Consider two things. First, consider the distinction between propositional and doxastic justification for some subject, \(\mathrm{S}\)’s, belief that \(p\). On one hand, \(\mathrm{S}\) is propositionally justified in believing that \(p\) when \(\mathrm{S}\) has sufficient reasons, \(\mathrm{R}\), to believe that \(p\). \(\mathrm{S}\)’s belief that \(p\) is justifiable, in other words. On the other hand, \(\mathrm{S}\)’s belief that \(p\) is doxastically justified when \(\mathrm{S}\) believes that \(p\), given (or on the basis of) \(\mathrm{R}\) (Silva and Oliveira 2024). That is, \(\mathrm{S}\)’s belief that \(p\) is justified (Korcz 2000).¹ The crucial difference is the basing relation: there is a difference between having (available, or at hand) \(\mathrm{R}\) to believe that \(p\) and actually believing \(p\) based on \(\mathrm{R}\) (Alston 1985). Second, consider the younger cousin of the popular Frege-Geach problem for non-cognitivist, expressivist meta-ethical views: the oft-neglected wishful thinking problem (Dorr 2002). While the latter is about validity, the former is about justification. How are these two things related?

The above distinction plays an important role in both characterizing the problem and evaluating its proposed solutions. Yet few (if any) explicitly acknowledge this. I remedy that here. Doing so is instructive, for as we will see, understanding both the problem and its solutions with the above distinction in mind reveals the ways that the responses fail. This is an interesting result, for the problem is thus not as dead as it seems.

In section 1, I recast the wishful thinking problem in terms of two kinds of justification. In section 2, I do the same for several prominent responses. I also argue that they fail. In section 3, I conclude.

1 The Wishful Thinking Problem: Recast

Consider the following moral-descriptive² modus ponens (Dorr 2002). Call it the Liar Argument (Long 2016).³

As we can see, moral-descriptive modi ponentes have as their major premise a conditional claim. That conditional’s antecedent is a moral claim, whereas the conclusion is a non-moral (descriptive) claim.

Now consider Edgar. Edgar is reasoning himself through the Liar Argument. The states of affairs as he does can be represented as follows (Dorr 2002, 98):

At \(\mathrm{T}_1\), it seems irrational for Edgar to believe (C) for two reasons. First, it is incoherent to believe that (C) given that he already justifiably believes (P1)—on the basis of reliable testimony—and also believes that \(\neg\)(P2). Second, any belief that (C) at this time lacks propositional justification. Edgar, in fact, right now has good reason to be ambivalent about (C) and is ambivalent precisely because of those reasons (Dorr 2002, 98).

Now suppose that Edgar then reads some moral philosophy. As a result, he reconsiders his moral beliefs and thereby comes to immediately, justifiably believe (P2). Thus:

He now does as attitude-coherence demands: He revises. Given that Edgar has doxastic justification for (P1) and (P2), he jettisons his previous ambivalence about (C). So, he comes to believe (C) on the basis of (P1) and (P2). Hence:

According to Dorr (2002), pure non-cognitivist expressivism⁵ struggles to explain cases like this. Why is that?

At \(\mathrm{T}_1\), it is irrational for Edgar to believe (C). This seems plausible. After all, Edgar has no justification to believe that (C). Moreover, he cannot justifiably believe that (C) on the basis of what he believes at \(\mathrm{T}_1\) on pains of incoherence. So, any belief that (C) of Edgar’s is neither propositionally nor doxastically justified.

Now, from \(\mathrm{T}_1\)–\(\mathrm{T}_2\), Edgar’s mental states changed. In particular, it seems like he underwent a change in beliefs; he gained a new one. He came to believe that (P2) for the first time at \(\mathrm{T}_2\), and justifiably so. Now, a diachronic change in mental states in general is compatible with non-cognitive expressivism. The trouble is that on non-cognitive expressivism, Edgar did not gain a belief. Rather, his non-cognitive states changed. That is, Edgar only gained a new non-cognitive state. This is because (P2), remember, is a moral claim. And that the state of accepting a moral claim is a non-cognitive state is part and parcel of non-cognitivist expressivism.

This is bad for non-cognitive expressivism. It means that on that view, Edgar still cannot justifiably believe that (C) at \(\mathrm{T}_2\). That is, at \(\mathrm{T}_2\), Edgar’s belief that (C) still lacks doxastic justification for him—just as did for him at \(\mathrm{T}_1\). This is because a new non-cognitive state cannot be that on the basis of which Edgar justifiedly believes that (C). So, for the non-cognitive expressivist, nothing changed from \(\mathrm{T}_1\)–\(\mathrm{T}_2\) that explains why it seems intuitively rational for Edgar to believe (C) on the basis of (P1) and (P2), or why Edgar seems rational to justifiedly believe that (C) on the basis of only accepting the premises.

Said differently, it is intuitively plausible that the following two states can obtain over time in the Edgar case:⁶

and

Yet on non-cognitivist expressivism, both cannot obtain. On that view, (ii) is not possible. This is because Edgar’s belief that (C) at \(\mathrm{T}_3\) is neither justified nor justifiable. Remember: this was also true at \(\mathrm{T}_1\). The only diachronic change in Edgar’s mental states was his attaining a new non-cognitive state at \(\mathrm{T}_2\). But non-cognitive states cannot be justifiers, things on which it is rational to base our beliefs; they are just the wrong kinds of things.⁷

Call this the wishful thinking problem for pure, non-cognitivist expressivism.⁸ It concerns justification. More precisely, given the doxastic/propositional justification distinction, one can see that it concerns whether Edgar’s coming to believe (C) over time is rational because (and insofar as) said belief becomes doxastically justified after having been not previously even propositionally justified, i.e., justifiable.

With this in mind, let us now re-evaluate some proposed solutions. As we will see, I find them all wanting, given this understanding of the problem.

2 Re-Evaluating Some Proposed Solutions

2.1 The Decalogue Proposal

Consider the following argument (Lenman 2003). Call it the decalogue argument (Schroeder 2011). Suppose that \(\mathrm{S}\) is reasoning through it over time, and we can represent how his beliefs seemingly rationally change over time like this:

Currently, for \(\mathrm{S}\) to believe the descriptive claim that “\(\mathrm{S}\) never looks at a woman with lustful intent” would be neither justified nor justifiable. It is irrational for \(\mathrm{S}\) to believe that at \(\mathrm{T}_1\). Now, suppose that on the basis of (P3) and (P4), \(\mathrm{S}\) comes to justifiably believe the following:

Thus, at \(\mathrm{T}_2\), the belief that (P6) is doxastically justified for \(\mathrm{S}\); it is justifiedly believed on the basis of a pair of claims, (P3)\(\:+\:\)(P4). However, \(\mathrm{S}\)’s belief in the descriptive claim “\(\mathrm{S}\) never looks at a woman with lustful intent” is still neither justified nor justifiable at \(\mathrm{T}_2\) and is irrational to believe.

Moving on, \(\mathrm{S}\) subsequently comes to believe two more things:

More precisely, here \(\mathrm{S}\) gains at least one new belief: (P8). \(\mathrm{S}\) believes it, given (or on the basis of) his belief in (P4)\(\:+\:\)(P7). So, the belief that (P8) is doxastically justified for \(\mathrm{S}\). Also, at \(\mathrm{T}_3\), notice that it is justifiable for \(\mathrm{S}\) to believe that “\(\mathrm{S}\) never looks at a woman with lustful intent.” This is because (P3)\(\:+\:\)(P7)‌⁹ is sufficient reason to believe it. Finally, after all this, \(\mathrm{S}\) comes to justifiedly, rationally believe that (C) on the basis of (P6)\(\:+\:\)(P8).

What is the point of all of this? In general, Lenman’s (2003) goal seems to be to show that it can be rational to infer the conclusion of a moral-descriptive modus ponens without wishful thinking. This is achieved by showing that \(\mathrm{S}\) is guaranteed to have evidence for (C1) that can justify \(\mathrm{S}\)’s coming to believe (C1) without wishful thinking; \(\mathrm{S}\)’s justification for believing (C1) is guaranteed to be overdetermined, in other words.¹⁰ For in the very act of accepting a moral claim like (P8), one is guaranteed to accept beliefs that support it and also support (C1) independently of (P6) and (P8).

I have said nothing yet about propositional/doxastic justification. But now I ask: how can one understand the decalogue argument and this proposed solution in general, given the propositional/doxastic justification distinction? Asked differently: how can this solution be recast with the distinction—with it, how can one explain the way in which this proposal vindicates non-cognitive expressivism from the wishful thinking problem?

To see, recall (C1):

Also, recall that the solution works, in part, by guaranteeing overdetermined justification for \(\mathrm{S}\)’s belief that (C1) by the very process of coming to believe it in the first place. Further, recall: this overdetermined justification is secured by finding two pairs of claims. One is (P6)\(\:+\:\)(P8):

The other is (P3)\(\:+\:\)(P7):

Both sets justify \(\mathrm{S}\)’s coming to believe (C1). Hence, with the relevant distinction between kinds of justification in mind, the proposal works by establishing one of two things. \(\mathrm{S}\)’s belief is guaranteed to have, for \(\mathrm{S}\),

(In which case, the decalogue case shows that \(\mathrm{S}\)’s belief that (C1) is always partly based on (P3)\(\:+\:\)(P7). So, said belief is doxastically justified, i.e., based on two sets of claims, one set of which lacks a moral claim. And thus, basing the belief on (C1) is always, in part, not wishful thinking on \(\mathrm{S}\)’s part.) or,

(In which case, two sets of claims always support (C1), and thus \(\mathrm{S}\)’s coming to believe (C1) is rational. This is because (and insofar as) while \(\mathrm{S}\)’s believe that (C1) is justifiedly inferred on the basis of a non-cognitive claim—(P8)—the inference is guaranteed to be justifiable given (P3)\(\:+\:\)(P7).) This puts the non-cognitivist in a dilemma.

If the proposal establishes (iii), then it relies on the following assumption: Namely, for all of \(\mathrm{S}\)’s doxastically-justified moral beliefs, whenever a moral belief is doxastically justified for \(\mathrm{S}\) on the basis of \(\mathrm{R}\), and that moral belief entails some non-moral claim, then \(\mathrm{R}\) doxastically justifies \(\mathrm{S}\)’s belief in the non-moral claim. This is dubious, though. \(\mathrm{S}\) can be unaware of what propositionally justifies the non-moral belief, which is the fact that it is entailed by the moral belief. This means that \(\mathrm{S}\) need not necessarily form the non-moral belief on the basis of the moral belief, in which case that non-moral belief is not doxastically justified.

Moreover, if the proposal establishes (iii), then \(\mathrm{S}\) is still basing their belief that (C1) in part on the basis of (P8). There is still wishful thinking present; there is just less of it on this proposal since it is also based on (P3) and (P7)—neither of which are non-cognitive claims given non-cognitive expressivism. The proposal would be better to establish the following: in coming to believe (C1) at \(\mathrm{T}_4\), \(\mathrm{S}\)’s belief is only based on that other pair of claims—the pair that lacks a moral claim, namely, (P3)\(\:+\:\)(P7)—and thereby makes the justified belief in (C1) not a case of wishful thinking.

If the proposal establishes (iv), then it does not help with the wishful thinking problem. It needs to be shown that \(\mathrm{S}\) is rational because (and insofar as) \(\mathrm{S}\) went from, at one time, having no justification for believing (C1) to having doxastic justification for it (and crucially without believing (C1) on the basis of a non-cognitive attitude). But if the proposal establishes (iv), this still happens: \(\mathrm{S}\) still comes to accept (C1) on the basis of (P8). So, \(\mathrm{S}\)’s belief is based on, problematically, a non-cognitive attitude. That \(\mathrm{S}\)’s belief is guaranteed some bonus propositional justification may lessen the sting of a charge of irrationality. But it stings nonetheless.

So, if the proposal showed that \(\mathrm{S}\)’s belief in (C1) is rational insofar as the belief is guaranteed to be always justified or justifiable for \(\mathrm{S}\) without wishful thinking, then it either relies on a false assumption or fails to address the problem. Either way, the wishful thinking problem remains.

2.2 The Modified Proposal

Another way to deal with the wishful thinking problem is to attempt to argue that, in reasoning through the Liar Argument, \(\mathrm{S}\) is guaranteed to be propositionally justified in believing (C). So, for example, whenever \(\mathrm{S}\) argues through the Liar Argument, \(\mathrm{S}\) has available to them at any time via introspection the following modified, companion argument (Enoch 2003):

This Modified Argument is always available through introspection. The idea is thus that the Liar Argument will never lead Edgar to irrationally believe (C) in the sense that \(\mathrm{S}\) believes it without that which sufficiently justifies it.

This line is also problematic. It fails to guarantee that \(\mathrm{S}\)’s belief that (C) is justifiable for \(\mathrm{S}\). Why is that? The Modified Argument is “available” to \(\mathrm{S}\) in a weak sense. It is guaranteed to be possible that \(\mathrm{S}\) can come to possess a Modified Argument. But this does not entail that \(\mathrm{S}\) actually has available (or is in the possession of) the relevant argument. The mere presence of the Modified Argument only propositionally justifies Edgar’s acceptance of (C) only if Edgar actually has accepted (P9) and (P10).

Suppose that I am wrong. Suppose that, on this strategy, any time \(\mathrm{S}\) accepts (P1) and (P2), \(\mathrm{S}\) will always possess (in some suitably strong sense) propositionally justification for (C) since there will always be other things that propositionally justify (C).

This proposal does not address the problem, either, for familiar reasons. In particular, the problem is that when \(\mathrm{S}\) forms the belief that (C) on the basis of both (P1) and (P2),

then \(\mathrm{S}\)’s subsequent belief that (C) is doxastically justified. And that means that \(\mathrm{S}\)’s belief that (C) is doxastically justified whether \(\mathrm{S}\) is also justified in having some other, auxiliary belief(s). Said differently: what generates the problem in the original case is that it seems rational to believe that (C) because the belief is doxastically justified for \(\mathrm{S}\) regardless of whether \(\mathrm{S}\) has some other, auxiliary claims available that are themselves doxastically justified.

2.3 The Entailment Proposal

Another way to think about the wishful thinking problem is that there is a condition that needs to be met for the states (i) and (ii) (section 1) to simultaneously obtain, and the pure non-cognitivist expressivist cannot satisfy it. As we saw, the constraint seems to be something like this: \(\mathrm{S}\)’s belief that \(p\) goes from not doxastically justified to being just that only if \(\mathrm{S}\) acquires a new cognitive state. Call this Dorr’s constraint (Mabrito 2013, 1072). Given this, the decalogue problem and modified proposals can be thought of as attempts to show that the non-cognitivist expressivist can meet Dorr’s constraint.

This is not the only way to vindicate non-cognitive expressivism, though. One can also attempt to argue that \(\mathrm{S}\) can meet the constraint and still be rational (Mabrito 2013). How might one show this? One idea is to say that while the Edgar case violates Dorr’s constraint, it is compatible with another, independently motivated constraint (Mabrito 2013). The obeyance of this constraint vindicates the intuitive rationality of Edgar coming to believe (C) in the Liar Argument. Here is the argument again:

This other constraint is called the entailment constraint (EC). With respect to the wishful thinking argument:

Edgar’s case seems compatible with this because the wishful thinking problem assumes that the Frege-Geach problem is solved: it concedes that moral-descriptive modi ponentes are valid. So, while Edgar initially lacks justification for believing (C), he later comes to accept two claims that entail (C)—securing the obeyance of the Entailment Constraint—and is thereby rational for believing (C) on the basis of them.

Again, though, we should think of how this solution goes in terms of the doxastic/propositional justification distinction. To illustrate, consider the Entailment Constraint itself. As a necessary condition, with respect to changes in what kinds of justification is the condition plausible? And obeying which of these various formulations also helps with the wishful thinking problem?

Suppose that the relevant change concerns a belief’s status with respect to being propositionally justified:

(EC2) is plausible but irrelevant. No one denies that Edgar’s belief that (C) of the Liar Argument is justifiable—i.e., denies that there are reasons to accept it. The worry is that Edgar’s belief seems justifiedly inferred from his belief in the premises. So, a case that obeys this constraint is explanatorily moot with respect to the relevant intuitive rationality that needs preservation on the wishful thinking problem.

Suppose that the salient change is from a belief being not justifiable to actually being justified. Thus:

(EC3) is false. One can both believe that \(p\) for no reason and then only later base one’s belief that \(p\) on very good reasons, and yet neither come to believe that which entails \(p\) nor acquire evidence for \(p\). This happens in cases where one dogmatically believes that \(p\) but only later comes to accept \(p\) on the basis of good reasons that one already had. This is because one can fail to recognize good reasons for beliefs when they have them. Hence, one can fail to base that which they already believe on the basis of those good reasons.

2.4 The Hopeful Proposal Proposal

Another proposed solution to show that wishful thinking is sometimes rational is accepting that the premises of arguments akin to moral-descriptive modi ponentes do make the belief in the conclusion justifiable (Long 2016). As an example, consider the following argument that \(\mathrm{S}\) is reasoning through. Call it the hopeful proposal argument.

This argument is wishful thinking as it is normally understood outside the seminar room: forming beliefs about how the world is, given one’s wants, desires, hopes, dreams, etc. (which tell us how the world is not). Now, how does one argue that it is rational to infer (C2)? The idea is that “accepting of the […] premises is often a reason to accept its conclusion, since paradigm cases of wishful thinking are often valid” (Long 2016, 3).

Once again, the propositional/doxastic justification distinction renders this proposal ambiguous. So, for example, either \(\mathrm{S}\)’s belief that (C2) is rational because (and to the extent) that it is doxastically or propositionally justified given the fact that modus ponens is valid.¹²

No one would deny that \(\mathrm{S}\) coming to believe (C2) is justifiable, given the validity of modus ponens. The issue, recall, is whether \(\mathrm{S}\) is justifiably rational in going on to believe (C2) on the basis of the premises—as it intuitively seems. So, if this proposal is to work, the very fact of the argument’s validity must be a part of that set of things on the basis of which \(\mathrm{S}\) believes that (C2).

Here is how this works with respect to the Liar Argument. \(\mathrm{S}\) rationally, justifiedly infers (C),

when the belief that (C) is based on the following:

The trouble is that this proposal misses the mark. The mere availability of (P13) does not mean that \(\mathrm{S}\) is doxastically justified in believing (C). \(\mathrm{S}\) can fail to base their belief in (C) on (P13). \(\mathrm{S}\) would need to be shown to, in every case, in fact base their belief in (C) on the basis of (P1)\(\:+\:\)(P2)\(\:+\:\)(P13).

Moreover, the wishful thinking problem remains. The non-cognitivist expressivist still cannot make sense of how \(\mathrm{S}\)’s belief that (C)—even in that case—is doxastically justified. For (P2), on their view, is still a non-cognitive state. And \(\mathrm{S}\) cannot, it seems, justifiedly believe (C) on the basis of (P1)\(\:+\:\)(P2)\(\:+\:\)(P13) for that very reason: (P2) is the wrong kind of thing for one to justifiedly believe something else on the basis of. It would help if (P3) was itself sufficient for \(\mathrm{S}\) to justifiedly believe (C) on the basis of. We saw attempts at this. But the problem with those attempts remains. It would have to be shown that in every case, \(\mathrm{S}\) in fact believes the conclusion on the basis of the stuff that is both guaranteed to be present; would, in fact, make the belief doxastically justified if the belief was based on them; and are the right kinds of things to base beliefs on.

3 Conclusion

The wishful thinking problem seemed dead, the recipient of several fatal blows. I hope to have shown that the distinction between propositional/doxastic justification helps clarify the nature of the problem, the nature of proposed solutions, and why those solutions are dubious. Perhaps, then, it is premature to ignore the wishful thinking problem.