THE EXPLANATION DEFINITION

It is not my claim that probability is irrelevant for evidence but only that the particular probability definitions (1) and (2) of the previous section— the standard definitions—will not suffice.

(1) e is potential evidence that h if and only if e is true and h would

correctly explain e if h were true.

This definition can be closely associated with at least two views. One is Hanson's account of retroductive reasoning, which takes the form

Some surprising phenomenon P is observed

P would be explicable as a matter of course if h were true

Hence, there is reason to think that h is true.^[12]

The fact that phenomenon P has been observed is then potential evi­dence that h; it is so because h would correctly explain P if it were true.

(1) is also closely associated with the hypothetico-deductive account of theories according to which if hypothesis h is a potential explanans of e (which, on this view, means roughly that h contains a lawlike sentence and entails e), then if e turns out to be true e is confirming evidence for h. Sympathy with (1) might then lead to the following simple definition of veridical evidence:

(2) e is veridical evidence that h if and only if h correctly explains e

(i.e., e is potential evidence that h, and h is true).^[13]

Despite the emphasis in recent years on the role of explanation in inference from evidence, neither (1) nor (2) provides a necessary or a sufficient condition for potential or veridical evidence. Neither provides a necessary condition since, as noted earlier, e may be evidence that h even if h does not, and would not if true, correctly explain e. The fact that Jones has the chest wound he does may be potential or veridical evidence that he will die, even though the hypothesis that he will die does not, and would not if true, correctly explain why he has that chest wound. The explanation condition if necessary at all should be changed to require only some explanatory connection between h and e.

Nor do these definitions provide sufficient conditions. Suppose my car won't start this morning. The hypothesis

h: At precisely 2.07 last night 5 boys and 2 girls removed the 18.9 gallons of gas remaining in my tank and substituted water

would if true correctly explain why my car won't start this morning; indeed, suppose that h is true and that it does correctly explain this. In either case the fact that (e) my car won't start this morning is not evidence that h is true. There is too much of a gulf between this e and h for e to be evidence that h, even if h does or would if true correctly explain e. What this gulf amounts to I shall try to say later.

It is worth noting here that the earlier principle of reasonable belief is violated. According to this principle, if the fact that my car won't start this morning were evidence that h, then this fact would be a good reason for believing h. But (given the “normal” background information one might imagine for such a case) the fact that my car won't start this morning is far too meager a reason to believe the very specific hypothesis h. Indeed, innumerably many hypotheses in addition to h can be invented which if true would correctly explain why my car won't start. The hypothesis that

h': At precisely 3.05 last night 2 monkeys removed the remaining 3.7 gallons of gas in my tank and substituted crushed bananas

if true would explain why my car won't start. Is the fact that my car won't start evidence that h' is true? Does this fact provide any reason to believe such a hypothesis?

4.