No hypothesis is tested in isolation, so a failed prediction never tells you which assumption to abandon.
The 1919 eclipse could have been read as refuting Newton, as an error in the instruments, or as an atmospheric artefact. Which assumption to abandon is a judgement no experiment settles.
In practice
- A dashboard shows no traffic and the team cannot tell whether the campaign failed, the tracking broke, or the definition of a session changed.
- A drug trial misses its endpoint, and the sponsor must decide whether the drug does not work, the dose was wrong, or the population was chosen badly.
- Further reading
- Wikipedia
- Subject
- Science & Epistemology
- Essay
- How knowledge grows, and how inquiry fools itself