An Introduction to the Is-Ought Problem

Of all the dialectical bludgeons, the alleged inferential gap between “ought” and “is” is ubiquitous. The thesis is used to rebut numerous normative claims, but few in popular circles are aware of its pedigree, and because of that, they’re prone to misunderstanding its significance and meaning. Usually, those who employ the is-ought problem in its orthodox guise don’t realize that it’s a double edged sword. If the thesis is true, its undermining effects do not discriminate.

Historical Background:

The is-ought problem can be traced back to A Treatise on Human Nature. There are several interpretations of Hume’s words, and some others will be investigated in future posts, but for now the dominant 20th century interpretation will be explored. The passage in which the is-ought problem makes its appearance goes as follow,

“In every system of morality, which I have hitherto met with, I have always remarked, that the author proceeds for some time in the ordinary ways of reasoning, and establishes the being of a God, or makes observations concerning human affairs; when all of a sudden I am surprised to find, that instead of the usual copulations of propositions, is, and is not, I meet with no proposition that is not connected with an ought, or an ought not. This change is imperceptible; but is however, of the last consequence. For as this ought, or ought not, expresses some new relation or affirmation, ’tis necessary that it should be observed and explained; and at the same time that a reason should be given, for what seems altogether inconceivable, how this new relation can be a deduction from others, which are entirely different from it. But as authors do not commonly use this precaution, I shall presume to recommend it to the readers; and am persuaded, that this small attention would subvert all the vulgar systems of morality, and let us see, that the distinction of vice and virtue is not founded merely on the relations of objects, nor is perceived by reason” (Hume 302).

The orthodox interpretation is that Hume believes that no evaluative judgment may be the conclusion of a valid deductive argument that only has non-evaluative premises. For any set of descriptive or factual premises, no evaluative conclusion can follow without additional evaluative premises or some inferential rule that is essentially evaluative in content. Another way of putting the thesis is that there is no rule of deductive inference that licenses the move from factual or descriptive premises to evaluative conclusions.

The dominant interpretation of the 20th century aligns Hume’s is-ought thesis with non-cognitivism. Since evaluative claims are distinct in kind (“ought”) from factual/descriptive claims (“is”), it seems natural to embrace some form of non-cognitivism. Statements of fact are truth evaluable, whereas other kinds of utterances don’t seem to be. Examples of such utterances are expressions of fright, such as screaming, and expressions of pain, such as saying “ouch” and groaning. These utterances express non-cognitive attitudes through such phrases or noises. The attitudes are distinct from propositional attitudes due to the fact that the latter are truth-conducive (your beliefs may be true or false).

The orthodox interpreters tend to point to Hume’s motivation arguments to establish the non-cognitivist reading. Given Hume’s non-cognitivism being established by such arguments, the standard reading of the is-ought problem makes more sense; it would just be an entailment of Hume’s anti-rationalism about morality.

A brief statement of the more well known argument is that reason alone cannot motivate action (given the soundness of Hume’s first anti-rationalist argument), but morality can, so morality cannot be grounded in reason alone (Cohon 2010). If the dominant 20th century interpretation of Hume is correct, then this argument aimed to establish some form of non-cognitivism, which, in essence, would carve an ontological gap between fact and value. So, on the non-cognitivist reading, there are facts, which are those things that our beliefs aim to represent in some way, and there are values, which are products of our faculty of moral sentiment.

It should be noted, however, that the non-cognitivist reading of Hume is something that is no longer taken for granted, given recent Hume scholarship (Radcliffe 2006).

In essence, this would carve an ontological gap between fact and value. So, on the non-cognitivist reading, there are facts, which are those things that our beliefs aim to represent in some way, and there are values, which are products of our faculty of moral sentiment.

The problem for people who employ this interpretation of the is-ought problem as an argument against evaluative positions they disagree with is that it is embedded in an interpretive framework that includes a non-cognitivist element as an overarching theme. For the most part, the people in popular circles that I’ve encountered using this argumentative strategy tend to have cognitivist sympathies, and with such sympathies come evaluative claims that are believed to be truth evaluable. Insofar as their opponents’ arguments are supposed to fall prey to this problem, so will theirs.

Logical Maneuvers:

There are various ways to respond to the orthodox version of the is-ought problem. The first way involves what could be deemed logical tricks. The first trick, originally from Arthur Prior, goes as follows:

P1. It’s raining outside.


C. Either it’s raining outside or you ought not to steal.

This is a valid deductive argument that employs disjunction introduction. Now, if you claim that this isn’t really getting an ought from an is because the conclusion isn’t really an statement with evaluative content, then this example will evade your concern,

P1*. It isn’t raining outside.

P2*. Either it isn’t raining outside or you ought not to steal.


C*. You ought not to steal.

If you claim that the conclusion of the first argument lacks evaluative content, then the second argument only employs non-evaluative premises (since P2* is assumed to be non-evaluative for the sake of argument) and gets you to the evaluative conclusion C*.

The second trick goes like this,

P1$. It is raining outside and it is not raining outside.


C$. You ought not to steal.

This is also a valid deductive argument that employs the principle that from a contradiction, anything follows. Start with (i) P and not-P, simplify to (ii) P, (iii) not-P. Apply disjunctive addition to (ii) and you can deduce (iv) P or Q. Disjunction elimination on (iii) and (iv) gets the conclusion Q (Joyce 153).

Another logical trick involves an appeal to authority, but it raises issues about what constitutes evaluative content. If you’re interested in that, you can read chapter seven of Armstrong’s book, Moral Skepticisms.

There are other, more complicated, attempts at bridging the inferential gap between is and ought, such as Toomas Karmo’s proof. Unfortunately, there is no space to give a sufficient explanation of his argument.

Robust Bridges:

A different sort of attempt at bridging the alleged gap involves more than just logical trickery. One problem people may raise about the previous solutions is that they don’t involve deductions from non-evaluative premises that produce genuine moral knowledge or justification. Mere disjunction introduction and explosion aren’t sufficient because one could replace the evaluative disjunct with any other evaluative disjunct without affecting the argument. So, a more robust inference from is to ought needs to be explored.

It could be useful to step back and rethink the framework in which we’re trying to understand the is-ought problem. If we think of the regress problem in the context of moral justification, then there are only a few non-skeptical solutions: (i) foundationalism, (ii) coherentism, (iii) infinitism, and (iv) inference from non-evaluative knowledge. Granting for the sake of argument that i-iii all fail to secure moral justification, we need to evaluative our fourth option. So, we can think of the is-ought gap as epistemic rather than ontic.

So, the problem becomes finding a way to make an inference from a body of non-evaluative knowledge to an evaluative conclusion such that the inference transfers justification or warrant from the body of knowledge to the evaluative conclusion.

There are quite a few attempts in the literature to provide such an inferential link, so the selection of views here is going to be, to some extent, arbitrary.

Searle’s Metalinguistic Argument:

The first attempt, a metalinguistic strategy, is found in the work of John Searle:

P1. Jones utters the words, “I hereby promise to pay you, Smith, five dollars.”

P2. Jones promised to pay smith five dollars.

P3. Jones placed himself under an obligation to pay Smith five dollars.

P4. Jones is under an obligation to pay Smith five dollars.


C. Jones ought to pay smith five dollars (Searle 1964).

Searle’s argument aims to bridge Hume’s chasm by employing what he calls, “institutional facts,” and other things being equal, it is an institutional fact that the utterance of, “I hereby promise to . . .” is a performative by virtue of which the utterer undertakes an obligation (Searle 1964). So, it seems as though the gap between is and ought is bridged by institutional facts. It also seems as though the moral regress problem is solved by the fourth option alluded to in the last section. We could infer evaluative conclusions with robust content from non-evaluative premises.

The main issue is that the argument fails to rule out moral skepticism and the error theory, so it fails as a means of obtaining moral knowledge and justification. To see how Searle’s argument fails, notice the distinction between purporting to undertake an obligation and actually undertaking an obligation. An illuminating analogy comes from Michael Huemer,

“. . . suppose you call me collect, and I agree to accept the charges, but due to a mistake, the phone company never actually charges me. My having “accepted” these charges does not entail that I ever actually receive the charges. In a similar sense, a person might conceivably “accept” an obligation without ever actually having the obligation” (Huemer 75).

Somebody may undertake an obligation in the weaker sense of just purporting to if there are no actual obligations or if nobody is ever actually obligated to do anything. In other words, if the error theory is true, then Jones only purports to undertake an obligation. The problem arises, then, because the argument fails to rule out that possibility; and any argument employed to vindicate moral knowledge that fails to rule out the possibility that we lack moral knowledge is a dialectical failure.

If one introduces additional premises to the effect that the argument rules out error theory and moral skepticism, then those premises will have evaluative content, which means that they must be justified, which then reintroduces the moral regress problem. The evaluative premises would also need to be justified by virtue of inference from some body of evaluative knowledge, given our rejection of the other non-skeptical solutions.

A second problem with Searle’s argument is the inference from P1 to P2. Such an inference presupposes a large body of background knowledge about the norms of the social context in which Jones is embedded (Huemer 75). On one reading, the constitutive facts of that knowledge would all be non-evaluative, but on another reading, they would not be.

There are two ways to understand the move from P1 to P2: the evaluative (internal) sense, and a non-evaluative (external) sense (Mackie 66-72). The external sense merely takes the move to be a descriptive account of a rule governing  Jone’s speech act under a linguistic institution. It would be as if an alien was evaluating the argument from outside the linguistic institution of promising. Such an alien would require an additional premise that says, if person utters, “I promise to X” within a particular linguistic institution, then that person has made a promise within that linguistic institution (James 154). But that rule would teach the alien a fact about the rules of a particular institution, such as a rule for moving chess pieces across a board. It becomes a description of what Jones ought to do within that particular institution, not what Jones ought to do, full stop (James 154-155).

Viewed from the inside, Jones is surely obligated to pay his debt, but from the outside, it’s merely a statement of fact about particular linguistic conventions. The problem is, though, that when viewed from inside the institution of promising, the argument requires an additional inferential rule to secure the conclusion.

In other words, there is a dilemma for Searle’s argument. If we view it from the outside, then the conclusion is merely a brute fact, and we haven’t really gotten an ought from an is. If we view it from inside the institution, then the argument requires an additional premise in the form of an inference rule that says if one utters, “I promise to do X,” then one ought to conclude that that person promised to do X (James 157). But that is a rule which tells us what we ought to do, and as such is subject to the same requirement of justification as any other normative statement, thus restarting the regress. Evaluating the argument from within the institution, then, requires accepting particular normative statements (James 157). So, on the internal reading, Searle’s argument fails to solve the moral regress problem, and the external reading doesn’t even interact with it.

So, we’ve gone over Searle’s argument and found that it doesn’t provide a means to end the regress problem; but that doesn’t render the entire strategy bankrupt, as there are more promising metalinguistic arguments, such as Jesse Prinz’s. Also, it should be noted that Searle did not intend for the argument to be employed as a solution to the moral regress problem. Searle saw his argument as an example of how one might bridge the deductive is-ought gap by way of institutional facts and the notion of speech acts (Searle 1964). I merely used it as a means of illustrating one way the metalinguistic strategy for ending the regress problem for moral justification may be developed.

Geach’s Attempt:

The next argument is from Peter Geach,

P1. If Evan were to promise to adopt some practice he would adopt it.

P2. If Evan were to utter sentence W, he would be promising to adopt the practice of doing something wrong twice a day.

P3. Nobody should adopt the practice of doing something wrong twice a day.


C1. If Evan were to utter W, he would adopt the practice of doing something wrong twice a day.


C2. Evan should not utter W (Geach 1977; Huemer 2005).

The first thing to note about this argument is that premise three has evaluative content. Geach claims that the third premise is analytically true (Geach 1977). So, if the argument was successful, it would bridge the is-ought gap by virtue of an analytic truth that would be known by grasping the content of the proposition being expressed; in other words, understanding the content would just be acquaintance with its truth conditions.

The problem is that this would be to smuggle in a moral epistemology insofar as there is some account of grasping the content of the analytic moral proposition that doesn’t rely on the fourth solution to the moral regress problem. If it relied on the fourth account of the structure of moral justification, then it would merely push the problem back a step.

If premise three requires an account of content-grasping for its justification that makes use of moral foundationalism, infinitism, or coherentism, then it merely assumes that there is no gap by virtue of adopting a solution in which it cannot be formulated.  In other words, the epistemic gap only makes sense given the fourth solution to the regress problem, so adopting another solution changes the subject.

Another problem is that the argument is invalid. The inference from P3 to C1 is valid just if,

P1*. S ought not to do A.

P2. If S did B, S would do A.


C*. S ought not to do B.

is a valid form of inference (Huemer 77). An example that illustrates the invalidity of the above inference form is provided by Huemer,

“John is a judge about to pass sentence on Mary, a convicted marijuana dealer. Mary’s crime is minor at worst; John, however, has an intense, irrational hatred for all drug users, as a result of which he is determined to sentence Mary to either life imprisonment or death. He could sentence Mary to only a brief prison term, but he would not in fact do so. Now consider the following inference:

P1$. John ought not to sentence Mary to life imprisonment.

P2$. If John were to refrain from sentencing Mary to death, then he would sentence Mary to life imprisonment.


C$. John ought not to refrain from sentencing Mary to death” (Huemer 77).

It seems obvious that P1$ is true given the fact that John could sentence her to a minor term, assuming we put aside the regress problem for the sake of argument. P2$ is stipulated as a fact about John’s psychology. However, the conclusion is clearly false (Huemer 77). So, the argument form is invalid.

A third problem with the argument is that P1 is evaluative (Huemer 77). The premise goes like this,

(∀x) (Evan promises to do x > Evan does x)

The evaluative nature of the premise can be revealed by substituting “act wrongly” for x.

Evan promises to act wrongly > Evan acts wrongly

Once more, I’ll draw from Huemer to illustrate this point,

“. . . once we fix the natural facts about how Evan would behave upon making such a promise, whether the whole sentence is true then depends on whether the behavior would be wrong” (Huemer 78).

Let’s say that if Evan promised to act wrongly, he would rob a bank and go for a walk. The facts about what Evan would do are fixed, but the whole sentence is true just if the things he would do if he made such a promise are actually wrong (Huemer 78). So, the truth conditions for P1 are such that P1 is true if and only if some evaluative state of affairs obtains.

It seems as though Geach’s argument fails to provide a means of overcoming the regress problem due to the problems explored above.

The Naturalist’s Alternative:

The next attempt at providing a means of making the fourth solution to the regress problem work is more indirect. So far, we’ve only looked at attempts that aimed at establishing evaluative knowledge through deduction from non-evaluative premises. But what if being unable to establish an ought by deducing it from non-evaluative premises isn’t that interesting? For instance,

P1. My glass contains the liquid H2O.

P2. I am about to drink the liquid in my glass.


C. I am about to drink water.

This argument is invalid by Tarskian standards because a translation of the argument into predicate calculus would allow for models where P1 and P2 are true but C is false (Joyce 154).

However, we don’t deny that we can have knowledge of the molecular structure of the liquid in my glass, despite the invalidity of such arguments. Perhaps something analogous holds for the nature of normativity. To further illustrate the point, the worldview that physics gives us clearly allows for biological facts without the need for us to be able to deduce such facts from propositions about fundamental particles, fields, and the laws of nature (Joyce 154). So, a moral naturalist who is of the non-reductivist or synthetic reductivist stripe has room to maneuver insofar as she can show that the is-ought gap is as uninteresting as the H2O-water gap and the physics-biology gap.

One avenue for the moral naturalist is inference to the best explanation. For example, perhaps the best explanation for Hitler’s behaviors during WW2 and prior is that Hitler had a morally depraved character. The fact that Hitler was morally depraved or vicious (along with non-evaluative background knowledge) best explains his actions, or so this form of reasoning goes. So, the explanation of Hitler’s actions makes them more probable than without the explanation, or with an alternative explanation.

Another way to think about it is in terms of a hypothesis making certain observations more expected than they otherwise would be. Some thinkers such as Nick Sturgeon adopt this strategy. The merits of abduction applied to evaluative knowledge cannot be assessed here, since the post would become too lengthy and disjointed. Suffice it to say, however, that the Sturgeon strategy involves conceptions of explanation that are controversial.


There are several other attempts to bridge the is-ought gap that were not explored here. I will assess them over the course of several future posts. Also, the Sturgeon explanatory strategy will be explored in the future, as will alternative ways to interpret Hume’s is-ought thesis, and my own solution to the moral regress problem. So far, the prospects for the fourth solution to the regress problem seem dim. However, there may be options for those who wish to endorse the fourth solution that are more satisfactory than the ones explored here.

Works Cited:

Cohon, Rachel, “Hume’s Moral Philosophy”, The Stanford Encyclopedia of Philosophy (Fall 2010 Edition), Edward N. Zalta (ed.), URL = <>.

Geach, P. T. “Again the Logic of ‘Ought’.” Philosophy: 473. Print.

Huemer, Michael. Ethical Intuitionism. Basingstoke England: Palgrave Macmillan, 2005. Print.

Hume, David, and Mary J. Norton. A Treatise of Human Nature: A Critical Edition. Oxford: Clarendon, 2007. Print.

James, Scott M. An Introduction to Evolutionary Ethics. Chichester, West Sussex, U.K.: Wiley-Blackwell, 2011. Print.

Joyce, Richard. The Evolution of Morality. Cambridge, Mass.: MIT, 2006. Print.

Mackie, J. L. Ethics: Inventing Right and Wrong. Harmondsworth: Penguin, 1977. Print.

Radcliffe, Elizabeth S. “Moral Internalism and Moral Cognitivism in Hume’s Metaethics.” Synthese (2006): 353-70. Print.

Searle, John R. “How to Derive “Ought” From “Is”” The Philosophical Review: 43. Print.