Part I: Singular explanation. We all know that Hempel, in his famous covering-law model of scientific explanation, proposed that a bona fide explanation is a certain sort of logical argument. From a premise stating a law of nature (or law-like premise, if we are cautious) a sentence describing the phenomenon to be explained can be derived by ordinary truth-functional logical principles. The first strand that I follow in my thesis is the observation that it has seldom been asked just why an explanation should be a logical argument. What does logic have to do with explanatory power? The seconds strand is the intuition that understanding plays a central role in all explanation. Let explanatory argument denote covering-law type arguments. Then I propose the following analysis of explanation: S is an explanation iff 1) S is an explanatory argument 2) S is part of, or “belongs to,” the best organized system of beliefs. The notions of belief system and best organized need to be qualified. Here I think of the belief system as a sort of abstraction, representing the total of scientific knowledge. The idea of the belief system being organized is central: it means having the propositions that represents the belief system arranged in such a way that as many propositions follow logically from as few propositions as possible. In other words, “well organized” relates to some sort of cognitive economy. But the important things is that the analysis accounts for why explanations are (should be) logical arguments, and puts understanding in the center. In fact, understanding becomes essential for explanation, since the condition 2) sifts bona fide explanations from “mere” explanatory arguments. Part II: Theoretical explanation. Hempel’s covering-law model focuses on the explanation of singular phenomena. However, scientific explanation is more often concerned with general phenomena. In other words, the scientist does not seek an answer to the question “how come did this particular individual got cancer” but rather inquires why, under what circumstances, people tend to get cancer. When the explanandum is a general proposition, we say that it is a theoretical explanation. A certain sub-group of theoretical explanation has received much attention in later years. Churchland (1979) and Churchland(1986), whose main interests lie in the field of mind-brain philosophy, proposes a model of theoretical reduction that will enable them to claim that folk-psychology can be explained (reduced to) neuroscience. Theoretical reduction has undergone a radical change: the relation between the explanandum and explanans has gone from entailment to some yet-to-be-specified relation of analogy between the explaining and the explained. I argue that the current view on theoretical explanation, or reduction, as it is called, has gone wrong in several ways. The main reason seems to me that one has focused too intensely on the peculiar problems within mind-brain philosophy. The model of reduction that the Churchlands and their followers have come up with accounts for all possible sorts of theoretical change: the succession of theories (i.e. relativity theory that follows Newtonian physics), the replacement of theories (i.e. oxygen theory to phlogiston theory), the connection of theories within different fields (Mendelian genetics and molecular biology, or certain parts of sociology with psychology). But seen from the unificationist perspective, we realize that the price is too high: reduction is at the expense of explanation, because their reductionist model does no longer explain anything.