BUT IS IT CIRCULAR?

A pair of circularity charges may be leveled against this account. One involves circularity in the account of explanation itself. Even if “explana­tion” is not defined by reference to evidence, it is defined by reference to concepts dangerously close to that of explanation itself, that is, by refer­ence to the concept of a content-noun, where the latter category includes the noun “explanation.” This charge of circularity I reject, since the cate­gory of content-nouns, and more generally of content-giving propositions, is characterized, as I have done in the previous section, without appeal to any concept of explanation.

The second and more serious charge of circularity, or at least triviality, concerns condition (5) in the previous section for being a correct explanation. On this condition, whether

(1) The reason that John's symptoms were relieved is that he took medicine M is a correct explanation of why John's symptoms were relieved depends not only on whether (1) is a complete content-giving proposition with respect to the question but also on whether (1) is true. But whether (1) is true is just what we want a definition of “correct explanation” to tell us how to determine. And condition (5) in the previous section doesn't do that at all!

My response to this is to agree that (5) does not provide a way of de­termining whether a proposition such as (1) is true. What (5) does, which is not trivial, is to say that if some proposition p is a complete content­giving proposition with respect to Q, then it provides an explanation of q that is correct if and only if p is true. This cannot be said for propositions generally, even if they are true and provide answers to Q. It is possible for the proposition “John took medicine M’’ to be both true and to provide an answer to “Why were John's symptoms relieved?" without that proposi­tion's being a correct explanation of why his symptoms were relieved.

There is an analogy between my procedure here and what Hempel does in his account of D-N explanation. He offers conditions for being a potential explanation (which may be a correct or an incorrect expla­nation). The conditions—such as the requirement of lawlike sentences in the explanans and a deductive relationship between explanans and explanandum—do not invoke any notion of explanation itself. To give the further condition for a correct D-N explanation he simply adds that the sentences in the explanans must be true. In an analogous fashion I define the notion of explanation in terms of content-giving propositions and so forth, without invoking any notion of explanation itself. The further con­dition I give for a correct explanation is simply that the relevant content­giving proposition must be true.

The difference between what Hempel does and what I do is this. Hempel's explanans sentences do not include ones such as (1) above. He banishes from an explanans any terms such as “reason," “causes," and “explanation" (itself). This idea is required by what (in The Nature of Ex­planation) I call the NES (No Entailment by Singular sentences) require­ment. (See chapter 8 in the present volume.) NES, to which Hempel is committed, precludes from an explanans any singular sentences (ones describing particular events), including (1), that by themselves, without the necessary laws, entail the explanandum.

Hempel's NES requirement, however, leads to serious counterexam­ples in which, although the explanans is true (and satisfies all of Hempel's conditions, including NES), it does not correctly explain the explanandum.

At time T, Ann ate a pound of arsenic

Anyone who eats a pound of arsenic dies within 24 hours

Therefore,

Ann died within 24 hours of T.

Suppose, however, Ann was killed not by the arsenic but by being hit by a truck, which had nothing to do with the arsenic. Then the above expla­nation is incorrect—it does not give a correct reason that Ann died within 24 hours of T—despite the fact that the sentences in the explanans are true, the explanans contains a true law, the explanans deductively implies the explanandum, and NES is satisfied. (The only singular sentence in the explanans, the first sentence in the argument, does not by itself deduc­tively imply the explanandum.)

This problem arises because of the occurrence of an intervening cause. The only way to avoid it is, I suggest, to include in the explanans a sen­tence such as

The reason that Ann died within 24 hours of T is that she ate a pound of arsenic at T

The cause of Ann's death within 24 hours of T is her eating a pound of arsenic at T^[23]

These sentences, if true, would correctly explain why Ann died within 24 hours of T. They express complete content-giving propositions with respect to the content-question “Why did Ann die within 24 hours of T?" So, in accordance with my condition (5) of the previous section, if they are true they provide correct explanations. But these sentences violate Hempel's NES requirement. They are singular sentence that by them­selves, without the need of laws, entail the explanandum sentence “Ann died within 24 hours of T."

I have described a concept of explanation that meets three criteria required for potential evidence. It is objective, noncontextual, and not itself defined in terms of evidence. Nor is the account circular in the sense of defining explanation by reference to itself. The general notion of (“potential") explanation is defined by reference to complete content­giving propositions, without appeal to any notion of explanation. But conditions for correctness in such explanations will need to include the requirement of the truth of propositions that violate NES: the truth of complete content-giving propositions that contain terms such as “reason" and “cause."