Tuesday, April 26, 2005

The Fact-Value Gap

[The following essay was written for my meta-ethics class, PHIL 236. I tried to cover far too much given the word limit. Oh well.]

The apparent dichotomy between fact and value strikes us as so commonsensical as to require little by way of explanation. This neat picture carves the world into two independent spheres: the descriptive and the evaluative. Within the former we find the objective facts of science, or the way things are. The latter realm comprises instead what ought to be, and is the province of value judgments, which are taken to be subjective or else mysterious in a way that sets them apart from natural facts. It might seem this dualism is enforced by an unbridgeable gulf – that no collection of facts could ever entail an evaluative conclusion. In this essay I aim to show why that assumption is mistaken.

The doctrine of the fact/value gap is essential to two opposing meta-ethical theories. Non-cognitivists use it to support their conception of ethics as subjective, something we project onto the world, rather than an objective part of reality. Intuitionists, by contrast, do believe that objective values exist, but they are held to be non-natural properties that must be detected through a mysterious faculty of ‘moral intuition’. Ethical naturalism might be seen as a ‘third way’ between these two extremes, ironically combining the objectivity of intuitionism with the metaphysical naturalism of non-cognitivism by denying the one doctrine that they share.

The ‘Naturalistic Fallacy’

Attempts to provide a reductive definition of ‘good’ have traditionally been charged with committing the ‘naturalistic fallacy’. Consider the following argument:

(1) X maximizes happiness.

(2) Therefore, X is good.

This is a clear example of an ‘ought’ conclusion being invalidly[1] derived from an ‘is’ premise, since there is no explicit connection between the two. But note that the argument “This shaker contains NaCl, therefore it contains salt” suffers the same logical problem. We are not thereby tempted to propose a ‘fact/chemical gap’![2] In either case, we require a theoretical account of the relation between the two concepts (e.g. “salt is NaCl”). This is the role of normative ethical theory: to relate the moral facts to the non-moral facts so as to enable reasoning between the two.

Our sample argument involved an implicit appeal to utilitarianism:

(0) What maximizes happiness is good.

Inserting this premise into the earlier argument yields a logically valid deduction. Yet proponents of the fact/value gap would nevertheless charge it with committing the ‘naturalistic fallacy’. Their target is not the argument’s logical form, but the definitional premise (0). As G.E. Moore defined it, “the naturalistic fallacy… consists in identifying the simple notion which we mean by ‘good’ with some other notion.”[3] But of course it is question-begging to charge reductionists with this, for whether their proposed definition is indeed of one and the same concept (rather than some other one) is precisely what is at issue. So the naturalistic fallacy is no argument against reductionism, it can only be the conclusion of a prior anti-reductionist argument.[4]

One might appeal to the ‘Open Question’ argument to play this prior role. If a proposed definition was correct, then anyone with a full grasp of each word’s meaning would recognise their synonymy, so it would be absurd to ask whether an object could have one property without the other. For example, it makes no sense to ask, “Bob is a bachelor, but is he unmarried?”. However, for any proposed reduction of ‘good’ to some natural property ‘X’ (e.g. pleasure), it seems that “is X good?” remains an open question. So, the argument goes, ‘good’ and ‘X’ cannot mean the same thing after all.

Three responses can be made to this. Firstly, the truth of even analytic statements is not always immediately obvious. So an analytically ‘closed’ question might yet appear open, pending further reflection and analysis.[5]

Secondly, two words might designate the same property even if they differ in meaning. The empirical methods of science uncover synthetic, rather than analytic, identities.[6] However, the naturalist must explain how moral facts could be identified with natural ones. Since Open Question arguments could be re-applied to any proposed criteria for identification, the justification for which must be a priori, this collapses into the first response.[7]

The third option is to hold that the common conception of morality is confused and in need of revision. Our aim then would be to capture the most important and coherent aspects of the folk concept in a naturalistic analysis, and abandon the rest.

Descriptive and Evaluative Meaning

Non-cognitivist defenders of the fact/value gap might seek to ground it instead on how we use language. When one describes a ‘good wine’, there appears to be an evaluative component to the expression which is independent of its descriptive meaning. Two people might agree on all the (descriptive) qualities of the wine – call these ‘X’ – and yet disagree as to whether these qualities made it good or not. What meaning ‘good’ has over and above ‘X’ is the evaluative meaning, that is, the commendation of the speaker.[8]

I think this picture is rather misleading, in that it conflates evaluation with conation. Evaluations need not be intrinsically action-guiding at all; ‘dispassionate evaluation’ is not a contradiction. One might evaluate an object against criteria which they care little for, as does the reluctant judge of some local competition.

Note that we may happily grant a fact/attitude gap, and locate value in the former realm. In doing so, however, we will need to provide some explanation of why it is that evaluations are often – if not always – action-guiding. The best explanation for this is found not in positing some special sort of evaluative meaning, but by recognising the import of the objects being evaluated, and noting the variety of uses to which language may be put. As McNaughton writes:

In telling my fellow-picnickers that there is a bull in the field I may not only be making a statement but also, in that context, warning them and advising them to take evasive action. This does nothing to show that the sentence ‘There is a bull in the field’ has a special kind of meaning. [9]

So even if sentences exhibit but a single sort of meaning, they may still be employed for various other communicative purposes. Philippa Foot points out that we often use the word ‘dangerous’ for the speech-act of warning, much like ‘good’ is often used for commending.[10] But of course we are not thereby tempted to incorporate the ‘warning function’ of ‘dangerous’ into the very meaning of the word, or suggest that anyone may, without mistake, assess as ‘dangerous’ whatever they have a fearful attitude towards.[11] That such language often happens to be action-guiding is more plausibly due to the nature of good or dangerous things than the mere force of the words.[12]

Contrary to the non-cognitivist’s suggestion, description and evaluation are not always entirely independent of each other. For example, the primary evaluative criteria for functional concepts, such as ‘knife’ or ‘pen’, are internal to the concept. Knowing that a good pen must write legibly is part and parcel of knowing what a pen is.[13] The same holds for intrinsically teleological notions like ‘belief’; since belief aims at truth, we can infer that a false belief is, in this respect, a bad one.[14] Pre-established criteria also apply to the assessment of roles, such as ‘father’, ‘farmer’, or ‘patriot’.[15] One could not appeal to just any old fact as evidence that someone is a good father. Descriptive meanings put constraints on what might legitimately feature in evaluations.

The entanglement of fact and value becomes even clearer in the case of thick ethical concepts such as ‘cruel’ and ‘brave’. Dichotomists must argue that such concepts can be ‘factored’ into strictly descriptive and evaluative components. It is not obvious how to go about this, however. Hilary Putnam argues that it is quite impossible to give the ‘descriptive meaning’ of ‘cruel’ without using the word itself or some synonym. Mastery of the concept, even in neutral (‘descriptive’) contexts, requires sensitivity to the evaluative point of view.[16] This exemplifies a fundamental intertwining of the descriptive with the evaluative.

The Facts About Value

It seems clear that we can be mistaken in our value judgments. But this implies there is something for us to be mistaken about. That is, the possibility of moral error requires that there be moral facts. We might naturally wonder what sorts of facts these could be. In what follows, I will approach this question through analysis of non-moral value.

Evaluation occurs relative to some presupposed criteria. Whether an object measures up to some particular standard is a matter of objective fact, about which we can be mistaken. We saw above that sometimes the criteria of evaluation will be internally fixed by a concept. Normativity may also arise within an institution: if you make a promise, you ought to keep it, and if the umpire says you are ‘out’, then you ought to leave the field.[17] One may engage in an external critique of institutions, but at least from within they create their own normative standards, against which ‘ought’ claims can be assessed.

A key question is whether there are determinate criteria for assessing well-being. Anscombe notes that we have no difficulty reasoning from natural facts to conclusions about what a plant ‘needs’ in order to flourish.[18] Similarly, there is little question that humans have various basic and complex needs, the deprivation of which makes us worse off. It might be further added that, in general, the fulfilment of our desires makes a positive contribution to our welfare. And note that whether our desires are fulfilled or not is also a matter of objective fact.

Once we have an account of human well-being,[19] it is easy enough to construct an agent-objective aggregate, which could plausibly be equated with moral value. All we need to grant is the eminently plausible claim: “Morality, by its very nature, is concerned with what is good from the perspective of the moral community”,[20] in conjunction with some aggregation principle (e.g. maximization or maximin) that allows us to define the good of the community in terms of the good of the individuals that compose it.

On the general account sketched above, value is assessed relative to presupposed criteria. We can say what ought to be done according to the rules of a game, or what is necessary for the flourishing of an individual or society, but there is no sense of value that transcends these various frameworks. One might object that there is no way to adjudicate between rival standards of evaluation: if morality and self-interest conflict, which is the more important consideration? But this question is either incomplete or nonsensical, as we must ask: ‘important’ according to what criteria?

The essentially relational nature of value is made clearer by the notion of an evaluative point of view. Absolute objectivity, or the absence of a perspective, leaves no room for value; the universe doesn’t care what we do. But this need not lead to nihilism. Viewpoints exist, and value exists in relation to them. Morality does not require some ‘absolute value’ that is good from the null view. On the contrary: evaluation is impossible when deprived of all criteria, and ‘ought’ becomes meaningless – “a word of mere mesmeric force”.[21]

This might be thought to undermine the demands of morality. Morality is, of course, categorical in the sense that it applies independently of our personal concerns. One is not released from a moral obligation by saying that they do not care about it. The viewpoint of morality is collective, so not subject to the whims of any sole individual. Yet this is consistent with morality being ‘hypothetical’ in the sense that it is a means to an end – that of human wellbeing – and will not be considered reason-giving by those who do not share this end.[22] Similar remarks could be made of epistemology, yet this doesn’t undermine our general concern for evidence and truth.[23]

The relational theory situates value firmly within the realm of fact. The facts about value are not made any less real simply by virtue of being relational. If something is obstructive to human ends, that is a matter of natural fact. The fact-value gap is founded on a failure to recognise human desires, and other such sources of teleology, as part of the real world. The ‘no ought from an is’ doctrine might be better restated as ‘no ought from non-teleological facts’,[24] where ‘teleological facts’ involve concepts that come with pre-established criteria for evaluation, as discussed earlier in this essay.

We saw various examples of non-moral ‘oughts’ being legitimately derived from teleological facts. This result is also supported by the generally-accepted thesis that values supervene on the natural facts. There is no possible world that is exactly like our own one in all factual respects, but where it was 'morally right' of Hitler to carry out the Holocaust. This suggests that the natural facts of the matter ‘fix’ or determine the moral ones, which would set up an entailment relation between them. This essay will now conclude with a moral counter-example to the fact/value gap, borrowed from James Rachels:[25]

P: The only difference between doing A and not doing A is that, if we do A, a child will suffer intense prolonged pain. Otherwise, everything will be the same.

Q: Therefore, it is better not to do A.

It seems clear that there is no possible world where P is true yet Q false. That is, P entails Q. The fact/value gap is thus crossed.


Anscombe, G., ‘Modern Moral Philosophy’ in W. Hudson (ed.) The Is-Ought Question. London: Macmillan, 1969.

Black, M., ‘The gap between ‘is’ and ‘should’’ in W. Hudson (ed.) The Is-Ought Question. London: Macmillan, 1969.

Chappell, R., ‘Morality as Means’, http://pixnaps.blogspot.com/2005/02/morality-as-means.html

Darwall, S., Philosophical Ethics. Boulder: Westview Press, 1998.

Foot, P., ‘Goodness and choice’ in W. Hudson (ed.) The Is-Ought Question. London: Macmillan, 1969.

Foot, P., ‘Moral Beliefs’ in P. Foot (ed.) Theories of Ethics. London: Oxford University Press, 1967.

Foot, P., ‘Morality as a System of Hypothetical Imperatives’ in S. Cahn and P. Markie (eds.) Ethics: history, theory, and contemporary issues. New York: Oxford University Press, 1998.

Frankena, W., ‘The Naturalistic Fallacy’ in P. Foot (ed.) Theories of Ethics. London: Oxford University Press, 1967.

Hare, R., ‘Descriptivism’ in W. Hudson (ed.) The Is-Ought Question. London: Macmillan, 1969.

McNaughton, D., Moral Vision. New York: B. Blackwell, 1988.

Putnam, H., The Collapse of the Fact/Value Dichotomy. Cambridge, Mass.: Harvard University Press, 2002.

Rachels, J., ‘Naturalism’ in H. LaFollette (ed.), The Blackwell guide to ethical theory. Malden, Mass.: Blackwell, 2000.

Railton, P., ‘Moral Realism’ in P. Railton, Facts, values, and norms. Cambridge: Cambridge University Press, 2003.

Searle, J., ‘How to derive ‘ought’ from ‘is’’ in W. Hudson (ed.) The Is-Ought Question. London: Macmillan, 1969.

Searle, J., ‘Objections and replies’ (Appendix) in W. Hudson (ed.) The Is-Ought Question. London: Macmillan, 1969.

Simpson, P., Goodness and nature. Boston: M. Nijhoff, 1987.

Smith, M., ‘Moral Realism’ in H. LaFollette (ed.), The Blackwell guide to ethical theory. Malden, Mass.: Blackwell, 2000.

[1] At least in the syntactical sense. Once the meanings are considered, perhaps it would be impossible for the premise to be true and yet the conclusion false. This will depend on which moral theory is true.

[2] H. Putnam, The Collapse of the Fact/Value Dichotomy, p.19, points out that it may be useful in some contexts to distinguish “between ethical judgments and other sorts of judgments”, or for that matter, chemical judgments and other sorts, but none of this entails any metaphysical conclusions.

[3] Quoted in W. Frankena, ‘The Naturalistic Fallacy’, p.56.

[4] Ibid., pp.58-59.

[5] P. Simpson, Goodness and nature, p.26. He suggests, for example, that the Euclidean definition of a circle as “a plane figure contained by one line, which is called the circumference, and is such that all straight lines drawn from a certain point within the figure to the circumference are equal to one another” (p.20) is far from self-evident to non-geometers. This response is elaborated by M. Smith, ‘Moral Realism’, p.31, who argues that the naturalist must aim at an analysis which “entails the complex set of constraints on the way in which we use the word ‘right’.” So long as it does this job, it is of no concern whether the analysis is “obvious”.

[6] S. Darwall, Philosophical Ethics, p.35. For example, “salt is NaCl” is a synthetic identity statement.

[7] M. Smith, ‘Moral Realism’, pp.29-30.

[8] R. Hare, ‘Descriptivism’, p.243.

[9] D. McNaughton, Moral Vision, p.53.

[10] P. Foot, ‘Moral Beliefs’, p.87.

[11] Of course you may justifiably feel fear because you believe something is dangerous. But the point is that such a belief may be in error. This suggests that we may be similarly mistaken about whether something really is good. It is a factual matter.

[12] Foot, ‘Moral Beliefs’, p.95, makes a similar point about the desirability of virtues.

[13] Foot, ‘Goodness and choice’, p.216.

[14] This idea is developed from J. Searle, ‘Objections and replies’, p.264, “it is internal to the notion of a statement (descriptive word) that a self-contradiction (descriptive word) is a defect (evaluative word).”

[15] Foot, ‘Goodness and choice’, pp.218-219.

[16] Putnam, pp.38-40. As noted earlier, we can distinguish the (whole) meaning of the word from the various uses it is put to; but this is not to factor the meaning into ‘descriptive’ and ‘evaluative’ components. This seems to support my suggestion of a fact/attitude gap in place of the fact/value gap.

[17] J. Searle, ‘How to derive ‘ought’ from ‘is’’, p.132.

[18] G. Anscombe, ‘Modern Moral Philosophy’, p.181.

[19] It would go beyond the scope of this essay to provide a full account here, but I hope the previous paragraph highlights some general grounds for thinking this can, in principle, be done. See P. Railton, ‘Moral Realism’, pp. 9-17, for a very plausible explication of this ‘non-moral value’.

[20] Darwall, p.125.

[21] The quote is from Anscombe, p.182.

[22] Foot, ‘Morality as a System of Hypothetical Imperatives’, pp.616-617.

[23] Railton, pp.6-8. The question of whether we have good reason to care about ‘the moral point of view’ is one I explore further in my online article, ‘Morality as Means’.

[24] J. Rachels, ‘Naturalism’, p.79, suggests something along the lines of ‘no ought from facts that make no mention of desires’, but this fails to account for other sources of normative criteria discussed earlier.

[25] Ibid., p.76.


  1. Is/ought is NOT a philosophical distinction, it’s a cultural one. It starts with the separation of the world of value from the world of fact (dualism). It continues with the reduction of value to bare moralistic “oughts”, given under divine command. It is completed with the development of naturalistic science, which uses a hypothetical “objective” or “not-in-the-world” viewpoint to dissect the world of facts without any reference to values.

    The philosophical distinction is not an innovation, it is just a final diagnostic stage, thick with irony, where the philosopher thinks he has discovered some ineffable truth about the universe. In reality, he is just unaware of the profound cultural changes that have given rise to the fact/value distinction, a distinction which most cultures that have ever existed would find puzzling.

  2. Isn't this problematic because of the implicit use of the 'is/ought' distinction by 'Anonymous' who pretty clearly thinks that the actual way philosophers think (a matter of fact or 'is') is not the way they ought to think (a matter of value or 'ought')? If everything dissolves into culture then we may as well give up; what for example would there be left to be 'ironic' about?

  3. Enjoyed your article: "It seems clear that there is no possible world where P is true yet Q false. That is, P entails Q. The fact/value gap is thus crossed." If human does A --> child suffers. Therefore, humans should not do A. This contains an unrecognized 'because' clause (or a missing premise) that a child suffering at the hands of any natural event is a bad thing (evaluation), and at the hands of a human actor is a morally bad thing. Tooth pulling can cause a child to suffer, and, as an evaluation, that may be a good thing. What science and observation reveals is that humans (and living organisms in general) are evaluative entities, structured to obtain non-teleological objective ends, such as food, a mate, or survival. I wrote this essay (in 1994):
    I appreciate the work you achieved in your post here. Thanks for posting it.


Visitors: check my comments policy first.
Non-Blogger users: If the comment form isn't working for you, email me your comment and I can post it on your behalf. (If your comment is too long, first try breaking it into two parts.)