If the theory accurately predicts what they [scientists] see, it confirms that it's a good theory. If they see something that the theory didn't lead them to believe, that's what Thomas Kuhn calls an anomaly. The anomaly requires a revised theory - and you just keep going through the cycle, making a better theory.
Ultimately, when you come up with a classification scheme that is collectively exhaustive and mutually exclusive, then the theory can become what Kuhn called a paradigm.