This essay argues that Quine's naturalized epistemology is unable to avoid epistemological relativism and even nihilism due to the lack of a potent theory-wide rule within the epistemic network of his “pragmatic holism” that (1) supervises the interrelations and revisability of all network beliefs including itself, and (2) survives its own revisability.
I will also show that his basic epistemological assumptions are the source of problems in his arguments concerning inquiry preference, assent-as-truth, sentences-as-objects, psychological reductionism, projection inputs, and the limits of prior theoretic knowledge. Finally, I will demonstrate that Quine does not succeed in reducing epistemology to psychology, and that this cannot be done without necessarily reverting to the epistemological nihilism, radical relativism, and extreme skepticism he wishes to avoid. The conclusion will be that a cryptic version of foundationalism is implicit throughout Quine's epistemology.
Epistemological theories pertain to, among other things, statements about knowledge. And epistemological theories themselves are statements or sets of statements about knowledge. Therefore, epistemological theories necessarily refer to themselves. Consequently, in the spirit of the necessary self-referencing of W.V.O. Quine's “naturalized epistemology”, the reasoning and assumptions of that epistemology must apply to the theory itself.
Quine conflates statements of logic and statements of fact, and he reduces all questions about ontology to questions about language. By conflating the distinction between logic and fact, he abandons the distinction between structure and content, between analytic and synthetic statements, thereby undermining the traditional epistemological view that there is a universal way of structuring human experience or a universal method of pursuing epistemological inquiry.
On the traditional view of the analytic-synthetic distinction, an analytic statement like “No unmarried person is married,” is logically true because of the occurrence of the logical components of the statement, independently of the statement's content. The statement's structure makes it true regardless of how the non-logical components are substituted. An analytic statement such as “No single person is married” can be rendered logically true, like the former statement, by analyzing the meaning of the word “single” and then replacing it, synonym for synonym.
But synthetic statements cannot be rendered logically true. Statements such as “No unmarried person is married” are usually called logical truths because their necessary truth does not depend on semantics but on the logical particles, so that they are true for all reinterpretations of their non-logical components. Quine objects to the analytic-synthetic distinction because analytic statements such as “No single person is married” depend on essential predication or synonymity or their reduction to logical truths.
Quine replaces the old empiricist epistemology, empiricism with the “dogmas”, with a version of pragmatism called field theory, network theory, or holism. The central idea is that we have a network of beliefs with epistemological relationships between them. So each person has an amorphous network of epistemic beliefs. Beliefs at the periphery of the network correspond to sense experience. Those beliefs are observation statements. The more centrally located beliefs are relatively distant from from sense experience. Consequently, the more central beliefs correspond to theoretic or logical statements. However, the location of any statement in the network is purely relative to the position of all other beliefs in the field.
So one encounters a new experience. Quine calls this a recalcitrant experience, an experience that cannot be dismissed or explained away. This experience is approached by the subject with an already-existing belief network and results in a statement that is then added to that network. The statements in this network differ only in how useful or pragmatically valuable they are in relation to all the other statements in the network. Even logical laws are merely additional statements in the system. Statements about the logical connections between statements in the network are just as subject to reevaluation, and therefore revision, as any other statements in the network.
Furthermore, the belief-evaluative criteria are also internal to the network, including the practical standard of minimizing network changes. And since logical laws are as revisable as any other statements in this network, rationality standards themselves are relative to other statements with a particular network, and relative to other networks. Consequently, any absolute notion of rationality, any defense of scientific method, and any preferred or unique objective method of inquiry, are equally relativized by this relational network theory.
Consequently, when a recalcitrant experience occurs and causes a formulation or a revision of some statement on the periphery of one's network, one must reevaluate other statements because of their logical interconnections. And this process of logically relational revision redistributes truth-values of the statements throughout the network because of a new or revised statement at the periphery.
But the logical laws or rules, which govern the reevaluation of statements, have the same epistemological value as all of the other statements in the network. The general rule for reevaluating statements in a belief network is what I will call the revisability rule:
If some recalcitrant experience occurs which causes one to regard some statement as true, and if this statement and some other statement are incompatible, then the other statement must be revised, and if this causes some other inconsistency with yet other statements, then those other statements must also be revised.
The most important thing in this reevaluation is the practical concern for preserving the network of statements. Therefore, the statements toward the center of the system, since they are the most important for the system of beliefs to operate, are the statements which resist reevaluation the most.
This revisability rule is what allows comparisons among different statements in the network. Hence the network includes statements on different logical levels which are not regarded as statements of different logical types. And since the revisability rule itself is simply an additional statement in the network, it also must be revisable within that same network.
But the network belief theory cannot survive the reevaluation of the revisability rule. If no statement is immune to revision, then neither is the revisability rule itself. If the revisability rule is not revisable, then that revisability rule is not a legitimate statement by the rules of the network. If the revisability rule is itself revisable, then the network belief theory is itself relative to different networks and might be considered useless to anyone who refuses to accept it on any grounds.
For Quine, the revision of the revisability rule must be just another statement in the network, and therefore must be revisable. But since the revisability rule is on the meta-linguistic level, a revision of the revisability rule must be even higher than the meta-linguistic level, since it makes a claim about the revision of the revisability rule itself. Whether one decides to revise the revisability rule or hold on to it, the decision of whether or not to revise the revisability rule requires an even higher meta-supervisory logical level containing a statement that says:
If the revisability rule and its revision are incompatible and one wants to keep the revisability rule, then one cannot revise the revisability rule.
But this statement itself must be revisable, too. Consequently, the entire process is viciously regressive because there would ensue an infinite series of different logical levels if the meta-linguistic rule-statements for reevaluating all statements including themselves are considered merely additional statements to be reevaluated in the overall network.
However, if one chose to revise or give up the revisability rule, the belief network theory would have no rule for reevaluating statements in the network. Yet the reevaluation of statements in the network, given the reevaluation of yet other statements, is supposed to be necessary to the network in order to revise statements in light of new recalcitrant experiences.
Since the statements in the network are related to each other because of their logical interconnections, the reevaluation of any individual statement in the network causes the reevaluation of one or more other statements in the network. But since such initial reevaluation could be of any statement in the network, including the statements concerning logical relations between beliefs and statements concerning belief revisability, there can be no reliable rule-statement for determining which statements are to be reevaluated on the basis of other equally-undeterminable statements.
But without reliable statements at these supervisory and meta-supervisory levels, reevaluation is either a random process or else cannot occur at all. If reevaluation is a random process, then there is no way to know how to reevaluate statements at all, or how to minimize network disruption, much less to even know that network disruption should be minimized. In fact, if the reevaluation is random, the entire network belief theory lacks the minimal conditions to merely qualify as a theory.
Therefore, to reevaluate statements, there must be some meta-linguistic rule operating within the network, along with logical laws such as the law of non-contradiction, to determine which statements are related to other statements, and in what way, so that the revisability rule and the logical laws can be applied non-randomly, and under appropriate conditions. Furthermore, to apply the revisability rule within a network, one must reason that, given the occurrence of some recalcitrant experience, some statement must be true; if that statement is true, then some other statement must be false; and if that other statement is false or in need of revision, then some third statement (or even others) must be false as well, and so on.
Moreover, there must be some definitions and semantic rules which fix the relationships of meaning among statements in the network in the first place. Otherwise, there would be no way to decide which statements to reevaluate in relation to any new or revised other statements.
In summary, for any reevaluation to change the truth-values of other statements in the network, it must be based on inferences grounded in the logical interconnections among those statements as well as rules for revisability. However, this can be determined only by the logical rules and laws of the network, and if one does not accept some logical rule or law and treat it differently from the inferences it generates, no justification of any statement is possible, no inference can occur, and no revision can take place. Consequently, one must use some such logical laws in order to reach the conclusion, in relation to some new or revised statement, that some other statement must be true or false or in need of revision.
Quine so much as admits this in the second edition of his Web of Belief, where he uses contradiction as a general law governing belief revision. "When a set of beliefs has accumulated to the point of contradiction . . . we can be sure that we are going to have to drop one of the beliefs in that subset." But if one recognizes that such laws are necessary, those logical laws are no longer merely additional revisable statements in the network.
Moreover, without such logical laws, there would be no way to decide which statements are to be considered false on the basis of some other statement being taken as true, and no way to ever infer that modus ponens must be given up as false or revised without invoking that same modus ponens as the basis of that inference. Consequently, any revision of modus ponens within a network using the revisability rule must itself use modus ponens. But in that case, modus ponens is not revisable and is not just an additional statement treated like all the other statements in the network. Hence, there must be logical laws on the meta-linguistic level which are unrevisable, according to which other statements in the network are evaluated.
Consequently, the necessity of such logical laws on the meta-linguistic level is not merely a practical one, as Quine claimed. Using Quine's revisability rule, there is no way to warrant abandoning the laws of logic without thereby using those same laws. Therefore, meta-linguistic logical rules are necessary for any belief network in which there is a general explanation of one's experience. Hence rationality cannot have a merely practical and relative status among one's beliefs.
Hilary Putnam tries to defend a sense-specific moderation of the revisability of logical laws, but does not even mention the priority of his own arbitration criteria for determining whether there are such things as prior truths. To decide whether there is any prior truth, one needs some meta-theoretic authority to decide the issue, an authority which must itself withstand the same kind of scrutiny as any proposed prior truth. Hence, Putnam's entire discussion has no standing to arbitrate the issue it addresses without assuming that his own theorizing about the subject has the authority of prior truth.
On the other hand, Putnam, in arguing against epistemological relativism, would, by analogy, argue that if Quine understands every utterance of belief that he uses as meaning 'it is true by the agreement of community members', then he must understand his own epistemological utterances, the utterances he uses to evaluate others' statements, the same way, no matter how many qualifiers of the 'according to the agreement of community members' type he accompanies them with. Other communities become, so to speak, psychological constructions based on the agreements of his community members. Moreover, since Quine characterizes all beliefs this way, he is making “the transcendental claim of a symmetrical situation,” a claim which cannot be understood if his epistemological theory is right.
Quine agrees with most of Hume's attack on induction and the scientific reasoning based on induction. He uses the same rigorous standards for epistemology by holding that grounding science in sense observation means deducing science from those observations. He cites the failure of any given program where every sentence of science “is equated with a sentence in strict observational and logico-mathematical terms.” Naturalized epistemology replaces traditional epistemology when we stop trying to deduce science from sense data. It is impossible to “strictly derive the science of the external world from sensory evidence.” Quine sets unreasonably high standards for epistemology, and while this may be because of his focus on logical positivism, it seems to ignore other empirical epistemologies. (How Quine, by the lights of his own theory, can even know what a deductive justification of science would be like, remains a mystery.)
How do i know anything at all about the world around me? is it possible for the world to be different in general from the way it is perceived to be? Quine intends for his naturalized epistemology to address the same basic issues of traditional epistemology. And even though he wants naturalized epistemology to supplant traditional epistemology, he never says that Hume and Carnap's attempt to provide some justification of our scientific claims about reality through sense experience is a different problem from the one he is addressing. “The Humean predicament is the human predicament". Trying to provide some justification of such claims is a universal problem, and we can't solve that problem any more than could Hume.
But naturalized epistemology does not resolve the predicament. When someone makes claims about their surrounding world, that person posits bodies and projects their physics from their data. To assess those claims' truth or falsity, one must be able to assess or judge those claims in addition to merely stating that certain facts are being asserted by those claims. And since we determine that some claims made by others are true and others are false, we must have some means of determining truth and falsity independently of simply examining what is said by other people.
But Quine says that in carrying through on the enterprise of epistemology we know that our position in the world is just like some other subject's position. Yet if we are to imagine ourselves in the same epistemological position as another subject, how is one to gain independent information about the world which will allow us to make any judgments about the status of those claims, or even our own in relation to theirs?
Stroud holds that we cannot compare our beliefs with the world they are about as we can in the normal experimental study of another person. One would find oneself with a set of beliefs and dispositions to assert things about the world, and one could have experiences that would strengthen or alter those dispositions, but those reinforced or newly acquired beliefs themselves would have to be seen in turn as at most some further projections from some new but minimal input. They could not be viewed as a source of independent information about the world against which their own truth or the truth of the earlier beliefs could be checked.
If all our beliefs about the external world are mere projections, then the belief that the input for those projections is the result of sense experiences must also be a mere projection as well as the belief that we have sense organs at all. On the naturalized theory of knowledge, “[M]y own 'output' would for me be no better than whistling in the dark.” Without the epistemological underpinning in which one's projections are based on sensory inputs, Quine would have no reason to prefer science to any other method of inquiry, and no reason to prefer psychology to theology. Without epistemology, there is no reason to prefer naturalized epistemology to theologized epistemology or even Cartesian foundationalist epistemology, nor is there any way to even tell them apart.
According to Quine's naturalized theory of knowledge, epistemology is reduced to psychology or eliminated altogether. But the notion of truth is one of those epistemological concepts which are eliminated or reduced to psychology. If the notion of truth has no epistemological normative content, then all that's left of truth is a semantic conception according to which to say that something is true is merely to assent to the assertion of that something. The Tarskian procedure is to define true so that saying that a statement is true is merely assenting to that statement. Tarskian truth is not a property of statements, but a syncategorematic notion that enables one to “ascend semantically”, which is to talk about sentences instead of objects.
Quine's Tarskian truth has no normative epistemological import. It's just a semantic tool for moving from object-language to meta-language. Without an epistemological notion of truth, one cannot access claims of others or even one's own claims. There is in that case no notion of rightness by which to make assessments. As Putnam says, if rightness is eliminated, then statements are indistinguishable from non-statements.
To eliminate the difference between epistemology and psychology and reduce epistemology to psychology, one must be able to eliminate justification for all knowledge and account for justification in purely psychological terms. Psychology tells one how one does arrive at beliefs and knowledge, while epistemology tells one how one ought to arrive at beliefs and knowledge. The concerns of psychology and epistemology are concerned with different spheres of inquiry. If justification is about providing reasons for beliefs and knowledge claims, then a psychological claim can never justify, or be a justifying claim, because it cannot by itself offer good reasons for believing that some claim is true.
Since psychology merely describes the method for arriving at beliefs, it cannot tell us what beliefs are good or true. Psychology cannot even be relevant to truth in traditional epistemology. It cannot explain notions such as justification and truth unless one counts whatever people think as justified or true, and conduct epistemological inquiry by merely observing and tabulating what people say and do. But that method would eliminate all content from justification, and reduce it to merely a factual question, an epistemological nihilism which Quine would like to avoid. Unqualified permissibility of all beliefs and knowledge claims is a denial of epistemology, resulting in epistemological anarchism and nihilism.
Naturalized epistemology offers no way to resolve the Humean and human predicament of gaining from warrant for claims about the external world. It cannot even address the same problems.
Nor can one refer to observation to anchor beliefs, since “One [person's] observation is another [person's] closed book or flight of fancy.” Since what counts as an observation statement varies with the speakers of a language, one can only arrive at something that seems like an objective and absolute standard for assessing beliefs by surveying most or all of the speakers of the language.
But that requires being able to make judgments about the statements of those speakers of the language. In other words, one must be able to judge the judgments, even though one is in the same epistemological position as everyone else, and everyone else is the standard of one's judgment. Consequently, one is not even in a position to begin the process of assessing beliefs. The result is relativism or skepticism.
For Quine, all questions about the justification of epistemology and science must be raised within some theory. The old epistemology tried to contain natural science, constructing it from sense data. Naturalized epistemology is contained within natural science, as a part of psychology. The old containment is still valid in a sense. One studies how one posits bodies and projects one's physics from one's data, and one's position in the world is just like anyone else's. Therefore, one's epistemological activity and the psychology it is an aspect of, is one's own projection from stimulations. If justification is treated as an internal problem of epistemology and science, circularity is no longer a problem, since one will have then given up on deducing science from sense data. The illegitimate circularity of the old epistemology arose supposedly from trying to validate knowledge of the world by using that same knowledge of the world.
But one is not in any better position with naturalized epistemology. The circularity of naturalized epistemology is itself held by Quine to be a part of psychology. Therefore this circularity is part of the same science of nature whose sources it seeks to understand. And even though, according to Barry Stroud, there is nothing vicious about this, allowing it is the same special pleading found in Nelson Goodman, Thomas Kuhn, and Hans Gadamer. Any attempt to exempt a theory from general constraints and requirements fails.
But the circularity is vicious after all. Neurath's tale of the mariner at sea who manages to rebuild his boat while staying afloat is seen by Quine to imply that one can examine and validate or justify epistemological claims one by one, without involving the illegitimate attempt to validate the entire enterprise. By examining and validating each claim, one can supposedly avoid any illegitimate circularity or inconsistency. But Neurath was still committed to the logical empiricist account of knowledge according to which the structure is qualitatively and epistemologically different from its content. Quine cannot embrace this parable, because for him analytic statements, including logical rules and laws, have the same epistemic status as synthetic ones such as observation statements.
Replacing each part of the boat for others to gradually rebuild the boat is strictly analogous to evaluating beliefs and redistributing truth values of statements to retain a belief network. Both processes require some principle for evaluation and validation (or refurbishing). If we invoke Quine's revisability rule to refurbish a boat while keeping it afloat, we could invoke the following floating-boat refurbishability rule:
Each part of the boat is examined and evaluated. If some evidence allows us to validate a part, it stays. But if some evidence or experience causes us to regard a part as unwarranted or undesirable, we make whatever adjustments are necessary to the other parts to keep the boat afloat, either by replacing that unwarranted part with a new part or by rearranging some or all of the other parts.
This rule sets forth a logical procedure for examining the parts of the boat (or theory) and deciding what to do with them. The process necessarily assumes some principles and inference rules (modus ponens). To replace some parts and keep other parts while keeping the boat afloat, the laws of physics, including archimedes' law, and laws of logic must be strictly observed. For Neurath, this refurbishability rule, the laws of physics, and inference rules are not parts of the boat. These rules and principles and laws cannot be evaluated, revised or replaced if one wants to keep the boat afloat. But for Quine, they are just other parts of the boat. But in that case, as with the belief network when the revisability rule is itself revised, the boat would sink.
Furthermore, the boat analogy is useful only if one makes an epistemological distinction between logical structure statements and descriptive content statements, a distinction which Quine rejects. Only because of the refurbishability rule can there be a proper method of repairing the boat, and there must be a proper and an improper method since one must repair it in a specific rule-based way, to keep the boat afloat. The rules and principles introduce a normative evaluative aspect into the procedure for evaluating beliefs. Without such rules and principles, anything goes, and the circularity will sink Quine's boat as well.
To make any progress over traditional epistemology in validating scientific claims, naturalized epistemology must pull psychology out of its epistemological rut by using only psychology itself. The warrant for empirical psychology must come from within psychology itself, not from epistemological claims. Unfortunately for Quine, there is no way to give empirical psychology a privileged status. If one begins an epistemological inquiry into how beliefs are related to the external world, there is no special condition that would make psychology immune from the inquiry. Unless epistemological problems are dealt with first, there is no way (or need) to construct an epistemological theory within psychology.
If one assumes certain things to be true about perception, and then asks on that basis how any knowledge of that world is possible, and then by reduction to absurdity arrives at the skepticism which Quine thinks is coherent, any alleged knowledge of the physical world would be suspect, and no scientific knowledge could be unproblematically introduced to refute the skepticism. Consequently, one would need an argument for the special status of empirical psychology which makes it immune to epistemological problems of justification and truth, and establishes it as the context within which all inquiry occurs, to prevent this bootstrap failure. But if the attempt to naturalized epistemology is successful, the claims of empirical science thereby get the epistemologically privileged position which provides a foundation for all justification.
Even though Quine wants to make epistemology a part of psychology, he still wants to continue talking about evidence and observation. He maintains that once we reject talk about consciousness and awareness, and focus exclusively on the scientific question of sensory receptor stimulation, we are freed of traditional epistemological problems. He then merely assumes that the currently dominant theories of empirical psychology and perception are immune to the usual epistemological questions.
Quine also assumes that the old epistemological problems are confused or logically inconsistent, and provides no arguments for this. He claims that what counts as observation can be settled in terms of the stimulation of sensory receptors, and thinks that this factor resolves the epistemological problems of evidence and how evidence is related to theory. Sure, if one chooses empirical psychology, some issues in the philosophy of mind concerning consciousness can be dismissed. But significant issues remain concerning the status and role of sense organs in perception. The stimulation of sensory receptors assumes some causal theory of perception, and there is no way to know that we even have sensory receptors except by yet other stimulations of those same receptors.
Quine's psychological discovery, according to Rorty, will be very remote from the foundations of science or the relation between theory and evidence, and that any alleged connection between psychology and epistemology comes from equivocating about key epistemological terms. Quine sees the problem of epistemology as the problem of trying to distinguish between causal stimulus and awareness. If we all just convert to naturalized epistemology, we will not have to worry about when or how or even if consciousness arises. We can view a human as a black box, with externally stimulation as input and external testimony about the external world as output. In that case, which inner workings of the black box are tinged with awareness is the only remaining issue. But this ignores the question of causality, or else arbitrarily assumes an answer to it. To make epistemological problems into problems of consciousness and awareness is nothing more than a self-serving distortion.
Talking about irradiated areas of the retina or else optic nerve pulses, Rorty says, is choosing a black box, not discovering criteria for inquiry. Quine tries to dismiss the dilemma by merely changing the motive of inquiry. If one is interested only in causal mechanisms, then sure, one can ignore awareness altogether. In fact, if there are no experimental criteria for data sources, then Quine's view that we give up on sense data and talk about nerve ending causation and observation sentences does not resolve the epistemological issue. It arbitrarily ignores it.
Rorty thinks Quine wants to retain observation by defining it as intersubjective agreement in the language community, defining an observation sentence as one all speakers of the language agree on, when given the same concurrent stimulation. He needs something to link epistemological claims to sensory stimulation, and wants observation sentences to be what are causally contiguous with sensory receptors. An observation sentence is the sensory periphery, and is the “minimal verifiable aggregate” which shows its empirical content.
However, Quine does not follow the implications of his positions to their logical conclusions, however, He develops what should lead to getting rid of epistemology, and then at the most crucial point, turns away from the process and reintroduces or refuses to part with observation sentences. Quine recognizes that an observation sentence is necessary to ground language and epistemology in experience, and holds that the observation sentence's relation to our knowledge of what is true is the traditional relation, and that observation sentences are the repository of evidence for scientific hypotheses. By retaining observation sentences and grounding that notion in our experience and intersubjective agreement about experience, Quine tries to avoid throwing the epistemological baby out with the relativist bath water. He admonishes those who emphasize cultural relativism and discredit the idea of observation, offering instead the absolute standard of community agreement for observation sentences.
The problem, however, is that observation sentences and community agreement are merely ad hoc exceptions to Quine's theory. He has same choice Carnap had in trying to develop protocol sentences. Carnap's move from a phenomenalistic interpretation of protocol sentences to a public intersubjective interpretation was required by the same concerns which are problematic for Quine. The same problems of verification, truth, and evidence occur when trying to determine intersubjective community agreement. Some way of handling these problems is necessary to even determine that there is a community to agree or disagree. Community agreement is supposed to provide an absolute standard for observation statements. Consequently, we must survey, observe, record, and compare the reactions of members of the language community to the same sensory stimulations under the same conditions. This is supposed to describe empirical psychology and ground epistemology. But what counts as a member of the language community? What counts as the same sensory stimulus? What counts as the same conditions? How do we recognize agreement or disagreement? Why should we trust our records of our earlier interviews with other community members? These questions have to be answered if naturalized epistemology and empirical psychology are to be seen as improvements over traditional epistemology.
How do we settle on the empirical content that an observation sentence must have, on Quine's view, to provide evidence of scientific theories? All of the same problems of verification that Quine says are the major stumbling blocks of traditional epistemology remain in his use of observation sentences. Quine's response is that the empirical content of an observation sentence is so obvious that we all see it and agree what it is. But this is no advance over Descartes' clear and distinct ideas. This is another area where Quine reverts to traditional empiricist epistemology. Consequently, Quine retains an aspect of empiricism and realism to avoid relativism.
As we have already shown, both holism and naturalized epistemology operate in a supervisory way, above the level of observation sentences. For Quine, individual sentences are meaningless once we get beyond observation sentences. Psychology replaces epistemology once we have psychology. However, we need observation sentences, with all of their epistemological problems, to develop psychology itself.
Rorty thinks Quine should either let the relativists say that observation is whatever we can agree on, or else show how psychological discoveries can make something more of the notion. If empirical psychology cannot do this, then defining dependence on present sensory stimulation in terms of intersubjectivity is just, again, invoking traditional epistemology without any psychological purpose.
The main points of dispute in these criticisms could be, in a sense, semantic ones about how terms are used and with what meanings. For Quine, the Humean predicament persists “[o]nly in the fallibility of prediction: the fallibility of induction and the hypothetico-deductive method in anticipating experience.” So in a sense, Quine is a mainstream epistemologist---except that he wants to call it science.
Quine claims that the things which occupy our world such as sticks and stones, electrons and molecules, are merely ways of speaking. And yet he holds that they are nonetheless real. But through naturalized epistemology, we deal with the apparent tension between such different claims within science rather than within epistemology. Quine views metatheoretic and meta-methodological questions as internal to the theory or method. The major differences are merely semantic, and the semantic issues are questions about method and evidence, not reality. Those issues “belong not to ontology but to the methodology of ontology, and thus to epistemology. And here still I recognize no first philosophy prior to science.” But the main issue is not merely semantic. Quine cannot develop the general theory of naturalized epistemology without doing first philosophy.
The general principle of naturalized epistemology maintains that meaningful epistemological inquiries can be pursued only within empirical psychology, and epistemology is an aspect of psychology. But this principle itself is not a claim of empirical psychology. It is a meta-scientific metatheoretic claim which cannot itself be reduced to or made a part of any scientific theory. But the principle is supposed to be a rationally held position. It is a philosophical theory, of interest to philosophers, not scientists. Quine offers a detailed rationale for naturalized epistemology, which is supposed to be an argument or reason for preferring it to traditional epistemology. He accepts Hume's critique of induction, and uses his rejection of the analytic-synthetic distinction and the resulting collapse of analytic statements into synthetic statements to broaden the scope of Hume's attack to include all knowledge claims. However, both of these arguments are philosophical arguments within traditional epistemology. At the same time, he denies that there is a first philosophy within which such questions as the justification of induction and the viability of the analytic-synthetic distinction can arise, as the main point of naturalized epistemology.
Naturalized epistemology must be unjustified, or else it is self-refuting. If all epistemological issues of justification and warrant are internal to the theory, then it is impossible to justify or give any reasons or evidence for that theory itself. Without extra-scientific criteria for epistemologically evaluating the reductive principle, there can be no basis for knowing it is a good theory or that it is preferable to competing theories or that there is any reason for adopting it. Naturalized epistemology is not a claim of empirical psychology. And if all evidence and reasons occur only within empirical psychology, then there can be no evidence or reasons for naturalized epistemology itself. To the extent that it is a justified theory which compels assent, naturalized epistemology is to that same extent inconsistent and self-defeating. Consequently, good reasons for naturalized epistemology are also good reasons against it.
As Harvey Siegel has observed, to rationally embrace naturalized epistemology “involves accepting that some meta-scientific theories are justified on the basis of extra-scientific reasons . . . but this recognition forces the rejection of naturalized epistemology, which renounces that kind of justification.” Therefore, naturalized epistemology does not eliminate first philosophy, but assumes it.
Naturalized epistemology simply cannot be justified as a scientific theory. Paul Roth argues that Quine's ambiguous use of the notion of theory gives the impression that naturalized epistemology is a scientific theory. But since naturalized epistemology is a meta-scientific theory about the nature of science, it cannot be justified by any appeal to natural science itself. Hence, Quine's theory about the nature of epistemology is part of his argument for adopting naturalized epistemology, and therefore cannot itself be a part of naturalized epistemology. As Roth explains, “since Quine's arguments with regard to the a priori limits to theoretical knowledge are what are to convince us to accept the new epistemological project, [naturalized epistemology], the relevant arguments cannot assume the prior acceptance of that epistemological stance.” Ironically, Quine's theory about naturalized epistemology is therefore a prior claim about the nature of knowledge. Consequently, he must justify his claims about the nature of knowledge as prior ones. And resorting to prior claims to justify or provide some rationale for naturalized epistemology means that traditional epistemology has been smiling in the shadows all along.
We have seen, in examining Quine's pragmatic holism, that one cannot revise or suspend logical laws to hold on to recalcitrant data of experience for the good of the system without using those logical laws at the meta-linguistic level to make inferences about the consequences for that system or to project the possible outcomes of different revision scenarios within the network. Moreover, without logical laws and semantic rules at the meta-linguistic level, there would be no basis for preferring one possible revision of beliefs to another. Any decision about change within a belief network must assume inviolate logical laws governing the network itself. Consequently, a notion of rationality based on universal and necessary logical laws remains intact as well as the method of fixing beliefs based on those logical laws. Quine's attack on traditional epistemology is internally inconsistent, and therefore does not actually address the questions of either epistemology or the philosophy of science. His own theory cannot survive the kinds of objections which he himself raised against traditional epistemology. Therefore, epistemology is not only not reducible to psychology---it cannot be. Epistemology may have much to learn from psychology. But that is an epistemological claim, not a psychological one.
