In contrast to many contemporary epistemologists, Quine takes epistemology to be a particular branch of the study of a particular kind of animal behavior. To understand his conception of epistemology, we need to begin by understanding how he conceives of the study of animal behavior in general, and of linguistic behavior in particular.
All animals engage in goal-directed behavior in order to satisfy their needs in a challenging environment. At any given point in time, we can think of an animal's behavioral repertoire at that time as a function from sensory stimuli to motor response. Animals can change their behavioral repertoire over time as a result of conditioning, but different animals are susceptible to different sorts of conditioning. For instance, dogs can be conditioned to salivate in response to noises that human beings cannot be conditioned to respond to at all. And, to take another example, human beings can be conditioned to respond to certain complicated patterns of noises, gestures, or inscriptions by producing other complicated patterns of noises, gestures, or inscriptions; but other, non-human animals cannot be conditioned to respond in this same way to such stimuli. In other words, human beings, unlike other animals, can be conditioned to respond to human language by producing human language themselves.
An animal's susceptibility to conditioning can itself be adaptively beneficial to the animal, or its species. For instance, when a dog is conditioned to salivate in response to a particular noise that has, in the past, been perceptually associated with feeding, the dog's salivation makes it easier for the dog to chew and digest the food that it may be about to receive. Again, when a dog is conditioned to respond to the sound "sit!" by sitting, that response makes it easier for the dog to acquire the desirable reward that has been associated with this response in the past. And when a human being is conditioned to respond to the perception of human language by the production of human language, then the human being acquires a whole new mechanism for altering its behavioral repertoire: communication. The person who hides in fear when she sees the Aurora Borealis can learn to stay and calmly gaze at the sight once she learns through communication that there is no basis for her fear. The leader who speaks in a frightened tone of voice to her followers can learn to speak in a calm and reassuring tone of voice once she learns through communication that this is more likely to produce the desired submissiveness in her followers. And finally, the scientist who describes a fire as the shedding of phlogiston can learn to describe it as the consumption of oxygen once she learns through communication what the mechanisms of combustion are.
So, because humans can be conditioned to respond to human language by producing human language, they can be conditioned to communicate, and communication provides them with a new mechanism for altering their behavioral repertoire, including their linguistic behavioral repertoire. Now, so far, I have described communication simply as a process of responding to certain complicated patterns of noises, gestures, or inscriptions by producing other complicated patterns of noises, gestures, or inscriptions. But because human beings can be conditioned to respond differentially to patterns of enormous complexity, the patterns of which we speak here can be extremely complex. And, as it happens, over the course of human linguistic history, these patterns have actually become extremely complex, and are now multi-layered. We can describe one layer of pattens by segmenting the gestures, noises, and inscriptions into a certain set of units that we call "words"; we can describe another layer of patterns by segmenting the gestures, noises, and inscriptions into a set of larger and more inclusive units that we call "sentences"; and we can describe yet another layer of patterns by segmenting the gestures, noises, and inscriptions into a set of still larger and still more inclusive units that we call "theories".
We said above that human beings can alter their linguistic behavioral repertoire either as a result of conditioning or as a result of communication. But when they alter their linguistic behavioral repertoire, they can alter any or all of these layers of patterns in the noises, gestures, and inscriptions they make. Sometimes, the only part of their behavioral repertoire that they alter is their tone of voice (as in the case of the leader who learns to speak to her followers in a calm and reassuring tone of voice). Sometimes, the only part of their behavioral repertoire that they alter is their words (as when someone learns to call someone else by a particular name). But sometimes, they alter very large scale units in their gestures, noises, and inscriptions -- they alter their theories. This process, whether it is brought about by conditioning or by communication, is what we will call "theory change". And epistemology, for Quine, is a branch of the study of theory change.
Of course, the study of theory change includes many different branches. Which of these branches is the one that Quine identifies as epistemology? Here is how Quine himself puts the point:
"From impacts on our sensory surfaces, we in our collective and cumulative creativity down the generations have projected our systematic theory of the external world. Our system is proving successful in predicting subsequent sensory input. How have we done it?
"Neurology is opening strange new vistas into what goes on between stimulation and perception. Psychology and more particularly psycholinguistics may be looked to for something to say about the passage from perception to expectation, generalization, and systematization. Evolutionary genetics throws further light on the latter matters, accounting for the standards of similarity that underlie our generalizations and hence our expectations. The heuristic of scientific creativity is illuminated also, anecdotally, by the history of science.
"Within this baffling tangle of relations between our sensory stimulation and our scientific theory of the world, there is a segment that we can gratefully separate out and clarify without pursuing neurology, psychology, psycholinguistics, genetics, or history. It is the part where theory is tested by prediction. It is the relation of evidential support, and its essentials can be schematized by little more than logical analysis." (The Pursuit of Truth, 1-2)
The segment of which Quine speaks in the preceding paragraph -- the segment that can be clarified without pursuing neurology, psychology, psycholinguistics, genetics, or history; the segment that involves the relation of evidential support -- that is that segment the study of which constitutes epistemology. In short, epistemology, for Quine, is the study of evidential support.
This characterization of epistemology is completely unremarkable for an analytic philosopher: Carnap, Ayer, or Russell could easily have offered just the same characterization. So what is distinctively naturalistic about Quine's conception of epsitemology then? And in what sense does epistemology, for Quine, become a "chapter of psychology", as Quine repeatedly says?
The answer to these questions is strongly suggested by the context in which Quine situates naturalistic epistemology throughout his writings. For Quine, the effort to naturalize epistemology is a response to the failure of a particular program that Carnap undertook. Recall that Carnap initially attempted to construct a so-called “phenomenalistic” language into which (along with the terms of logic and set theory) all our talk of material things and events could be translated. (A “phenomenalistic” language is one in which there are no expressions that refer to material things and events, and all reference is directed towards features of our subjective experience.) This program failed, even by Carnap's own lights: Carnap was never able to specify a language that was both phenomenalistic and also adequate for translating all our talk of material things and events. Carnap concluded, as did most other philosophers, that our talk of material objects and events is not simply a means for stating facts that could be more elaborately stated using phenomenalistic language. But Carnap still thought that he could at least use his phenomenalistic language as a device for stating the evidential basis for all our talk of material things and events. In attempting to naturalize epistemology, Quine was simply calling into question the central presupposition of this latter program, viz., that there is a phenomenalistic language in which the evidential basis for all our talk of material things and events could be stated. If our talk of material things and events is not itself phenomenalistically statable (which it isn't, given the failure of the earlier program), then why suppose that the evidential basis for our talk of material things and events is phenomenalistically statable? To naturalize epistemologize is to give up the presupposition that our evidential basis is phenomenalistically statable, and to treat the question of what our evidential basis is as itself an empirical question -- one to be answered by looking and seeing what our evidential basis is. That is what is distinctively naturalistic about Quine's epistemology.
This interpretation of Quine's naturalistic program can help us to understand the otherwise puzzling paragraph of "Epistemology Naturalized" in which Quine first explicitly introduces the idea of a naturalized epistemology. The paragraph that I have in mind here is the third paragraph of the following famous passage:
"...Two cardinal tenets of empiricism remained unassailable... and so remain to this day. One is that whatever evidence there is for science is sensory evidence. The other, to which I shall recur, is that inculcation of meanings of words must rest ultimately on sensory evidence. Hence the continuing attractiveness of the idea of a logischer Aufbau in which the sensory content of discourse would stand forth explicitly.
"If Carnap had successfully carried such a construction through, how could he have told whether it was the right one? The question would have no point. He was seeking what he called a rational reconstruction. Any construction of physicalistic discourse in terms of sense experience, logic, and set theory would have been seen as satisfactory if it made the physicalistic discourse come out right. If there is one way there are many, but any would be a great achievement.
"But why all this creative reconstruction, all this make-believe? The stimulation of his sensory receptors is all the evidence anybody has had to go on, ultimately, in arriving at his picture of the world. Why not just see how this construction really proceeds? Why not settle for psychology? Such a surrender of the epistemological burden to psychology is a move that was disallowed in earlier times as circular reasoning. If the epistemologist's goal is validation of the grounds of empirical science, he defeats his purpose by using psychology or other empirical science in the validation. However, such scruples against circularity have little point once we have stopped dreaming of deducing science from observations. If we are out simply to understand the link between observation and science, 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." ("Epistemology Naturalized", 75 - 76)
Many philosophers have wondered what Quine could have in mind here when he invites epistemologists to "see how this construction [of one's theory of the world] really proceeds", to "settle for psychology", and to "understand the link between observation and science". These claims seem to suggest that Quine wants epistemology to cease studying how we ought to think, and confine itself to studying how we actually do think – and this is just how some philosophers have interpreted Quine’s program (e.g., see Jaegwon Kim’s widely influential interpretation of Quine in “What is Naturalized Epistemology?”). But this interpretation of Quine has two problems. First, it makes Quine’s view look very unattractive: clearly, there is some worthwhile project of studying how we ought to think, and if epistemology isn’t the study of how we ought to think, then what is? And second, this interpretation of Quine is very hard to square with the passage quoted above from The Pursuit of Truth, in which Quine says that we can clarify relations of evidential support "without pursuing psychology". Isn't this latter claim simply inconsistent with Quine's claim that epistemology -- the study of evidential support -- can become a chapter of psychology?
The interpretation that I have offered dissolves the apparent inconsistency, and also helps us to understand why Quine is not rejecting the epistemologist’s effort to understand how we ought to think. I’ll take up each of these points in turn.
When Quine says that we can clarify relations of evidential support without pursuing psychology, what he has in mind is this: relations of evidential support are, broadly speaking, logical or statistical, and we can understand what those relations are by doing logic or statistics, not by doing psychology. But when Quine says that epistemology can become a chapter of psychology, what he has in mind is this: in order to understand what evidence supports our theories, and to what extent, what we need to do is to locate our actual evidential bases, and determine to what extent those evidential bases do stand in the relevant logical or statistical relations to our theories. Of course, in figuring out what our actual evidential bases are, or even in figuring out what our theories are, we may apply a principle of charity, and allow our determination of those bases or those theories to be partly guided by consideration of which evidential bases would successfully support which theories. This is a way in which logic and statistics can themselves help to guide the empirical study of psychology. But still, the attempt to figure out what our actual evidential bases are, and what our theories are, is a thoroughly empirical enterprise, and this is the sense in which epistemology -- the study of evidential support relations -- is a chapter of psychology.
But, to say that epistemology is a chapter of psychology is not to deny that epistemology has a normative dimension, i.e., that it concerns how we ought to think. As Quine writes:
"To emphasize my dissociation from the Cartesian dream, I have written of neural receptors and their stimulation rather than of sense or sensibilia. I call the pursuit naturalized epistemology, but I have no quarrel with traditionalists who protest my retention of the latter word. I agree with them that repudiation of the Cartesian dream is no minor deviation.
"But they are wrong in protesting that the normative element, so characteristic of epistemology, goes by the board. Insofar as theoretical epistemology gets naturalized into a chapter of theoretical science, so normative epistemology gets naturalized into a chapter of engineering: the technology of anticipating sensory stimulation." (The Pursuit of Truth, 19)
So there is theoretical epistemology, which is the study of what evidence we actually have for whatever theories we actually hold. And there is normative epistemology, which is the study of how to adjust our theories in order most effectively to anticipate sensory stimulation. And both of these pursuits can be guided by logic and statistics, though both of them are heavily empirical. In this respect, the different branches of epistemology are like the different branches of the study of human action: we can study the reasons for which people do what they do, and we can also study what sort of action proves most successful in achieving its goals. But both of these branches of empirical study are guided by our theory of what it is to act rationally (or what is typically called “rational choice theory”).
For Quine, then, epistemology is the study of evidential support, and this study has three branches. First, there is the empirical, psychological study of what evidence we actually have, and what theories we actually hold. Second, there is the empirical, normative study of what strategies of theory change we should employ, in order to optimize or at least improve our ability to anticipate future sensory stimulation. And third, there is the logical and statistical study of evidential support relations, i.e., of the logical or statistical relations that obtain between the sentences our assent to which is most directly and universally keyed to particular sensory stimulations ("observation sentences", as Quine calls them) and our theories. The first two of these branches are obviously empirical, and the second of those two empirical branches is obviously normative. In describing only the first two of these branches as obviously empirical, I do not mean to suggest that the third branch is non-empirical. But empiricality is a matter of degree for Quine. Our beliefs about logic and statistics are less immediately or directly impacted by our sensory stimulations than our beliefs about psychology are. Or, to put the same point using Quine's famous metaphor, logic and statistics are farther from the sensory periphery of our web of belief than psychology is. (An accessible discussion of this metaphor is in Quine and Joseph Ullian’s co-authored book The Web of Belief, published in 1970.)
Quine himself devotes most of his efforts in epistemology to the second of these three enterprises. Specifically, he outlines a number of principles that should (and, he thinks, for the most part do) guide our theory change so that it can successfully anticipate future sensory stimulation. There is, for instance, the principle of simplicity, according to which we should hold to the simplest theory that predicts and explains all our evidence. And there is the principle of minimum mutilation, according to which we should change our theories as slightly as possible in order to predict and explain all our evidence. (Again, see The Web of Belief for discussion of these principles of theory revision, as well as others.) These principles can conflict with each other, and sometimes conflict with other principles that guide our efforts to change our theories so as most successfully to anticipate future sensory stimulation. In case of such conflict, we must simply strive to do the best we can, and there is no recipe for that. All we have are principles to guide our theory choice, and good judgment to guide us in cases in which the principles conflict.
I have so far ignored a complication in Quine's account of evidence and of evidential support. For Quine, the logical or statistical relations involved in evidential support must be relations among sentences; some sentences imply or probabilify other sentences. But when Quine describes our evidence, he frequently speaks of sensory stimulations. How do the sensory stimulations that Quine takes to constitute our evidence relate to the sentences that Quine takes to stand in relations of evidential support (e.g., implication or probabilification) with our theories? For Quine, this relation is causal. The sentences that Quine takes to stand in relations of evidential support with our theories are what Quine calls "observation sentences". Observation sentences are those sentences which, for each person in a given community, have the following property: no matter what else that person believes, when she receives a sensory stimulation of a particular kind, then she will assent to that sentence. Assent to observation sentences is causally keyed to the occurrence of a particular kind of sensory stimulation within a given community. This makes observationality relative to a community of speakers, but it also insures that the same evidence can be had by multiple speakers within that community. In short, it insures the publicity of evidence, at the cost of making it community-relative.
Two things remained constant throughout Quine's writings in epistemology. The first is that he held steadfastly to the conception of epistemology that I have outlined above: I see no sign of his deviating from this conception -- at least at the level of abstraction described above -- at any point in his career. Quine changed his mind about a number of things in philosophy, but he did not change his mind about the points mentioned above. The second is that, at no point in his career did Quine have anything to say in response to some of the questions that form the focus of so much epistemological attention since the 1963 publication of Edmund Gettier’s agenda-setting paper “Is Knowledge Justified True Belief?”, e.g., what is knowledge? What is it for a belief to be justified? What is it for someone to be justified in believing a proposition? Does knowledge require justification? Does knowledge require some especially strong relation between the believer and the world? It is these questions to which Alvin Goldman has devoted most of his attention in epistemology over the past four decades, and it is to the history of his various views that I now turn.