On the possibility of wholesale moral error

The moral error theory, it seems, could be true. The mere possibility of its truth might also seem inconsequential. But it is not. For, I argue, there is a sense in which the moral error theory is possible that generates an argument against both non- cognitivism and moral naturalism. I argue that it is an epistemic possibility that morality is subject to some form of wholesale error of the kind that would make the moral error theory true. Denying this possibility has three unwelcome consequences such that allowing for and explaining it is an adequacy condition on meta- ethical theories. Non- cognitivism and moral naturalism, I argue, cannot capture the epistemic possibility of wholesale moral error and so are false. My argument additionally provides independent reason to accept Derek Parfit's claim that if moral non- naturalism is false then nothing matters. I conclude that whether wholesale moral


| INTRODUC TI ON
The perceived threat of the moral error theory motivates much meta-ethics. Many assume that morality could be subject to some form of error of the kind that would make the moral error theory true. Moral error theorists think this error actual, others do not. Regardless, it does seem that it is in some sense possible that the moral error theory is true.
Many will think this uncontroversial; a possibility acceptable by all. But I argue that there is a sound argument from the epistemic possibility of the moral error theory to the falsity of both moral non-cognitivism and naturalism.
This argument also provides reason to believe, with Derek Parfit (2011), that if moral non-naturalism is false, then nothing matters.
I proceed as follows. In §2, I present my argument. I respond to three objections in §3. I argue, in §4 and §5 respectively, that, if true, both non-cognitivism and naturalism cannot capture the epistemic possibility of the moral error theory, which completes the defence of my argument. I conclude in §6 by drawing out its aforementioned consequence and explaining that even if non-cognitivists and naturalists remain unmoved by my argument this only helps to highlight its significance.

| THE ARG UMENT FROM THE EPIS TEMI C P OSS IB ILIT Y OF WHOLE SALE MOR AL ERROR
Call this the argument from the epistemic possibility of wholesale moral error, (P1) It is an epistemic possibility that morality is subject to some wholesale error that would make the moral error theory true.
(P2) If (P1), then it is an adequacy condition on meta-ethical theories that they allow for and explain the epistemic possibility of such error.
(P3) Non-cognitivism and moral naturalism cannot allow for and explain the epistemic possibility of such error.
(C) So, non-cognitivism and moral naturalism are false.
By wholesale error, I mean some error that permeates a discourse such that all of the claims within that discourse are false or that they are all subject to a false presupposition. This kind of error is what afflicts astrology: astrology is committed to some claims, such as the causal influence of celestial bodies on human personalities, which are false and so render the entire discourse defective. By wholesale moral error, I mean some error that runs through morality such that the moral error theory is true and either all (atomic) moral propositions are false (Mackie, 1977) or that they are all subject to a false presupposition (Kalf, 2018).
The epistemic possibility of wholesale moral error would mean that, for all we know, an error sufficient to make the moral error theory true obtains. On this view, some proposition p is epistemically possible just in case p's truth is consistent with all we know. My argument does not depend upon any account of epistemic possibility.
It can be unproblematically recast with other accounts, such as a probabilistic one on which p is epistemically possible when, given our evidence, p has a non-zero evidential probability of being true. On that account, (P1) is the plausible claim that, given our evidence, the probability of the moral error theory being true is above zero.
I will occasionally talk in terms of this account to help illustrate my claims below. Returning to astrology: it was once an epistemic possibility that astrology was a non-defective practice. For example, before coming to know its falsity-and thus the epistemic impossibility of astrology being non-defective-the causal influence of celestial objects on human personalities could have been true.
For all we know, moral practice is subject to the kind of error that would make the moral error theory true.
Denying this has three unwelcome consequences. The first: it ascribes to moral error theorists a simple epistemic failure, namely, that they fail to realise that we already know that the moral error theory is false. Not only is this uncharitable, it seems prima facie false to think either that the error of the error-theorist is epistemic in this sense, or that the error theory is immediately refuted by reflection on what we know.
Second, it suggests that morality is somehow self-protecting from sceptical worries about itself. For to suggest that wholesale moral error is epistemically impossible is to claim that morality is guaranteed to evade wholesale error. But this is, prima facie, implausible. Mere appeal to the fact that moral judgements are made, criticised, and the like-or our knowledge of this-for example, does not show that there are no plausible error-theoretic worries to be had about morality.
This self-protection is especially objectionable for a reason that is the third unwelcome consequence of denying (P1). For if moral practice is taken to be self-protecting against error-theoretic worries, then it is unclear how to distinguish between clearly confused domains of discourse (such as astrology) and morality. The clearest way to do so is to show how one (epistemically) could formulate an error-theory for either domain, before showing why one domain seems clearly subject to wholesale error whilst the other is disputable.
For at least these reasons, we should accept (P1). But they are also reasons to accept (P2). If a meta-ethical theory does not allow for and explain how it epistemically could be that morality is subject to wholesale error, then it faces the above worries. Of course, many meta-ethical theories entail that this epistemic possibility is not realised and that the error theory is false. But that is consistent with a requirement upon them to show how it epistemically could be that the error theory is true.
The moral error theory obviously allows for the epistemic possibility of wholesale moral error. And it explains how such error could obtain by maintaining that there are some alleged commitment(s) of moral discourse-such as commitment to irreducibly moral phenomena-which could be, and which they argue are, false.
Moral non-naturalism also allows for and explains the possibility of wholesale moral error. For the non-naturalist asserts that moral practice has certain commitments, such as to irreducibly moral phenomena, which epistemically could be false but, the non-naturalist argues, are not. And crucially, as we'll see, non-naturalism allows for an explanation of how the error theory epistemically could be true even if non-naturalism is true. I argue that no form of non-cognitivism or moral naturalism can do so. So, they are false. To be clear: my argument is not that these theories, if true, entail moral error theory false. Rather, it is that, if true, they cannot make sense of the epistemic possibility of wholesale moral error (whilst moral error theory and moral non-naturalism can).

| THREE OBJEC TI ON S
One might suspect that non-cognitivists and moral naturalists can easily allow for and explain the epistemic possibility of wholesale moral error. How? By being fallibilists about their own theories: suggesting that their theories might be false, and, if so, then the error theory could be true. If such an easy explanation is available, then (P3) is false.
But no such explanation is available. For a meta-ethical theory to allow for and explain the epistemic possibility of wholesale moral error, their explanation must be one they can provide by the lights of their own theory.
For example, for a naturalist qua naturalist to have an explanation of the epistemic possibility of wholesale moral error requires them to provide an explanation that is consistent with the truth of naturalism. If they must withhold assent from, or deny, their own theory to explain this possibility, then they are providing no explanation of it qua naturalist.
Whilst its truth would entail that there is no wholesale moral error, moral non-naturalism nicely illustrates how a non-error-theory, even if true, can have the resources to show how wholesale moral error is epistemically possible. As we'll see, this is by maintaining that moral practice has certain commitments the falsity of which would have entailed wholesale moral error, and that the reason why these commitments are true preserves the epistemic possibility that they are false.
A second objection: suppose fallibilist naturalists and non-cognitivists cannot provide an explanation of the epistemic possibility of wholesale moral error, qua naturalist or non-cognitivist, by pointing out that their theories might be false. Does withholding full commitment not allow such theorists some explanation of how, for all we know, it could be that moral error theory is true, namely, that their own theories might be false? If so, this might suggest it is not an adequacy condition on meta-ethical theories that they allow for and explain our target possibility and that (P2) is false.
Rather, perhaps it is meta-ethicists as epistemic agents who should do this. If there is some epistemic possibility, E, and if an agent, x, who endorses a theory, T, can make sense of E-despite T's entailing E is not realised-by withholding full confidence in T, then why think T itself must be able to account for and explain E? Why would x simply withholding full confidence in, and from claiming to know, T not suffice?
Because, as I've argued, meta-ethical theories that do not allow for and explain the epistemic possibility of wholesale moral error, if true, implausibly imply that the error theorist commits a simple epistemic mistake, suggest that morality is self-protecting from external sceptical worries, and rob us of the clearest way to distinguish morality and clearly confused domains of discourse. To avoid such consequences, it is meta-ethical theories themselves that must allow for and explain the epistemic possibility of wholesale moral error.
But, finally, perhaps any non-error theorist can explain the epistemic possibility of wholesale error without denying their theory this way. Whilst their theory entails the error theory is false, the epistemic possibility of wholesale moral error remains and remains explainable because, as fallibilists about their own theory, any given non-error-theory is consistent with the epistemic possibility of its falsity. If so, (P3) is false.
To explain: let 'T' be any non-error-theory, 'E' be wholesale moral error, and 'EP(x)' be that x is epistemically possible and that this possibility is explainable. We are asking if some non-error-theory, T, can capture EP(E), given that T and that T entails ¬E. And this could be done without denying T, so long as we assume that T is consistent with the epistemic possibility of its falsity: that is, that [T & EP(¬T)] is consistent.
This assumption may be intuitively plausible, and T and EP(¬T) are obviously logically consistent. But to assume that they are consistent beyond mere logical consistency is question-begging here. For the question now becomes whether T is consistent with EP(¬T) given T and that T entails ¬E, for some non-error theory T. And if the reason why a specific non-error-theory entails ¬E also entails that ¬EP(E), then this would entail that T and EP(¬T) are inconsistent, since the EP(¬T) requires EP(E).
To illustrate, let 'NN' be non-naturalism. NN entails ¬E. NN can explain the epistemic possibility of wholesale moral error if [NN & EP(¬NN)] is consistent. So, we must ask if NN entails ¬E in a way that also entails ¬EP(E). And it does not: NN entails ¬E because it entails that there are non-natural moral facts. But NN also allows that there are certain non-naturalist presuppositions about morality which, if false, would entail the error theory. And, crucially, why it entails ¬E is consistent with EP(E) because, for all we know, there may be no non-natural moral facts (or: given our evidence, the probability of non-natural moral facts not existing is above zero). So, NN is consistent with EP(¬NN), since NN is also consistent with EP(E). 1

Let 'N' be naturalism and 'NC' be non-cognitivism. This objection to (P3) succeeds if [N & EP(¬N)] or [NC &
EP(¬NC)] are consistent. So, we ask: assuming N or NC, and given that they both entail ¬E, do they do so in a way 1 Of course, if non-naturalism entailed that we know that non-naturalism is true, then it would entail wholesale moral error is epistemically impossible since we would know that no such error obtains. But the truth of non-naturalism does not entail that we know that non-naturalism is true. So that thought does not show that the non-naturalist cannot consistently say that even if their theory is true, for all we know it might be false (or that, given our evidence, the probability of the error theory being true is non-zero even if we accept non-naturalism). that also entails that ¬EP(E)? If so, then this provides an argument that both [N & EP(¬N)] and [NC & EP(¬NC)] are individually inconsistent. Why? Because the truth of N or NC would entail ¬EP(E), and ¬EP(E) itself entails that ¬EP(¬N) and ¬EP(¬NC) since, as non-error-theories, it is a necessary condition of the epistemic possibility of their falsity that wholesale moral error is epistemically possible.
In what follows, I argue that naturalism and non-cognitivism cannot allow for and explain the epistemic possibility of wholesale moral error on the supposition that they are true. This is because, if true, they entail wholesale moral error not just false but, I argue, epistemically impossible and/or inexplicable. This shows that they not only cannot capture the epistemic possibility of wholesale moral error if their theories are true, but they cannot do so by suggesting their theories (epistemically) could be false. For if true-in denying a necessary condition for the epistemic possibility of their own falsity-these theories actually entail that it is epistemically impossible for them to be false.

| NON -COG NITIVIS M AND THE P OSS IB ILIT Y OF WHOLE SALE MOR AL ERROR
Now, for well-known reasons, there remains no uncontroversial account of the cognitivist/non-cognitivist divide.
But on standard ways of distinguishing between them, non-cognitivists are unable to allow for and explain the epistemic possibility of wholesale moral error.
Risking nostalgia for halcyon days past, (Truth-apt): Cognitivism is the view that moral judgements are truth-apt beliefs. Non-cognitivism denies this, maintaining instead that moral judgements are non-truth-apt desire-like states.
During the time of Ayer's (1936) emotivism and Hare's (1952) prescriptivism, (Truth-apt) sufficed. And it suffices to show why these older forms of non-cognitivism rule out the epistemic possibility of wholesale moral error. For if moral claims are desire-like states such as emotional responses, and are not truth-apt, then such a view cannot explain how it could be that there was some kind of error in moral practice that would render all moral claims false.
Contemporary non-cognitivists claim they are able to consistently make sense of moral claims being true and expressing beliefs by appeal to minimalism: the sentence '"p" is true' expresses the same attitude as the sentence 'p', and a sentence expresses a belief iff it can be true. This allows them to claim that a sentence like 'abortion is wrong' expresses some non-cognitive attitude, like disapproval of abortion, whilst also expressing a true belief.
This suggests what is perhaps the most common way of casting our divide, (Representation): Cognitivism claims that moral judgements aim to represent the world, and that whether they succeed depends wholly on whether the world has the moral features such judgements ascribe to them. Non-cognitivism denies this, maintaining instead that moral judgements do not aim to represent the world. (Streumer, 2013, p. 452) Something like this seems to distinguish cognitivism from the non-cognitivism that quasi-realists like Blackburn (1984,1993,1998) and Gibbard (2003) accept. 2 On, for example, Blackburn's view, whilst moral judgements can be true and express beliefs in these minimal senses, moral judgements are expressions of non-cognitive attitudes of approval or disapproval.
These quasi-realist forms of non-cognitivism cannot allow for and explain the epistemic possibility of wholesale moral error. We might think that allowing moral judgements to be (minimalistically) true or false would suffice for this. But that is a mistake.
For what is needed to explain how a domain of discourse could be in wholesale error is more than just allowing that claims within that domain could all be false (or subject to a false presupposition). Rather, one needs to show that there are some commitments of the domain itself that allow for, and whose falsity would explain, the kind of wholesale error that would entail an error theory for that domain.
The same holds of morality. It does not suffice to allow for and explain the epistemic possibility of wholesale moral error for one to maintain that moral claims are all capable of (truth or) falsehood. Rather, we need an explanation for how and why they (epistemically) could all be false (or subject to a false presupposition).
Both the non-naturalist and the error theorist can provide this. The non-cognitivist of the quasi-realist variety cannot. For their view does not ascribe any commitments to the practice of morality the falsity of which would entail all moral claims are false or are subject to a false presupposition. Take Blackburn's view: moral judgements express attitudes of approval and disapproval. Moral judgements can be minimalistically true and involve some form of projection of our non-cognitive attitudes onto actions and other morally evaluable things. This eliminates anything that could explain how and why it could be that all of our moral judgements are false, since they are only expressing our attitudes-even if, in principle, they could all be false.
One might object that this is too quick against contemporary non-cognitivism in quasi-realist garb. In particular, perhaps it is false that a theory needs to identify some commitment(s) of a discourse whose falsity would entail wholesale error in order to explain the possibility of such error.
An alternative: one's theory may have counterfactual implications which, if true, would explain how there could be wholesale moral error. For example, Blackburn's quasi-realist could suggest that if moral agents failed to express any attitudes of approval or disapproval, then all moral claims would be false or subject to some false presupposition.
This counterfactual strategy, even if workable, is of no help to the quasi-realist. This is because if quasi-realism accepted the counterfactual above, this would turn it into a response-dependence theory. But quasi-realists emphatically deny their view entails this, fearing the relativistic implications of response-dependence theories. 3 So, this move is unavailable to them.
The quasi-realist may now complain that I am assuming they think the existence of moral attitudes suffices for moral truth. But quasi-realists allow that not all attitudes are equal. Some attitudes are correct when approving of what is good, right, and the like, and others like that of the bigot or the racist based on inconsistency, cruel mindedness, 'immaturity, lack of imagination, bias, coarseness, and so on' are incorrect (Blackburn, 1998, p. 313).
How then can quasi-realists understand mistaken moral attitudes? Unfortunately, quasi-realists refuse to provide any account of moral error. Responding to Egan's (2007)  Moral p is mistaken or erroneous if and only if ¬p-and that is all that can be said. (Blackburn, 2009, p. 207) This leaves the quasi-realist with no explanation of how and why all moral attitudes epistemically could be false, even if it explains how some are. 4 Still, Blackburn does suggest that quasi-realists can accept this consistently with their truth-minimalism: (Improvement) 'If something is entrenched in anyone's outlook, in such a way that nothing that is an improvement would undermine it, then it is true.' (Blackburn, 2009, p. 206) By contraposition, then, quasi-realists can apparently say: (C-Improvement) If something entrenched in anyone's outlook is false, then there is something that is an improvement that would undermine it.
Giving us: (All-Improvement) If all moral attitudes entrenched in all of our outlooks are mistaken, then there is an improvement or set of improvements that would undermine all of our moral attitudes.
So, if it is epistemically possible that there is an improvement or set of improvements that would undermine all of our attitudes, perhaps quasi-realists can appeal to (All-Improvement) to explain the epistemic possibility of wholesale moral error.
What might such improvements be that epistemically could undermine all of our attitudes? There are two options: (External) All moral attitudes may be undermined by some meta-ethical consideration(s).
(Internal) All moral attitudes may be undermined by some moral consideration(s).
(External) requires some feature or presupposition of morality that, for example, metaphysical inquiry could undermine, such as commitment to non-natural facts. We've seen that quasi-realists deny this. And whilst quasi-realists think some attitudes better than others, they deny that the explanation for this involves anything beyond normative theorising (Blackburn, 1998, pp. 311-313;317-318).
The quasi-realist, however, might appeal to (Internal) and say: it is epistemically possible that we are systematically wrong about the correct normative ethics. That possibility might undermine all of our moral attitudes.
Since quasi-realism is consistent with that epistemic possibility, it can allow for and explain wholesale moral error.
But suppose we all have mistaken moral attitudes because we are all deluded about what the correct normative ethical theory is. This alone would only show that our current moral attitudes are false. It would not show that all adoptable moral attitudes are false, since, assuming quasi-realism, there are still some correct moral attitudes one can adopt in line with the best normative ethics even if we are wrong about which attitudes those are. But without that quasi-realism is not explaining the possibility of wholesale moral error, of all moral attitudes being mistaken.
Quasi-realists must, then, suggest that it is epistemically possible that our moral attitudes would be mistaken even after alignment with the best normative ethical theory. This would be straightforward if, like non-naturalism, quasi-realism recognised some presuppositions that, say, moral metaphysical enquiry could show mistaken. By denying this, quasi-realists can only appeal to normative considerations, saying: for all we know, the best normative ethical theory is one on which, for example, both racists and non-racists and bigots and non-bigots alike are mistaken.
That, I submit, is an epistemic impossibility. For we know that the attitudes of the non-racist or non-bigot are not ones that can be improved upon as moral attitudes. This is not to say, inconsistently with (P1), that it is epistemically impossible that there is wholesale moral error. Rather, it is to say that we know non-bigot and non-racist attitudes are not ones that can be morally improved upon, even if morality-for reasons quasi-realists deny-is subject to wholesale moral error and, say, wholly fictional. Since this appeal to a moral improvement that would undermine all our moral attitudes relies upon the epistemic impossibility that we can improve some clearly morally unimprovable attitudes, quasi-realism cannot explain the epistemic possibility of wholesale moral error.
Blackburn's quasi-realist takes themselves to 'stand on one part of the (Neurath) boat and inspect the other parts' when doubting a moral attitude (Blackburn, 1998, p. 318). Quasi-realism does place us firmly on Neurath's boat. But, I've argued, it is mistaken to think that we need not explain how it is epistemically possible that we could be doing something wrong by being on the boat at all. That, I've argued, the quasi-realist cannot do.
This failure generalises to other non-cognitivist theories, such as sentimentalism or pragmatism. For example, take a pragmatism on which moral practice is best understood as an attempt to create and adopt norms to aid in meeting some practical goal(s), such as solving problems of interpersonal interactions. 5 Such a theory cannot appeal to anything about the practice of morality itself which, if mistaken, would entail wholesale moral error.
At best, the pragmatist could maintain that sincere engagement in moral discourse will frustrate the practical goals of moral practice. But to think this allows them to capture the epistemic possibility with which we are concerned conflates two distinct forms of wholesale error: (Veritistic): Error such that all claims within a domain are false or subject to some false presupposition.
(Non-Veritistic): Error such that sincere engagement in some domain of discourse or practice will lead to failing to achieve some practical goal(s) or aim(s) of that discourse or practice.
Our pragmatist and other non-cognitivists who insist the only goals or aims of moral practice are practical ones can explain how engaging in moral practice could meet (Non-Veritistic). Perhaps, for example, moral practice will frustrate fruitful coordination of human action, a claim cognitivists and non-cognitivists alike could accept. But I have suggested there is another sense of wholesale error epistemically possible regarding morality: (Veritistic). And if morality is claimed to be concerned only with the creation and adoption of various norms to help achieve some practical goals, then these forms of non-cognitivism cannot allow for and explain veritistic wholesale moral error. 6

| NATUR ALIS M AND THE P OSS IB ILIT Y OF WHOLE SALE MOR AL ERROR
Moral naturalism also fails to allow for and explain the epistemic possibility of wholesale moral error.
To begin to see why, take a simple form of moral naturalism. Suppose moral rightness is identical to the natural property of happiness maximisation. If so, then it is epistemically impossible for there to be wholesale moral error.
Since whilst some moral judgements about which actions maximise happiness may be false, the view lacks any explanation as to how and why all moral judgements could be subject to a kind of error that would make them all false-or subject to a false presupposition-because we know that at least some judgements concerning which acts maximise happiness are true. Now consider, for example, Railton's (1986)  Railton's theory, if true, entails that it is epistemically impossible that there is wholesale moral error. To not entail that, it would have to be epistemically possible for there to be nothing that is non-morally good for an agent (to 5 Max Khan Hayward (ms) advances a version of this view. 6 Henceforth, 'wholesale moral error' refers to (Veritistic). explain why all moral claims are false or subject to a false presupposition). But this is implausible: whilst there may be cases where nothing is non-morally good for an agent, we know that at least some things will be non-morally good for agents in some circumstances.
But does naturalism not entail that morality has certain presuppositions, which, if false, would entail wholesale moral error? Namely that if certain natural phenomena did not exist, then there would be wholesale moral error.
And would this not explain how, on naturalism, the moral error theory could be true?
Naturalist theories can allow that there are some presuppositions of moral practice which, if not met, would in principle entail that all moral claims are false (or subject to a false presupposition). But the problem for naturalists is that the way they argue these constraints are met in fact entails, if true, that it is epistemically impossible that they are not met, and so epistemically impossible that there is wholesale moral error if naturalism is true.
Identifying moral rightness with natural facts about happiness maximisation would make it epistemically impossible that there is wholesale moral error, since it is epistemically impossible that no judgements about happiness maximisation are true. And if true, Railton's theory must allow for the epistemic possibility of no choice an agent's counterpart can make being non-morally good for an agent. And we know this is false: at least some facts about one's constitution and circumstance will make at least some choices good for an agent. So, in trying to explain the epistemic possibility of wholesale moral error by appeal to what we already know to be epistemically impossible, this response fails. 7 Railton may object: why think that some facts about one's constitution and circumstances will always make at least some choices good for an agent? Is an agent with a certain constitution and circumstance such that none of their available options is either good or bad for them not epistemically possible?
Suppose that such an agent was epistemically possible. 8 This would mean that it is epistemically possible that nothing can be good or bad for that agent. For them, then, nothing is morally good or bad. But this falls short of wholesale moral error. For that, Railton's naturalist needs to show that all agents epistemically could be such that their constitution and circumstances makes nothing either good or bad for them, as that is what would be required for no moral claim to be true (and not just true or false as entertained by some agent).
But that is epistemically impossible. We know that agents in our world are such that some things are (nonmorally) good or bad for them. Railton can allow for and explain how all moral claims could be false only if he can show that it is epistemically possible that all agents are so constituted or circumstanced such that they have no needs. That latter claim, however, is false. So, Railton cannot capture the epistemic possibility of wholesale moral error on the supposition that his naturalism is true.
This problem generalises to any form of moral naturalism. 9 In principle, naturalists can suggest that the (simple or complex) natural phenomena they take the moral to be identical with, reducible to, or fully grounded in may not exist so that moral claims are false. But as realists they maintain that such phenomena do exist, and some moral claims are true. Their identification/reduction of the moral to the natural is made partly to fend off the error theorist. If true, however, their theory lacks any explanation of how moral practice could be subject to wholesale moral error, since, as we've seen, realist candidates for such identification/reduction are ones that (and we know do) exist.
And they must be. For if the moral naturalist identified or reduced the moral to some natural phenomena that do not exist, then the moral naturalist would be an error theorist in disguise. But moral naturalists are not error 7 Response-dependence theories also face this problem. For they must appeal to actual or hypothetical, ideally-informed agents failing to make any moral judgements to try to capture the epistemic possibility of veritistic wholesale moral error on the supposition their theory is true. But this is an epistemic impossibility; we already know that at least some actual agents make moral judgements and that at least some ideally-informed ones would too.
theorists, and to respond to my argument they need to accommodate the epistemic possibility of wholesale moral error without collapsing into an error theory. I have argued they cannot do so.
It is instructive to recall how non-naturalism avoids this problem. For the non-existence of the non-natural is epistemically possible since, for all we know, they may not exist (or: given our evidence the probability of nonnatural phenomena not existing is above zero). So even if true, alleged non-naturalist presuppositions of morality epistemically could be false, since the non-existence of non-natural moral phenomena is epistemically possible.
The naturalist, by contrast, appeals to natural phenomena the non-existence of which, as we've seen, is an epistemic impossibility. 10 This completes the response to the final objection to (P3) considered in §2. That objection suggested that naturalism and non-cognitivism can allow for and explain the epistemic possibility of wholesale moral error by claiming their theories (epistemically) could be false even if they are true. This assumed that both theories are consistent with the epistemic possibility of their own falsity. But since both theories, if true, entail that there is no wholesale moral error in a way that also entails that wholesale moral error is epistemically impossible or inexplicable, and since the epistemic possibility of the falsity of their theories (as non-error-theories) requires the possibility of wholesale moral error, neither theory is consistent with the epistemic possibility of their own falsity.
So, naturalists and non-cognitivists cannot explain the epistemic possibility of wholesale moral error, qua naturalists or non-cognitivists, either by saying such a possibility remains and is explicable if their theories are true, or by supposing their theories epistemically could be false even if they take them to be true. (P3), then, is true. So, naturalism and non-cognitivism are false.

| CON CLUS ION
Derek Parfit famously argued that if moral non-naturalism is false, then nothing matters (Parfit, 2011, pp. 303-304, 367). And some prominent non-naturalists, such as David Enoch, take the non-naturalist's 'most respectable opponent' to be the moral error theorist (Enoch, 2011, p. 121). Parfit's claim has been much discussed. 11 It is noteworthy that my argument provides independent reason to accept Parfit's claim and justifies Enoch's attitude.
For if neither non-cognitivism nor moral naturalism can allow for and explain the epistemic possibility of wholesale moral error, then they are false. Both families of theories, I have argued, simply fail to recognise that morality epistemically could be subject to veritistic wholesale error. Because such error is epistemically possible, and if only non-naturalism and the error theory can capture this, then if non-naturalism is false nothing matters. 12 The question of whether it is an epistemic possibility that there is wholesale moral error is, I suspect, in Richard Rorty's words: one of those issues which puts everything up for grabs at once-where there is no point in trying to find agreement about "the data" or about what would count as deciding the question. (Rorty, 1982, p. xliii) 10 And this is part of the naturalist anti-error-theory strategy: deny that moral truth depends upon the existence of irreducibly moral phenomena and offer extant natural phenomena in their place. 11 For an excellent recent challenge to it, see Hayward (2019). My argument thus provides an indirect response to Hayward, and others like Bedke (2019), concerned about claims like Parfit's. For an excellent direct response, see Enoch (2020). 12 This raises the interesting question of whether so-called 'relaxed' or 'quietist' non-naturalist realists such as Parfit (2011Parfit ( , 2017, Scanlon (2014), Dworkin (1996), and Kramer (2009 can explain the epistemic possibility of wholesale veritistic moral error. I argue in other work that they cannot. For the committed non-cognitivist or moral naturalist will likely be unmoved by my argument. They might instead bite the bullet and accept that it is a consequence of their theories that veritistic wholesale moral error is epistemically impossible. This, they may think, is all the better for them and a nice argument against the error theory. To grant (P1), they may continue, would be to already assume a way of thinking about morality they deny, perhaps motivating a charge of question-begging against me.
But my argument can be accused of question-begging only in a weak sense that it includes a premise someone already committed to the truth of non-cognitivism or moral naturalism will deny. 13 Most philosophical arguments unavoidably contain such a premise, and the important question becomes whether we find non-cognitivism or moral naturalism, or the claim that there is an epistemic possibility of wholesale moral error, more plausible. The committed non-cognitivist or moral naturalist will, unsurprisingly, think the former. We face, then, a familiar Moorean standoff.
But as Rorty continues: 'the messiness of the issue is not a reason for setting it aside.' (Rorty, 1982, p. xliii) I have provided reasons to think the denial of the epistemic possibility of wholesale veritistic moral error is implausible. But even if these will not convince the unrepentant, my argument can serve other crucial functions.
First, it may convince those as yet undecided over these theories to reject non-cognitivism and moral naturalism. Second, it reveals a consequence of said theories that may convince those previously attracted to them but unaware of this consequence to reject them. Third, arguments that lead to Moorean, intuitional standoffs reveal some of the bedrock issues that divide us within philosophy. And sometimes these bedrock issues are also ones where everything is up for grabs at once. It is an important form of philosophical progress to identify such doublytroubling issues. What to do once we do so will have to wait for another time.

ACK N OWLED G M ENTS
Some of the main ideas in this paper emerged during my thinking through written correspondence with Max Khan Hayward. I am exceptionally grateful to him for those discussions, his comments on a version of this paper, and his encouragement. I am also very grateful to Roger Crisp, David Enoch, Matthew Hewson, Geoff Keeling, Kacper Kowalczyk, Lucy Mellor, Michael Ridge, Richard Rowland, Bart Streumer, and Timothy Williamson for their very helpful written feedback on and/or conversations over versions of this paper. I gratefully acknowledge the invaluable funding I have received from the Arts and Humanities Research Council, St Anne's College, Oxford, and the Aristotelian Society that supported this work.