problem: break the material into small steps and reinforce each one at once. Hence the teaching machines.

Two things stayed even though the boxes did not sell. First, the assumption that the observable behaviour is the whole of the learning — that what cannot be counted did not happen. Second, that a person's difference is a schedule problem: with correct contingencies, any behaviour can be brought to the target.

Notice this is the exact inverse of the personal equation. Bessel's constant said the difference is stable, so accommodate it. Skinner's schedule says no difference is stable, so remove it.

documented  operant conditioning → Lovaas

This is where the chain stops being history, because from here it is a description of a great many of our childhoods.

O. Ivar Lovaas applied operant conditioning to Autistic children at UCLA. His 1987 paper in the Journal of Consulting and Clinical Psychology reported that after intensive long-term treatment, 47% of the experimental group (n=19) achieved "normal" intellectual and educational functioning, against 2% of the controls (n=40).

Read the outcome measure, because the outcome measure is the argument. What was scored as success was normal-range IQ and passing an ordinary first grade. Not health. Not communication. Not whether a child could ask for what they needed, or was in less distress, or was happier. The endpoint was indistinguishability.

Now both rails are visible in one room. The diagnostic rail — Galton, Kraepelin, the DSM — supplies what this child is. The behavioural rail — the puzzle box, the item, the machine, the schedule — supplies what will be done about it, in trials, with reinforcement, at a table.

We are not making the naturalistic move here — nothing about stars or transit times settles what is owed to a child. The point is narrower and harder: the apparatus that produced this endpoint was never designed to choose an endpoint at all. It was designed to correct readings. Somebody had to supply "be indistinguishable," and it did not come from the physics.

contested  behaviourism → the dashboard

The teaching machines failed as products and won as an idea, which is the more thorough kind of victory.

Behaviourism was supposedly displaced by the cognitive turn decades ago. The apparatus was not. It moved into software, where points, streaks, badges, levels, immediate feedback and a progress bar toward a fixed target are simply what a learning platform is — reinforcement schedules with better graphics, rarely named as such by the people shipping them.

Papert's alternative was not a softer schedule. It was a different question — children building things they cared about, with the machine as material rather than as monitor. He lost, and Dewey lost before him, and both losses are the same loss: the thing that could not be scored was not defended.

We mark this joint contested deliberately. It is a strong argument by a historian who read the archives, and it is an argument — a claim about intellectual descent, not a documented handoff. If we marked it solid we would be doing the thing this zine is about.

Two disciplines met the same discovery — that human beings differ from one another in perception, timing and attention, stably, and not by choice. They answered it in opposite directions.

Astronomy said: the difference is real, so publish the constant and keep the observer. The variation went into the model. The science got better — more accurate, not less — precisely because it stopped pretending its observers were interchangeable.

Schooling said: the difference is real, so bring the observer to the standard. Rank, then remediate, then reinforce, then score how close to indistinguishable we got.

And here is why the chain form matters rather than a villain. Nobody authored this road. Maskelyne wanted accurate transits. Bessel wanted good positions. Cattell wanted a science of mind. Kelly wanted children protected from arbitrary marking. Pressey wanted teachers freed from drill. Each solved their own problem competently, and could not have seen what their work would join up with a century downstream.

That is not an excuse — the harm is not smaller because it was unintended. It is a route. A thing that was accumulated can be dis-accumulated, one competent adjustment at a time, in the other direction. Nobody has to win an argument with a conspiracy. We have to keep supplying the endpoint the instrument was never able to choose.

We are the assistants who got dismissed for eight tenths of a second. And we are also, now, the ones holding the pen — writing our own constants down, and refusing to be corrected away.