Measurement and observation
The Error Belongs to Us
Every measurement is a number plus an admission of how wrong it might be. Improve the instrument, and the irregularities you were blaming on the object often turn out to have been yours.
Established ScienceWhat we actually know
A measurement without an uncertainty is not a measurement. Saying a table is two metres long is a different claim from saying it is 2.000000 metres long; the extra zeros are promises about the quality of the apparatus, the calibration and the corrections for temperature, pressure, ground motion and the observer. Real manufactured objects have dents, welds, thermal expansion and gravitational sag. Refining your instruments normally reveals more of that structure, not less.
Where the extrapolation begins
So the diagnostic question about any anomalous object is simple: as your uncertainty shrinks, does the object get messier or cleaner? Messier is the ordinary outcome. Cleaner — the residuals collapsing toward an ideal mathematical form survey after survey — is the outcome that has no natural precedent, because nothing in nature is manufactured to a limit.
What the novel invents
Dr Maseko's team ran the object from a twelve-centimetre survey to a four-centimetre survey to thirty-two stations synchronised on three atomic clocks, in rain, wind, day and night. The surface approached an ideal form every time. Her line is the whole of the scientific method compressed into a sentence: the errors belonged to us, not to the object. Asked whether she trusts figures quoted to nine decimal places, she says no — she trusts them within their stated uncertainty, and adds that trusting a number without stating what might be wrong with it is faith, not science.
Questions to think about
- — Why is 'approximately 108 metres' a stronger scientific claim than '108 metres'?
- — List four things that could be wrong with a measurement of a large object made from a kilometre away.
- — Better instruments made the object look more perfect. Why is that the disturbing result rather than the reassuring one?
Where to read next
- — NIST guidance on expressing measurement uncertainty
- — Any introductory laboratory text on error budgets and systematic versus random error
Continue in the archive
