VAN FRAASSEN'S PRAGMATISM

In the chapter entitled “The Pragmatics of Explanation” in his book The Scientific Image, van Fraassen seems to be arguing in direct opposition to Hempel's nonpragmatic theory of explanation-sentences of form

(2) Account A explains fact X.

Van Fraassen writes:

The description of some account as an explanation of a given fact or event is incomplete. It can only be an explanation with respect to a certain rel­evance relation and a certain contrast-class. These are contextual factors, in that they are determined neither by the totality of accepted scientific theories, nor by the event or fact for which an explanation is requested. (p. 130)

I shall briefly characterize van Fraassen s position by using as an ex­ample some Baltimore lore. By the dawn's early light Francis Scott Key is able to see the flag atop Fort McHenry. And he asks:

Q: Why is our flag still there?

This interrogative, van Fraassen will say, can be used to pose different questions depending on the contrast intended. For example, Key might be asking:

Why is our flag (rather than some other flag) still there?

Why is our flag still there (rather than somewhere else)?

Why is our flag (rather than something else) still there?

And so forth. The contrast class includes what is presupposed by the question (our flag being there) together with the alternatives (there being some other flag there, our flag being somewhere else, etc.). More generally, van Fraassen claims, in the case of any why-question there is a contrast­class that is usually implicit in the context:

In general, the contrast is not explicitly described because, in context, it is clear to all discussants what the intended alternatives are. (p. 128)

For Key the context will tell us that the likely contrast is between our flag being there and the British flag being there.

Now let's turn to the second contextual concept van Fraassen men­tions, the relevance relation.

Q: Why is our flag still there?

might be construed (in van Fraassen's terms) as a request for the “events leading up to its being still there.” If so, we might answer by appeal to the battle raging throughout the night and the failure of the British to cap­ture Fort McHenry. However, there is another possible (though perhaps less likely) interpretation of this interrogative, that is, as a request for the function or purpose of our flag's being there. What we need to know, says van Fraassen, is what “relevance relation” is being requested—“events leading up to,” “function,” or something else. And this, as in the case of the contrast class, is to be determined by looking to the context. “Looking to the context” in our example means invoking the intentions, beliefs, and puzzlements of Francis Scott Key. And this is pragmatic.

Now let's apply this to explanation-sentences of the Hemplian type (2). For our example consider:

(3) The hypothesis that the British failed to capture Fort McHenry during the night’s battle explains the fact that our flag is still there.

Recall van Fraassen’s words:

The description of some account as an explanation of a given fact or event is incomplete. It can only be an explanation with respect to a certain relevance relation and a certain contrast-class.

And the latter are contextual, requiring reference to some particular per­son. Well, if (3) is incomplete, let us complete it by specifying some rel­evance relation and contrast-class. We can do so, van Fraassen tells us, by understanding the question being raised as having three components: the topic (in this case “our flag is still there”), the contrast class (in this case let’s say: “our flag is still there,” “the British flag is there”), and the relevance relation (in this case let’s say: “events leading up to”). Although van Fraassen does not do it quite this way we might now reformulate (3) above by writing:

(4) The hypothesis that the British failed to capture Fort McHenry during the night’s battle explains (by citing “events leading up to”) why our flag is still there (rather than the British flag being there).

We now have an explanation-sentence which provides the sort of infor­mation van Fraassen wants. Is it pragmatic?

It is not explicitly pragmatic, since it contains no terms for an explainer or audience. Is it implicitly so? Do its truth-conditions contain terms for an explainer or audience or others defined by reference to these? Van Fraassen points out, correctly I think, that to determine what relevance relation and contrast-class are being requested appeal is made to the con­text. We look to the explainer, Francis Scott Key, and what intentions and beliefs he had. But this is not sufficient to show that the truth-conditions for (4) must contain terms for an explainer or audience.

Indeed, Hempel—presumably van Fraassen’s arch-foe—could agree that in order to determine what question someone wants to answer, or what event someone wants to explain, essential reference to the intentions and beliefs of the questioner will need to be made. This is no damaging admission for the nonpragmatist, Hempel will say. The important issue is whether once the question being asked has been iden­tified, it can be determined whether the explanation explains without invoking any term for an explainer or audience. So far van Fraassen has offered no reason why this cannot be done. All he has said is that (3) is incomplete. By analogy, Hempel might say, the following sentence is incomplete:

The hypothesis that the British failed to capture Fort McHenry during the night's battle explains.

Suppose we find this incomplete sentence in a history book. To com­plete it appeal is made, let us say, to the historian's likely intentions and beliefs, and/or perhaps to those of Francis Scott Key. That won't make the resulting completed sentence “pragmatic” in what I have so far taken to be Hempel's sense. Suppose we complete the sentence by identifying the explanandum as

why our flag (rather than the British) is still there.

Just because we have appealed to pragmatic considerations in identi­fying the explanandum, Hempel will ask, how does that show that the truth-conditions for the completed explanation-sentence must contain terms for an explainer or audience? Indeed, Hempel will urge us to accept his own truth-conditions for the completed sentence—say those of the D-N or I-S model—which contain no terms for an explainer or audience.

What about van Fraassen's truth-conditions? I find his intentions a bit cloudy at this point. He seems to present two sets of conditions, one set (perhaps) for the concept of a (merely, or minimally) correct explana­tion, and another for the concept of a good explanation. To give the first set of conditions we have a question Q determined by the topic P, the contrast class X, and the relevance relation R. And we have an answer of the form

P in contrast to X because A.

Van Fraassen asks: what is claimed in this answer (p. 143)? He gives four conditions. First, that P is true. Second, that the other members of the contrast class are not true. Third, that A is true. And fourth, that A does bear the relevance relation R to P and X—e.g., that the answer A does give the events “leading up to” the event in P. I'm not sure if this is supposed to be a set of sufficient conditions, or only necessary ones, or, indeed, if it is supposed to be a set of conditions for the truth of sentences of the above form (the latter is suggested by van Fraassen's question “What is claimed in this answer?”)^[59] In any case, these conditions, let it be noted, contain no terms for an explainer or audience. Nor does their application to sentences of the form “P in contrast to X because A” require any refer­ence to explainers or audiences once the question Q is given. Nor do the definitions of van Fraassen's technical terms in these conditions (“topic,” “contrast class,” and “relevance relation”) appear to require the concept of an explainer or audience.

What about van Fraassen's second set of conditions for (as he puts it) “evaluating” answers? Again, we have a question Q determined by the topic P, the contrast class X, and the relevance relation R. How good is the answer

P in contrast to X because A?

Van Fraassen proposes three things that must be determined (pp. 146­147):

1. We must determine whether proposition A is “acceptable” or

“likely to be true.”

2. We must determine whether A shifts the probability toward P

more than toward other members of the contrast class X.

3. We must compare “because A” with other possible answers to the

explanatory question in three respects:

a. Is A more probable than other answers given the background information K?

b. Does A shift the probability toward P more than other answers do?

c. Does some other answer probabilistically “screen off” A from P? (Is there an answer A' such that p(P/A'&A) = p(P/A')?)

This evaluation of explanations introduces two important new factors: a set of background beliefs K relative to which probabilities are to be determined, and a set of answers to the question Q with which the answer A is being compared. Both factors might be deemed pragmatic or contextual. To determine what background beliefs should be used, and what alternative answers proposition A should be compared with, refer­ence will be made to intentions and beliefs of the explainer or perhaps of the evaluator of the explanation. (Indeed, van Fraassen insists that only part of the background information K is to be used in the evaluation, and that which this is “must be a further contextual factor” (p. 147).)

I don't propose here to assess van Fraassen's conditions. (For criticisms see my 1983, chapter 4.) I will simply note what I believe the nonprag­matist's response is likely to be. Just as van Fraassen earlier accused the nonpragmatists of focusing on an incomplete explanation-sentence, so the nonpragmatists will retort “tu quoque” to van Fraassen. All van Fraassen is arguing, the nonpragmatist will say, is that sentences of the following form are incomplete:

“P in contrast to X because A” is a good explanation of q.

The (more) complete form of such explanation-sentences is

(5) “P, in contrast to X, because A” is a good explanation of q relative to alter­natives A₁,...,A, and relative to background information K (or relative to such and such a subset of K).

Now that we have completed the explanation-sentence by relativization to a specific set of alternative hypotheses and to background information we are in a position to use the three conditions van Fraassen presents.

I conclude that van Fraassen ought not to view his position as a prag­matic one—at least with reference to complete explanation-sentences such as those of forms (4) and (5). To be sure, to obtain such complete sentences to begin with, reference may have to be made to explainers. With this Hempel could agree. But once the sentences are complete no reference to any (particular or type of) explainer or audience needs to be made to un­derstand what they mean, or to determine whether or not they are true.

3.