CORRECT EXPLANATION

Since the original paper in Mind was published in 1978, my thoughts on evidence (and its relationship to explanation) have expanded, particularly in the Nature of Explanation and in The Book of Evidence.

The concept of an explanatory connection (used for “potential evi­dence”) is defined by reference to that of a correct explanation: There is an explanatory connection between h and e if and only if either h cor­rectly explains why e is true, or e correctly explains why h is true, or some true hypothesis correctly explains why h is true and why e is true. What is this concept of correct explanation?

First, whatever else it is, it is (or, for potential evidence we need) a concept of correct explanation that is objective, rather than subjective. Whether some h correctly explains some e does not depend upon what anyone knows or believes. In this respect the concept needed is like that which Hempel attempts to explicate by means of his deductive- nomological (D-N) model of explanation.^[18] A D-N explanation of a particular event is a deductive argument in which the premises contain laws and statements describing “initial conditions,” and the conclusion is a sentence describing the event to be explained. Whether some ar­gument is a correct D-N explanation does not depend upon whether anyone knows or believes the truth of the premises or conclusion or knows or believes that the conclusion follows deductively from the premises. Without accepting Hempel's D-N model, I am accepting his idea that there is an objective concept of correct explanation.

Second, the required concept of explanation is that of a correct one, not necessarily one that is good or appropriate for one context of inquiry but not another.

Third, on pain of circularity, the required concept of correct explana­tion must not itself be understood or explicated in terms of evidence. In this respect also the concept of correct explanation needed is like that of Hempel's D-N model. Whether an argument satisfies the D-N conditions for being a correct explanation does not depend on the satisfaction of any requirement of evidence. For example, it is not required that there be evi­dence that supports the premises or conclusion of a D-N argument.

Unfortunately, Hempel's D-N model of explanation is subject to nu­merous counterexamples and other objections that make it impossible to employ for purposes of defining evidence.^[19] In any case the model requires a deductive connection between the explanatory sentences and the sen­tence describing what is to be explained. As noted in section 2 [omitted here], we want to consider cases where e is evidence that h, yet there is no such deductive connection between h and e, or between some other hypothesis H and both h and e, where H correctly explains h and e.^[20]

What can we say further, then, about the concept of correct explana­tion needed for potential evidence? One option is to say nothing further, but to treat “correct explanation” as a more basic concept than evidence, simply asserting that it has the three features noted above: it is objective, noncontextual, and not to be understood in terms of evidence.

In other writings I have proposed an account of explanation that can be used for present purposes. The account is complex, but I will attempt to sim­plify appropriate parts of it for presentation here.^[21] I will then show that it generates a concept of explanation of a type needed for evidence, and I will contrast this concept with Hempel's D-N explanation and indicate its advantages. It is not my claim that my particular account of explanation is required to understand evidence. Everything I have said so far about evidence could be accepted without endorsing this account of explanation. My claim is only that the account supplies a concept that will suffice for evidence.

9.