Elliott Sober’s Contrastive Empiricism

There has traditionally been a tension between empiricism and scientific realism.  Elliott Sober aims to combine important elements of the two in his Contrastive Empiricism.   In his paper “Contrastive Empiricism”, Sober defends a modest realism where “science properly aims to identify true theories about the world.”  Sober’s Contrastive Empiricism can be contrasted with Van Fraassen’s Constructive Empiricism where the goal of science is to find theories that are empirically adequate.

To give an example of the reasoning behind Contrastive Empiricism, Sober applies it to the no miracles argument.  Before Putnam changed his mind, he presented a common argument for scientific realism: “realism is the only philosophy that doesn’t make the success of science a miracle.”  Sober uses Bayesian reasoning to show how this argument doesn’t work.   Testing is a contrastive activity between a theory and its competitors. Consider:

  • Observation: Theory T is predictively successful.
  • Hypothesis 1 (scientific realism): Theory T is true or approximately true.
  • Hypothesis 2 (constructive empiricism): Theory T is empirically adequate, though false.


  • Bayes’s theorem: Pr (H/O) = Pr (O/H)Pr (H)/Pr (O)

It follows from Bayes’s theorem that Pr(H1/O) > Pr(H2/O)—that is, we should choose H1 over H2 given O–when Pr (O/H1)Pr(H1) > Pr (O/H2)Pr(H2).  In this case Pr(O/H1) = Pr(O/H2), since they are empirically equivalent, so the only thing we can appeal to are the prior probabilities Pr(H1) and Pr(H2).  But this means that observation is irrelevant and the no miracles argument becomes an a priori argument.  Sober doesn’t think a priori arguments of this type are justifiable.

Contrastive empiricism states:

For any two theories T and T’, if it is possible to say that T is more plausible than T’, then this will be because there exists a set of observation statements such that T and T’ make incompatible predictions as to which members of that set are true.

The main departure that Contrastive Empiricism makes from previous Empiricisms, including both Logical Empiricism and Constructive Empiricism, is that it is about problems, not propositions.  Previous empiricisms, as I’ve said, have tried to discriminate one set of statements from another. Van Fraassen, like earlier empiricists, wants to say that science ought to treat some statements differently from others. Contrastive empiricism draws no such distinction. Rather, it states that science is not in the business of discriminating between empirically equivalent hypotheses.

Here Sober is referring to the fact that he doesn’t distinguish between observables and unobservables and their relation to meaning.

Sober gives two examples to illustrate his synthesis of empiricism and scientific realism in Contrastive Empiricism.

  • (X1) There is a printed page before me.
  • (X2) There is no printed page before me; rather, an evil demon makes it seem as if there is one there.
  • (Y1) Space-time is curved.
  • (Y2) Space-time is not curved; rather, a universal force makes it seem as if it is curved.

Empiricist aspect of Contrastive Empiricism: Science can not discriminate between the empirically equivalent propositions of (X1) and (X2) (and between (Y1) and (Y2)).  On Bayesian reasoning, any difference between empirically equivalent hypotheses would have to come from different prior probabilities.   On empiricist principles no justification can be given for different prior probabilities of empirically equivalent propositions.  This means Descartes evil demon hypothesis is equally confirmed with the regular non-demon hypothesis.  Justification of the non-demon hypothesis by appeals to parsimony is ungrounded.

Realist aspect of Contrastive Empiricism: There is no difference in the treatment of observable (X1) and the theoretical (Y1).

Some thoughts: I find it hard to reject the empirically equivalent hypothesis of Descartes’ demon by appeals to parsimony.  It should be noted that Sober doesn’t reject all uses of parsimony (see his “Prediction vs Accomodation).  For more on contrastive testing, readers should look to the second chapter in Sober’s book “Evidence and Evolution.”

Sober, Elliott. Contrastive Empiricism.

This entry was posted in Philosophy of Science. Bookmark the permalink.

1 Response to Elliott Sober’s Contrastive Empiricism

  1. tyrel771 says:

    Excellent article; succinct, clear and useful. Thanks for it.

    I have a question to throw your way, perhaps just a note of clarification. I wonder why two hypotheses H1 and H2 cannot be said to fail to be empirically equivalent if, all other things being equal, H1 is more parsimonious than H2. Perhaps the intuition here is that parsimony is pragmatically useful, but that it has no predictive/explanatory power. That, in other words, something’s being a more parsimonious explanation does nothing at all to make it more likely to be true. I have two thoughts about this though. First, it seems prima facie as though parsimony is a good reason, ceteris paribus, to think that one story is more likely true than some rival. Even in metaphysics Occam’s razor should raise hairs. Thus, if some view denies that prima facie intuition then it plausibly loses some degree of plausibility for itself which it may otherwise have had. Second, on some metaphysical theses, like that of certain versions of Theism, doesn’t it seem as though we might very well have good reason to think that parsimony is, ceteris paribus, a feature of a hypothesis given which it is more plausibly true than it otherwise would have been?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s