Naturalism and Rationality


Christopher Hookway


The paper examines a common objection to naturalized epistemology: since it only provides a scientific explanation of cognition, it fails to address the normative issues that are fundamental to epistemology. Quine and other naturalized epistemologists such as Laudan try to meet this objection by arguing that all methodological or epistemological norms are instrumental, concerning means to the achievement of cognitive ends. The paper considers how far this strategy can be sustained, exploring whether there are issues of epistemic evaluation which concern ends rather than means to ends, and considering whether we need explanations of concepts such as knowledge and justification which cannot be understood in means-ends terms. It concludes with a discussion of an objection to naturalized epistemology due to L. BonJour.


1.  Introduction: naturalised epistemology and normative epistemology.

It is a familiar objection to Quine�s programme of naturalised epistemology that it fails to address some of the most fundamental questions that motivate epistemological inquiries in the first place. The objection takes several forms. Some would object that epistemology is obliged to take Scepticism seriously, its fundamental task  being to show whether (and, if so, how) we can satisfy ourselves that we are not brains in vats who are systematically misled in our view of the world by wicked scientists.  Quine would reply that if that is what epistemology is, then his programme is not epistemology, and none the worse for that. [1]   A more worrying objection is that a naturalised epistemology cannot deal with normative questions about how we ought to carry out inquiries and revise our beliefs.  Bearing in mind Quine�s proposal that epistemology should fall into place as �a chapter of psychology�, the objection runs:

Cognitive psychology can only describe and explain the ways in which our cognitive faculties actually do work.

Epistemology is concerned with the evaluation of our cognitive practices; among other things, it describes epistemic norms and attempts to settle whether they are legitimate or illegitimate.

Granted the additional premise that a descriptive- explanatory account of a practice cannot provide it with a legitimation, or defend the norms used in pursuing it, the standard anxiety is easily understood. The issue raised here is obviously related to a much older tradition of debate about naturalism in ethics.  The naturalistic fallacy, supposedly the fallacy of attempting to account for norms (�oughts�) in terms of natural properties of agents and their environments, would, it seems, be as much to be avoided in the case of epistemic norms as in the case of ethical norms.  Any description or explanation of our practices can, it seems, invite the normative question which it is good or desirable that our beliefs and activities should be structured in such a way.

            Biting the bullet is a possibility here too.  Quine sometimes suggest that he calls his work �epistemology� because it attempts to explain and understand how beliefs and theories are constructed on the basis of experience, an aim it shares with the work of philosophers such as Carnap, Locke and Hume. The failure to address normative issues may not be an obstacle to the claim that he belongs to the epistemological tradition. However more often he resists the suggestion that naturalised epistemology would commit the naturalistic fallacy: there is no difficulty in using a naturalised epistemology to settle fundamental normative issues. This rests upon a distinctive view about the kinds of normative issues that arise in epistemology, and it is this view that will be considered in this paper.

            The core idea is straightforward.  Suppose I wish to measure the temperature of the fluid in a flask.  How should I go about it? It may be proposed that I should use a particular thermometer which is ready to hand: is this a good method for finding the information I require? Perfectly straightforward factual information seems adequate to meet my needs: we can test the reliability of the thermometer; and if it turns out to be unreliable, then I ought not to use it.  Results from the natural sciences are commonly used to provide information about how we can most efficiently meet our ends. Changing habits of medical diagnosis as our knowledge of the causes of diseases grows provides a further example of how growing scientific knowledge alters our views of what we ought to take as giving us a reason to believe something. (Elgin, 1996, p. 97) And there seems to be no feature of cases where our ends involve the gathering of reliable information to prevent our saying the same thing about these cases too. The growth in empirical knowledge provides invaluable information about how best to pursue our cognitive ends. Since reasoning about means and ends is often grounded in empirical or scientific information, then the possibility emerges that all the normative issues of epistemological importance are of this kind:

         They concern the relations of means to ends.

         The sorts of means/ ends issues with which they deal can be settled by appeal to empirical information. [2]

In that case, as Quine has put it, normative epistemology �gets naturalised into a chapter of engineering�. (1990, p.19) Nobody could seriously doubt the relevance of empirical information to questions about how we should form our beliefs in the sorts of cases described above.  But the suggestion that such examples are typical of the normative issues that arise in epistemology is very controversial. [3]  


2. Three grades of naturalistic commitment

            For reasons that will become clear below, I shall not say much about what I mean by �naturalism�.  The position can encompass a broad range of views - as can be seen from the fact that Donald Davidson, who is presumably committed to a kind of non-reductive realism about norms, is happy to describe his own work as �naturalised epistemology�.  It will be useful to map out three naturalistic themes, of differing degrees of strength. 

a) The first minimal grade of commitment holds simply that an adequate philosophical account of human life and rationality should make no claims that conflict with the supposition that we are material objects and that our faculties and capacities are the products of natural selection: there is nothing �supernatural� about our functioning that requires the belief that we are partially exempt from familiar biological or physical explanatory models. 

b) The second grade of commitment holds that an understanding of human life and rationality can be obtained through investigations that are a posteriori, that are �scientific� in a very broad sense. This might allow for history as an indispensable source of knowledge, and it also leaves open the possibility that ranges of knowledge which do not belong to the special sciences may have a role.  For example, everyday descriptions and characterisations of our practices may use classifications that are invaluable for ordering our inquiries and understanding our lives even if those classifications are not employed in any of the special sciences. [4]

c) A  stronger position (sometimes defended by Quine) holds that the philosophical needs that prompt us to think about mind, cognition and rationality can be met entirely by research in the special sciences - for example, in psychology.  Much of Quine�s rhetoric about his naturalism fits this position. 

It will be apparent that the differences between these grades of naturalistic commitment have implications for how we understand the normative side of naturalised epistemology.  Are normative epistemological issues entirely a matter for psychology? For psychology and the other sciences? Or do some of the issues relate to everyday common sense observation?


3.Methodological norms.

How should we move beyond the examples concerning the diagnosis of illness and the reliability of thermometers that do not seem fundamental to epistemology? Another defender of a naturalistic account of epistemic norms offers a way ahead.  In his paper  �Progress or rationality? The Prospects for Normative Naturalism� [5] , Larry Laudan has explored the nature of �methodological norms�.   He has in mind pieces of advice such as:

Propound only falsifiable theories.

Avoid ad hoc modifications.

Since they are commands, he continues, �they appear decidedly not to be the sort of utterance which could be true or false, but at best only useful� (Laudan, 1987, p.202).  Moreover, he worries, �To enquire concerning their truth conditions seems a mistake, for they appear to be quite unlike ordinary statements.  Yet if they have no truth conditions, what would it even mean to ask about their warrant?�.(p. 203)

            His answer is that the grammar of methodological rules obscures their underlying logical form: they are hypothetical imperatives whose form is:

If one�s goal is y, then one ought to do x.

Thus the second of the Popperian maxims listed above should properly read:

If one wants to develop risky conjectures, then one ought to avoid ad hoc hypotheses.

The final twist of the argument is that such imperatives �always assert a relation between means and ends. Specifically, every such rule presupposes that �doing x� will, as a matter of fact, promote y, or bring us closer to the realisation of y.� (p. 203)

The conclusion Laudan draws from this is confusedly expressed.  He initially suggests that methodological rules are statements about instrumentalities.  Later formulations weaken this a little: the warrant of the rule is determined by the warrant for the underlying statement; and anyone who asserts the rule is committed to the truth of the means-ends claim.  The general point is close that these methodological norms are hypothetical imperatives: they concern means to ends. that a naturalised epistemology can provide extensive information about means to ends, and that methodological rules (and, perhaps cognitive norms more generally) are fixed by these facts about means to ends.

As we have seen, the idea that epistemic norms are all hypothetical imperatives is found in Quine too:

Insofar as theoretical epistemology gets naturalised into a chapter of theoretical science, so normative epistemology gets naturalised into a chapter of engineering: the technology of anticipating sensory stimulation. (Quine, 1990,  p.19)

Since Quine normally tells us that epistemology falls into place as a chapter of psychology,  normative epistemology is presumably applied psychology.  And, this quotation suggests, epistemic norms are hypothetical imperatives, concerned with means to the end of �anticipating sensory stimulation�. So we have a standard naturalistic response to the challenge of normative epistemology.  It consists of two claims:

I. All epistemological or methodological norms are hypothetical imperatives; they concern means to ends.

II. The instrumental truths upon which these imperatives depend can all be investigated within a naturalised epistemology.


4. Norms and normativity

            It would be hard to deny that there are epistemic norms of the sorts described: they are hypothetical in form and they can be supported by reference to empirical information of the kind investigated in the natural sciences.  Anyone unsympathetic to naturalised epistemology is likely to feel that this does not settle the matter.  For present purposes, we should note two forms that such unease can take.

            A norm is a proposition which formulates a standard to be followed in conduct or inquiry: it specifies the sorts of things we should take as reasons for action or belief; they state what we ought to do or believe.  As we have seen, there are questions about which norms we ought to accept, and both Laudan and Quine argue that choice of norms can be determined by empirical information about the reliability of instruments and the effectiveness of strategies of inquiry. It may be doubted whether this is true of all epistemological norms.  Are there true epistemic �ought� statements whose truth cannot be traced to instrumental facts about nature? We might suppose that there are fundamental principles of inductive logic which have normative status but which could not be settled empirically without circularity.  A second concern would be that we can make use of norms specifying how best to achieve our cognitive ends only if we have identified the norms that tell us what our cognitive ends should be.  Thirdly: even if the naturalist can make sense of the content of the norms that govern scientific inquiry, she will not be able to satisfy our questions about how we ought to carry out investigations of ethical matters or in mathematics.

            Even if the naturalist could satisfy her critics on all of these counts, a further deeper worry remains, one that Quine seems not to address at all.  The problem of identifying correct norms, it will be suggested, does not exhaust the normative concerns of epistemology.  The ability to use, and reflect upon norms is a characteristic feature of human rationality, one that distinguishes us from animals.  A naturalistic philosophy must not only provide us with materials for understanding which norms we should adopt.  It must also explain what is involved in being subject to norms, what it is to follow norms or to be subject to normativity.  This seems to require an account of how reasons function in the ordering of inquiries, of how active beings can take responsibility for their beliefs and opinions and so on. The deeper worry is that even if naturalised epistemology can account for what makes one normative standard better than another, it provides a model of the human agent that cannot make sense of how there can be responsible active users of norms at all.  It cannot make sense of normativity.  This charge underlies Davidson�s anomolous monism: our practice of thinking and evaluating through the use of propositional attitudes is subject to a �constitutive ideal of rationality� which cannot be fully explained in naturalistic terms.  It is also the basis of McDowell�s charge that �bare naturalism� is forced to reject the insight that thought and judgement involve �active employment of capacities that empower us to take charge of our thinking� (McDowell, 1994 p. 66).  Conceptual activity manifests �spontaneity�, a form of freedom which is linked to rational self control and responsibility. Since Quinean naturalised epistemology can make no sense of this, it provides no account of what is involved in being subject to norms. There is more to being normatively guided than knowing which norms to accept.

            Textbooks in epistemology are typically organised round a discussion of the proper analysis of some fundamental terms of epistemic appraisal: they discuss the nature of knowledge, justification, warrant, epistemic entitlement which belong essentially to reflection about the normative status of beliefs.  These are very broad terms of epistemic appraisal; many would hold that once we have shown that knowledge or justified belief are possible, we have done all that is required for the defeat of Scepticism and the vindication of our practice of inquiry. So closely are these terms linked to our ideas of epistemic normativity that one may think that if naturalism can make sense of these, then it will be able to make sense of epistemic normativity in general.  One strategy for a naturalised epistemology, in that case, would be to show that these �evaluative kinds� correspond to what can be seen as �natural kinds�, as kinds which can be described and understood using a naturalistic vocabulary. Some defenders of naturalistic epistemology take up this challenge.  Alvin Goldman, for example, in offering a reliability account of justified belief, acknowledges the central role of this concept in our practice of epistemic evaluation, and tries to make sense of it in naturalistic terms.

            Laudan seemed to think that the only �evaluative� term used in expressions of methodological norms was �ought�.  And Quine�s position, too, does not seem to recognise any need for a systematic explanation of these general evaluative concepts. This may support the view that more is needed to make sense of normative epistemology than an account of how instrumental truths can provide guidance about how to improve our skills as inquirers. Or, as often in the literature, it may be taken as a sign of how radical Quinean naturalised epistemology is.  He may be turning his back on problems as traditionally formulated; perhaps the focus on explaining knowledge and justification is an error. Perhaps we just say that a belief is �justified� when we think that it has been formed in a normatively sound way. There may be no fundamental forms whose expression requires the use of this vocabulary. Either way, we may see this as another focus for the anxiety that something important has been missed.  The remainder of this paper introduces some themes in Quine�s work which address these challenges. Before doing this, one further introductory observation should be made.


5. Normative epistemology: history,  psychology, science, common sense

Although I have presented them as adopting similar strategies in dealing with epistemic norms, there is a large apparent difference between Laudan and Quine.  For the former, the naturalistic basis for methodological morals is history: we learn how to do science better from seeing what has worked, and what has impeded progress, in the past.  Quine�s official view appears to be that theoretical epistemology is part of cognitive psychology and normative epistemology is applied theoretical epistemology.  Normative lessons, we would expect, will be obtained from psychology.

This expectation, it turns out, is misplaced. When Quine offers examples of how science can ground epistemic norms, his examples  rarely come from psychology.  Having noted how �Podiatry, appendectomy. and the surgical repair of hernias are technological correctives of bad side effects of natural selection�, he urges that �such also in essence is normative epistemology in its correcting and refining of our innate propensities to expectation by induction�.

A vest-pocket specimen of this is the exposure and correction of the gambler�s fallacy: the insidious notion that a run of bad luck increases the likelihood that the next try will win. (Quine, 1994a, p.50)

Such correction, we learn, �is the therapeutic side of statistical theory, a substantial branch of applied mathematics that is part and parcel of normative epistemology� (ibid). In that case, Quine�s naturalism turns to science for epistemological insight, but his focus on psychology is not an essential part of the position.

            Other comments make clear that Quine is not opposed to gaining help from the History of Science as well as other non-psychological sciences such as mathematics. In The Pursuit of Truth, he  recalls the list of five relevant epistemic virtues he had stated in The Web of Belief  (conservatism, generality, simplicity, refutability, and modesty), and adds: �Further counsel is available anecdotally in the history of hard science.� (Quine, 1990, p.20) [6]   I shall discuss the importance of �anecdotally� below.  We might also note that the merits of simplicity and modesty et al do not appear to be discoveries from neuroscience either.  These seem, rather, to be expressions of common sense, the results of ordinary non scientific reflections upon our practice. In his claims about normative epistemology, �naturalism� is taken in a relatively weak sense: we are to abandon first philosophy and the Cartesian dream, but all manner of a posteriori information may be grist to our epistemological mill.

            Indeed, we might think that psychological information will only rarely be what is required.  Worries about the reliability of various cognitive strategies or the value of different forms of belief formation will be expressed using a distinctive vocabulary, one that describes states and events as beliefs, inferences, methods and uses concepts such as inquiry, evidence, prediction and the like.  Although these have a place in our everyday talk and thought about epistemic matters, it is debatable whether they will have much role within cognitive psychology, especially if we share Quine�s apparent taste for connectionist approaches. In order to apply psychological results (and other results from theoretical science) to our everyday practices of inquiry, we need some kind of bridge between two vocabularies: the theoretical categories used in formulating scientific theories; and the terms we ordinarily use in describing and regulating our beliefs, inferences and investigations. Much a posteriori information should be available and relevant to assessing our beliefs and methods without taking a detour through theoretical psychology.

            The matter just discussed receives very little attention in Quine�s work. The reader who is impressed by his radical naturalism might suppose that he is one of those eliminativists who would propose that we change the vocabulary in which we describe ourselves and our cognitive products as our theoretical understanding of nature grows. [7] The picture I have sketched, it may be objected, accords more with the anomolous monism of Davidson than the physicalist naturalism of Quine.  Many texts support this, but we need to take account of those recent writings in which Quine insists that he too is an anomolous monist (1990, pp. 71-2). He does not relate this directly to epistemological themes, nor is the idea developed in detail. But we should take seriously the possibility that he thinks that many of our everyday concerns are guided by an intentional vocabulary which is not reducible to physical language but which is used in reflecting on beliefs, actions and inquiries.  Using this vocabulary, we express norms which exploit information acquired in an a posteriori manner, either in the sciences of on the basis of everyday common sense information. If that is right - if we are correct to take seriously Quine�s claim to be an anomolous monist - then we may wonder whether part of his response to worries about normativity may involve reference to the governing ideals of the use of intentional language.

            What does Quine say about this �anomolous monism�?  He insists that there are �irreducibly mental ways of grouping� events: �grouping a lot of respectably physical perceptions as perceptions that p, and grouping a lot of respectably physical belief instances as the belief that p�. (Quine, 1990, p, 71). Moreover there is �no dismissing� this way of talking since it �implements vital communication and harbors indispensable lore about human activity and motivation, past and expected.� Since it is irreducible, we have �all the more reason to treasure it�.  All that it provides would be lost if it were abandoned and �there is good reason not to try to weave it into our scientific theory of the world to make a more comprehensive system.� (p.71). It is difficult to escape the conclusion that we possess a lot of �indispensable lore� about belief, inquiry and reasoning, which is invaluable in ordering and evaluating our beliefs, and which we should not try to embed in our scientific theory of the world.  Cognitive psychology will not solve all problems about the normativity of cognition.

We have noticed a tension (albeit an unimportant one) in Quine�s characterisations of this naturalised epistemology.  Although he often insists that epistemology should be transformed into a chapter of theoretical and experimental cognitive psychology and he urges that normative epistemology should thus stand to this discipline as engineering stands to the hard sciences, his examples of normative epistemology owe little to cognitive psychology  Mathematics and probability theory, anecdotal history of science, observations cast in our everyday intentional vocabulary, seem to have more to contribute to regulating our inquiries than does scientific psychology. [8]   Given Quine�s broad understanding of �science� this need not manifest any inconsistency in his views, but it does make mysterious the central role he attaches to psychology.

Division of labour may provide part of the explanation.   Explicit �heuristic� norms govern the process of theory construction - �thinking things up�.  Theory testing is less susceptible to explicit normative control - he has said that induction does not present a body of normative standards, because we have no choice over whether to rely upon induction.  Perhaps psychology is especially relevant to understanding our inductive procedures, to understanding how evidence is related to theory.  In that case, psychology may have special relevance to the first of our two normative questions - that of grounding our confidence in our cognitive and rational capacities. Quine�s emphasis on psychology (and on stimulations) may reflects his heavy commitment to empiricism. Were it not for his tough minded empiricism, psychology would be less central to Quine�s characterisation of his practice.

            This is supported by a passage from �Epistemology naturalised� in which Quine explains why insights from the history of science should be treated with care.

The dislodging of epistemology from its old status of first philosophy loosed a wave, we saw, of epistemological nihilism.  This mood is reflected somewhat in the tendency of Polanyi, Kuhn and the late Russell Hanson to belittle the role of evidence and to accentuate cultural relativism.  Hanson ventured even to discredit the idea of observation, arguing that so-called observations vary from observer to observer with the amount of knowledge that the observers bring with them.  The veteran physicist looks at some apparatus and sees an x-ray tube.  The neophyte, looking at the same place, observes rather �a glass and metal instrument replete with wires, reflectors, screws, lamps and pushbuttons�.  One man�s observation is another man�s closed book or flight of fancy.  (Quine, 1969, p.88)

Hence Quine�s central interest in psychology over history depends upon the belief that �evidence� and �observation� have to be made fundamental if we are not to collapse into �nihilism� or �relativism� and the belief that those who have turned to history for a naturalistic basis for epistemology have failed to avoid this danger.  Indeed, this emphasis was already present in his description of how epistemic norms are to be understood as hypothetical imperatives: the end to which we seek efficient means was stated explicitly in terms drawn from Quinean armchair psychology - �anticipating sensory stimulations�.  Even if he is abandoning �first philosophy�, he wants to retain psychology as a kind of �first epistemology�.

            Quine holds that if we want to understand how theories are developed on the basis of our sparse experiential input, then we have no alternative but to turn to cognitive psychology and social psychology: let us say, to cognitive science.  There is an intelligible cognitive task which can best be carried out scientifically. In each case, we might think that Quine is too Cartesian, focusing on individual psychology and ignoring the social and co-operative nature of these activities.  Moreover such investigations are likely to throw up information which we can use to improve our performance: we can be alerted to unexpected disrupting influences or to ways in which we can be led into error.  But these inquiries are largely theoretical, concerned with enabling us to understand how we form our beliefs or make our evaluations. [9]   Since Hume, Locke and (presumably) Carnap were largely concerned with these explanatory ventures when they studied cognition, it is not surprising that Quine saw continuities between scientific studies of cognition and various enterprises within the earlier epistemological tradition.

            Since some of these earlier studies of cognition were largely introspective, dealing with observable relations between metal states that were assumed to be open to the introspective gaze, it is unsurprising that they fed directly into claims about normative epistemological standards.  Psychological theory was not very different from careful introspective reflection. Studying the nature of cognition and studying the normative structure of the ways in which we reflect upon our beliefs, actions, plans and deliberations were not wholly separate investigations.  If psychological categories were grounded in introspection, it is unsurprising that they are well suited for the descriptions of reasoning and beliefs that we use when trying to regulate our inquiries.  Once we move beyond an introspective psychology, there is no reason to expect this to continue to be true. One can be a naturalist and allow that all standards of rationality can be reassessed in the light of experience, without adopting the belief that normative epistemology is simply applied cognitive science. Descriptions of cognition which are cast in vocabularies well suited to the attempt to reflect upon, and improve, our ways of thinking about the world are more likely to be found elsewhere.


6. Scientific inquiry: aims and checkpoints

            When we introduced the idea that epistemic norms concern means to cognitive ends, we considered the objection that there must also be normative issues about the ends we should adopt. Since Quine treats epistemology as co-extensive with the epistemology of science (however narrowly or widely construed), this becomes a question about the goal or goals of scientific investigation. Recall the passage from Pursuit of Truth:

Insofar as theoretical epistemology gets naturalised into a chapter of theoretical science, so normative epistemology gets naturalised into a chapter of engineering: the technology of anticipating sensory stimulation. (Quine, 1990, p.19)

It would be natural, but mistaken, to infer from this that the aim of science is �anticipating sensory stimulation�. The most common goal for scientific inquiry is �technology and understanding�.

�Anticipating sensory stimulation� is the checkpoint, of science, not the goal. A theory should be rejected if its predictions fail, and it can be accepted so long as its predictions succeed: the test of an acceptable theory is that it enables us to anticipate sensory stimulation. It does not follow from this that my goal in seeking theory is to be able to make accurate predictions of future experience. But whatever my goal is, I will not succeed in achieving it unless I reach a theory that is predictively adequate. My goal may be understanding; theories which fail to pass the checkpoint will not provide the right kind of understanding.  Or my goal may be a practical one, building a bridge or curing a disease; once again, my theory will not serve that purpose unless it successfully anticipates sensory stimulation. Perhaps better, I will not accept a theory unless it passes this test, and it is for this reason that it will not enable me to achieve my goal.

Whether �anticipating sensory stimulation� is the goal of science of not, there is a question about why I should accept this as a checkpoint.  It seems to be a fundamental norm that cannot be expressed merely as a means to some further end. In fact, Quine insists that it �is not normative�: �I see it as defining a particular language game, in Wittgenstein�s phrase�, �the game of science, in contrast to other good games such as fiction and poetry.� (Quine, 1990, p.20) He seems to hold that it is an a priori truth, indeed analytic (!), that a form of inquiry is scientific if and only if it is governed by this criterion of success.  In a telling analogy, we learn that prediction �is what decides the game, like runs and outs in baseball�. To seek scientific understanding is to seek understanding that is provided by a theory that has met this checkpoint. [10]   It is compatible with this that �empiricism� is normative: it decrees that the best way of arriving at theories that anticipate sensory stimulation is to rely upon the familiar senses, rejecting the deliverances of clairvoyance or (perhaps) mathematical intuition.  We could conceivably learn that science should abandon empiricism, so understood, and accept these further kinds of interaction with the world as sources of �sensory stimulation�.

In that case cognitive ends, whether to seek understanding of this matter or that, whether to use the scientific method to tackle this problem or answer that question, may be matters of practical rationality.  At least they seem to raise no normative issues about ultimate epistemic ends which need     disturb the Quinean programme. There does remain a question about how �anticipating sensory stimulation� can serve as a checkpoint for those who simply lack this concept: he must hold that we can use this as a checkpoint even if we are not conscious that this is what we are doing.  I shall return to this topic below.


7. Limiting the scope of normative concerns

            It is natural to suppose that our engagement with norms in epistemic matters can take a number of forms.  First we sometimes address normative questions: we do not know how we should proceed in our investigations, so we ask what we ought to do. Second, we are aware of the force of norms when we find ourselves to be rationally constrained. When I believe that Socrates is a man and also believe that all men are mortal, I have no choice but to believe that Socrates is mortal.  This is not experienced as a mere physical compulsion; the fact that I am bound to form this belief coincides with it�s being the right thing to do.  The judgement that, in the light of my other beliefs and commitments, I ought to believe that Socrates is mortal has no role in guiding my inquiry.  As we saw above, there are manifestations of �normativity� which do not involve the acceptance and application of action or believe guiding normative principles.

            I have already suggested that Quine says little about this second kind of phenomenon.  When he says that normative epistemology is concerned with means to ends, he means that whenever we raise a normative question, whenever it is an open question what we ought to do, information about instrumentalities will provide us with a solution. Indeed, he constrains the scope of normative epistemology even further.  Claiming that �naturalised epistemology on its normative side is occupied with heuristics generally - with the whole strategy of rational conjecture in the framing of scientific hypotheses�, he notes that in The Pursuit of Truth, he has been focused on �the testing of a theory after it has been thought up� and has thus �passed over the thinking up, which is where the normative considerations come in.� This seems to add to the above interpretation the thought that genuine normative questions are all concerned with heuristic matters of  arriving at new theories for inductive testing.  This seems rather implausible.  We would expect there to be norms of theory testing as well as norms for theory testing. In that case, we should consider in more detail some things Quine says about induction.

            Let us consider a passage where Quine considers what naturalised epistemology has to say about our inductive habits.  He rejects the need to explain why we are �entitled� to rely upon induction as resting upon a false presupposition: we are not �entitled�, but we are �bound to�:

We are bound by the innate wiring of our nervous system, conferred by natural selection.  Our tendency is to expect similar stimulations to have sequels that are similar to each other.  the similarity here is subjective, and our scale or standard of similarity is innate but subject to modification through experience.  Natural selection has favoured in us a similarity standard that has made for largely successful induction by meshing pretty well with the regularities of nature over the past few dozen millennia.

            This, plus the happy circumstance that nature has apparently persisted pretty well in her old ways right down to the present day, accounts for the continuing success by and large of induction.  Our genes account for our still continuing to rely on it..  But all this is compatible with a major change, right now, in the course of nature, so I see no entitlement [to rely upon induction].  Such a change would be contrary to our firmest scientific laws, but to argue thus is to argue inductively, begging the question.

            Where evidence can be decisive is rather on the negative side: refutation of an observational categorical by an observed counter-instance.  This is how some of our false inductions get weeded out, and how science keeps a grip on reality.  (Quine 1994a, p. 502f)

A paragraph or so later comes what might be Quine�s key move: �But traditional epistemology enjoys no evident advantage over naturalism in these normative matters� (ibid p. 503).

            This passage invites some comments.  Quine resists the poor argument that we are warranted in relying on induction because natural selection would not provide us with inductive habits if they were not adaptive.  Since our cognitive needs differ from those of our evolutionary forebears, the evolutionary argument does not provide any reason for believing that our habits and capacities are well fitted to our current refined cognitive needs.  Striking, although not emphasised as much as may be appropriate, Quine is committed to the importance of our having the capacity to revise and develop our inductive habits in the face of predictive failure - our similarity standard is innate �but subject to modification through experience� and, he should have added. education.  So we possess a body of inductive habits, which we are able to modify and develop in the light of our other knowledge and experience, which we have never been given any reason to doubt.  Indeed those inductive habits confirm us in trusting them.  If our reliance on induction does not meet the strong degree of support that he takes to be required for �entitlement�, we can see how reflection does not threaten our confidence in it.  And we can see how empirical information - about our successes and about our flexibility, as well as our growing knowledge of the mechanisms involve all contribute to reinforcing that confidence.

            If the issue is one of legitimating our confidence in our cognitive capacities, Quine�s worries about circularity seem out of place [11] .  Our confidence legitimately grows with success.  The possibility that all might change tomorrow may not be answerable by an argument that provides any sort of guarantee, but nor does it give rise to any strong sense of epistemic vulnerability.  At worst, it brings out that we are fallible, and dependent upon the grace of nature.  Thus far it seems reasonable to agree: there is no reason in principle why a naturalized approach to the study of epistemic rationality should be any worse placed than �traditional epistemology� to legitimate our confidence in our cognitive skills.  The burden of proof lies with someone who wants to deny this: they must point to some particular aspects of our reflective practice where our confidence would be shaken by the information that only �naturalistic� information was available.

            When we turn to explicit norms, the response is essentially the same.  We are aware of many examples of mathematical or psychological investigations bring out inadequacies in our reasoning habits. Once we are informed about the gambler�s fallacy, we shall see the force of norms which call for care and caution before placing bets.  Psychological studies of inference - for example, the debate that has grown up around the Wason selection task - alert us to areas where mistake is likely; and we can learn to avoid these errors by adopting norms to govern our deliberations and conversations.  So there is not problem in accepting that a naturalized study of knowledge can ground norms of epistemic rationality.  An adequate criticism on this basis would have to show that there are more fundamental norms that cannot receive this kind of vindication.  Once again, the burden of proof lies with someone who would persuade us that this is actually the case.  So far, we have no reason to think that naturalized epistemology is any worse off than traditional epistemology.


8. An objection to naturalised epistemology

We might clarify things a little further by considering an argument by Laurence BonJour.  It aims to show that a naturalistic approach to epistemology is somehow self-defeating: in general, an adequate epistemology must acknowledge that some beliefs about what is a reason for what are knowable a priori.  The argument assumes that there is a distinction between those beliefs which are immediate reports of the contents of experience and those which �cannot be construed as strictly observational or experiential in any sense that has the slightest plausibility: beliefs about the remote past, beliefs about the future, beliefs about present situations where no observer is present, beliefs about general laws, the vast majority of the beliefs that make up theoretical science, and perhaps others� (BonJour, 1994, p. 295f).  He then asks what sort of belief can count as a reason for one of these non-observational beliefs.   The reason must either �(i) depend on an inference of some sort from some of he directly observational beliefs or (ii) be entirely independent of direct observation. A reason of sort (ii) is plainly a priori�.  A reason of type (i) can obtain only if a conditional statement of the form:

If p1, p2, p3, etc are true then q is true  (where p1 - p3 are observational and q is non-observational).

If all the relevant observational material is included in the antecedent, then a conditional of this kind �can only be a priori.�  Thus if any beliefs about non-observational matters are to be justified, there must be a priori statements of warrant, and this conflicts with the claim that all normative matters can be examined a posteriori.  BonJour finds this argument �as obvious and compelling as anything in philosophy�.  His use of the argument is allied to his identification of the real gap in naturalized approaches to epistemic issues: the naturalist cannot provide any real reason for thinking that his beliefs are true. He believes that this is something that cannot be provided without engaging with sceptical arguments, and he thinks (plausibly) that we can only confront Scepticism with the aid of a priori reasoning.  Since we have already expressed the hope that a naturalised approach to epistemology can meet the normative demands that actually arise in the course of reflective inquiry, I will not say anything more here about the latter move.  The question is rather: what should a defender of naturalism say about BonJour�s �obvious� and �compelling� argument.

            The applicability of the argument to Quine�s position is clear.  Testing theories, he argues, involves deriving �observation categoricals� from collections of more theoretical beliefs and sentences.  Such derivations must be licensed by conditional propositions which can be used to assert that if the more theoretical characterisation is true, then experience will be as described in the observation categoricals.  BonJour�s challenge concerns the status of the conditional propositions which license these derivations: what kind of �warrant� do they possess? If Quine�s position is as I have described, then two lines of response (at least) are available to him.

            First, we can object that a priori is a vague and unclear term. A naturalist would object to a notion of a priority which traced such knowledge to our intellectual acquaintance with the contents of some platonic realm.  Equally unsatisfactory for Quine would be an explanation which saw them as analytic in a very strong sense: possessing determinate meanings which ensure that they are absolutely unrevisable in principle. But there are notions of a prioricity which are less problematic to a Quinean epistemologist.   For a moderate holist like Quine, a belief can be held firm by its role in sustaining or integrating a substantial body of information even if the believer cannot point to the �reasons� that warrant him in holding it.  Suppose that the agent is reflectively confident in the habits of memory, inference and reflection which lead to the disposition to assent to the proposition that q.  Since the weighing of evidence is `a largely passive affair�, no specific doubts or questions are (or, perhaps, even can be) raised by reflection on that belief.  In a very loose sense, this may be described as a priori: it is not a candidate for empirical assessment or re-assessment and it is taken for granted in providing reasons for believing other propositions. It is currently unnegotiable, and we could not identify particular observations which would count for it or against it.  Alternatively it may be analytic, in Quine�s favoured sense: taught and learned in the process of instructing someone in the use of the words contained in its expression.  The belief is warranted, and its warrant cannot be traced to the support of particular experiences or experiments.  Does that mean it has a �a priori� warrant?  That depends upon what �having a warrant� means.  BonJour�s argument appears to assume that for a belief is justified or warranted only if there is something that �constitutes its warrant�. Whether that is compatible with use of a loosened notion of a priority may not be fully clear.

            BonJour seems to believe that the relations of epistemic dependence that hold our beliefs together have a clear and articulated structure which we can identify (he calls this the �doxastic presumption�).  If this were the case, then I can see that reflection would lead us to identify these structures of argument and raise questions about their soundness.  BonJour�s challenge may then be a pressing one.  Once this assumption is abandoned (as surely it must be) then the focus of epistemic evaluation becomes more restricted: it is a feature of our cognitive economy that particular propositions and arguments emerge out of the murk and call for reflective assessment.  My suggestion has been that it is not yet evident that the questions that arise in this way cannot be answered using normative standards whose soundness can be assessed in an a posteriori manner.

            The second moral I wish to draw from this discussion of BonJour�s remark can be put in slightly different terms.  Contrast two pictures of epistemic rationality.  One, which is characteristic of much recent �epistemology� holds that we carry out these evaluations using a general evaluative concept of justification.  Epistemology must tell us what it is for a belief to be justified; unless each of our beliefs is justified our overall cognitive position is flawed; and so responsible reflective inquiry requires us to be able to tell, for each of our beliefs, whether it possesses this virtue of �justification�.  BonJour�s challenge seems to be a call for us to identify the kind of �justification� that his conditional possess.  This view treats justification as a sort of �evaluative kind�.  The second picture is sceptical that we possess clear substantive general epistemic virtues such as �justification�.  However in the course of carrying out inquiries and testing our opinions, questions can arise about how we ought to go about answering particular problems, how we ought to respond to a surprising experience, whether we ought to trust our instincts in some particular case and so on.  The Quinean epistemologist that I am considering holds that these specific issues can be addressed in a naturalistic manner.  Although we might say that a belief is justified when it has been arrived at in a normatively adequate manner, we do not need to ground our practice of epistemic rationality in any sort of substantive account of �justification�.  It is not implausible that if you have the former picture, naturalism may turn out to be unattractive. The naturalistic picture that I am examining resists this picture.


9. Conclusion: epistemic means and ends

. We should conclude by raising an obvious difficulty.  Both Laudan and Quine suggest that methodological norms all concern means to ends.  But surely there are normative questions about what our ends should be: can a naturalised approach to epistemic rationality say anything sensible about these questions? What should our cognitive ends be? [12]  

            It is tempting to view this as a question about the form to be taken by the hypothetical imperatives that express the epistemic norms: what are the prescribed standards of belief and inference means to?  Is it truth, empirically adequate belief, or what?  Since both Quine and (especially) Laudan [13] insist that the goals motivating scientific activity can vary over time, there would seem to be genuine questions about what these aims should be.  In that case, we need normative guidance with the task of choosing them.  It is surprising that neither seriously addresses the issue.  In the case of Quine, this is especially surprising because he take the presence of competing �ends� for morality to be a source of a major disanalogy between science and ethics, one which makes ethical realism impossible.  We have seen that in the scientific case, he believes that the variety of goals which can motivate scientific activity have it in common that they can be pursued only by seeking theories that pass the checkpoint of experience.  Hence norms that relate to that goal, together with ideals of simplicity etc, will be common to all versions of the �scientific language game.� Quine�s confidence that normative epistemology is wholly concerned with means - ends matters appears to depend heavily on his controversial claims that an activity is scientific only if its success depends upon our finding beliefs and theories that pass the checkpoint of experience and that issues of normative epistemology are wholly concerned with how this checkpoint can be passed.

            Does reflection upon our beliefs, deliberations and inquiries give rise to any questions about the rationality of ends which cannot be transformed into questions about means to ends? If so, does this raise a serious difficulty for any naturalistic treatment of the matter?  I assume that no interesting questions arise from questions about which language game to play - whether to be a scientist or a novelist.  I also assume that no particular problems arise about which particular theoretical issues we should occupy ourselves with - whether to be philosophers or psychologists, which problems to work on and so on. The remaining question, to be pursued elsewhere, concerns whether the scientific research requires us to take account of other explicit normative issues, ones that cannot be tackled in this way.



BonJour, L. (1994). Against Naturalised epistemology. Midwest Studies in Philosophy, XXIX, 283-300.

Elgin, C. (1996). Considered Judgement. Princeton: Princeton University Press.

Haack, S. (1993). Evidence and Inquiry. Oxford: Blackwell.

Laudan, L. (1987). Progress or Rationality: The Prospects for Normative Naturalism. American Philosophical Quarterly, 24, 19-31.  Reprinted in: Papineau (ed) (1996) Philosophy of Science. Oxford: Oxford University Press, 194-214.  References to this reprint.

Laudan, L. (1990). Normative Naturalism. Philosophy of Science, 57, 44-59.

McDowell, J. (1994). Mind and World. Cambridge MA: Harvard University Press.

Nozick, R. (1993). The Nature of Rationality. Princeton: Princeton University Press.

Quine, W.v.O.  (1969) Ontological Relativity. New York: Columbia University Press

Quine, W.v.O.            (1990). Pursuit of Truth, Cambridge MA: Harvard University Press.

Quine, W.v.O.            (1994a). From Stimulus to Science. Cambridge MA: Harvard University Press

Quine, W.v.O.            (1994b) Responses.  Inquiry, 37, 495-505..




University of Sheffield

[1] There are many passages in which Quine allies himself with those who �repudiate the Cartesian dream of a foundation for scientific certainty firmer than the scientific method itself�.  But, he announces, �I remain occupied � with what has been central to traditional epistemology, namely the relation of science to its sensory data. (Quine 1990, p. 19)

[2] The suggestion that means ends issues can be settled by using empirical information was made plausible by appeal to examples - science can tell us how reliable a thermometer is.  Such examples do not suffice to establish that all issues about the relations of means and ends can be settled empirically.  Normative naturalism in epistemology is committed to the claim that all the means ends issues that arise in epistemology can be settled empirically - even if there are apparent means issues arising elsewhere that do not have this property.

[3] I should note one further complication here. The suggested strategy meets the challenge of explaining how naturalism can handle normative issues in epistemology only if a naturalistic account can be provided of the rationality of means/ ends reasoning.  It is not clear just how problematic means/ ends rationality is. We should take it that the view under discussion is that questions of epistemic normativity raise no additional problems once we understand the nature of instrumental norms.

[4] This possibility might be incorporated in the pluralistic outlook found in Davidson�s Anomolous Monism.

[5]   No significance probably attaches to the fact that in his latest book, Quine he used the name �normative naturalism� for his own position.  (Quine 1994a, pp. 49f)

[6]             However he does seem to suggest that �thinking things up� may resist very precise or rigorous sets of heuristics.  �Conservatism� and simplicity appear to have �no general calibration�, �much less any comparative scale of the one against the other�: �For this reason alone - and it is not alone -there is no hope of a mechanical procedure for optimum hypothesising,  Creating good hypotheses is an imaginative art, not a science. It is the art of science�. (Quine, 1994a, p. 49)

[7] That such revision should occur has been proposed by Paul Churchland, whose work is cited by Quine with admiration.

[8] Susan Haack makes this observation in Evidence and Inquiry (Haack, 1993) and urges that the attempt to cast epistemology in scientific terms is an error.

[9] In saying that an investigation within cognitive science can answer theoretical questions in these areas, I am not committed to the claim that such investigations can answer all relevant theoretical issues about how we carryout inquiries or make evaluations. This seems extremely unlikely.

[10] This may be relevant to Quine�s new found anomolous monism: do our ordinary idioms of intentional description and explanation (described in (Quine, 1960) as a �dramatic idiom�) form part of a different �game� subject to different standards of success?

[11] Elsewhere, Quine makes similar remarks about induction and natural selection.  He addresses the question �Why should our innate subjective spacing of qualities have a special purchase on nature and a lien on the future� and after pointing out (memorably) that �Creatures inveterately bad in their inductions have  a pathetic but praiseworthy tendency to die before reproducing their kind� he concludes that �These thoughts are not meant to justify induction.  ..What natural selection contributes ... is a reason why induction works, granted that it does.  (Quine, 1969, p. 126f)

[12] The view that �principles of rationality� should all be defended in means-ends terms is also defended in Robert Nozick�s The Nature of Rationality (Nozick, 1993, chapter 1).  It is part of  his view that we justify adopting something as a principle by showing that doing so can strengthen our motivation for doing what it requires.  It accords with the picture being discussed here that principles are adopted because of the practical benefits of doing so; they are not adopted because they are somehow implicit in the foundations of the systems of evaluation that we are concerned with.

[13]    A major concern of his papers is to show that we can gain methodological insights from (say) Newton�s achievements in spite of he fact that one of his aims was �show the hand of the Creator in the details of his creation� and to �discourse of [God] from the appearances of things�.  His conception of the aim of science was strikingly different from ours.