Measurement and Method

This archive has a quiet thesis, and it stands in nearly every file: it is not the better idea that prevails but the more testable one. What can be put into numbers and reproduced by strangers gets tested, corrected and taken into guidelines; what hangs on the person of its inventor gets admired and dies with him. This theme collects the evidence.

At the beginning stands a wager. Wundt built the laboratory in Leipzig and confined it to what could be measured at the apparatus; the higher processes he declared inaccessible. Ebbinghaus took the wager, invented the meaningless syllable and the savings method — and his one-man experiment of 1885 passed its literal repetition a hundred and thirty years later. The lesson is uncomfortable for any arithmetic of sample sizes: replicability is not a question of case numbers alone but of transparency. A single man with a metronome can measure more durably than a field with funding — if he writes down what he does, and does what he writes down.

That measurement has its price is shown by developmental psychology. Ainsworth’s Strange Situation owes its reliability to a coding discipline that cannot be picked up on the side — weeks of training, test sets, drift checks. Precisely this requirement makes the studies expensive and their samples small: the quality of the measurement and the size of the data stand in a trade-off the literature seldom declares. Whoever set out to simplify the procedure has regularly made it worse.

The clinic supplies the counterpart with the sign reversed. Beck built a measure and a manual to go with his therapy — and precisely for that reason cognitive therapy became the most-tested talking treatment in the world, while Ellis, who was there earlier, shrank to a paragraph in the textbooks. The priority question is settled and beside the point at once: whoever measures writes the history of influence. The Beck Depression Inventory has outlived its originator as Ebbinghaus’s curve outlived his.

Measurement does not protect against error — it only makes error findable. The shrinking effects of the early therapy trials, publication bias, the career of undersized samples: all of it is on record under the replication crisis theme, and it is no objection to measuring but its result — a field auditing itself with its own tools. In this archive things are overturned almost always by better methods, seldom by better arguments.

The rule the theme comes down to was demonstrated by Ebbinghaus and institutionalised by Ainsworth: write it down so that others can refute you. Everything further — preregistration, shared data, many-lab studies — is a footnote to that sentence.