The Quartz Piezoelectrometer We Built

Sèvres, or rather my writing desk facing the rue d’Ulm, the sixteenth of a cold April - I write to you, Debierne, because today you will tell me I am wrong about the radium standard, and I want to answer before you do.

You recall the quartz piezoelectrometer we built together, the one Pierre and I used to weigh the faintest currents the rays could summon. Its calibration rested on two things: the charge deposited on the plates, and the elastic constant of the quartz itself. The constant was measured against external standards, traceable to the Bureau. The confidence we placed in our ionization readings was therefore earned, inch by inch, from outside sources. Nothing in the instrument confirmed itself.

Now to the matter that keeps me at the desk past midnight. When I first measured the activity of pitchblende, the evidence supported a certainty: a new element, intensely active, present in small quantity. When I later refined the separation, fractional by fractional, in the isolation vessels you know well, the same instruments, the same method, supported the opposite certainty in a weaker claim: that the activity was carried by barium fractions and required longer reasoning to attribute. Both conclusions drew on the same electrometer readings. The instrument did not change. The activity of the residue did not change. What changed was which reading I chose to treat as the signal and which as the barium’s interference.

Here is my error, stated plainly. I believed that identical evidence accumulating could only strengthen a single conclusion, the way dose accumulates in the body - always in one direction. This is false. Identical evidence accumulating against a refined protocol can weaken a conclusion. Each new fraction dissolved was not more confirmation; it was a fresh calibration check, and a calibration check can pass or fail. I had confused repetition with replication. Repetition is internal consistency. Replication requires the external standard, the held-back sample, the fraction I do not let myself see until the attribution is fixed in the notebook.

So I amend my position: where last month I wrote to you that the radium determination stood at high confidence, I now hold it at moderate confidence, not because the substance faltered but because my reading of the fractions was not blind. Rewrite the entry in the Annales with the downgrade included. The record that admits its own correction is the only record worth binding.