Kitcher's unification theory of explanation seems to suggest that only the most reductive accounts can legitimately be termed explanatory. This is not what we find in actual scientific practice. In this paper, I attempt to reconcile these ideas. I claim that Kitcher's theory picks out ideal explanations, but that our term “explanation” is used to cover other accounts that have a certain relationship with the ideal accounts. At times, “versions” and portions of ideal explanations can also be considered explanatory.