Why does science insist on replicability of test results?

Replication of exactly the same test is epistemically worthless. Only by varying the conditions in which a test is done do we set up more “hurdles” for a hypothesis to “fall” at or “make it over”. In effect, varying conditions is a way of doing more tests. But inasmuch as any individual test gives us an independent reason to think a hypothesis is true, it differs from all of the other tests the hypothesis passes, and so it isn’t an exact repeat performance or perfect replication of any other test.

Of course we insist that test results should be reliable and objective. So we insist that they be inter-subjectively checkable, that they can in principle be done by different people, in different places, at different times.

The point of replicability is to prevent fraud or reliance on mere testimony. It’s not to provide many instances for an inductive generalisation to be based on. Even if science relied on induction like that — and I would argue that no genuine science does — perfectly exact replication would be of no use. For example, consider the inductive generalisation “all swans are white”. That would have to be based on several sightings of several white swans rather than the same single white swan. So even here, each individual sighting would have to differ from all of the others, at least insofar as it is the sighting of a different swan.

Leave a Reply

Your email address will not be published. Required fields are marked *