Lecture Notes, UC Davis Philosophy 102, Theory of Knowledge

Naturalized Epistemology

G. J. Mattey, Senior Lecturer

Jaegwon Kim

Text: Jaegwon Kim, "What is 'Naturalized Epistemology'?" (1988)

In the last lecture, we saw that Fumerton objected to any attempt by externalists to analyze justification or knowledge in terms of causal laws. He claimed that epistemic concepts are not definable at all.

Jaegwon Kim, professor at Brown University, looks for an acceptable way to understand epistemic concepts naturalistically, which puts him at odds with Fumerton. Kim approaches the issue through the lens of a famous 1969 paper by W. V. O. Quine, "Epistemology Naturalized." In that paper, Quine had argued that theory of knowledge should be treated as a branch of empirical psychology.

The Two-fold Task of Epistemology

Kim begins by reciting the "textbook history of philosophy," which he says helps explain the present tasks in theory of knowledge. The history begins with Descartes's First Meditation, where he asks the question "What propositions are worthy of belief?" In Kim's terminology, this question concerns the "criteria of justified belief." Philosophers ever since Descartes, including many that we have already studied, have been searching for appropriate criteria to regulate what we should believe.

The second question asked by Descartes, according to our "textbook history," is "to determine what we may be said to know according to those criteria." Kim re-describes the project as that of "coming to terms with the skeptical challenge to the possibility of knowledge." It is not clear that these are the same thing.

Suppose we follow Descartes himself and say that those propositions worthy of our belief are the ones which are backed up by clear and distinct perception. Then the only question remaining is which propositions have the proper backing. For Descartes, these would be propositions such as that God exists, that God is no deceiver, and that a material world exists.

In my own view, the skeptical challenge comes earlier. It is a challenge to our ability to find criteria of justified belief in the first place. If the criterion is to be clear and distinct perception, then the skeptic asks why beliefs conforming to that criterion are justified. This is, again in my own view, the toughest challenge for Descartes's whole program.

It seems that what Kim has in mind as the "textbook" account is that the criterion is indubitability. If so, then showing which beliefs are justified does require coming to terms with the skeptical challenge to the possibility of knowledge. Any candidate belief could be challenged by a skeptic as being subject to some doubt, no matter how slight it may be. This is why overcoming the evil demon possibility is considered such a great obstacle to developing a theory of knowledge given this criterion.

Bringing these two themes together, the problem for the "real" Descartes is connecting the criterion of justification, backing by clear and distinct perception, with demand for indubitability. Descartes must show that what meets his criterion is not subject to doubt. The fundamental problem for his approach is that he needs to prove that God exists in order to show this, but to do so he needs to rely on some beliefs that are justified according to his criterion. This is the "Cartesian circle."

Criteria of Justified Belief

Note that Descartes's program does not call for an "analysis" of knowledge. Descartes indeed operated with assumptions about what knowledge is. But he was more interested in whether we know and how it is that we know. We know, he claimed, when our belief is based on clear and distinct perceptions. And many of our beliefs are backed up in this way.

Descartes did not say in the Meditations what a clear and distinct perception is. But it can be safely said that his account of clear and distinct perception meets the condition set up by Kim, that it be "stated without the use of epistemic terms." Roughly, clear and distinct perception exists when there is an unimpeded mental vision, by a properly-functioning and attentive mind, of an idea. There are no epistemic terms such as "justified," "ground," "reason," or "doubt" lurking in this account.

This meets Kim's first condition for a criterion of justified belief. But it does not meet his second, that it contain no normative or evaluative terms. For the term "properly-functioning" is normative. It is an essential part of the criterion, however. A mental vision by a dysfunctional mind could well lead to a "clear and distinct perception" that I am made of glass, to adapt one of Descartes's own examples.

Kim insists that what is needed in giving a criterion of justified belief is that it be "formulated on the basis of descriptive or naturalistic terms alone." He grants that a description using normative concepts may be informative but "it will not do," unless the normative terms themselves can be understood in non-normative terms.

In fact, Descartes understood the condition of proper functioning in descriptive terms. It is functioning according to God's "design plan," as Alvin Plantinga might put it. A problem for this approach is that one ought first to know that there is a God who has a design plan, and this can only be established if we already have the criterion in hand: the Cartesian circle again.

Chicken and Egg

With this constraint (which has yet to be defended) in hand, we can ask about the relationship between the criterion of justified belief and what is known. Which one should be established first? If you set down a criterion which dictates what is known, you run the risk of skepticism. It may be that the criterion cannot be met in most, nearly all, or all cases. On the other hand, if you tailor the criterion to what you assume that we know (which most theorists do nowadays), then it may be that some kind of question is begged against skepticism.


Whatever the history of the concept, justification is central to theory of knowledge today. One explanation is that it is the only epistemic concept in the "traditional" analysis of knowledge as justified true belief. Truth and belief, that is, are not epistemic, unless they are somehow linked in the background with the concept of justification or at least with something normative. Justification itself "is manifestly normative," and it is what makes knowledge normative as well.

To see this, note that a justified belief is one that is "permissible" and "reasonable for us to hold, from the epistemic point of view. And it would be "epistemically irresponsible" to hold beliefs contradicting what is justified. We could think of justification as vindicating an action, that of believing. Epistemology is a normative discipline like ethics.

To summarize:

Kim takes it that these general points are uncontroversial, though there may be dispute about details.

The Foundationalist Strategy

Epistemologists have tended to follow the approach of Descartes when carrying out the program of giving conditions under belief is justified. The "textbook" version of Descartes is that he was a foundationalist. Some beliefs, the basic ones, are justified directly, while others, the non-basic ones, are justified indirectly or "inferentially" "by standing in an appropriate relation to those already justified."

The directly justified beliefs include beliefs "about our own present conscious state." They are justified, without help from other beliefs, because they are indubitable, "that the attentive and reflective mind cannot but assent to them." Indirectly justified beliefs are "deduced" by a series of inferential steps. At each step there is an "intuition" that is indubitable.

We can take indubitability to be a non-normative concept. That is, it could be purely descriptive of a person's psychological state that doubt is impossible. On this interpretation, Descartes would have given a naturalistic account of justification that meets Kim's stronger condition on such an account.

Earlier, I noted that clear and distinct perception is another candidate for the criterion of justification for Descartes. This looks like the best candidate given the following consideration. The psychological inability to doubt, while it is naturalistic, is not a good criterion for justification. One may be unable to doubt because one is brainwashed, for example. This can be fixed by requiring that there is no reason for doubt. But then "reason" is an epistemic term, and so this substitute description of indubitability does not meet Kim's stronger condition. So, it is perhaps best to stick with clear and distinct perception as a criterion.

The empiricists generally inherited this strategy, particularly the "mentalism," i.e., the claim that foundational beliefs are about one's own present conscious states. This approach had not been seriously challenged until the twentieth century. The chief disagreements were over whether anything besides beliefs about conscious states are basic and how to view the inferential move from the basic beliefs to the non-basic beliefs.

The "Logical Positivists" of the early twentieth century were foundationalists in this way, though they came eventually to reject mentalism. But the basic beliefs are still "observational." Moreover, the tie to observation is required not only for justified belief, but for the very meaning of terms. So Kim calls them "foundationalists twice over."

"Epistemology Naturalized"

Epistemologists today who call themselves "naturalists" generally point to Quine as their inspiration. Quine's paper "Epistemology Naturalized" contains a unique argument in favor of naturalism. The first step is to say that Descartes's "quest for certainty" is hopeless. The second, original, step is to infer from this that the project of normative epistemology itself is hopeless and should be replaced by a naturalized epistemology. The connection between the two is not obvious, but it is "something that any serious student of epistemology must contend with."

The Positivist Project

Quine criticizes not Descartes's own project, but rather the project of some of the Logical Positivists. Let us call this the "positivist project." He apparently thought that the positivist project was the most promising development of the Cartesian project, presumably because it is not rationalistic and does not require appeal to God.

The positivist project is then divided into two parts. First is the "conceptual reduction," in which theoretical terms of science are defined by purely observational terms. Second is the "doctrinal reduction," in which truths about the physical world "are appropriately obtained from truths about sensory experience." What makes the connection appropriate is at least that it preserves truth, so the relation ought to be one of deduction. "Quine assumes that if deduction doesn't fill their bill, nothing will."

Both reductions are failures, according to Quine. No one has succeeded in carrying out the conceptual reduction, and no one ever will. The reason is that meaning is "holistic." Quine once said that "the unit of empirical significance is the whole of science" ("Two Dogmas of Empiricism"). But definition proceeds on a term-by-term basis, as in a dictionary.

Hume had shown that the doctrinal reduction is impossible. The evidence of our internal state will always "under-determine" beliefs about the physical world. It is always possible to imagine being in a psychological state without any particular physical state holding.

The Successor Project

Kim rightly notes that this is no surprise to anyone. It had been accepted for quite a while that the positivist project ("translational phenomenalism") was a failure. But no one else had drawn the extreme consequences of this failure that Quine did. He said that we should give up trying to "validate" empirical science and instead use empirical science, specifically psychology, to see how sensory stimulation actually gives rise to a picture of the world.

This is not circular reasoning, for we take it for granted that science is valid, since "we are well advised to use any available information, including that provided by the very science whose link with observation we are seeking to understand" (Quine, quoted from "Epistemology Naturalized").What we get from this is a successor to classical epistemology: naturalized epistemology.

Kim rightly notes that this response by Quine overlooks alternative successors to the positivist project, ones that remain within the basic framework of classical epistemology. We could replace the foundational structure with one of coherence. We could reduce the level of certainty of basic beliefs. We could find a logical relation weaker than deduction, perhaps probability, to supplement deduction. Or, like Chisholm, we could develop special "principles of evidence."

The main difference between the classical program and Quine's program is the former is supposed to be "validating" while the latter is merely descriptive. The culprit, for Quine, is the normative character of the classical program, though he does not put it this way. "Epistemology is to get out of the business of justification."

Is This Still Epistemology?

If justification is taken out of episteimolgy, knowledge goes with it. The reason is that the two are tied "inseparably," since knowledge is "a normative notion." So, "Quine's nonnormative, naturalized epistemology has no room for our concept of knowledge." Quine even replaces talk of "knowledge" with that of "science," "theories," and "representations." Instead of justification of beliefs, he talks of stimulation leading to theories and representations, which Kim takes to correspond to beliefs or belief-systems.

The "leading" should be understood as a causal or nomological process. We begin with "meager input" from sensory stimulation and end with "torrential output" of theory. Quine wants to see what causal process is involved, not how the input "justifies" the output.

But although Quine pays lip-service to the idea that "evidence relates to theory," he is not talking about the evidential relation, i.e., how a theory is justified. One way to see this is that because this is a causal process, it will vary from species to species, while the evidential relation does not vary. All it is concerned with is "the degree to which evidence supports hypotheses." And the concept of evidence is inseparable from justification. Kim claims that "we simply do not understand" a concept of evidence which is strictly non-normative.

It is not as though it is not valuable to study the causal process Quine has identified. The question is whether this is still doing epistemology. Why not make the study in addition to doing classical epistemology? Quine's answer is that the project is misguided, and we would be better off replacing it with a non-normative process. Kim's objection is that there is no sense to be made of "replacing" one project with another, since they concern two different things.

Belief Attribution and Rationality

Perhaps it could be argued that there is something in common between the two projects. Both could be said to be about "beliefs" or "representations." The classical project seeks to understand how they can be supported by evidence, while the naturalistic project seeks to understand how they come about. Then it can be said that the naturalistic project is feasible, while the classical project is not.

If we are to make the comparison, we would have to understand "belief" as the output, the "theory" or "picture of the world" which is the response to the stimulation of the senses. But the raw, so to speak, output of sensory stimulation is a torrent of words. These words have to be interpreted as beliefs if we are to take belief as the ultimate output. This requires the use of an "interpretive theory" that not only assigns beliefs but also assigns meanings to the words that come out of a person's mouth, are written down, etc.

An interpretive theory could never get off the ground unless we postulate that the output is "largely and essentially rational and coherent." There is no other basis for thinking that what we hear or see on paper is belief and not some other kind of output. This point was emphasized by Donald Davidson. Rationality and coherence, in turn, are regulated by norms of evidence and justification. Any attribution of belief requires that the output in question be evaluated according to standards of evidence and justification. So any investigation of belief essentially involves the project of classical epistemology.

Kim considers two possible objections to this anti-Quinean argument. The first is that normative considerations only bear on the question of when to attribute a belief to a subject. But this does not make the concept of belief normative. Kim dismisses this objection on the grounds that "the point has already been made." The Quinean project requires that we find out what beliefs a person has, and this in turn requires embedding the output in a framework of rationality and coherence.

The second objection is that we can think of the output not as beliefs, with propositional content, but instead in some neutral way. Kim suggests that they could be taken as "appropriate neural states" of an organism. Then the normative element would be eliminated altogether. But which neural states are "appropriate?" Either they are those which are identified with beliefs, in which case the objection is moot, or else they are eliminated entirely. In the latter case, "it is uncear in what sense we aer left with an inquiry that has anything to do with knowledge."

The New Naturalism

Quine is not the only philosopher to advocate that the classical program in epistemology be abandoned in favor of a more "naturalistic" or "scientific" approach. People of this type generally criticize the traditional epistemology as "aprioristic" and out of touch with the psychological and evolutionary facts about the human organism. My own colleague, Robert Cummins, often makes this charge. His motto is: "No Armchair Epistemology," and he displays his attitude graphically.

As Kim notes, these new naturalists tend to think of classical epistemologists as backward and out of touch with the latest developments in our understanding of the human being.

Kim takes as representative of the psychologistic "new naturalism" Philip Kitcher, who characterizes classical epistemology as "apsychologistic." He asks what needs to be added to true belief go get knowledge. The apsychologistic answer is that it is a logical relation between beliefs, and no appeal is made to the psychological processes that produce belief. The naturalist, on the other hand, will appeal to these processes.

In response, Kim notes that there is nothing to stop the classical epistemologist from making an appeal to psychological processes, any more than there is to stop the naturalistic epistemologist from appealing to logical relations. (Indeed, if it does not, it is hard to see how it could succeed as a theory of justification.)

Goldman's reliabilist theory in "What is Justified Belief?" is psychologistic in Kitcher's sense, but there other forms of reliabilism that are not. Specifically, David Armstrong's account of non-inferential knowledge in terms of a "law-like connection" between a belief and its truth is reliabilist but makes no reference to any psychological process. But these differences can be papered over.

The new naturalism, Kim argues, is closer to traditional epistemology than to Quine's naturalized "epistemology." Most advocates of naturalism do not want to make epistemology exclusively descriptive. "Justification in its full-fledged normative sense continues to play a central role in their epistemological reflection." The difference with the classical project lies in which elements they think comprise the criteria for justification.

Kitcher's characterization of the difference reveals a way in which both want to state their criteria of justification in non-evaluative terms. The traditional way is to appeal to logical relations among propositions, while the "new" way is to appeal to causal properties of events or states.

Another difference is that traditional theorists want to grant the cognizer "a position of special privileged and responsibility with regard to the epistemic status of his beliefs." They are (at least in one sense of the term) "internalists." The new naturalists deny this special position and so can be classified as "externalists." But these differences arise from "within the familiar normative framework, and are consistent with the exclusion of normative terms in the statement of criteria of justification."

Normative Ethics

A useful comparison is between normative epistemology and normative ethics. As noted in the last lecture, G. E. Moore denied that normative ethical concepts such as "good" and "right" can be defined in non-normative terms. To do so it to commit the "naturalistic fallacy." But this does not prevent the criteria of goodness and rightness from being non-normative. Moore himself allowed that being good "follows from the fact that it possesses certain natural intrinsic properties."

Kim interprets Moore's remarks as allowing that these properties constitute "naturalistic criteria of goodness, or at least pointing to the existence of such criteria." So with respect to justification, we can be naturalist without committing any kind of "naturalistic fallacy."

If Quine is seen as rejecting classical epistemology because it is normative, then there are only two ways one might embrace his descriptive epistemology. Either one must reject "the entire justification-based epistemological program" or else one must claim that epistemic concepts are naturalistically definable, thus committing an epistemic version of Moore's "naturalistic fallacy." Kim concludes that, "I doubt that very many epistemological naturalists will embrace either of these alternatives."

Epistemic Supervenience

So far, Kim has been operating under the assumption that the normative concept of justification must have non-normative, naturalistic, criteria for application. Now he explains why he makes this assumption. It is because of a general underlying belief that normative properties "supervene" on naturalistic conditions. The more specific belief is that epistemic properties supervene on naturalistic properties.

How are we to understand this relation of "supervenience?" Kim gives an to illustrate it. We think of a car as good "because it has a certain contextually indicated set of properties having to do with performance, reliability, comfort, styling, economy, etc." Being good is not simply a "brute fundamental fact." Similarly, a belief is justified because it is indubitable (Descartes), is caused in an appropriate way by the object of belief (early Goldman), or what have you.

Consider this thought-experiment. Suppose there were two cars, both of which perform equally well, are equally reliable, comfortable, well-styled and economical. Could we say that one is good and the other is not? If we cannot, then the goodness of the car is supervenient on just these properties. But suppose we were to say that one of the two is not good. Perhaps the good car is far less polluting than the one that is not good. Then add environmental friendliness to the list, and ask whether two cars equal with respect to all the above-listed properties are both good. If so, then the goodness of a car is supervenient on the expanded set of properties.

Now apply this test to justified belief. Suppose there are two people who have all the same perceptual experiences, are equally rational, and apply their rationality equally well to a given belief. Are they both justified? If you say they are, then their being justified supervenes on these properties. If you do not (say, because the perceptual experiences of one might be the product of an evil demon), then you go looking for other properties. Kim's claim is that different accounts of justification yield different lists of properties, but they have in common that they think some (naturalistic) set must be found.

Why do epistemologists think this way? The first reason Kim gives is that values (in this case, justification) must be "consistent" with facts in the sense that "objects that are indiscernible in regard to fact must be indiscernible in regard to value." That is, if two things are valuated differently, there must be some fact with respect to which they differ. This is the requirement of "weak supervenience."

The second reason is stated rather quickly and is not so clear as the first. The basic idea is that there must be some necessary connection between the facts and the values which supervene on them. This is what Kim calls "strong supervenience." In order for justification of a specific belief to supervene strongly on some facts, there must be some reason tying all facts of this kind to the justification of all beliefs of this kind. So what ties the facts and the justification has to be "reasons" or "norms." (It should be noted that while the facts which serve as criteria can be naturalistic, the relation of supervenience itself is valuational, as Kim states it here.)

Kim makes the stronger claim that the belief that values supervene on facts "arguably, is fundamental to the very concepts of value, and valuation." First, we must grant that any valuational concept must have criteria of application. If the criteria are valuational, then they must have their own criteria of application, generating a regress. So the series of criteria for valuations must end in non-valuational facts, to avoid the account's being "deeply incoherent."

The conclusion Kim draws from this is that the need for supervenience on non-valuational facts is enough to promote a form of naturalism in epistemology. There is no need to find "new inspirations form the sciences to acknowledge the existence of naturalistic criteria for epistemic and other valuational concepts." The same holds for normative ethics, and so if we want to allow that there is such a thing as normative ethics, we should allow that there is normative epistemology as well. And if we can naturalize one, we can naturalize the other.

Another parallel is with the degree to which scientific results influence our choice of criteria. It may be that psychological studies of human action and motivation are highly relevant to the development of criteria of goodness or rightness. Or they may have little or nothing to do with it. Similarly, the results of "cognitive science" or neuro-physiology may have a greater or lesser role in normative epistemology.


In 1986, Quine claimed that his naturalized epistemology is in fact normative. He wrote, "Naturalization of epistemology does not jettison the normative and settle for the indiscriminate description of ongoing procedures" (The Philosophy of W. V. Quine, ed. Hahn and Schilpp, p. 664). He compared normative epistemology to "the technology of truth-seeking," where the value involved is only relative to the goal of true belief, just as all values in engineering are relative to some ends. Specifically, the value in epistemology is "efficacy for an ulterior end, truth or prediction."

[ Course Home Page | Menu of Lectures ]