EVIDENTIAL HOLISM: DUHEM, QUINE, WHEWELL

Duhem, the most famous exponent of the view, summarizes it more expansively:

In sum, the physicist can never subject an isolated hypothesis to experimental test, but only a whole group of hypotheses....

A few pages earlier, Duhem explains the doctrine by showing how evidence disconfirms a hypothesis:

A physicist decides to demonstrate the inaccuracy of a prop­osition; in order to deduce from this proposition and institute the experiment which is to show whether the phenomenon is or is not produced, in order to interpret the results of this ex­periment and establish that the predicted phenomenon is not produced, he does not confine himself to making use of the proposition in question; he makes use also of a whole group of theories accepted by him as beyond dispute. The predic­tion of the phenomenon, whose nonproduction is to cut off debate, does not derive from the proposition challenged if taken by itself, but from the proposition at issue joined to that whole group of theories; if the predicted phenomenon is not produced, not only is the proposition questioned at fault, but so is the whole theoretical scaffolding used by the physicist. The only thing the experiment teaches us is that among the propositions used to predict the phenomenon and to establish whether it would be produced, there is at least one error; but where the error lies is just what it does not tell us.

In the first quotation, Duhem formulates the idea that only a theoretical system as a whole is tested, not an iso­lated proposition within that system. In the second quota­tion, he spells this out for testing that involves disconfirming evidence. His general view is that you obtain evidence, whether confirming or disconfirming, only for a theoretical system, not for an individual statement within that system. And you obtain such evidence by deriving consequences (“predictions”) from this system and determining by ex­periment and observation whether these obtain. If they do, then they constitute confirming evidence for the theory as a whole. If they don't, they constitute disconfirming evidence for the theory as a whole.

Unlike Duhem and Quine, William Whewell, whose doc­trine of evidence I summarized in chapter 1, does not explicitly endorse a holist position. But this, I think, is the most reason­able way to understand his view. For him, some fact or set of facts e is evidence for some H only if H not only explains e but also explains and predicts facts of types different from e (“con­silience”), and only if, as H develops over time and new or re­vised assumptions are made in the light of new observations, the revisions “tend to simplicity and harmony” (“coherence”). Whewell clearly has in mind here a system of hypotheses H,

rather than an individual one. (His two favorite examples are Newtonian mechanics and the wave theory of light.) And his criterion of “coherence” involves groups of hypotheses that “run together” (his phrase) in giving explanations of phe­nomena. When these phenomena are explained by a theory that is consilient and coherent, they constitute evidence that provides a good reason for believing the theory, or even stronger, evidence that establishes or proves it.

Now, let me formulate a general holistic doctrine suggested by these ideas. Suppose we seek to obtain ev­idence that provides a good reason for believing a set or system of hypotheses H. In order to do so, we need to de­rive consequences from this set that can be subjected to ex­perimental test. These consequences are ones explained or predicted by H. In general, they are derivable not from just one hypothesis in H but from many, if not all of them, in H—from the whole system H. These consequences, when confirmed or disconfirmed by experiment and observation, provide evidence sufficiently strong to accept or reject the whole system H, not isolated parts of it.

By contrast, suppose we seek to obtain evidence for a single (“isolated”) hypothesis h. How can we do so? On the present holistic view, we can do so only by deriving testable consequences from h and determining by experiment and observation whether they are true. But, the holist insists, in general in order to derive testable consequences from h, you will need to add further assumptions to h—e.g., laws gov­erning the entities postulated in h, assumptions about the instruments used in testing the consequences, and many others. Therefore, what you are testing by means of these consequences is the entire set of assumptions used to gen­erate them, not any individual assumption in that set.

Why doesn't the holist add that if e is evidence for the set H, then e is evidence for each hypothesis in H? The ho­list doesn't want to do that because that destroys the holist's central idea that it is only systems, not individual hypotheses, that receive confirming or disconfirming evidence. Adding the present idea would turn the doctrine into a type of par­ticularism. More important, the holist, at least in my version of him, wants to restrict the idea of evidence for (or against) a theory to tested consequences derivable from the theory itself. But adding the idea that if e is evidence for a system H and if H contains, or otherwise entails, a hypothesis h, then e is evidence for h, is to accept what Hempel called the “spe­cial consequence condition.” And this is a considerable de­parture from the holist's idea of evidence for h as something derivable from h itself. (e could be derivable from h, and h could entail h', without its being the case that e is derivable from h').

3.