Objectivist Epistemology: Strengths and Weaknesses
(Summer 1999; last revised, August 2001)


1. Where Objectivist Epistemology is Right

First, some good things: Objectivist epistemology aspires to be realist, non-skeptical, non-authoritarian and non-mystical. In my view, these are all good things. However, they do little to distingush Objectivism from much other work in epistemology. If there is something distinctively good about Objectivist epistemology, it must be sought in the more detailed working out of the themes.

Two more detailed areas in which I think it is also correct are its rejection of the sharp analytic-synthetic distinction and its adoption of direct realism with regard to sensory perception. However, these also do not amount to anything especially distinctive about Objectivism – at least, not distinctively good. Let me elaborate.


1a. Analytic-Synthetic

Rejection of the sharp analytic-synthetic distinction is reasonably common among contemporary epistemologists and may be the majority view. (This is impressionistic – I haven’t done a head-count.) It is certainly not something that Objectivists have a special claim to. In my view, the distinction was blown out of the water close to fifty years ago by the work of people like Quine, Goodman and White. Since then, many epistemologists have come to agree.

The main treatment of it in Objectivist literature is Peikoff’s essay from the mid-60s that appears in ITOE. Apart from the fact that it was something of a late-comer to the debate, it is disgracefully bad. One indication of how bad it is can be found in Peikoff’s assertion that the distinction “is accepted, in some form, by virtually every influential contemporary philosopher.? (ITOE, p. 89) Making a claim like that – at the time he made it – bespeaks stupidity, incompetence or dishonesty. It simply was not true in the mid-60s that it was accepted by virtually every influential contemporary philosopher. The only “contemporary? philosophers whom he cites as endorsing the distinction are Wittgenstein (who wrote the material quoted around 1918 and later radically revised his epistemological views) and Ayer (writing in the mid-30s). (ITOE, p. 94) Had Peikoff cited more recent figures, he would hardly have been able to avoid mentioning people like Quine and Goodman as influential contemporary philosophers who did not accept the distinction. A bit of relevant background: When Peikoff wrote, Quine was probably the single most influential contemporary American philosopher, and he had rejected the distinction in his most influential article, “Two Dogmas of Empiricism? – which was delivered as a paper at a conference in 1950, published in a journal in 1951, and reprinted inFrom a Logical Point of View in 1953. Since then, it has been reprinted in numerous collections, and there is – and already was in the 60s, when Peikoff wrote – a mass of literature discussing both the credibility of the distinction and Quine’s article in particular. Since Peikoff was in graduate school in philosophy in the early 60s, the case is quite strong that he was deliberately dishonest.

Perhaps all that might be excused if Peikoff had really done anything important in the way of developing, elucidating or clarifying the argument. In fact, he did not. He runs together several different distinctions – including analytic vs. synthetic, a priori vs. a posteriori, necessary vs. contingent, and certain vs. uncertain – and muddles them all together. Further, his actual argument against the distinction is irretrievably weak (for some details, see below). Though I think the sharp analytic-synthetic distinction should be rejected, no reasonably well-informed philosopher who defended the distinction would have found Peikoff’s arguments persuasive. If Peikoff knew what he was doing, he was dishonest; if he didn’t, he was incompetent.


1b. Direct Realism

Objectivism’s direct realism about perception can, thankfully, be discussed much more briefly. It is worked out reasonably well and at a philosophically competent level in David Kelley’s The Evidence of the Senses.  This is the best work I’ve seen by a philosopher who acknowledges being an Objectivist. Though there are debatable points in it, it does not egregiously pursue straw men, demonize philosophers who disagree or try to pass off good rhetoric as good argument. Moreover, the writing is graceful, clear and mercifully free of Objectivese.

I also think direct realism is correct. The reason for bringing it up is just to point out that it (again!) is hardly distinctive of Objectivism. Direct realist theories are quite popular among epistemologists working on perception and have been, increasingly, over the last half-century or so. The version that Kelley works out may ring minor changes on the theme, but it does not amount to a major contribution. The Objectivist position on perception is a minor variant in a popular family of theories: It’s no big deal. Anthony Quinton’s 1955 article, “The Problem of Perception,? for example, which I read about the same time as Kelley's book, develops a closely parallel account. On some issues, Quinton was much the clearer of the two.


2. Where Objectivist Epistemology is Weak

Chris Roberson has pointed especially to the theory-ladenness of observation and the under-determination of theory by evidence as areas in which Objectivist epistemology, especially as it applies to the sciences, is weak. I agree with him that these are troublesome areas, and that Objectivists haven’t adequately addressed them, but I want to mention a couple of others.


2a. Counterfactuals

Objectivists seem allergic to considering hypothetical or counterfactual cases. They are apt to accuse anyone who employs arguments involving counterfactuals of being “rationalistic? or “not grounded in reality.? Apparently, they think this charge, which is usually unexplained, is sufficient to get them off the hook of having to actually address those arguments. The arguments are dismissed as arbitrary or meaningless or some such thing. Oddly, I find little warrant for this particular bit of lunacy in Rand’s own writings. (Is she to be accused of rationalism when she talks about indestructible robots?) Nonetheless, the attitude is quite common among Objectivists.

I could speculate that this is defensive maneuver to keep from having to deal with hard questions that Objectivism is ill-equipped to handle, but, for now, I will place little weight upon that speculation. I will confine myself to pointing out that reference to counterfactuals is nearly unavoidable if you want to think clearly about issues in philosophy of science such as causation and laws of nature. The claim, for example, that it is a law of nature that (say) unsupported rocks fall to the ground does not just report a regularity in the behavior of rocks. It also implies things like “if this rock were unsupported (which it isn’t), then it would fall? – and that can be true even if the rock never is unsupported. Note that this is not just a material conditional (which is always true when the antecedent – the if-clause – is false) because we also want to deny claims like “if this rock were unsupported (which it isn’t), then it would turn into a bird and fly away.?

A very similar point can be made about causation. To say that an event, A, causes another event, B, is not just to report that A is followed by B, but also implies (if there are no other causes of B or other possibilities for the occurrence of B about) that if A had not occurred, B would not have occurred.

Whether we speak of laws of nature or of causation, an adequate understanding of what is meant is not possible without reference to situations that do not actually occur – that is, without reference to counterfactuals. If we refuse to deal with counterfactuals, then the most that we will be able to manage in this area is some kind of Humean regularity analysis of causation and of laws of nature.

Once we admit counterfactuals, as we must to address important issues in the philosophy of science, it will be much harder for Objectivists to dismiss other arguments as merely hypothetical or rationalistic word-spinning. They will have to offer specific reasons that particular arguments that appeal to counterfactual states of affairs go wrong (if they do). General objections to counterfactual arguments won’t do the job.


2b. Induction

Another and related area in which Objectivist epistemology is weak has to do with understanding induction – and this despite the fact that Objectivist polemic is filled with references to the importance of understanding the philosophy as inductive!

Philosophical work on induction can be divided more or less neatly into two not wholly independent parts. On one level, we can consider the kinds of inductive reasoning in which people actually engage and attempt to elicit common features of good inductive reasoning – that is, of inductive reasoning that we take to be reliable or more likely to be reliable – including canons of simplicity, accuracy, experimental design and so on, as well as trying to identify inductive fallacies, ways in which inductive reasoning is liable to go wrong. So far as I know, Objectivists have done very little on this, but I do not wish to focus upon it.

On another level, what is commonly called the problem of induction has to do with how inductive reasoning is justified or legitimate at all. Basically, the problem of induction is one of justifying inference from experienced to unexperienced cases or, as I prefer to put it, from sample to population. Why is that a problem? First, statements about the whole population don’t deductively follow from statements about the sample (except for trivial cases, e.g., from “all the ducks I’ve seen were in Central Park? it follows that some ducks – that is, some members of the population of ducks – were in Central Park). This is what’s meant by saying that induction is non-demonstrative reasoning – it’s not just a special case of deduction or “demonstration.? Hence, it’s not contradictory to suppose that cases outside your sample will be different from cases inside your sample. So, what could justify you in thinking that it’s true or probable that the population resembles the sample in the respect you’ve identified? It seems that what would be needed is something like “populations always (or usually) resemble samples.? But then comes the big question: You haven’t investigated all or even most cases of alleged resemblance between sample and population – so how would you justify “populations always (or usually) resemble samples?? That’s another case of reasoning from a sample, namely, the cases of resemblance you’ve investigated, to a population, namely, all the cases of resemblance or non-resemblance that there are. So, you could have a justification of induction only if you were already justified in inductive reasoning.

Now, induction is a hard problem and one to which there is no generally accepted solution. Accordingly, I would not be especially inclined to criticize Objectivists for failing to solve it were it not that Rand speaks so contemptuously of the failure of other philosophers “to prove the validity of scientific induction.? (FNI, p. 30)

The problem, of course, is that she didn’t solve it either – nor did she even come close. Judging from her fragmentary comments on it, it’s hard to believe she even understood the problem.


3. Where Objectivist Epistemology is Wrong

3a. Foundationalism

Objectivist epistemology is a version of foundationalism, one of a number of views that holds that knowledge has foundations, that there are privileged starting points for knowledge, that justification runs uni-directionally from foundations to superstructure, that nothing is justified unless it is connected in the right way to the foundations and that nothing “superstructural? can count against or provide a reason for revising or modifying anything taken to be foundational.

I consider that deeply confused and, though most foundationalists did not so intend, an invitation to skepticism. Since I’ve discussed foundationalism elsewhere (in Foundationalism, Skepticism, Coherentism), I won’t pursue it here.


3b. The Confusion of Meaning with Reference

Meaning is confused with reference. The meaning of a concept is supposed to be the existents it refers to. This leads to or encourages several further mistakes.

It helps support the false claim that a concept must be based on two or more instances. That would rule out the possibility of a concept of something that is or may be unique, for example, a black hole if there turned out to be only one, or the origin of life on earth. It also runs into problems with theoretical concepts from the sciences, such as black holes (again) or neutrinos, which were formulated before any instances were known.

A different problem is with concepts that have no referents. We do not, for example, mean the same thing by “centaur? and “mermaid.? However, if meaning were the same thing as reference, we should conclude either that they have the same meaning (because they have the same referents, i.e., none) – which is absurd – or else that they have no meaning (because they have no referents) – which is likewise absurd.

Nor is meaning the same as reference when there are referents. “The animal that, species-typically, has the highest brain-to-body-mass ratio? has the same reference as “human being? but not the same meaning. We could discover, contrary to current scientific opinion, that some other animal had the highest, species-typical, brain-to-body-mass ratio, but that discovery, if made, would not be a discovery that those animals were human beings or that we are not.

This confusion of meaning with reference is also at the core of Peikoff’s argument against the analytic-synthetic distinction. Since one way of expressing that distinction is by way of the claim that some truths, the ones that are analytic, are true by virtue of their meaning alone, while others, the synthetic truths, are not, but depend also upon the referents of synthetic claims (compare, e.g., ‘all bachelors are unmarried’ and ‘some bachelors are lonely’), identifying meaning with reference surely rules out the possibility of an analytic-synthetic distinction. If, however, that were the only way to avoid the sharp analytic-synthetic distinction, then the right thing to do would be to accept the distinction, since the identification of meaning with reference is demonstrably untenable. (In fact, it is not the only way to avoid the distinction, and there are good reasons for rejecting it, but those reasons are at best obscured by the standard Objectivist approach.)


3c. The Confusion of Universals with Concepts

Rand confused universals with concepts. Roughly speaking, universals are features in the world which may be multiply exemplified, while concepts are our means of classifying things. The neatest and simplest connection between them would be if a correctly formed concept always picked out a universal, but there is little reason to suppose that’s true, and it’s entirely possible to have a workable system of classification that fails to pick out universals (either sometimes or perhaps ever). This might be because the real universals are beyond our cognitive grasp (suppose we would have to think in terms of a 26-dimensional space in order to get any universal exactly right) or because there just aren’t any (maybe, every generalization we can formulate has a few exceptions – and any rule we can formulate about the exceptions also has exceptions, and so on!).

Whatever the right answer, you need to distinguish concepts from universals to get at it – and Rand didn’t.

Rob Bass

Comments? I'd love to hear.