Jump to ContentJump to Main Navigation
Women, Culture, and DevelopmentA Study of Human Capabilities$

Martha C. Nussbaum and Jonathan Glover

Print publication date: 1995

Print ISBN-13: 9780198289647

Published to Oxford Scholarship Online: November 2003

DOI: 10.1093/0198289642.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (oxford.universitypressscholarship.com). (c) Copyright Oxford University Press, 2021. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use. date: 15 June 2021

Pragmatism and Moral Objectivity

Pragmatism and Moral Objectivity

(p.199) Pragmatism and Moral Objectivity
Women, Culture, and Development

Hilary Putnam

Oxford University Press

Abstract and Keywords

Putnam engages with moral objectivity and the question of ethical truth in this paper, in which he combats the idea that there is no intellectual structure worth taking seriously to the arguments of American pragmatists such as Charles Peirce, William James, and John Dewey. As a countercurrent to contemporary analytic philosophy, Putnam interprets and builds on the work of Dewey so as to yield the conclusion that there can be a rational basis for adopting ethical positions and that democratic processes are necessary constituents of social rationality.

Keywords:   American pragmatism, John Dewey, William James, indispensability arguments, instrumental rationality, objectivity, Charles Peirce, practice, truth

The papers in the present volume resulted from the happy idea of bringing together philosophers used to dealing with issues of justice, issues about what constitutes flourishing for an individual or for a society, feminist issues and development issues, and also some people who, like Martha Chen, have firsthand knowledge of what those issues mean in the lives of the women and men of developing countries. (Some of us, like Amartya Sen, fell into both of these categories.) This paper, like the others, was and is intended to contribute to the discussion of those issues. If it has ended up being more ‘abstract’ than most of the others, that is not because the author got ‘carried away’ by a particular line of abstract thought. Rather, it is because it was my conviction, as well as the conviction of the organizers of the conference, that positions on the ‘abstract’ question of moral objectivity have real world effects. In one of his major works,1 Jürgen Habermas repeatedly expresses the conviction that Max Weber's conclusion, that modern scientific rationality requires us to accept it as a fact that there can be no such thing as a rational foundation for an ethical position, and that we are ‘committed to a polytheism of ultimate values’, is both wrong on intellectual grounds and disastrous in its effect on the ‘lifeworlds’ of ordinary women and men. Of course no one can tell just how much influence a wrong philosophy has in contributing to such real world effects as the instrumentalization of human beings and the manipulation of their cultures. Nevertheless, to show that the justifications which are offered for ethical scepticism at a philosophical level will not stand up to examination, that the foundations of the idea that there is no rationality beyond purely instrumental rationality are in trouble, may help to combat that instrumentalization and that manipulation. This was also the conviction of John Dewey—whose philosophy I discuss in the fourth section of this paper—throughout his long philosophical life. And it was in that spirit that I decided (and was in fact encouraged) to devote my paper to the ‘abstract’ issue of moral objectivity.2

But not every defence of moral objectivity is a good thing. We live in an ‘open society’, a society in which the freedom to think for oneself about (p.200) values, goals, and mores is one that most of us have come to cherish. Arguments for ‘moral realism’ can, and sometimes unfortunately do, sound like arguments against the open society; and while I do wish to undermine moral scepticism, I have no intention of defending either authoritarianism or moral a priorism. It is precisely for this reason that in recent years I have found myself turning to the writings of the American pragmatists.

In my case, turning to American pragmatism does not mean turning to a metaphysical theory. Indeed, the pragmatists were probably wrong in thinking that anyone could provide what they called a ‘theory of truth’, and Peirce was certainly wrong in thinking that truth can be defined as what inquiry would converge to in the long run. What I find attractive in pragmatism is not a systematic theory in the usual sense at all. It is rather a certain group of theses, theses which can be and indeed were argued very differently by different philosophers with different concerns, and which became the basis of the philosophies of Peirce, and above all of James and Dewey. Cursorily summarized, those theses are (1) antiscepticism: pragmatists hold that doubt requires justification just as much as belief (recall Peirce's famous distinction between ‘real’ and ‘philosophical’ doubt); (2) fallibilism: pragmatists hold that there are no metaphysical guarantees to be had that any of our beliefs will never need revision (that one can be both fallibilistic and antisceptical is perhaps the unique insight of American pragmatism); (3) the thesis that there is no fundamental dichotomy between ‘facts’ and ‘values’; and (4) the thesis that, in a certain sense, practice is primary in philosophy.

The pragmatists did not only have an interesting set of very general theses, however. Contrary to the way James and Dewey in particular are often painted, they all had intricate arguments: arguments which are still not well known among analytic philosophers, and in the present essay I wish both to describe and also to build upon some of those arguments.

If one sees the four theses I listed as the fundamental theses of pragmatism, then one will not be surprised to learn that the several pragmatist arguments for ethical objectivity were related from the beginning to the effort to preserve and perfect the open society, and to preserve what was right in the epistemological critiques of authoritarianism and a priorism that appeared on the scene together with modernity.

1 ‘Indispensability Arguments’ In Science and in Ethics

As I just remarked, pragmatists believe that, in ‘a certain sense’ practice is primary in philosophy. It is necessary to say a little more about what that means, and doing so will bring us to the first of the arguments I have in mind. In philosophy of science, the most celebrated example of an appeal (p.201) to practise in the course of a metaphysical argument is due to Quine: I am referring to what I have elsewhere3 called his ‘indispensability argument’ for the acceptance of the conceptual scheme of set theory. Quine ignores the problem which goes back, in part, to Plato, as to how we can know that abstract entities exist unless we can interact with them in some way. Instead of arguing that we can causally interact with abstract entities, as one or two philosophers have,4 or contending that we have some sort of supercausal or noncausal interaction with abstract entities (perhaps with the aid of the faculty of ‘reason’), Quine deliberately displaces the discussion to a more humanly accessible region. Quine replaces questions about the ‘mystery’ of interaction with abstract entities with the question ‘Why do we need the conceptual scheme of sets?’5 Quine's argument is that the the conceptual scheme of set theory is indispensable to mathematics, and indeed to physical science6 as well, and what is indispensable our best paradigms of knowledge cannot, Quine argues, be criticized from some supposedly ‘higher’ philosophical viewpoint; for there is no ‘first philosophy’ above and outside of science.

To be sure, Quine is in many ways a most atypical pragmatist. Although Morton White has argued early and late7 that given Quine's own doctrines, in particular Quine's rejection of the ‘dualism’ of analytic and synthetic, Quine ought to reject the fact/value dichotomy, Quine has stubbornly refused to ‘go along’ with his old friend White. Indeed, Quine seems to have rejected one dichotomy (the dualism of analytic and synthetic) only to replace it with another (the dualism of science, or rather science as formalized in the notation of symbolic logic, and everything that Quine regards as of only ‘heuristic’ value, including intentional idioms, and value idioms). Although these elements of Quine's philosophy are more reminiscent of Viennese positivism than of American pragmatism, within the sphere of science Quine does give an account which is strongly influenced by pragmatism. I am thinking not only of the ‘indispensability argument’, but also of Quine's insistence8 that the scientific method is not an algorithm, but an informal matter of ‘trade‐offs’ between such desiderata as preservation of (p.202) past doctrine, predictive efficacy, and simplicity. This account is strikingly reminiscent of, for example, William James.9

In any case, the classical pragmatists—Peirce, James, and Dewey—whatever their disagreements, did not hesitate to apply to the topic of values exactly the kind of indispensability argument that Quine applies to the topic of scientific ontology. Just as Quine displaces the traditional worry as to the mystery of our ‘interaction’ (if any) with abstract entities such as sets and numbers, the classical pragmatists displaced the equally ancient worry as to the nature of our ‘interaction’ (if any) with ethical properties and with value properties and normative properties generally. Like Quine, they found it fruitful to replace metaphysical conundrums with the more humanly accessible question, ‘What function does this discourse serve?’ According to the pragmatists, normative discourse—talk of right and wrong, good and bad, better and worse—is indispensable in science and in social and personal life as well. Even relativists and subjectivists frequently concede this; we make and cannot escape making value judgements of all kinds in connection with activities of every kind. Nor do we treat these judgements as matters of mere taste; we argue about them seriously, we try to get them right, and, as philosophers have pointed out for centuries, we use the language of objectivity in our arguments and deliberations: for example, we use the same laws of logic when we reason about an ethical question that we use when we reason about a question in set theory, or in physics, or in history, or in any other area. And, like Quine, the classical pragmatists do not believe that there is a ‘first philosophy’ higher than the practice that we take most seriously when the chips are down. There is no Archimedean point from which we can argue that what is indispensable in life gilt nicht in der Philosophie. Here at least, pragmatism rejects revisionary metaphysics.10

Of course, there is an obvious difference between set theory and ethics: there is a wide body of accepted doctrine in set theory (even if nominalists and intuitionists reject it), and there is a vast amount of disagreement about values in the modern world. The fact that pragmatists offer indispensability arguments in both areas does not mean that pragmatism fails to take moral (p.203) disagreement seriously. For pragmatists like John Dewey, moral disagreement of the kind that we find in an open society is not a metaphysical problem but a political problem—a problem and a challenge. The problem is to keep moral disagreement within the bounds of community and productive co‐operation, and the challenge is to make moral disagreement serve as a stimulus to the kind of criticism of institutions and values that is needed for progress towards justice and progress in enabling citizens to live in accordance with their various conceptions of the good life. Pragmatism anticipated an idea that has become a commonplace in contemporary moral philosophy,11 the idea that disagreement in individual conceptions of the good need not make it impossible to approximate (even if we never finally arrive at) agreement on just procedures, and even agreement on such abstract and formal values as respect for one another's autonomy, non‐instrumentalization of other persons,12 and such regulative ideas as the idea that in all our institutions we should strive to replace relations of hierarchy and dependence by relations of ‘symmetric reciprocity’.13 Ruth Anna Putnam and I have written elsewhere14 about the ways in which the classical pragmatists attempt to justify these ideas, and I shall also speak about this in what follows.

2 An Important Objection

Before we proceed any further, however, it is necessary to face a very important objection to the whole idea of ‘the primacy of practice in philosophy’. The objection is most easily presented in the form of a parody, a parody of the ‘indispensability argument’. Here is the parody:

If we find the belief that p ‘indispensable’, for any reason, then that is all the justification we need for saying that p is true—never mind that in our more reflective moments we may realize that p cannot possibly be true.

The parody carries the accusation that pragmatism does not care about truth, but only about ‘successful practice’. A well‐known variant of that accusation is that pragmatism dodges the whole issue by simply identifying truth with successful practice. To see how this accusation can be rebutted it is useful to look again at Quine's indispensability argument.

(p.204) The problem that Quine ignored—I said that he ‘displaced’ it—is that (one hears it said) even if there were such things as sets, we could not possibly know of their existence (in a variant of the argument,15 one could not so much as refer to them) because (it is said) knowledge (and reference) require causal interaction. Since we do not interact with sets causally, we cannot know anything about them (and we cannot refer to them). While Quine does not explicitly discuss this argument, I think I can guess what he would say. He would say that the premise of the argument is, in fact, the whole argument. And the premise is a ‘dogma’. The philosopher who advances such an argument claims that whatever does not fit a certain paradigm is not knowledge (or not reference) at all. But this claim is not supported by anything one might call evidence; it is simply a metaphysical dogma. If we do not let ourselves be trapped by an appeal to what is ‘evident to reason’, we shall see that we do have evidence of a kind for set theory. Pragmatists do not urge us to ignore sound arguments against what we believe, when such arguments are advanced; they do urge us not to confuse the ‘intuitions’ of metaphysicians with genuine arguments.

It is pretty easy to see, on reflection, that Quine is right about set theory.16 Philosophical arguments against set theory do depend on some ‘fishy’ metaphysical premises. But can we say the same about the well‐known materialist arguments17 against the very possibility of ethical knowledge? As those arguments were set out by John Mackie, in particular, they are strikingly similar to the arguments against set theory alluded to above. According to Mackie, moral properties would be ‘ontologically queer’. What he meant by this was that they would not fit in to the picture of the world provided by modern science; and Mackie assumes that the picture of the world provided by modern science is not only approximately correct (which the pragmatists would not challenge) but that it is approximately complete (an assumption that they would challenge). Moreover, even to show that the existence of ethical properties (or of value properties in general) is incompatible with the world view of physics, Mackie needs to employ some pretty substantial premises, not all of which are made explicit. Value properties are, according to Mackie (this is supposed to be a conceptual (p.205) truth) such that if something has a positive value property, then anyone who fully recognizes that it does is thereby moved to approve of the thing or action in question. They are, in a very strong sense, action guiding. But it is part of the world view of physics (according to Mackie) that one can know that something has a property without feeling either approval or disapproval. (In effect, Mackie seems to read the world view of economics into modern physics.)

Mackie's argument, like the argument against the truth of set theory, rests on a number of metaphysical ‘intuitions’ which look more and more suspicious as one begins to probe them. For example, is it really a conceptual truth (that is, a conceptual truth viewed from within moral language) that if I know that an action is good, then I will approve of it? After all, a person may perform a good action, but I may disapprove, because I think the person could have done something much better. It is also no contradiction at all to think that somebody's action is good, and not to approve of a similar action in another case (in particular, not to be moved to perform a similar action myself). This is so both because different cases require different actions, and because, as Agnes Heller observes,18 there are different ways of being good, and another person's way may simply not be my way. (And there is also the rare but unfortunately not non‐existent case of the person who does evil because it is evil; such a person knows that he is an enemy of the good, but he does not approve of the good.) But these are not the grounds on which I wish to criticize Mackie's argument.

Let us reformulate Mackie's incautious claim by saying that positive moral properties, if any such really exist, are conceptually required to be such that when one knows that something possesses one of them, then ceteris paribus that very knowledge will produce in one feelings of approval, and negative moral properties are conceptually required to be such that when one knows that something possesses them, then ceteris paribus that very knowledge will produce in one feelings of disapproval. Mackie's argument (whether or not we modify the premise by inserting ceteris paribus clauses as I just have) is that these conceptual requirements show that moral properties (if there be such are ‘ontologically queer’. But do they really?

It is easy to see that some properties whose existence we do not find at all incompatible with the world view of physics are, in fact, ‘action guiding’ in the way Mackie claims moral properties (and value properties generally) are conceptually required to be action guiding. I am thinking in particular of pain. The knowledge that something is extremely painful may not keep me from wanting it, or wanting to do it, under special circumstances, but it is certainly true that ‘other things being equal’ the very knowledge that something is painful will lead me to disapprove of it, to recommend against (p.206) it, etc. In some sense, this seems to be ‘conceptually linked’ to pain, even if there are no ‘analytic truths’ in the old metaphysical sense in this area. And we don't find this at all ‘queer’. On the positive side—and this example brings us closer to moral properties—consider the property of being intelligent. People who are themselves intelligent value their intelligence, other things being equal. If an intelligent person were offered a million dollars provided he or she agreed to undergo a brain operation which reduces IQ by 60 points, then he or she would have to be seriously disturbed or depressed to even consider saying ‘yes’. Again, we don't find this fact particularly remarkable.

Of course, an unintelligent person might not care about losing or gaining in intelligence; but an unintelligent person does not have a good idea of what intelligence is. What makes this example relevant to the moral domain is that an ancient tradition in ethics, the Greek tradition, did in fact think that virtue was in these respects like intelligence; it held, on the one hand, that only the virtuous have an adequate idea of what virtue is, and that those who have an adequate idea of what virtue is do not ceteris paribus wish to lose their virtue. Perhaps most people do not believe any longer that the Greek tradition was right; but it is not clear that this disbelief is based on some piece of ‘scientific knowledge’ that we have gained in the meantime.19

Mackie would have replied that even if it is true, as a matter of empirical fact, that those who are highly intelligent do not (unless they are disturbed or depressed) wish to lose their intelligence, still there is no logical contradiction in the idea of a highly intelligent (and not mentally disturbed or depressed) person who has an adequate idea of what intelligence is and simply values something more than keeping his intelligence. On the other hand, there is, according to Mackie, a logical contradiction in saying that one recognizes that something is good, but one does not approve of it and one would not under any circumstances recommend acting in that way (if the thing in question is an action). The problem with this argument is that it assumes the metaphysical notion of analytic truth. Mackie himself may have been a firm believer in a sharp distinction between conceptual truth and empirical truth (and may have further believed that these are the only sorts of truths there are); but pragmatism was in large part an attack on the analytic/synthetic dichotomy. Once again we see that what was supposed to be a rational argument against the objectivity or cognitivity of ethics turns out to depend on a lot of suspect metaphysical baggage.

Perhaps a better line for a defender of Mackie's position to take today would be to argue that the value of intelligence (to animals possessing it) (p.207) was originally a purely instrumental value. It may indeed be, that as a result of evolution, intelligence has acquired a quasi‐instinctual value for members of our species, or for members who possess it to a high degree, but there is no corresponding story to be told about virtue, and that is why the claim that there are action‐guiding virtue properties is, or should be, suspect from a naturalistic point of view. But this argument is extremely ‘blurry’. The story about the process by which intelligence was transformed from an instrumental value to a terminal value20 is itself speculative, and, moreover, there is no reason to think that a parallel story could not be told in the case of at least some of the basic virtues; indeed, some writers have suggested that one can. More important, the supposed fact that intelligence was originally an instrumental value, and that it became a terminal value as a result of evolution, says nothing about whether that process was a rationally laudable process or one that we should rationally criticize and, if possible, try to reverse. Of course, it could be a premise of the argument that I just suggested that the transformation of values from instrumental values into terminal values is merely a ‘psychological’ process, that is, a process which cannot be rationally criticized; if so, then the premise is as strong as the desired conclusion. If we assume a sharp fact/value dichotomy as one of our premises, it is not surprising that we can pull out a sharp fact/value dichotomy as our conclusion.

In summary, I have just replied to the charge that pragmatism was anti‐intellectual by trying to show that pragmatists do not seek to suppress or play down genuine intellectual difficulties with any of our beliefs, or with the justifications of any of our actions. What pragmatism does argue is that criticisms of our beliefs and actions which are associated with various kinds of intellectualistic metaphysics,21 and with equally intellectualistic scepticism, will not stand up to close scrutiny. This is not something the classical pragmatists asked us to take on faith; on the contrary, every one of them was concerned to carry out a detailed criticism of the better‐known metaphysicians of his own time. Pragmatism goes with the criticism of a certain style in metaphysics; but the criticism does not consist in wielding some exclusionary principle to ‘get rid of metaphysics once and for all’. Indeed, as Strawson22 has reminded us (and as Dewey remarked many years earlier)23 revisionary metaphysics is not always bad, and we owe many insights, in science and in morals and in politics, to various kinds of revisionary metaphysics. But if revisionary metaphysics is not always bad, (p.208) neither is it always good. It does spawn more than one intellectual bugaboo; and the criticism of these intellectual bugaboos is a permanent task of the pragmatist philosopher.

So far, then, I have rehearsed a negative argument for the objectivity of values. The negative argument is that the belief in the cognitive status of value judgements underlies an enormous amount of our practice—indirectly, perhaps, the whole of our practice. Metaphysical arguments that say that even though we adopt a ‘realist’ stance towards value disputes in practice, realism with respect to values cannot possibly be right, are themselves unsupported by anything more than a collection of metaphysical dogmas. It is time now to consider more positive arguments.

3 Instrumental Rationality and Topic‐Neutral Norms

I now want to present and discuss an argument to the effect that instrumental rationality would be impossible if there were not topic‐neutral norms whose claim to rational acceptability is not derived simply from the fact that they help us to achieve particular goals a certain percentage of the time. This argument is an important part of the case for the view that objective values of a kind (though, so far, not ethical values) are presupposed by science itself.

First a word on my choice of terminology. I have chosen to speak of ‘topic neutral norms’ rather than to employ some traditional language, for example, the language of ‘intrinsic values’, for a number of reasons. For one thing, talk of ‘intrinsic values’ is associated in the history of philosophy with many doctrines that pragmatists have always viewed with suspicion. For example, the pragmatists have typically been hostile to the natural law tradition, and especially to the form of that tradition according to which there are norms somehow built into the very structure of reality, norms which ‘the wise’ (as opposed to ‘the many’) have always been able to discern. That tradition is riddled with appeals to ‘what is agreeable to reason’, and pragmatists have always distrusted such appeals unless they are tempered by a healthy dose of fallibilism and experimentalism. Dewey, in particular, loved to point out how the discourse of ‘natural law’, ‘what is agreeable to reason’, and so forth, has provided rationalizations for the interests of privileged groups. Again, talk of ‘intrinsic values’ has sometimes gone with ethical intuitionism, in the style of G. E. Moore, and that is also a tradition that pragmatists distrust. Since I am trying to argue in the spirit of American pragmatism, even if the arguments I am giving are only ‘in the spirit’ of American pragmatism (they are not arguments found in the pages of American pragmatists in the form in which I give them, in spite of their obvious debt to the arguments of Peirce, Dewey, and James), I do not wish to use a terminology which presupposes just the sort of metaphysics that (p.209) pragmatism was instrumental in overthrowing. There are topic neutral truths, to be sure (the truths of deductive logic are a well‐known example), but even topic neutral truths emerge from experience, or better from practice (including reflection on that practice) and belong to conceptual schemes which have to show their ability to function in practice (which does not mean that they are ‘empirical’).24 In the same way, I shall argue, there are topic neutral norms; but they too emerge from practice and have to show their ability to function in practice, as do all norms and values. The topic neutral norm I wish to discuss applies to behaviour that is directed towards a concrete goal, and that applies in circumstances in which it makes sense to think of the alternative actions as having known or fairly well‐estimated probabilities of reaching the goal. The norm is, moreover, one whose justification at first blush seems to be entirely instrumental. It is simply this: Act so as to maximize the estimated utility. This norm is, of course, a famous rule of decision theory; and while there is controversy as to whether it applies in every situation (of goal oriented behaviour with known probabilities), there is no question but that it applies to the great majority of such cases. In particular, it applies to cases of the kind that Peirce discussed in ‘The Doctrine of Chances’.25 These are cases in which I must, on one single occasion, perform one of two actions, and one of the actions has a high probability of giving me an enormous benefit and the other has a high probability of giving me an enormous loss. In these situations, the rule of course directs me to perform the action which is associated with a high probability of an enormous benefit. But why is the justification of this rule not entirely instrumental (‘if you follow the rule, you will get an enormous benefit a high percentage of the time’)? Note that I say that the justification of the rule is not entirely instrumental. Obviously, part of its justification is instrumental. Whenever I (or a group to which I owe loyalty) have to engage in a series of repeated wagers (with known probabilities) we can say (or it looks as if we can say—Peirce raises an important problem here too) that obeying this norm will yield a better ‘payoff’ in the long run, and in many cases payoff in the long run is exactly what we are interested in. But, as Peirce pointed out, there is a very special problem about the single case. If what I am interested in is not success in some long run, but success in one, unrepeatable, single case,26 I cannot justify performing the action with the higher estimated utility by saying that (p.210) I know that that action will succeed, because I do not know which action will succeed (if I knew which action would succeed, I wouldn't need to talk of probabilities). And if I justify it by saying that I know which action will probably succeed, then I am simply saying that I am justified in performing the action which has the higher estimated utility because it has the higher estimated utility; for ‘will probably succeed’ is just an imprecise way of saying ‘has the higher estimated utility’. (Indeed, Peirce develops his argument directly in terms of the rule that, in cases where the benefits are as described, one should perform the action with the higher probability of gain, rather than in terms of the notion of estimated utility. Peirce's point is that, for a self‐interested player who has no long run to look forward to, the justification of this rule cannot be ‘if you follow it in the long run, you will succeed in a large percentage of the cases’.)

Nor is the problem restricted to the unrepeatable27 single case. Peirce points out28 that when we say that the practise of betting on the action with the higher estimated utility will succeed in a high percentage of the cases in the long run, we had better not mean any given finite long run. For if we do, then what we are doing is, again, betting on an unrepeatable single case. If, for example, the entire future experience of the human race is contained within a time period of, say, a million years, then the statement that ‘the percentage of the time that the recommended strategy (of betting on actions with a high estimated utility, or of performing actions whose probability of success is high) will succeed will not, in this one single run of cases, be improbably far below its expected value’ is itself a statement about a single case. And what was just said about the justification of betting that the action with the higher estimated utility will probably succeed in a single case applies here too. Knowing the probability with which a policy will succeed in one particular unrepeatable finite ‘long run’, is not the same as knowing the percentage of the time (that is, the frequency) with which the policy will succeed in that one finite period of time; it is only knowing the expected value of that percentage. Betting that the actual value of that percentage will be close to the expected value is precisely to rely on the norm whose justification we are considering.

In The Many Faces of Realism, I said29 that I found Peirce's own solution to the problem he raised to be unbelievable. Peirce believed that there was no problem about the infinitely long run, because he accepted the frequency theory of probability, according to which the probability just is the limit of the relative frequency in the infinite long run; he dealt with the problem of the single case (which is also the problem of any given finite ‘long run’) by arguing that the purpose of a rational person, in any action, (p.211) should not be to benefit himself, but to act in accordance with the policy that would benefit all rational beings in the infinitely long run. A person who is not interested in the welfare of all rational beings in the infinitely long run is, according to Peirce, ‘illogical in all his inferences’.30 Even if I am facing a situation in which the alternatives are ‘eternal felicity’ and ‘everlasting woe’, on Peirce's view my belief that I, in this one unrepeatable situation, am somehow more rational if I perform the action that will probably lead to felicity than if I perform the action that will probably lead to everlasting woe, is fundamentally just a fictitious transfer31 of a property which has only to do with what happens in the infinitely long run to a single case. What is true, and not fiction or projection, however, is that my fellows, the members of the ongoing community of inquirers with which I identify myself, will have eternal felicity 24 times out of 25 (or whatever),32 if they follow the recommended strategy; or more generally, that even if this one particular situation is never repeated, if in all the various uncorrelated cases of this kind or of any other kind that they find themselves in, they always perform the action with the higher estimated utility, then in the infinitely long run they will experience more gains and fewer losses. But, as I asked in The Many Faces of Realism, is it really true that the reason that I would make the choice that is likely to give me eternal felicity and not everlasting woe is that I am altruistic?

Peirce's argument is that I ought to choose the arrangement that will probably give me eternal felicity for what one might describe as ‘rule utilitarian’ reasons: in choosing this arrangement I am supporting and helping to maintain a norm which will benefit the community of rational investigators in the infinitely long run. But is this really what is in my mind when what I am facing is torture? Obviously, it isn't. What Peirce himself attempts is to give a completely instrumental justification for the norm. In order to push through his instrumental justification, he needs not only an incorrect theory of probability (very few students of the subject would today subscribe to the claim that probability can be defined as the limit of a relative frequency in the infinitely long run); but he also needs two further dubious assumptions: (1) the assumption that the motive for all of my actions, insofar as they are rational, is entirely altruistic; and (2) the assumption that there will be rational investigators in the indefinitely far future, that is, that the community of rational investigators is literally immortal. But even if Peirce failed to provide a satisfactory solution to his (p.212) own problem, it remains the case that he was the philosopher who called our attention to its importance and its depth.

It is just because the problem Peirce raised is so deep that I want to examine it again, and to go further than I did when I discussed it in the closing pages of The Many Faces of Realism. In those pages, I discussed only the part of Peirce's argument that refers to extreme choices in unrepeatable situations. I did not discuss Peirce's claim that the problem of relying on knowledge of probabilities to guide our actions in the finite long run poses the same problem; and, indeed, I think that there are objections to this claim of Peirce's. One possible objection is this: while we can easily imagine that in any ordinary ‘single case’ an improbable event happens, imagining that improbable events happen ‘across the board’ for a very long time (a million years) may, in fact, go beyond what is conceivable. Can we really make sense of the notion that almost all events might be improbable ‘flukes’ for the next million years?33 I won't try to decide this question here, but even if we assume, for the sake of Peirce's argument, that the answer is ‘yes’, there is still a further objection. For one might plausibly claim that the fact is that we firmly believe that this will not happen; that is, we don't just think it is improbable that the whole statistical pattern of events in the next million years (or even in the next ten years) will be a ‘fluke’, perhaps we find it inconceivable that in a run that long, the frequency will not be near its expected value. (When the time period is large enough, it may be that the very notion of ‘probability’ crumbles if we assume that all or most of the statistics we encounter are ‘flukes’.) But, if that is so, relying on the norm of expecting what is probable to happen, and on the norm of performing the action with the higher estimated utility, does have an instrumental justification when our interests really are very long run success, and not in success in some small number of unrepeatable cases, or even in one moderately long series of cases. On the other hand—and this speaks in Peirce's favour—even if this is true, the problem of the single case is still worse than it looks at first. The problem does not arise only in cases which are literally unrepeatable.

To see this, consider any case whatsoever—and such cases are familiar enough—in which I am tempted to perform an action whose success is ‘against the odds’. Let us say that I have a strong hunch that I am going to be lucky, so strong that I am tempted to stake my life savings or my family's savings on this hunch. I may perfectly well agree that if everyone followed the policy of relying on such hunches, then in the long run this would cause an enormous amount of suffering; but I am not recommending that other people follow that policy in the future. I do not even intend to follow that policy myself in the future, I only intend to follow it in this single case, let us assume.

(p.213) It may be true that if I follow this policy in this single case, then I am bound to follow the same policy in the future, and that even if my hunch is correct this time, sooner or later ‘the odds will catch up with me’. But it could be that I know myself well enough to say truthfully, ‘I am going to gamble just this once; win or lose, I am not going to risk going against the odds again.’ If that is my fixed intention, then even if I know that I would lose money if I made this gamble many times in the future, or that other people would lose money if they made such gambles many times in the future, that knowledge is relevant to the question, ‘Will this risky gamble achieve my goal?’—only on the assumption that probability is relevant to single actions, that is to say, only if the topic neutral norm is in place.

Some theorists would, no doubt, try to get around the whole problem in the following way: they would argue that deciding what to do can be broken into two steps, namely the fixation of belief and the choice of an action. With respect to the first, the fixation of belief, it might be said that the operative norm is simply a norm of rationality: a rational man, it is said, proportions his subjective degrees of belief (his ‘subjective probabilities’) to the relevant objective statistical probabilities when those are known or assumed, which is the case that Peirce is discussing. (Note that Peirce is not discussing ‘the problem of induction’, that is to say, the problem of whether we can know the objective statistical probabilities when those relate to future events.) In the Peirce thought experiment, the agent knows that the objective probability that he will achieve ‘eternal felicity’ and avoid ‘everlasting woe’ if he follows the recommended policy is 24/25ths. If he is rational, he will therefore proportion his subjective degree of belief in the proposition ‘I shall obtain eternal felicity if I follow the recommended policy’ to this objective probability; that is, he will almost believe that if he performs the recommended action (choosing a certain deck of cards) he will gain eternal felicity. He does not absolutely believe that he will obtain eternal felicity if he performs the action, but he believes it to the very high degree of 24/25ths. Given that he believes (almost) that if he performs the action he will be successful, he would be irrational not to perform the action. In this way, performing the action is instrumentally justified.

In fact, however, this justification is not a completely instrumental justification because it assumes a norm of rationality which is itself not instrumentally justified: the norm that one should ‘proportion one's subjective degrees of belief to the objective statistical probabilities when they are known’. That norm, like the norm that one should perform the action with the higher estimated utility, has a partial instrumental justification; in situations in which what we are interested in really is long run success, then one can argue that conforming to the norm will achieve the goal. (However, to do so, as we observed above, we must assume that in that particular ‘long run’, frequencies will be close to their expected values, which—Peirce would say—presupposes the very norm we are ‘justifying’.) (p.214) But in any case, to say that it is ‘rational’ to conform to the norm even when what we are interested in is not very long run success, is just another way of saying that one ought to conform to that norm in those situations, and this, for reasons we have just rehearsed, is a belief which cannot be instrumentally justified.

At this point, if I am not mistaken, two objections to what I have been maintaining will occur to various of my readers, though probably not to the same readers. The first objection, which I shall call ‘the naturalism objection’ runs as follows: ‘The norms you describe, like all norms, have evolved because they have a high degree of long run utility, to communities if not always to individuals. [When I use the word “evolved” in stating the objection, philosophers who are prone to sociobiological speculation may even think that Darwinian evolution is involved here, but in any case, social evolution is certainly involved.] But if we can understand perfectly well what the origin of these norms is, and what the origin is of the compulsion that each of us feels to conform to them, even when we don't know that that conformity will in fact bring us success in any particular individual case, then what's the problem?’

The objection is serious because it rests on an assumption which is undeniably correct; to say, as I have, that the norms cannot be completely justified instrumentally is not to deny that they do have long run instrumental value to communities. And the pragmatists certainly believed that the metaphor of ‘evolution’ is an appropriate one, in cases like these. The pragmatists also believed—and here I am thinking more of James and Dewey than of Peirce, perhaps—that one must take the agent point of view seriously, and not look at everything from the third person descriptive point of view. It is quite true that from a third person descriptive point of view there is no difficulty in understanding the origin of topic neutral norms like the ones I have been describing; that is, there is no difficulty in understanding the existence of such norms as a mere psychological fact. But if I as an agent put to myself the question ‘What should I do?’ in a situation of the kind we have been discussing, then the knowledge that I feel some compulsion to act one way, as a psychological fact about myself, does not answer my question. Indeed, if I also have a strong compulsion to ‘be irrational for once’, to act on a hunch, I may also be able to find some naturalistic explanation of the existence of that compulsion. Each compulsion is a brute fact; that I have both compulsions doesn't tell me what to do. The factual descriptive questions are certainly legitimate questions. But so are the normative first person questions.

Here is the second of the objections that I mentioned: ‘Your question is really just why it is rational to do certain things, to act in accordance with probabilities (or, more precisely, estimated utilities), or to proportion one's subjective degrees of belief to probabilities. But it is just a tautology that these are rational things to do. To be rational is just to obey certain rules, (p.215) for example, the rules of deductive logic and the rules of rational betting. So what's your problem?’

One problem (a decisive one) with this objection is that saying that something is rational is not merely to describe it as in accordance with some algorithm or quasi‐algorithm or other.34 If I say that believing something or acting in a certain way is rational, then, other things being equal, I am recommending that belief or that course of action. To say that something is rational, say that it is rational to bet in accordance with the probabilities (more precisely, the estimated utilities) is to say that, other things being equal, one ought to bet in accordance with the probabilities (the estimated utilities). In short, the fact that something is rational is precisely the sort of fact that Mackie regarded as ‘ontologically queer’; it is an ought‐implying fact.

Instead of creating spurious philosophical problems for ourselves by insisting that the causal‐descriptive world view must be ‘complete’, in a sense which is never satisfactorily defined,35 the pragmatists urged that the agent point of view, the first person normative point of view, and the concepts indispensable to that point of view, should be taken just as seriously as the concepts indispensable to the third person descriptive point of view.

In closing this part of my argument, it may be useful to compare once more the conclusion I have reached with the conclusion Peirce reached. What our arguments have in common—and, of course, mine is built out of the very materials Peirce provided—is that we both agree that the norm of acting in accordance with the probabilities or estimated utilities cannot be justified instrumentally if the only goals considered are the goals of the agent himself. Peirce concludes that all rational goal directed activity must be altruistic and personally disinterested, a conclusion that fits well with his own moral outlook, which tended to stress the ‘Buddhistic’36 virtues of self‐abnegation (p.216) and freedom from personal interest. As I have said, I cannot ‘go along’ with this conclusion; we are, after all, talking about goal oriented activity, and in this case our aims are usually not disinterested. The conclusion I draw from Peirce's problem is, rather, that even when I seek to attain a goal (in a situation in which risk is involved either way), the rational decision as to what I must do to attain my practical goal depends upon my acknowledging the binding force of norms which do not possess a satisfactory instrumental justification in terms of my own goals. This conclusion is not incompatible with Peirce's; but it is not as metaphysically daring.

Finally, to the inevitable question, ‘how would I justify norms like the one we have been considering, if a complete instrumental justification cannot be given?’, I would reply that the alternative to instrumental justification here is not transcendent knowledge but reflection. Norms like the estimated utility rule were not discovered by mere trial and error, but by normative reflection on our practice. Kant was right in urging us to realize that reflection on the possibility of gaining knowledge from experience is itself a source of knowledge, even if he was wrong in considering it to be an infallible source (when properly conducted). Kant was further right in supposing that we must reflect not only on the presuppositions of learning from experience, but also on the presuppositions of acting in the ways in which we do act. Most puzzles about the very ‘possibility’ of normative knowledge spring from a too narrowly empiricistic picture of how knowledge is gained and how actions are justified.

4 ‘Oughts’ and Belief Fixation in Pure Science

I have described the facts that certain actions and beliefs are rational as ‘ought‐implying facts’: by including beliefs here I mean to indicate that this allegedly ‘ontologically queer’ phenomenon is one that does not appear only in the context of goal oriented activity, and the context of moral activity, but also in the context of ‘pure’ scientific activity. In Reason, Truth and History,37 I described some of the terms we use in appraising scientific theories, e.g., the terms ‘simple’, elegant', and ‘well confirmed’, as having just this character; for to describe a theory as simple, elegant, or well confirmed is to say that, other things being equal, we ought to accept the theory or at least consider it seriously. Although these ought‐implying characteristics are not ethical characteristics, they share many of the properties of the so‐called38 ‘thick’ ethical characteristics. Just as no non‐ethical (p.217) term seems to have exactly the same descriptive range as the term ‘cruel’, so no non‐normative term applied to theories seems to have exactly the same descriptive range as the term ‘confirmed’.

It might be claimed, however, that even though there are ‘oughts’ connected with rational belief and rational action, that (at least in the case of rational belief) these ‘oughts’ are of quite a different kind from the ‘oughts’ that we encounter in connection with justice and in connection with morality. As far as I know, only one reason is ever given for suggesting that the ontological status of pure cognitive values is completely different from the ontological status of normative values, and that is, once again, that it is thought that at least the sorts of cognitive values I just mentioned—elegance, simplicity, predictive power, the conservation of past doctrine, and so on—are justified purely instrumentally (relative to truth as the goal of inquiry), and that the ‘oughts’ connected with them are thus purely hypothetical ‘oughts’. To say that one ought to consider seriously theories which have striking predictive power or simplicity is, on this view, just to say that if you follow the policy of taking theories with these virtues seriously, in the long run you will arrive at a higher percentage of true or approximately true theories. Roderick Firth devoted his Presidential Address to the Eastern Division of the American Philosophical Association39 to pointing out a major difficulty with this argument. One way of putting Firth's difficulty is the following: if the fact that the practice of accepting theories with these virtues has led to the discovery of truths is supposed to be an empirical justification of that practice, then it is flagrantly circular. It is circular, because the way we recognize that we have discovered truths in the past by using that practice is by relying on that very practice. It is not, after all, as if we had some way of telling that we had discovered a true (or approximately true) theory in the past independent of relying on the fact that acceptance of the theory is justified (in terms of successful prediction, simplicity, conservation of past doctrine, etc.), that is, relying on the rule that one should posit that theories which are sufficiently superior in these respects deserve to be provisionally accepted as true or approximately true. Of course, Firth did not wish to deny that theories which are ‘justified’ (as determined by the norms mentioned and other norms like them) are more likely to be true or approximately true; the point is, rather, that we could not possibly first discover that this is the case, and only the come to adopt the policy of accepting the justified theories. We cannot test the claim that acceptance of justified theories leads to acceptance of true (or approximately true) theories empirically; the most we can check empirically is the (p.218) frequency with which justified theories tend to remain justified later. The ‘hold’ of justification on our thought, the ‘oughtness’ of positing that justified beliefs are (probably, approximately) true, cannot be assimilated to the discovery of a useful empirical correlation, such as the correlation between building a fire and getting warm.

We can also see that it is not the case that we believe we ought to accept justified theories in each single individual case only because we have some kind of a non‐empirical belief that doing so leads to the acceptance of a higher percentage of truths ‘in the long run’; for if that were the complete justification for our reliance on these cognitive norms, then we would run into exactly the ‘problem of the single case’ that we ran into in connection with choosing an action. For both of these reasons, then, the strategy of saying, ‘Oh yes, there are “oughts” connected with the fixation of belief, but those are purely hypothetical imperatives’ is a failure. There are ‘ought‐implying facts’ in the realm of belief fixation; and that is an excellent reason not to accept the view that there cannot be ‘ought‐implying facts’ anywhere. If their sole reason for believing that ‘there cannot be “ought‐implying facts” in ethics’ is that ‘such facts cannot exist anywhere’, then the opponents of moral objectivity (or ‘moral realism’) had better rethink.

4.1. But What Does Any of This Have to Do With Ethics?

Perhaps the most detailed case for the view just defended, the view that all inquiry, including inquiry in pure science, itself presupposes values, is made by Dewey, in his Logic. Ruth Anna Putnam and I have analyzed Dewey's argument40 elsewhere; here, I only want to discuss one aspect of Dewey's view, the insistence on a very substantial overlap between our cognitive values and our ethical and moral values.

We have already examined the claim that there is a fundamental ontological difference between cognitive or ‘scientific’ values and ethical values, and found that the reasons offered for believing that claim fail. But even apart from the ‘ontological’ question, a view like Dewey's will not be intelligible if one starts with what I may call a ‘Carnapian’ view of the scientific method. For this reason, I need to say a few words about the differences between the way in which a philosopher like John Dewey sees the scientific method and the way in which a philosopher like Rudolf Carnap did. First of all, it is noteworthy that in Carnap's great work on inductive logic41—the work to which he devoted almost all of his energy in the last two decades of his life—there is virtually no reference to experiment:—the word does not even occur as an entry in the index to The Logical (p.219) Foundations of Probability. Scientific theories are confirmed by ‘evidence’ in Carnap's systems of inductive logic, but it is immaterial (that is to say, there is no way to represent the difference in the formalism) whether that evidence—those ‘observation sentences’—is obtained as the result of intelligently directed experimentation, or it just happens to be available. Passive observation and active intervention are not distinguished, and the question as to whether one has actually tried to falsify the hypotheses that have been ‘highly confirmed’ is not a question which can be asked or answered in the languages Carnap constructed. Even more important, for our purposes, is that fact that the term that Carnap used to characterize his own stance in the Aufbau, the term ‘methodological solipsism’, could also be applied, though in a different sense, to this later philosophical work. For just as it makes no difference from the point of view of Carnapian inductive logic whether our observation is passive or active, whether we just look or whether we intervene, it also makes no difference whether observation is co‐operative or not. Fundamentally, the standpoint is that of a single isolated spectator who makes observations through a ‘one‐way mirror’ and writes down observation sentences. Appraising theories for their cognitive virtues is then simply a matter of using an algorithm to determine whether a sentence has a mathematical relation to another sentence (the conjunction of the observation sentences the observer has written down), on this picture. The scientific method is reconstructed as a method of computation, computation of a function like Carnap's famous ‘c*’.42

Dewey's picture is totally different. For Dewey, inquiry is co‐operative human interaction with an environment; and all aspects, the active intervention, the active manipulation of the environment, and the co‐operation with other human beings, are vital. The first aspect, the aspect of intervention, is connected with pragmatist fallibilism. Of course, Carnap was also a fallibilist, in the sense of recognizing that future observation might disconfirm a theory which is today very well confirmed; but for the pragmatists this was not fallibilism enough. Before Karl Popper was even born, Peirce emphasized that43 very often ideas will not be falsified unless we go out and actively seek falsifying experiences. Ideas must be put under strain, if they are to prove their worth; and Dewey and James both followed Peirce in this respect. In what follows, however, I want to focus on the other aspect of Dewey's thought, that is, the conception of scientific inquiry as a form of co‐operation.

For the positivists—e.g., for both Carnap and Reichenbach—the most primitive form of scientific inquiry, and the form that they studied first when they constructed their (otherwise very different) theories of induction was induction by simple enumeration. The model is always a single scientist (p.220) who determines the colours of the balls drawn successively from an urn, and tries to estimate the frequencies with which those colours occur among the balls remaining in the urn. For Dewey, the model is a group of inquirers trying to produce good ideas and trying to test them to see which ones have value.

One more point must be mentioned at the very outset of any discussion, however brief, of Dewey's conception of inquiry: the model of an algorithm, like a computer program, is rejected. According to the pragmatists, whether the subject be science or ethics, what we have are maxims and not algorithms; and maxims themselves require contextual interpretation. Not only that, the problem of subjectivity and intersubjectivity was in the minds of the pragmatists from the beginning. They insisted44 that when one human being in isolation tries to interpret even the best maxims for himself or herself and does not allow others to criticize the way in which he or she interprets those maxims, or the way in which he or she applies them, then the kind of ‘certainty’ that results is always fatally tainted with subjectivity. Even the notion of ‘truth’ makes no sense in such a ‘moral solitude’ for ‘truth presupposes a standard external to the thinker’.45 Notions like ‘simplicity’, for example, have no clear meaning at all unless inquirers who have proven their competence in the practice of inquiry are able to agree, to some extent at least, on which theories do and which theories do not possess ‘simplicity’. The introduction of new ideas for testing likewise depends on co‐operation, for any human being who rejects inputs from other human beings runs out of ideas sooner rather than later, and begins to consider only ideas which in one way or another reflect the prejudices he has formed. Co‐operation is necessary both for the formation of ideas and for their rational testing.

But—and this is the crucial point—that co‐operation must be of a certain kind in order to be effective. It must, for example, obey the principles of ‘discourse ethics’.46 Where there is no opportunity to challenge accepted hypotheses by criticizing the evidence upon which their acceptance was based, or the application of the norms of scientific inquiry to that evidence, or by offering rival hypotheses, and where questions and suggestions are systematically ignored, then the scientific enterprise always suffers. When (p.221) relations among scientists become relations of hierarchy and dependence, or when scientists instrumentalize other scientists, again the scientific enterprise suffers.47 Dewey was not naïve. He was aware that there are ‘power plays’ in the history of science as there are in the history of every human institution. He would not have been surprised by the findings of historians and sociologists of science; but he differs from them (or from some of our contemporary ones) in holding that it makes sense to have a normative notion of science.

Moreover, it is not just that, on Dewey's conception, good science requires respect for autonomy, symmetric reciprocity, and discourse ethics—that could be true even if scientific theories and hypotheses were, in the end, to be tested by the application of an algorithm, such as the inductive logic for which Carnap hoped—but, as we already observed, the very interpretation of the non‐algorithmic standards by which scientific hypotheses are judged depends on co‐operation and discussion structured by the same norms. Both for its full development and for its full application to human problems, science requires the democratization of inquiry.

What I have just offered is, in part, an instrumental justification of the democratization of inquiry. But Dewey opposes the philosophers' habit of dichotomization. In particular, he opposes both the dichotomy ‘pure science/applied science’ and the dichotomy ‘instrumental value/terminal value’. Pure science and applied science are interdependent and interpenetrating activities, Dewey argues.48 And similarly, instrumental values and terminal values are interdependent and interpenetrating. Science helps us to achieve many goals other than the attainment of knowledge for its own sake, and when we allow inquiry to be democratized simply because doing so helps us achieve those practical goals, we are engaged in goal oriented activity. At the same time, as we saw above, even when we are engaged in goal oriented activity we also are guided by norms of rationality which have become terminal values for us, and which cannot be separated from the modern conception of ‘rationality’ itself. Moreover, we are not—nor were we ever—interested in knowledge only for its practical benefits; curiosity is coeval with the species itself, and pure knowledge is always, to some extent, and in some areas, a terminal value even for the least curious among us. And just as the norm we discussed in connection with Peirce's problem has become inseparable from the modern conception of ‘rationality’, so the norms Dewey discusses have become inseparable from the modern conception of proceeding ‘scientifically’, inseparable that is, from our normative conception of what it is to be scientifically rational. It is not, for us, any longer just a sociological‐descriptive fact that choosing theories for their predictive power and simplicity, and fostering democratic co‐operation and (p.222) openness to criticism in the generation and evaluation of theories, are part of the nature of scientific inquiry; these norms describe the way we ought to function when the aim is knowledge. Saying this is not the same as saying that inquiry which follows these norms produces knowledge in the way fire produces warmth; as Firth showed, we are not dealing with mere empirical correlation here. Nor is it saying that it is ‘analytic’ that inquiry which does not meet these standards does not produce justification and knowledge; ‘knowledge’ and ‘justification’ are not the sorts of words that can be analytically defined once and for all. Concepts of knowledge are essentially contested concepts; they are always open to reform. What we can say is that the applicability of our present conception to practice is constantly being tested (it is not a priori that, for example, the concept of ‘probability’ can be successfully used in practice) and that conception itself is partly constituted by norms which represent values which are now terminal but not immune to criticism. And this ‘messy’ state of affairs is one that, I would say, Dewey wanted us to see as typical.

4.2. Was Dewey Trying to Derive Ethics from the Logic of Science?

It would be a mistake to see Dewey's project in the Logic as ‘deriving an ethic from science’ (or from the theory of inquiry). What Dewey's argument does show is that there is a certain overlap between scientific values and ethical values; but even where they overlap, these values remain different. Scientific values are not simply instrumental (the relation between the ‘means’, scientific justification; and the ‘goal’, knowledge, is too ‘internal’ for that story to work), but they are relativized to a context—the context of knowledge acquisition—and knowledge acquisition itself is something that can be criticized ethically.49 Yet there was a close connection between inquiry and ethics, in Dewey's view, for more than one reason.

First, all co‐operative activity involves a moment of inquiry, if only in the ongoing perception that the activity is going smoothly/not going smoothly. What is essential to the rational, or, to use the word that Dewey preferred, the intelligent conduct of inquiry is thus, to some extent, essential to the intelligent conduct of all co‐operative activity. Perceiving this can help us to understand why, as the modern conception of rationality gains the upper hand, one sees an insistence on extending democratization to more and more institutions and relationships.

Secondly, ethics itself requires inquiry. Experiencing ourselves as ethical fallibilists, as persons who do not inherit values which cannot be questioned, as persons who have, in fact, criticized many inherited values (even if we have not all criticized the same inherited values), we more and more see ethical disputes as disputes to be settled, if possible, by intelligent (p.223) argument and inquiry, and not by appeals to authority or to a priori principles. Now nothing Dewey wrote, and nothing James wrote, or Peirce wrote, can prove that there are objective ethical norms, ideals, rules of thumb, or even situation specific values; but—and here I state the argument in a more ‘Kantian’ way than Dewey himself would—if there are ethical facts to be discovered, then we ought to apply to ethical inquiry just the rules we have learned to apply to inquiry in general. For what applies to inquiry in general applies to ethical inquiry in particular.

If this is right, then an ethical community—a community which wants to know what is right and good—should organize itself in accordance with democratic standards and ideals, not only because they are good in themselves (and they are), but because they are the prerequisites for the application of intelligence to the inquiry. It is true that some inquiries can be conducted with only partial democratization; a tyrant, for example, may allow his physicists a freedom of discussion (in certain areas) that is generally forbidden in the society. But—and this is an empirical presupposition of Dewey's argument—any society that limits democracy, that organizes itself hierarchically, thereby limits the rationality of those at both ends of the hierarchy. Hierarchy stunts the intellectual growth of the oppressed, and forces the priviledged to construct rationalizations to justify their position.50 But this is to say that hierarchical societies do not, in these respects, produce solutions to value disputes that are rationally acceptable.

At this point it may look as if Dewey is ‘pulling himself up by his bootstraps’. For even if we assume that inquiry into values should be democratized, that the participants should, qua seekers after the right and the good, respect free speech and the other norms of discourse ethics, not instrumentalize one another, etc., what criteria should they use to tell that their inquiry has succeeded?51

This objection overlooks another feature of the pragmatist position; we do not, in fact, start from the position of ‘doubting everything’. As long as discussion is still possible, as long as one is not facing coercion or violence or total refusal to discuss, the participants to an actual discussion always share a large number of both factual assumptions and value assumptions that are not in question in the specific dispute. Very often, participants to a disagreement can agree that the disagreement has, in fact, been resolved, not by appeal to a universal set of ‘criteria’, but by appealing to values which are not in question in that dispute. (Recall again Peirce's distinction—which all the pragmatists accepted—between real and ‘philosophic’ doubt.) It may, of course, happen that the criteria accepted by even one of the participants are in conflict with themselves; but in that case, rational reconstruction is called for. Pragmatism is not against rational reconstruction; but it sees rational reconstruction as one among other tools to be used (p.224) in resolving problems, not as a route to a set of universal principles applicable to all situations.

Moreover, Dewey stressed52 that one does not always need a set of criteria to tell that a problem has been resolved. We have all had the experience of discovering that the satisfactory resolution to a problem in our own lives was one that would not have counted as satisfactory by any of the ‘criteria’ we had in mind at the beginning of our search for a solution. (A corollary to this is that when we have solved a collective or an individual problem we may not, at the end, know whether it was an ‘ethical’ problem or some other kind of problem that we solved; and is that a bad thing?) And finally, believing that ethical objectivity is possible is not the same thing as believing that there are no undecidable cases, or problems which, alas, cannot be solved. The point is that, once we give up the metaphysical claim that there cannot be such a thing as ethical objectivity, and once we observe that objectivity in other areas is strongly connected with values, we can begin to see not just that ethical objectivity might be possible, but, more importantly, that investigating ethical problems requires just the values that have come to be linked with the open society.

5 Conclusion

In closing, I want to say just a word about my reasons for writing this paper. Besides my long‐standing interest in the problem of moral objectivity there was, as I indicated, another reason: to combat the idea that there is no intellectual structure worth taking seriously to the arguments of the American pragmatists. I have laid out some of that structure here in my own way, and without a lot of scholarly apparatus, because my interest was in the living relevance of what American pragmatism bequeathed to us, and that is something one can only explain in one's own words, because they are the only words one can take full responsibility for.

In particular, I hope to have shown that the appeal to the primacy of practice (e.g., the ‘indispensability arguments’) in pragmatism is always accompanied by critique of those metaphysical criticisms of practice that make it look ‘irresponsible’ to take practice as seriously as pragmatists do.53 I have tried to give some sense of the range, the complexity, and the depth of the investigations of the pragmatists into such questions as the value presuppositions of goal oriented activity and pure science, and the overlap (though never the identity) between science and ethics. If this leads other students of these questions to take the pragmatists seriously, I shall feel this all too brief account of some of their thought was well worth the effort it took me to present it.


(1) See J. Habermas, The Theory of Communicative Action trans. T. McCarthy (Boston: Beacon Press, 1984).

(2) I have chosen to speak of ‘moral objectivity’ rather than to use the currently fashionable term ‘moral realism’ because ‘realism’ is entangled in so many metaphysical and language‐philosophical disputes at this time. Of course, talk of ‘objectivity’ is also open to metaphysical construals of various kinds; when I speak of objectivity here and in what follows, what is at stake is ‘objectivity humanly speaking’, the objectivity of what is objective from the point of view of our best and most reflective practice.

(3) Cf. my ‘Philosophy of Logic’, reprinted as the final chapter in the second edition of Mathematics, Matter and Method (Cambridge: Cambridge University Press, 1979).

(4) This view has been defended in a series of papers by Penelope Maddy.

(5) Quine identifies all the abstract entities needed in science, including numbers, with sets.

(6) This has been challenged by H. Field in his Science Without Numbers (Oxford: Oxford University Press, 1980) and elsewhere. Field's controversial arguments, even if we concede that they are correct as far as they go, do not apply at all to modern quantum mechanics (for reasons unconnected with the ‘measurement problem’, by the way), and thus do not really show that physics can dispense with quantification over numbers and sets, however.

(7) In Towards Reunion in Philosophy (Harvard: Harvard University Press, 1956) and in ‘Normative Ethics, Normative Epistemology, and Quine's Holism’ in L.E. Hahn and P. A. Schilpp (eds.), The Philosophy of W. V. Quine (LaSalle, IL: Open Court, 1986) 649–62.

(8) Cf. ‘Two Dogmas of Empiricism’ in Quine's From A Logical Point of View (Harvard: Harvard University Press, 1953).

(9) In Pragmatism, James repeatedly mentions usefulness for prediction and control of experience (‘extraordinary fertility in consequences verifiable by sense’, p. 91), conservation of past doctrine (‘we keep unaltered as much of our old knowledge, as many of our old prejudices and beliefs, as we can.’, p. 83; ‘A new belief counts as “true” just in proportion as it gratifies the individual's desire to assimilate the novel in his experience to his beliefs in stock.’, p. 36), and simplicity (‘what fits every part of life best and combines with the collectivity of experience's demands, nothing being omitted’, p. 44). That success in satisfying these desiderate simultaneously is a matter of trade‐offs rather than formal rules is also a Jamesian idea: ‘Success in solving this problem is eminently a matter of approximation. We say that this theory satisfies it on the whole more satisfactorily than that theory; but that means more satisfactorily to our selves, and individuals will emphasize their points of satisfaction differently. To a certain degree, therefore, everything here is plastic.’, p. 35. (All quotations are from Pragmatism and the Meaning of Truth (Cambridge, Mass.: Harvard University Press, 1978)).

(10) For the distinction between ‘revisionary metaphysics’ and ‘descriptive metaphysics’ see Strawson's Individuals (London: Methuen, 1959).

(11) e.g. J. Rawls, A. Heller, J. Habermas.

(12) Although I describe this as a ‘formal value’, A. Heller, in A Philosophy of Morals (Basil Blackwell, 1990) derives a number of quite substantive ‘rules of thumb’ from the idea of non‐instrumentalization.

(13) This is one of Heller's ‘rules of thumb’ that I referred to in n. 12.

(14) See my ‘A Reconsideration of Deweyan Democracy’ in The Southern California Law Review, 1991; ‘Epistemology as Hypothesis’ (with R. A. Putnam) in Transactions of the C.S. Peirce Society, Fall 1990, xxvi, 407–33, and ‘William James’ Ideas' (with Ruth Anna Putnam) reprinted in my Realism with A Human Face (Cambridge, Mass.: Harvard University Press, 1990).

(15) This variant is due to Paul Benacerraf, cf. his ‘What Numbers Could Not Be’, reprinted in P. Benacerraf and H. Putnam (eds.), Philosophy of Mathematics, Selected Readings, 272–94 (Cambridge: Cambridge University, 1964).

(16) I ignore here a difference between Quine's position and my own; my own view is that we should not speak of evidence for the truth of set theory, but rather of evidence for the applicability of the concepts of set theory in practice. I argue in a forthcoming paper that the falsity of set theory or arithmetic (apart from the possibility of its turning out to be inconsistent, omega inconsistent, or something of that kind) is not something that has any clear present content. But discussing this difference between my position and Quine's would lead us away from the present topic.

(17) Cf. J. Mackie, Ethics: Inventing Right and Wrong (Harmondsworth: Penguin, 1977) and B. Williams, Ethics and the Limits of Philosophy (Cambridge, Mass.: Harvard University Press, 1985).

(18) Heller, A Philosophy of Morals, 28 ff.

(19) The claim that the person who does not engage in the practice of virtue cannot have a fully adequate idea of what virtue is may seem queer to many contemporary sensibilities, but the similar sounding claim that the person who does not engage in musical practice cannot have a fully adequate idea of what great music is does not sound so queer. In fact, something like this view has been defended in a well‐known series of papers by J. McDowell.

(20) The terminology of ‘instrumental’ and ‘terminal’ values, and the idea that the line between these two constantly shifts in the course of inquiry, are both J. Dewey's, of course.

(21) The pejorative notion of intellectualistic metaphysics is one I take from W. James, who employed it early and late.

(22) Strawson, Individuals.

(23) e.g., in Reconstruction in Philosophy (reprinted by Beacon Press, 1948) and in Ethics (with James Hayden Tufts), reprinted as vol. 5 of John Dewey, the Middle Works 1899–1924, J. A. Boydston, ed. (Carbondale: Southern Illinois University Press, 1978).

(24) For the reasons why it is a mistake to identify a scheme's showing its ability to function in practice with its being ‘empirical’ see ‘It Ain't Necessarily So’, in my Mathematics, Matter and Method, Philosophical Papers, vol. 1, and ‘Language and Philosophy’ in my Mind Language and Reality, Philosophical Papers, vol. 2. (both, Cambridge: Cambridge University Press, 1975).

(25) Reprinted in C. Hartshorne and P. Weiss (eds.), Collected Papers of Charles Sanders Peirce, vol. II, 389–414 (Harvard: Harvard University Press, 1931).

(26) I review Peirce's argument briefly in The Many Faces of Realism (La Salle, Ill. Open Court, 1987), 80–6.

(27) Peirce ensures unrepeatability by imagining a case in which the alternatives are ‘eternal felicity’ and ‘everlasting woe’; in The Many Faces of Realism, I modify Peirce's argument by imagining a case in which the choice is between an easy death and a hard death.

(28) Peirce, ‘The Doctrine of Chances’, 396 ff.

(29) p.84

(30) ‘The community . . . must reach, however vaguely, beyond this geological epoch, beyond all bounds. He who would not sacrifice his own soul to save the whole world, is, as it seems to me, illogical in all his inferences, collectively. Logic is rooted in the social principle.’ Peirce, ‘The Doctrine of Chances’.

(31) The term ‘fictitious transfer’ comes from Reichenbach's discussion of single case probabilities. Cf. The Theory of Probability (Berkeley, Calif.: University of California Press, 1940), 372 ff.

(32) 24/25ths is the probability involved in Peirce's example.

(33) The discussion of the limits of conceivability in Wittgenstein's On Certainty, G. E. M. Anscombe and G. H. von Wright, eds. (Oxford: Basil Blackwell, 1969) is on my mind here.

(34) Although for Carnap it was! Carnap at one time held the view that to say that a hypothesis is confirmed to a certain degree r is just the same as saying that a certain algorithm yields the number ‘r’ when applied to the evidence. On this view, anyone who knows that a certain mathematical function has the value r on a certain argument ipso facto knows something about rational belief; but this is clearly wrong.

(35) I am reminded here of Wittgenstein's distrust of philosophical ‘musts’. When I say that the metaphysical notion of a ‘complete’ description has never been satisfactorily defined, I am thinking, among other things, of Quine's claim that the description given by physics is complete because no change of any kind can take place without a physical change taking place. However, it is also true that no change of any kind can take place without a change in the gravitational field taking place; but not even Quine would maintain that I have ‘completely’ described the world if I have only described the gravitational field! The combination of a ‘must’ with a concept that only looks as if it has been defined is characteristic of the philosophical phenomenon Wittgenstein wanted us to attend to.

(36) In a letter to James dated 13 March 1897, Peirce wrote ‘Undoubtedly, its [philosophy's] tendency is to make one value the spiritual more, but not an abstract spirituality. It makes one dizzy and seasick to think of those worthy people who try to do something for “the poor”, or still more blindly, “the deserving poor”. On the other hand, it increases the sense of awe with which one regards Gautama Booda.’ This letter will be published as part of the preface to the text of C.S. Peirce, Reasoning and the Logic of Things, the 1898 Cambridge Conference Lectures, K. L. Ketner and H. Putnam, eds. (Cambridge, Mass.: Harvard University Press, 1992).

(37) Reason, Truth and History (New York: Cambridge University Press, 1981) 136.

(38) On the notion of a ‘thick’ ethical characteristic cf. B. Williams, Ethics and the Limits and my discussion in ‘Objectivity and the Science/Ethics Distinction’, in M. Nussbaum and A. Sen (eds.), The Quality of Life (Oxford: Oxford University Press, 1992). This essay is also reprinted in my Realism with a Human Face.

(39) ‘Epistemic Merit, Intrinsic and Instrumental’, in Proc. and Addresses of the American Philosophical Society, Sept. 1981, 55, 5–23.

(40) In ‘Epistemology as Hypothesis’, supra n. 14.

(41) R. Carnap, The Logical Foundations of Probability (Chicago: University of Chicago Press, 1950), and The Continuum of Inductive Methods (Chicago: University of Chicago Press, 1952).

(42) Cf. The Logical Foundations of Probability, 294 ff.

(43) Cf. ‘Pragmatism and Pragmaticism’ in Collected Papers of Charles Sanders Peirce (vol. V, ch. 6, pp. 399–430 esp. p. 419).

(44) Cf. The omnipresence of this theme in Peirce's philosophy is the subject of K. O. Apel's C. S. Peirce, from Pragmatism to Pragmaticism (Amherst: University of Massachusetts Press, 1981). See also James in ‘The Moral Philosopher and the Moral Life’ in The Will to Believe and other Essays in Popular Philosophy (reprinted by Harvard University Press, 1979), and Dewey, ‘Nature, Communication, and Meaning’ in Experience and Nature (Mineola, NJ: Dover, 1958) ch. V (first published in 1925 by Open Court).

(45) James in ‘The Moral Philosopher and the Moral Life’.

(46) This is the approach to ethics made famous by Habermas, and Apel. Cf. Habermas' The Theory of Communicative Action (in two volumes) (Boston: Beacon Press), and Apel's Diskurs und Verantwortung; Das Problem des Übergangs zur Postkonventionellen Moral (Frankfurt am Main: Suhrkamp, 1985). For a comparison of discourse ethics and pragmatism see ‘A Reconsideration of Deweyan Democracy’, supra n. 14.

(47) I have used the vocabulary of A. Heller's A Philosophy of Morals in this sentence to bring out the ‘ethical’ tone of the norms governing scientific inquiry.

(48) This is discussed in R. A. Putnam's and my ‘Epistemology as Hypothesis’ supra n. 14.

(49) Cf. My ‘Scientific Liberty and Scientific License’, in Realism with a Human Face (Cambridge, Mass.: Harvard University Press, 1991).

(50) Cf. ‘A Reconsideration of Deweyan Democracy’.

(51) This objection was suggested by Harry Frankfurt.

(52) This is gone into in ‘Epistemology as Hypothesis’.

(53) This may be a part of what Dewey meant when he defined philosophy as ‘the critical method of developing methods of criticism’ in Experience and Nature, 437.