It is expected that there will be discrepancies between models and observations. However, why these arise and what one should conclude from them are interesting and more subtle than most people realize. Indeed, such discrepancies are the classic way we learn something new.