Models and the limits of knowledge
Where the Model Breaks
An irregular result is not new physics. It is a suspect — and it must survive interrogation before it becomes evidence.
Established ScienceWhat we actually know
Anomalies usually die. A damaged sensor, a contaminated sample, a mistaken assumption, a software sign error, an unmodelled systematic — these explain the overwhelming majority of surprising results. The 2011 faster-than-light neutrino measurement at OPERA was traced to a loose fibre-optic connector and a clock oscillator. Particle physics demands five-sigma before calling a discovery precisely because the field has learned how often three-sigma evaporates.
Where the extrapolation begins
Only when an anomaly survives every attempt to eliminate it, in independent hands with independent instruments, does it become evidence that the model may be incomplete. This is slow by design, and the slowness is the safeguard.
What the novel invents
CROWN's measurement programme is built on exactly this discipline: ninety-three years of trying to explain the object away with instrument error, and failing. The elongation was dismissed as noise for decades before the noise refused to move.
Questions to think about
- — Why does science treat its own surprising results as guilty until proven innocent?
- — What is lost if the standard of proof is set too high? Too low?
Where to read next
- — OPERA faster-than-light neutrino anomaly and its resolution
- — Statistical significance conventions in particle physics
Continue in the archive