Science · Level 5 · 212 words
Sealed Until the End
Original passage © Studio AM, written for Fluency.
Plot measurements of a physical constant by year, and the points can tell an awkward story. The classic case is the charge of the electron: Millikan's celebrated value was slightly low, yet later measurements did not jump to the true value. They crept toward it over twenty years, because a result that strayed far from the famous number was rechecked until it behaved, while one that agreed was published as it stood.
No one was faking data. The bias worked through a smaller door: an analysis involves hundreds of judgment calls (which runs to discard, when to stop hunting for errors), and researchers who know the expected answer stop hunting the moment their number matches it.
Modern collaborations defend against their own expectations with blind analysis. Before anyone looks, a computer adds a secret offset to the quantity being measured, so every intermediate result is deliberately meaningless. The team debates every cut and correction blind to what each choice does to the answer. Only when the whole procedure is frozen in writing is the offset removed, and the number that appears is the number published.
The method concedes something uncomfortable: training does not remove the thumb from the scale. What removes it is making the scale unreadable until the weighing is done.
Source: Written for Fluency. Original passage © Studio AM, written for Fluency.