Open Question Arguments

In online discussions about meta-ethics, terms like, “open question argument” are tossed around a lot. In this post, I’ll discuss open question arguments in detail.

One common misconception about open question arguments is that they are only directed at ethical naturalism; this is false. Open question arguments are supposed to be problems for any kind of analytic reductionism. G.E. Moore used the naturalistic fallacy and open question argument as ammunition against any reductive accounts of moral predicates and properties.

Before getting deeper into these topics, a few things need to be explained. Analytic and synthetic reductionism are two kinds of meta-ethical reductionism. The former not only aims to reduce normative properties to non-normative properties, but also to analyze normative predicates (or concepts) in terms of non-normative predicates (or concepts). The latter merely aims to reduce normative properties to non-normative properties, while maintaining that no analysis can be given that entirely eliminates normative predicates. In this post, I’ll be dealing primarily with analytic reductionism.

The open question argument is intended to support the charge that it is fallacious to give a conceptual analysis of normative predicates in terms of non-normative predicates (the naturalistic fallacy). Here is a reconstruction of the classic version of the argument (Miller 12-13):

1a. Suppose the predicate ‘good’ is synonymous with, or analytically equivalent to the naturalistic predicate ‘N’.


2a. It is part of the meaning of the claim ‘x is N’ that ‘x is good’.


3a. Someone who seriously asked ‘Is an x which is N also good?’ would betray some conceptual confusion.


4a. For any given natural property N it is always an open question whether an x which is N is good. That is to say, it is always a significant question, of any x which is N, whether it is good: asking the question ‘Is an x which is N also good?’ betrays no conceptual confusion.

5a. It cannot be the case that ‘good’ is synonymous with, or analytically equivalent to ‘N’.


6a. The property of being good cannot as a matter of conceptual necessity be identical to the property of being N.

For any non-normative predicate ‘N’, it is always an open question whether or not anything possessing such a property also falls within the extension of ‘is good’. The reason this isn’t a problem just for ethical naturalists is that predicates like ‘is commanded by God’ and ‘would be assented to by an ideal observer’ can be substituted for the same result.

What are the responses to this argument? One famous reply was made by William Frankena, The basic objection is that the argument is question-begging due to premise 4a (Miller 14-15). Our belief that any statement leaves an open question can only be employed against a reductionist if it is justified, which is the very issue at hand. However, this objection employs a conception of question-begging that is far too liberal. Imagine a philosopher who holds to the justified true belief analysis of ‘knowledge’. She hears a Gettier counterexample and responds by saying that Gettier has begged the question, since if JTB is true, Jones does know that P, whereas if Gettier claims otherwise, he is assuming that JTB is false. Clearly something has gone wrong here. Providing a counterexample against an analysis that directly contradicts its central thesis is not begging the question. The JTB theorist is not entitled to directly appeal to her own theory as support against counterexamples. She must find flaws in the counterexamples, or ways to neutralize the intuitions elicited by them. Their dialectical force is not deflected by digging in your heels and claiming that they only disprove your theory if your theory is false. Analogously, contra Frankena, the analytic reductionist’s charge of question-begging is dialectically inappropriate (Huemer 69-70).

Another response is that ‘N and ‘goodness’ may have the same reference but they have different senses (Miller 16). The fact that the two terms have different senses, modes of presentation, or rules of use that determine their referents doesn’t preclude the fact that their referents are identical. This is a lesson learned from Frege’s example of the names “Hesperus” and “Phosphorus” referring to the same planet through different senses or modes of presentation [1]. I can refer to something by virtue of one mode of presentation without being aware of the fact that another mode of presentation determines the same referent as well. An example is, “John wants to read a book by Mark Twain” but, “John doesn’t want to read a book by Samuel Clemens,” because John believes that Mark Twain is a famous author who wrote ‘Huck Finn’ but he has never heard of Samuel Clemens.

One issue with the Fregean response is that it actually concedes that the two predicates aren’t conceptually identical. That would mean that this response is just a move from analytic to synthetic reductionism. So, it seems to just be an admission of defeat if the reductionist shifts from an analytic to a synthetic slant because of the classic open question argument. After reflecting on this objection, it should become clear why open question tests are still used for concept identity, but not property identity. We could mistake one referent for two when we employ two different senses to identify it (as is the case with ‘water’ and ‘H2O’), but when it comes to cases conceptual or definitional identity, competent speakers of the language thinking that it’s an open question that concept X and concept Y are identical is prima facie evidence that they are not identical. This is clear in cases of analytic or conceptual truths, such as bachelors are unmarried men. How else would we test supposed cases of analytic or conceptual truths?

The last response is that the argument presupposes that any true conceptual analysis will  be uninteresting and uninformative. But surely this isn’t true; mathematical analysis uncovers interesting and informative truths. So there must be something wrong here (Miller 15). Moore could answer that he is merely exploiting the paradox of analysis, where for any predicate ‘N’, it’s impossible to give an informative and true conceptual analysis of it; if the contents of ‘N’ are analyzed in terms of ‘N*’, and we already understand the contents of ‘N’ by virtue of being able to competently use it in our conceptual repertoire, then we must already know that ‘N’ is equivalent to ‘N*’. This generalizes to any instance of conceptual analysis, and therefore it applies to analytic reductionism about moral properties.

The paradox can be answered by making a distinction between ‘knowledge that’ and ‘knowledge how’. To know about a true analysis of ‘N’ is to possess propositional knowledge (knowledge that), and to grasp a concept such that one can competently use it is to possess know-how. One can know how to use concepts without possessing propositional knowledge of the means by which one does so and the ways in which it can be analyzed into simpler concepts. It is possible to gain knowledge upon hearing about a true reductive analysis of ‘goodness’ in terms of ‘N’ even if one already knows how to employ these predicates in ordinary circumstances (Miller 15-16). Another example is learning that ‘water’ is identical to ‘H2O’. One can know how to use ‘water’ but also gain knowledge upon reading a science textbook about the molecular composition of water. So the analytic reductionist can maintain that premise 4a may be true, but 5a doesn’t follow.

While it may seem as though the last objection is decisive against the argument, there are other versions which do appear to be more successful, but at the cost of narrowing the argument’s scope.

A way to avoid the last objection is to add an extra clause to what qualifies as properly seeing an open question (Miller 18-19):

1b. If ‘good’ and ‘N’ are analytically equivalent, then ceteris paribus competent speakers should, after conceptual reflection, come to find it natural to guide their evaluative judgments by the analysis.

2b. After conceptual reflection, the conviction that ‘Is an x which is N also good?’ is an open question persists among competent speakers. So after conceptual analysis they don’t come to find it natural to guide their evaluative judgments by the analysis of ‘good’ in terms of ‘N’.


3b. We can conclude that ‘good’ and ‘N’ are not analytically equivalent, unless ceteris aren’t paribus (i.e. there is some explanation of why competent speakers do not come to find it natural to guide their evaluative judgments by the analysis).

This is not as ambitious an argument as the classic version. All this argument establishes is a shift of the dialectical burden unto the analytic reductionist to provide an explanation of why, on conscious reflection, people don’t find it natural to be guided by whatever analysis the reductionist puts forward.

The analytic reductionist could reply by saying that if somebody doesn’t grasp the conceptual connection between ‘good’ and ‘N’, then she is the victim of some sort of confusion (Miller 19). However, if the analytic reductionist has no other reason to think a normally competent speaker is displaying such confusion, then this response seems ad hoc at best, and question-begging at worst (if there’s no other evidence of linguistic/conceptual confusion in other areas).

So, without a viable explanation from the analytic reductionist that neutralizes premise 1b, the reconstructed open question argument can be regarded as successful. Even if we embrace the paradox of analysis objection, there would still be a problem, since this argument takes the distinction into account by virtue of the ‘conceptual reflection’ clause. The paradox merely maintains that one can be ignorant of a proper analysis of ‘N’ while being a competent user of ‘N’. The ‘conceptual reflection’ clause takes the proper analysis into account, so the paradox objection has no force.

One response to this argument is to deny judgment internalism. The argument presupposes that judgment internalism is true, since the ‘conceptual reflection’ clause maintains that because somebody still finds it unnatural to guide their behavior by a given reductive analysis, there must be a gap between the two predicates. On judgment externalism, one can recognize and genuinely make normative judgments without being motivated to guide her actions by them (amoralists aren’t conceptual impossibilities).

Another reconstruction of the open question argument that wears its internalism on its sleeves goes like this (Miller 19-20):

1c. There is a conceptual or internal link between making a moral judgment and being motivated, ceteris paribus, to act as that judgment prescribes. Absent some weakness of the will or other psychological affliction, judging that a type of action is morally good entails being motivated to perform actions of that type. Someone with no psychological afflictions who apparently judges that a type of action is morally good but consistently claims that he sees no reason to perform actions of this type doesn’t grasp the concept of ‘moral goodness’.

2c. Competent and reflect speakers are convinced that they are able to imagine clear-headed (and otherwise psychologically healthy) beings who judge that ‘N’ obtains but who fail to find appropriate reason or motive to act in accordance with that judgment.

3c. If there were a conceptual link between judging that ‘N’ obtains and being motivated to act accordingly, we would expect competent and reflective speakers of English to have some conviction described in 2c.


4c. Unless there is some other explanation of the conviction in 2c, we are entitled to conclude that there is no conceptual link between judging that ‘N’ obtains and being motivated to act accordingly.


5c. Unless there is some other explanation of the conviction described in 2c, we are entitled to conclude that the judgment that ‘N’ obtains isn’t a moral judgment (from 1c).


6c. Unless there is some other explanation of the conviction mentioned in 2c, we are entitled to conclude that the property of being morally good is not identical or reducible to the property of being N as a matter of conceptual necessity.

Just like the prior reconstruction, this one escapes the charge of question-begging (if it was legitimate to begin with) and the paradox of analysis objection.

How can an analytic reductionist reply? One way is to deny judgment internalism, as mentioned above. This is a plausible move if one finds amoralists conceptually possible. However, this move may also qualify as a concession that analytic reductionists can’t be judgment internalists, thus narrowing the conceptual space amenable to analytic reductionism.

A second move is simply to adopt synthetic reductionism, which is basically just the sense-reference objection. The issue with this is that it isn’t an objection an analytic reductionist can make, as it constitutes an admission that their thesis is false. This may be attractive, though, to people who find judgment externalism unattractive, and are more willing to give up analytic reductionism than judgment internalism.

While the open question argument is clearly not a knockdown argument against analytic reductionists, it does establish a strong presumption against analytic reductionist theories that include judgment internalism as a component. While there are ways for the analytic reductionist to respond, most moves constitute an abandonment of the very thesis being defended.



[1]. Senses and modes of presentation are distinct but related concepts in Frege’s philosophy of language. The theoretical differences between the two ideas are immaterial to my point in this section, but ought to be taken into account if one wants to understand Frege’s theory of meaning.

Works cited:

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

Miller, Alexander. Contemporary Metaethics: An Introduction. Second ed. Cambridge, UK: Polity, 2013. Print.