One Third Smaller Than March

Dear Professor Bessel,

The actinometer has been giving me trouble at the Cape, and the trouble has turned out to be more instructive than the successes. On the fourth of March last, under a sky that any observer would have called uniformly clear, the instrument recorded a fall in the blackened thermometer that was a full third smaller than the reading of the second of March, conditions appearing identical. I spent a week hunting the error - the lamp-black, the bulb, the exposure interval, the meridian altitude of the sun. The instrument was blameless. The sky was not. There was a haze, invisible to the eye, that no instrument of mine could see and no question I had asked could detect.

Here is what I keep turning over. The answer came easily once the question was right: ask whether the atmosphere transmits equally on all clear days, and the anomalous reading becomes a datum rather than a nuisance. But arriving at that question cost me the week, and I begin to suspect the disproportion is general. An answer, once found, can be checked - recomputed, reproduced by any competent hand from the same inputs. A question carries no such test. There is no way to verify, before the fact, that the question is the one the phenomenon will actually answer. My father’s sweeps of the heavens produced catalogues by the thousand, but the questions that organised them - is the nebula resolvable, do the stars cluster - those were found by accident, in the margins, and sometimes not found at all for decades.

The catalogue of my errors grows faster than the catalogue of measurements, and I am minded to keep it deliberately. A gap in a sweep log is a load-bearing gap: it tells you where the instrument was blind, and therefore where the next question must be aimed. The answer fills a cell in the table. The question determines the shape of the table itself, and no table can audit its own shape.

I have resumed the actinometer series with a second bulb, of different glass, to see whether the invisible haze divides the two. The discrepancy between them, if it persists, will be the result. I am still watching it.

Yours, J. F. W. Herschel