THE “MULTIPLICITY” OBJECTION

A second objection is that, given Perrin's experimental results, various hypotheses can be shown to receive just as much support as the particular molecular one that he accepts. And these alternatives need not be com­mitted to the existence of molecules.

A. Parallel Antirealist Argument

The idea, a generalization of one offered by Bas van Fraassen,¹³ is that, for each argument used by Perrin whose conclusion is that molecules exist, another argument, at least as good, can be constructed from the same premises which does not conclude that molecules exist, but only that the molecular theory “saves the phenomena.” Indeed, an argument of the latter sort is simpler and stronger than the former, since it commits one to much less than does any argument of the former sort. It commits one only to the “empirical adequacy” of the molecular theory, not to its truth, whereas an argument of the former sort is committed to both.

For example, an eliminative-causal argument of the sort attributed to Perrin in section 2 is replaced by the following:

Antirealist eliminative-causal argument

(1 ) Given what is known, of the theories that attribute a cause for effect E (for example, Brownian motion), the possible candidates for theories that save the phenomena are ones that invoke causes C, C₁,...,C. (In probabilistic terms, given what is known, the probability is high that at least one of the theories cited saves the phenomena.)

(2) Theories that invoke causes C₁,...,C_> do not save the phenomena (since these theories predict that, under certain observable conditions, effect E will continue, which new information shows to be false).

So probably,

(3) The theory that invokes cause C (for example, molecules) saves the phenomena.

The conclusion is not that the theory that claims that C causes E is true, or that the entities it postulates (for example, molecules) exist or cause E, but simply that it is empirically adequate.

A similar claim is made for any argument that concludes with the truth of a theory postulating the existence of unobservables. For example, with respect to parallel “inference-to-the-best-explanation” arguments, van Fraassen offers two objections (pp. 20-21). The first is that the claim that scientists reason in accordance with realist versions of such arguments (or analogous eliminative-causal ones) is an empirical hypothesis “to be confronted with data, and with rival hypotheses” (p. 20). Van Fraassen proposes a rival hypothesis, namely, that scientists reason in accordance with the antirealist version of inference-to-the-best-explanation (or an analogous version of the eliminative-causal argument given previously). Both hypotheses, he seems to be saying, are compatible with the empirical fact that scientists employ explanatory (or eliminative-causal) reasoning. Van Fraassen's second objection is that the realist needs an extra premise for his argument, one that van Fraassen regards as false. For inference-to- the-best-explanation, the extra premise is that “every universal regularity in nature needs an explanation” (p. 21). For causal reasoning, the extra premise would be that every phenomenon in nature has a cause.

Reply

(1) To be sure, the fact that Perrin invokes eliminative-causal reasoning does not by itself establish that he is employing a realist or an antirealist version of such reasoning. Here one needs to examine what he in fact says. His claim is not (simply) that of various theories that attribute a cause for Brownian motion, the ones he cites are the possible candidates for saving the phenomena; it is the stronger claim that these theories are the pos­sible candidates for truth. Nor is his conclusion (simply) that the theory invoking molecules as causes in fact saves the phenomena. It is the stron­ger claim that such a theory is true, that molecules exist (“the objective reality of the molecules therefore becomes hard to deny”), and that mol­ecules cause Brownian motion.

(2) Perrin does not need an extra premise asserting that every phenom­enon in nature has a cause (or that every universal regularity needs an explanation). All he needs is the assumption that Brownian motion has a cause. To be sure, he does not explicitly defend this assumption, although he clearly makes it.^[121] Presumably, if required to defend this assumption Perrin would have appealed to the idea that Brownian motion involves accelerations of bodies—both changes in speed and direction—which, in the classical physics that he was assuming—require causes. Perrin had no need to introduce some very general assumption that every universal reg­ularity in nature has a cause or needs an explanation (an assumption not even made in classical mechanics). He simply needed to assume that the Brownian motion of the observable particles of gamboge has a cause.

B. Multiplicity-of-Causes Objection

The idea behind this second form of the multiplicity objection is to attack Perrin's particular eliminative-causal argument, which assumes that, given that known observable causes of Brownian motion are eliminated, one can infer an unobservable cause. The problem is that not all possible observable causes have been considered. The objection could be raised as a general criticism of eliminative-causal reasoning (“In general, how do you know you have cited all possible causes of a phenomenon?”). Or it could be raised as a specific one against Perrin (“What reason was there for supposing that Perrin had considered all possible causes of Brownian motion?”).

Reply

(1) The general criticism is based on the assumption that one is justified in employing an eliminative-causal argument only if all but one of the pos­sible causes of the phenomenon have been listed and eliminated.

(2) The specific criticism is that, even if eliminative-causal reasoning can be reasonable, if properly employed, Perrin did not properly employ it. He did not cite all the (plausible) causal candidates for Brownian motion, given information available to him. To be sure, he cites several candidates and argues that experiments eliminate these. But why suppose that these are the only possible or plausible causes permitted by his background information? And even if they are the only possible observable causes, why suppose that the motion of particles is the only possible unobservable cause?

Here, I suggest, the burden of proof is on the critic. Perrin cities various known causes of motion in a fluid and argues that experiments, particu-

15. Wiener demonstrated that Brownian motion is not caused by infusoria (one-celled animals found in exposed bodies of water), by electrical forces, by temperature differences, or by evaporation of the fluid. In addition to these, Gouy’s experiments eliminated causes pertaining to the size, composition, and density of the Brownian particles. For a discussion of various observable causes of Brownian motion postulated by scientists opposed to molecular theory, see Nye, Molecular Reality (London: Macdonald, 1972), pp. 22-27. Nye concludes (p. 27) that in his experiments, “Gouy [in 1889] refuted, point by point, all previous theories of Brownian movement other than the [molecular-] kinetic.”

16.

larly those of Christian Wiener in 1863 and Gouy in 1888 and 1889, show that Brownian motion continues unabated when these causes are altered or eliminated.¹⁵ Perrin, quoting Wiener, concludes that the motion “does not originate either in the particles themselves or in any cause external to the liquid, but must be attributed to internal movements.”¹⁶ Since Perrin cited and eliminated various possible causes, it is, I think, up to the critic to say what other possible causes he should have eliminated, given his information (and why these should be assumed to be observable).

(3) Both the general and the specific criticisms cited here, if valid, could equally well be used against the antirealist version of the eliminative- causal argument. That version cites theories that invoke causes C, C_{,...,C_n as the possible candidates for saving the phenomena and then eliminates all but one of these theories. But, the critic can ask, why sup­pose that these are the only possible causal theories that will save the phenomena? And the critic can ask Perrin the specific form of this ques­tion for Brownian motion. So an antirealist version of eliminative-causal reasoning is no better off than the realist version. My response is simply to reject the critic's claims for the reasons given in replies 1 and 2.

4.