Why Self-Defeat Is Not a Problem for Normative Error Theory

A reply to Spencer Case (and Bart Streumer)

The Self Defeat Argument

In 2018, the professional philosopher Spencer Case published a paper defending the self defeat argument against normative error theory. The argument goes as follows:

P1. The error theorist is committed to the self-defeating proposition, ‘Error theory is true, but I have no reason to believe that.’

P2. If adopting any philosophical position commits us to a self-defeating proposition, then we should reject that position.

C1. We should reject error theory.

P3. If we should reject error theory, then error theory is false.

C2. Therefore, error theory is false.

Why the Self Defeat Argument Fails

My response to this argument takes the form of dilemma. When Spencer Case talks about normative error theory being “self defeating” in that it entails that there are “no reasons to believe it”, he could mean one of two things. On the one hand, he might simply mean that normative error theory entails that there is no normative reason to believe it. On the other hand, he might mean something more robust, like that normative error theory entails that there is no reason to believe it in any sense, including purely descriptive epistemic senses where all we mean by “reasons for belief” is something like “evidential support relations”.

The dilemma for Spencer’s argument, then, is this: if, when Spencer talks about “self defeat”, he means in the first sense, then premise 1 of his argument is true, but premise 2 is more or less question begging. If, however, Spencer means to talk about “self defeat” in the second sense, then premise 2 is not question begging but premise 1 is false.

In other words, while a normative error theorist is committed to there being no normative reasons to accept their view, it is unclear why this would be concerning to the normative error theorist. A normative error theorist can, I hasten to point out, acknowledge the aforementioned commitment without denying that their view is likely to be true, and that this fact can be persuasively established via philosophical argumentation.

By contrast, the normative error theorist simply is not committed to there being no reasons to accept their view in a wider sense which would force them to deny, say, the probable truth of their view and the existence of good arguments for it. At any rate, if the normative error theorist is committed to this, then I have yet to see it established why, including by Spencer Case, whose entire paper I have now read in hopes of receiving such an explanation.

Perhaps one would argue that even if an error theorist can consistently say that there are identifiable truth conducive methods of inquiry and that normative error theory is probably true and well supported, the fact that they’re not able to say that we ought to believe what’s probably true and well supported, or that we ought to form beliefs in accordance with the truth conducive methods of inquiry, still constitutes a reductio to their view.

But if the self defeat argument just amounts to yet another appeal to “isn’t it intuitive that we have normative reason to do X?” Where X is just some supposedly common sensical normative obligation, then it’s likely to be dialectically inert against the error theorist. As Christopher Cowie points out, the error theorist is knowingly committed to it being the case that you have no normative reason not to torture babies for fun. Do you expect them to recoil at saying that we don’t have normative reason to form beliefs according to what’s true?

I have long thought that the dilemma which I have just posed represents a decisive rebuttal to the self defeat argument. I was therefore intrigued when I read the title and abstract of Spencer Case’s paper, which gave me the impression that Spencer would be offering a way for the self defeat argument to escape this objection. The abstract reads:

Many philosophers have noted that normative error theorists appear to be committed to saying ‘Error theory is true, but I have no reason to believe it’, which seems paradoxical. In defence of error theory, some have claimed that the word ‘reason’ in that statement is ambiguous between ‘normative reason’ and a purely descriptive sense of ‘reason’ that the error theorist can accept. I argue, however, that there is no descriptive sense of ‘reason’ that can prevent the paradox from re-emerging. Moreover, these implications of error theory probably provide good grounds for rejecting the view.

In what follows, I will assess the arguments which Spencer Case makes in his paper in an attempt to figure out whether any of them succeed at saving the self defeat argument from the dilemma which I’ve just laid out. My claim, as it should come as no surprise, is that none of them do.

Spencer Case on Why Error Theorists Cannot Avoid Self Defeat

In the section of his paper titled “Why the Error Theorist Cannot Avoid Self-Defeat”, Spencer Case mentions something like the aforementioned reply to the self defeat argument. He writes:

In order to reject (1), the error theorist must be able to avoid saying that he has no reason to accept error theory. The only way to do this is to distinguish between normative reasons, which he must reject, and some other form of reasons— presumably, epistemic reasons—that he can accept.

Spencer rejects this reply on the following grounds:

The problem is that reduction is not elimination. Indeed, reductionistic views preserve entities instead of eliminating them. To say that As are reducible to Bs is to affirm the existence of As as Bs. An eliminative materialist about minds, in the mould of Paul S. and Patricia M. Churchland [1998] who maintain that there are no minds or folk psychological states, must reject the view that the mind is reducible to the brain and that folk psychological states are reducible to brain states. That is because reductionism about the mind and about folk psychological states leaves these things in the picture instead of eliminating them. The error theorist must reject the view that normativity is reducible to convention-based norms for the same reason

The obvious problem with this, however, is that a normative error theorist can accept the existence of certain descriptive epistemic notions (such, as, for example, evidential support relations), and even identify such notions as “epistemic reasons”, all without identifying “epistemic reasons” as a normative notion, and thus reducing epistemic normativity to the aforementioned descriptive epistemic notions. In other words, a normative error theorist need not say “normative epistemic reasons just are evidential support relations”- instead, they could say “normative epistemic reasons don’t exist, but there are epistemic reasons, which just are evidential support relations, and thus my view isn’t self defeating in any meaningful sense”.

Of course, the normative error theorist could also drop talk of “epistemic reasons” altogether and simply believe in certain descriptive epistemic notions such as evidential support relations. This would be sufficient to ward off the charge of self defeat, at least insofar as the charge of self defeat is being made in a way which is supposed to be concerning for the normative error theorist.

The problem with this reply, so says Spencer, is that it does not actually avoid self defeat. As Spencer explains:

The error theorist could say, ‘there is no normativity, there are only conventions.’ That is consistent with error theory, but it does nothing to defuse the worry about self-defeat. So, it seems that error theory is either self-defeating or contradictory. If a reductive view of reasons is assumed, then the error theorist contradicts himself. If normative reasons are eliminated, then the view is self-defeating.

Again, this response only works if by “self defeating” Spencer just means that the view implies that there are not normative reasons to accept it. But as I have pointed out, this will be of little concern to the error theorist. This response does nothing to show that error theory would be self defeating in the stronger sense which would worry the normative error theorist.

Spencer Case then goes on to say a few other things in this section of his paper, none of which grapple with the issue for his argument that I have raised in the foregoing discussion. For instance, he challenges the idea that one can give a non normative construal of “epistemic reasons”, as the notion of an epistemic reason necessarily involves a kind of normative authority. He writes:

The essence of epistemic norms is that they are genuinely authoritative over how we should think. They are not merely epistemically authoritative over what we ought to think, which would be trivial. Rather, they tell us how we Just Plain Ought to think—at least in ordinary circumstances.

The problem with this response is that if Spencer is right that the notion of an “epistemic reason” is necessarily normative, then all this shows is that “epistemic reasons” are not necessary to avoid self defeat in the meaningful sense, and that the normative error theorist can happily abandon them in favor of some other purely descriptive epistemic notions, such as “evidential support relations”.

Again, if the category of “epistemic reasons” is not necessarily normative but includes the aforementioned descriptive epistemic properties as well, then it’s true that the normative error theorist would be troubled by the loss of “epistemic reasons”, but it is false that the normative error theorist must incur that loss. If, on the other hand, the category of “epistemic reasons” is necessarily normative, and thus the loss of “epistemic reasons” does not entail the loss of, for instance, evidential support relations, then the normative error theorist simply will not be troubled by this loss. Spencer has again seemingly opted for the latter horn of this dilemma, but without realizing that doing so renders his argument toothless.

Spencer then argues that even if we grant that a normative anti realist can offer a non normative construal of “epistemic reasons”, further problems present themselves. Let’s suppose, then, that there are norms relative to the epistemic domain (ie norms which specify proper epistemic conduct), but that the epistemic domain possesses no normative authority. The normative error theorist must, then, in order to say that they have reasons to believe normative error theory, be “partisan” towards the epistemic domain. If they were not partisans of that domain, they couldn’t have epistemic reasons, just as one who isn’t a baseball player doesn’t have reasons to steal bases and hit home runs.

But, Spencer asks the error theorist: does your partisanship towards the epistemic domain make a normative difference, or doesn’t it? If it does make a normative difference, then that contradicts error theory. If it doesn’t make a normative difference, then your view is self defeating. The closest thing I could find to an explanation of what Spencer means when he talks about something “making a normative difference” is when he says:

If partisanship toward a domain makes a normative difference for an agent, then it makes the agent have normative reasons that she did not previously have, which is inconsistent with error theory.

At risk of repeating myself: given that all Spencer means in asserting that normative error theory is “self defeating” is that the normative error theorist is committed to there being no normative reason to accept their view, then I agree that the normative error theorist denying that their partisanship to the epistemic domain “makes them have normative reasons that they did not previously have” entails that normative error theory is self defeating. But again, this is a sense of the term “self defeat” which will not be concerning to the normative error theorist.

Of course, on the flip side, insofar as we mean something more robust by “self defeat”, then it is just not true that the normative error theorist’s denial that their partisanship towards the epistemic domain makes a normative difference entails that normative error theory is self defeating, and certainly Spencer provides no explanation for why it would.

Spencer concludes this section of his paper by once again attempting to show that even if “epistemic reasons” can be construed non normatively, this does not allow the normative error theorist to avoid self defeat. He asks us to distinguish between two senses of self defeat (finally). The two contrasting senses that he has in mind are as follows:

Self-Defeat 1. Error theory is true, but there is no reason for anyone to believe that.

Self-Defeat 2. Error theory is true, but there is no reason—of a kind that anyone need take the least bit seriously, all things considered—for anyone to believe that.

Case points out that an error theorist who runs with the reply that he was just considering (wherein they offer a non normative construal of “epistemic reason”) would reject self defeat 1, but remain committed to self defeat 2. The reason the normative error theorist is committed to self defeat 2, according to Spencer, is that the normative error theorist, while acknowledging that there are standards of reasoning within the epistemic domain, must reject the normative authority of the epistemic domain.

According to Spencer Case, this is a big problem for normative error theory. He writes:

I think that if Self-Defeat 1 is problematic or paradoxical, then Self Defeat 2 is, as well. The error theorist’s having a reason to accept error theory only resolves the paradox if that reason is in some way binding on what he ought to think. The error theorist cannot avoid the self-defeat objection simply by appropriating the word ‘reason’ while stipulating that the reasons are normatively vacuous.

Now, when Spencer writes, “I think that if Self-Defeat 1 is problematic or paradoxical, then Self Defeat 2 is, as well”, does he mean that there is some sort of logical entailment? If so, he does not show how. I suspect, rather, that what he means to say is “self defeat 1 and self defeat 2 both strike me as problematic.” That might well be true, but what Spencer needs to establish, in order to be able to claim to have presented a dialectically useful argument, is that the normative error theorist would find self defeat 2 problematic.

Now, if, when Spencer Case says “there is no reason—of a kind that anyone need take the least bit seriously, all things considered—for anyone to believe that”, he just means something like “there is no reason, of a kind that anyone has a normative demand to be responsive to, all things considered, to believe that”, then in my estimation the normative error theorist will be perfectly happy to accept self defeat 2.

Again, keep in mind that for the normative error theorist to accept self defeat 2 essentially amounts to them accepting that, while normative error theory is probably true and this can be established via philosophical argumentation, we are under no normative demand to form beliefs according to what is probably true, or to believe a particular proposition (“normative error theory is true”) in virtue of its truth. Maybe Spencer recoils from this conclusion, but the normative error theorist likely will not, as I earlier explained.

Again, if all Spencer is saying is that the normative error theorist cannot account for the intuition that we are under X common sense normative obligation (in this case to form beliefs according to what’s true, or to believe a particular proposition because of its truth), then it seems to me that all Spencer has done in showing that the normative error theorist is committed to self defeat 2 is provide another version of the generic argument against error theory, but with an even less compelling reductio at the helm than is typically used.

Spencer Case on Why Error Theorists Cannot Embrace Self Defeat

To recapitulate, given that Spencer has made it relatively clear that when he talks about “self defeat” he means to invoke the narrower sense of the term, my response to him is to agree that normative error theory is self defeating, but to deny that this is problematic. Establishing that normative error theory is self defeating in the intended sense simply does not do any dialectically useful work for the normative realist.

In the last section of his paper, titled “Embrace Self Defeat?”, Spencer Case supposedly attempts to address precisely this response. I will now reproduce and consider what Spencer has to say, and determine whether any of his remarks successfully assuage the concerns that I have raised in the foregoing discussion.

Spencer proceeds by considering a defense of error theory from the self defeat objection by the philosopher Bart Streumer. Spencer writes:

According to him, error theory entails that there is no normative reason to believe error theory, but that this is no reason for thinking that error theory is false. Streumer further thinks that we must take ourselves to have reasons for our beliefs, and that in order to fully believe anything—meaning, roughly, that no one can believe it with complete conviction—we must understand the obvious implications of that belief. Since it is an obvious implication of error theory that there are no reasons to believe error theory, it follows that no one, including Streumer, can fully believe it. Streumer insists that this puzzling feature of error theory gives us no reason to think that it is false.

Spencer seemingly endorses part of Streumer’s reasoning here, accepting the premise that error theory cannot be “sincerely endorsed” or “fully believed”, but arguing, contra Struemer, that this is a problematic feature for error theory to possess.

Given this implication of normative error theory’s self defeat, Spencer rejects my earlier charge that the self defeat argument is just a less compelling repackaging of the generic argument against error theory, and is thus dialectically superfluous.

Spencer acknowledges that the normative error theorist is already committed to more unintuitive claims than that their own view cannot be fully and sincerely endorsed (such as the claim that one has no normative reason not to kill and eat their children for fun), but argues that this doesn’t show that the self defeat argument is irrelevant. As Spencer points out:

Even if the intuitive costs of error theory are already high, an argument that raises those costs further may nonetheless be worth making especially if it establishes a different kind of intuitive cost.

Spencer further notes that:

Biting the self-defeat bullet weakens the error theorist’s dialectical position in a way that other appeals to moral intuition do not. If the error theorist acknowledges that such things as sincerity or ‘faithful participation in the argumentative enterprise’ is irrelevant, then he has no grounds on which to object if his interlocutor argues in bad faith (by, for example, simply refusing to consider objections to his arguments).

Now, I will say first of all that I find the claim that the normative error theorist “has no grounds on which to object if his interlocutor argues in bad faith” bizarre. A normative error theorist could, upon being met with bad faith argumentation, point out that their interlocutor’s conduct precludes truth conducive dialogue and inquiry, and condemn them as pernicious for engaging in such conduct. It’s not at all clear to me why one would think that the normative error theorist couldn’t have “grounds” for reacting in such a way.

But perhaps this doesn’t count for Spencer as an “objection”. Perhaps, for Spencer, to “object” to your interlocutor behaving in bad faith just means to claim that they have normative reason to not act in this way. In that case, I just take it that Spencer is using language in a proprietary and misleading way which, once disambiguated, reveals his statement to be dialectically empty (the astute reader may be starting to notice a pattern here).

It’s also not clear to me how this point relates to the observation that normative error theory cannot be “fully believed”. That fact seems wholly irrelevant to whether the normative error theorist has “grounds” for complaining when their interlocutor argues in bad faith. But in the spirit of not drawing out this article too much, I will not dwell on this point any further.

In any case, I actually agree with Spencer that normative error theory entailing that it cannot be fully believed or sincerely endorsed would constitute a new kind of intuitive cost to the view. Therefore, if the self defeat argument successfully establishes this fact, then I would have to concede that it is a forceful argument. However, I emphatically reject Streumer and Spencer’s conclusion that the relevant entailment holds.

Remember, Streumer’s argument for this claim is that, in order to fully believe something, we must take ourselves to have reasons for believing it. Since normative error theory implies that we do not have reasons for believing normative error theory, then it follows that we cannot fully believe normative error theory.

Again, though, what we mean by “reasons for belief” must be disambiguated. If by “reasons for belief” we mean in a strictly normative sense, then I agree that normative error theory entails that there are no reasons to believe it, but I emphatically reject that this fact entails that one cannot “fully believe” normative error theory. Why would it be impossible to “fully believe” a view while simultaneously taking it that I’m under no normative demand to do so?

If, however, Streumer means “reasons for belief” in a broader sense, then I am inclined to agree that one cannot fully believe something while holding that they have no reason to believe it. But in that case, I do not accept that normative error theory entails that there is no reason to believe it. Once again, we are faced with the original dilemma for self defeat arguments against error theory.

Spencer does nothing to address this concern in his paper. In hopes of some sort of justification for Streumer’s reasoning, then, I decided to read his own paper, titled “Can We Believe The Error Theory?”

Thankfully, Streumer seemingly does attempt to address my concern in his article, albeit briefly and unpersuasively. He writes:

It may be objected that the belief that there is a reason for a belief is not a normative judgement, but is instead the belief that there is evidence for this belief. But reasons for belief are considerations that we base our beliefs on, and we cannot base a belief on a consideration without making at least an implicit normative judgement. Suppose that I base my belief that Socrates was mortal on evidence about human beings’ mortality. In that case, I cannot see this evidence as merely causing me to have this belief, or as merely explaining why I have this belief. I must also make at least an implicit normative judgement about the relation between this evidence and this belief: I must take this evidence to support, or to justify, or to count in favor of this belief.

As far as I understand Streumer’s argument here, he is saying that the kind of “reason for belief” which I must take myself to have if I am to fully believe something is a normative kind. Any other construal of “reason for belief”, such as the purely descriptive epistemic type of construal which I have offered, will not be able to do the job. The reasons that Streumer provides for this supposition, however, are less than impressive.

Streumer notes that if I am to “fully believe” some proposition, then I must take it that my reasons for believing the proposition in question serve to support, or justify, or count in favor of the belief- I cannot merely take it that the considerations on which my belief is based caused my belief, or serve to explain why I have the belief.

The perceptive reader should already be see the problem with this. When Streumer says that I must take the considerations on which my beliefs are based to justify those beliefs if I am to “fully believe” anything at all, does he mean to invoke a necessarily normative construal of the term “justification”? If so, then I agree that the normative error theorist cannot take their beliefs to be “justified”, but I deny the claim that taking oneself to be “justified” is a necessary condition for fully believing anything. Certainly, Streumer has not provided any argument for thinking that it would be.

On the other hand, if Streumer means to invoke the notion of “justification” in a broader sense which is not necessarily normative, then I am happy to agree that taking oneself to be justified is a necessary condition for fully believing anything, but it is then patently false that the normative error theorist cannot take their beliefs to be “justified”.

Moreover, Streumer argues that if one is to fully believe something, then one cannot take the relation between ones belief and the considerations on which their belief is “based” to be “merely causal”. But it does not follow that the relation must therefore be normative, and so the normative error theorist will have no problem accepting this claim. One could, for instance, as I have repeated ad nauseum throughout this article, take the relevant relation to be one of evidential support or entailment.

If, however, we interpret Streumer as arguing that no purely descriptive relation (or set of descriptive relations) is sufficient to give one grounds for “fully believing” things, then I agree that the truth of this claim would be problematic for the normative error theorist. The problem, of course, is that this claim is clearly false, and Streumer has yet to provide an argument for it that doesn’t blatantly beg the question.

To close out his paper, Spencer Case seals the deal by stipulating an imaginary neutral arbiter, and subsequently stipulating that they would agree with him over the normative error theorist regarding whether or not his argument is forceful. He writes:

Streumer, or another hard-bitten error theorist who is willing to embrace self-defeat, might be happy to accept this. How should we adjudicate the dispute with him, then? Whenever we find an intuitive stand-off like we can find in philosophy, it is worth asking how things are likely to seem to a neutral arbiter—in this case, someone who is neither committed to accepting error theory, nor committed to rejecting it. A neutral arbiter would presumably be a reasonable arbiter, someone who is committed to upholding basic norms of discourse, such as sincerity, that constitute philosophical best practice. Such a person could be expected to disfavour the error theorist in this exchange.

To be honest, I find myself somewhat at a loss regarding how to respond to this. With such robust epistemic backing as the figments of his own imagination, who am I to disagree with Spencer’s conclusion?

Conclusion

To be honest, when I began writing this article I was pretty sure that self-defeat style arguments against normative error theory rest on what amounts to more or less a straightforward conceptual confusion. Proponents of the argument claim that normative error theory entails that there are “no reasons to believe normative error theory”, which is only true if “reasons for belief” is meant in a strictly normative sense, but then they go on to pretend that affirming this is something akin to affirming global skepticism or the denial of any epistemic principles or truths whatsoever, which clearly doesn’t follow from normative error theory. Having now read and written about Case and Streumer’s papers on the matter extensively, I am soundly convinced that this assessment was accurate.

I did not particularly enjoy reading or responding to these articles. The experience of doing so, for the most part, went as follows. Firstly, they will present an argument which rests on the aforementioned conceptual confusion. Then, they will say “if you think this argument is just a conceptual confusion, then not to worry, because I am about to provide an argument proving that it is not.” They then proceed to provide an argument which exemplifies literally the exact same conceptual confusion as before, but put in a slightly different and perhaps more elaborate way. Consequently, their arguments did not require any creative or novel reasoning in order to surmount, and I am left feeling mostly frustrated, as though I put in hours of time to ultimately wind up exactly where I was when I began the endeavor.

I want to of course leave open, however, the possibility that the pervasive error is on my end. I have been so unimpressed with Case and Streumer’s reasoning that I have begun to feel like perhaps I am the one missing something basic. I lean against this possibility, but I inquire anybody who’s read this far into the article to point it out if so.