Where we have good, testable explanations, they then have to be tested, and we drop the ones that fail the tests.