Larry Solum has a typically informative and thought-provoking legal theory lexicon post on “it takes a theory to beat a theory.” (Let’s abbreviate the claim as TBT.) But perhaps he should have said something different — he should have said “this statement is sheer nonsense, and anyone who utters it hasn’t reflected on what it might mean to have, and beat, a theory.”
Well, that’s a little strong. But I’m extremely skeptical of the notion expressed by the sentence, and I don’t think we can even pin down very well what that notion is.
Let’s start with basics. What’s a theory? At the most abstract level, I think we can say that a theory is a way to organize sets of believed-true sentences at a higher level of generality, with the idea being that the theory accounts for those sentences (e.g., by making them consistent, by improving our understanding of them, by predicting future sentences) in some useful way. That will sound like a very unfamiliar notion of what a theory is, and there’s good reason for it to sound unfamiliar: we don’t usually speak of theories in general. Pretty much the only people who speak of theories as such, rather than, e.g., scientific theories, normative theories, etc., are people who think things like TBT. In reality, the various sorts of theories are very different, and may not even fit well into a single well-defined concept. At best, I think we can say that there’s a family resemblance (in Wittgenstein’s sense) between normative, scientific, and conceptual theories.
Scientific theories tend to be lawlike statements about empirical phenomena, which are judged on their simplicity, their consistency with observed phenomena, and their power to predict future observations accurately. (Of course, I’m not trying to give an account of what a theory is here, so please don’t charge me with one, but the idea I’m working with here is something like a looser version of Hempel and Oppenheim’s.) Normative theories tend to be organizing accounts of intuitions and considered judgments, and they are judged on, loosely, their ability to bring about reflective equilibrium. Conceptual theories are organizing accounts, again of intuitions, and of known features of concepts, and are judged on their able to organize those things consistently and (again) their survival in reflective equilibrium. Why should we think that the same requirement for “beating” a theory is applicable across all these different sorts of theory?*
Now let’s talk about what it means to “beat” a theory. There seem to be four possible things that one could mean by “beating” a theory, that arguably apply to all these classes of theory (which is what someone who utters TBT must want, as it’s uttered in all three of the classes of theory I’ve identified, and those who utter it show no recognition of the differences between them).
1. One could show that some other theory dominates that theory, in that it accounts for everything the first theory accounts for, and additional things besides. Paradigm case: the relationship between Einsteinian and Newtonian physics.
2. One could show that a theory fails to accurately account for something — that is, that there is an inconsistency between the set of sentences we take to be true and the theory. Paradigm case: voting and rational choice theory.
3. One could show that the theory is deficient on some other ground by which we judge theories, i.e., that it is insufficiently simple, has no real predictive power, etc. Paradigm case: garbage can theory in organizational sociology.
4. One could show that a theory fails to accurately account for anything — that the theory is strongly false, there is an inconsistency between all (or most, or extraordinarily many) of the sentences we take to be true and the theory. Paradigm cases: lamarckian evolution, phlogiston theory, divine command theory.
Let’s consider each of those senses of “beating” a theory in order.
In the first sense, TBT is true but vacuous. Of course it takes a theory to… be a better theory. Interpreting TBT to refer to that kind of case would be like taking the claim “it takes a dog to beat a dog” to be true in general because it’s true for dogfights.
Not only that, but one might reasonably be skeptical about whether cases of the first type really count as beating a theory at all. Is Newtonian physics “beaten” just because Einstein generalized it? Newtonian physics still works on the day-to-day level…
In the second case, TBT is just false. It doesn’t take a theory to show that an extant theory fails to account for something. It just takes an inconsistent observation.
However, this might be a little unfair to TBT believers because part, I take it, of what TBT is meant to express is that the second sense is not a case of a theory being beaten. (This is Solum’s point about the unacceptability of nit-picking.) Is this right? Well, it’s certainly right in some cases. As Nancy Cartwright has pointed out, fundamental scientific theories are all, strictly speaking, false in this sense, because they all come with ceteris paribus clauses and can never fully capture the complexity of the various influences on phenomena (though I suppose this might cease to be true if they ever unify physics). In such an environment, an anomaly ought not to count as “beating” a theory, just as pointing out the need for more elaboration, and, we might think, a sufficiently comprehensive new elaboration is what would beat the old theory, by replacing it with a new as in the first sense. This, the TBT believers have right. But the TBT believers are wrong to think that there are no other ways to beat a theory without articulating a new one. That is, there’s no dispute that a new theory is sufficient to beat a theory, but is it necessary?
Also, anomalies might not be beaters for scientific theories, but are they beaters for normative and conceptual theories? Often, I think they are. Many, many, theories in the philosophy of action, for example, have foundered on the shoals of their inability to account for akrasia. And any moral theory that commands something that is strongly against a single deeply held intuition (imagine a moral theory that commands us to murder our mothers) is, for that reason, seriously objectionable. Big anomalies can beat normative and conceptual theories, even without a replacement theory.
Consider now the third sense. Here, again, it seems like a replacement theory is sufficient but not necessary. If theory A and theory B explain the same phenomena, but theory A is substantially more general, or simple, or something like that, then, sure, theory A beats theory B. But there are some theories that are so bad that we say they’re beaten because they fail to meet the minimum standards for a good theory. Garbage can theory is like that. For those who haven’t had the pleasure, garbage can theory is the claim that, in organizations, problems, solutions, and people are matched basically at random. Garbage can theory is aptly named, for it belongs in the garbage can — it’s absolutely scientifically useless for the very simple reason that it has no predictive power whatsoever. Garbage can theory predicts random noise. You simply cannot do science that way, by having theories with the sole prediction “shit happens.” We don’t need a theory to beat garbage can theory. All we need do is point at its worthlessness.
Now let’s consider the fourth sense. Yet again, I’d say that a replacement theory is sufficient, but not necessary, to beat a theory. Consider the paradigm cases I mentioned. Lamarckian evolution was beaten by a better theory, viz., Darwinian evolution. However, phlogiston theory didn’t need a new theory to refute it — it got smashed by the fact that it consistently made false predictions about the mass of burned objects. Admittedly, it didn’t die completely until Lavoisier discovered oxygen’s role, so perhaps TBT believers can take some solace from that? I don’t think that’s quite right, since it was refuted beforehand, whether the scientific community recognized that fact or not — at a certain level of constant wrong prediction, a theory has simply got to die. But let’s give TBT the point, arguendo.
Let’s shift back to normative theories and consider the third paradigm case. Divine command theory is, for all intents and purposes, completely dead, refuted, utterly slaughtered. There are two independent grounds on which I so confidently proclaim its demise. 1) The nonexistence of god. 2) the Euthyphro problem. Each of those is totally sufficient on its own to refute divine command theory, but neither is a theory on its own. The first is a simple negative existential claim. The second is a simple argument. One doesn’t need an alternative theory of the sources of normativity to refute divine command theory, it is refuted because falsity pervades everything about it.
In sum, in the first sense of beating a theory, TBT is true but vacuous. In all the others, it’s false. And to the extent it fails to distinguish between wildly different types of theory (scientific, conceptual, normative) it risks incoherency. Never let its utterance go unchallenged.
* There are other sorts of theory too, of course. For example, there are literary and hermeneutic theories. I don’t know enough about those to be able to say what it might take to “beat” them. I’ll bet it’s different from what it would take to beat a scientific, conceptual, or normative theory, though.