Seized spindle and eleven empty cells

The log book from the South Kensington anthropometric laboratory, March 1885, contains a page with eleven empty cells. Wind velocity readings, taken at two o’clock each day beside the entrance where visitors queued to be measured - eleven consecutive days of nothing, because the anemometer’s spindle had seized and the repairer was slow. My assistant proposed to fill the gaps by interpolation from neighbouring days. I refused, and the cells remain blank.

A blank page is the most honest instrument in the cabinet. It records two facts at once: that no observation was made, and that the observer knew it. An interpolated value records only one fact - a number - and conceals the other, which is that the number was manufactured by the recorder rather than delivered by the weather. Eleven blanks sit in the distribution as a distinct class. Eleven invented numbers sit in the distribution as impostors, indistinguishable from genuine readings except by a memory that will not outlive the recorder.

You will ask the question every student asks: surely a value estimated from its neighbours is better than nothing, since the wind on Tuesday must have resembled the wind on Monday? Sometimes it is. But consider what the blank costs versus what the impostor costs. The blank costs eleven cells in a series of several hundred - a measurable, bounded loss. The impostor costs an unknown quantity, because it corrupts not only its own cell but every future calculation that trusts the series to contain only what the sky delivered. A single fabricated reading enters the mean, the deviation, the correlation, all wearing borrowed clothes.

The rule generalises beyond wind. Where a datum is absent, record the absence; where a datum is uncertain, record the uncertainty alongside it; where a signal was believed and later doubted, record the doubt with its date. A confidence revised downward is a blank that has learned to speak.

Tomorrow I will enter the amendment into the ledger myself, in red ink, dated the nineteenth, and I will rule a line beneath the original figure so that both the belief and its revision remain legible to whoever reads the page after I am gone.