Theories are pretty easy to come by. Even with the demand for genuinely explanatory theories, there are lots of competing claims about the unseen objects and events that genuinely explain the phenomena we see. But at most one of those theories is true. The urgent responsibility of science, then, is to weed out the bad from the good, that is, to expose the failures in false theories and to champion only those theories that have endured a process of testing that rewards the true and scorns the false. This means that, as much as is possible, a theory must be tested against the observable data. For all a theory's theoretical elegance and pragmatic virtue, there is no reasonable claim to accurate representation of the world until it confronts the world. You've got to do the experiments.

This requirement of empirical testing is probably one's first and most intuitive answer to the question of what makes science believable, and it is a good answer. If the theories claim to be about the world, then they should be accountable to information from the world. Confirmation by testing against observations is the most straightforward external event in science, and achieving a positive test is the most obvious of external virtues. But it is important to be clear on what is meant by “testing” a theory and by “comparison with the observable evidence.” Theories usually describe objects and events that are not amenable to observation.