Sidestepping the Frege-Geach Problem

: Hybrid expressivists claim to solve the Frege-Geach problem by offloading the explanation of the logico-semantic properties of moral sentences onto the belief-components of hybrid states they express. We argue that this strategy is undermined by one of hybrid expressivism’s own commitments: that the truth of the belief-component is neither necessary nor sufficient for the truth of the hybrid state it composes. We articulate a new approach. Instead of explaining head-on what it is for, say, a pair of moral sentences to be inconsistent, expressivists should “sidestep” and explain what it is to think that a pair of moral sentences is inconsistent. To think so is to think they cannot both be true – a modal notion. Since expressivists have given accounts of such modals, we illustrate how sentences like ‘“lying is wrong” and “ lying is not wrong ” are inconsistent ’ express sensible – and rationally compelling – states of mind.


INTRODUCTION
Moral expressivists have been wrestling with the Frege-Geach problem for the best part of a century, with mixed success. 1 In recent years, however, so-called "hybrid" expressivists have claimed to solve the problem 'on the cheap' (Ridge 2006: 309).Their strategy is to 'offload' (Ridge 2014: 144) the explanation of the logico-semantic properties of moral sentences onto beliefs that are components of hybrid states they express (see §1).This paper has a critical component and a constructive component.In §1, we show that the simple "offloading" strategy does not work, and by the hybrid expressivist's own lights.However, in §2, we articulate a new strategy, which employs the 'expressivist sidestep' (Dreier 2015: 273).Rather than explaining what it is for a set of moral sentences to be, say, inconsistent, we suggest that the expressivist should aim in the first instance to explain what it is to think that a set of moral sentences is inconsistent.We argue that, by the hybrid expressivist's lights, sentences like 'the sentences "p" and "¬p" are inconsistent' express states of mind that are not only perfectly coherent, but in fact cannot be rationally rejected.We thus argue that, by approaching the Frege-Geach problem side-on, the hybrid expressivist stands to gain a transcendental solution to it.
We proceed as follows.§1.1 introduces hybrid expressivism.§1.2 introduces the Frege-Geach problem.In § §1.3-1.4 we explain the simple "offloading" strategy and argue that it does not work.§1.5 concludes our critical discussion.The scene is thus set for our positive proposal.§2.1 introduces the expressivist sidestep; § §2.2-2.4 applies it to talk of inconsistency.§2.2 observes that inconsistency is a modal notion.§ §2.3-2.4 introduce two expressivist approaches to modals.We argue that, on either approach, the expressivist is entitled to talk of moral inconsistency, and further that sentences like '"p" and "¬p" are inconsistent' cannot be rationally rejected.§2.5 generalises the strategy to logico-semantic properties beyond inconsistency and responds to objections.We conclude the paper in §3.

Hybrid expressivism
Moral judgement notoriously displays both belief-like features and desire-like features (Smith 1994).Pure cognitivists, who maintain that moral judgements are simply beliefs (that are not composed, even in part, by desire-like attitudes) can explain the belief-like features, but struggle to explain the desire-like features.Pure expressivists, who maintain that moral judgements are desire-like attitudes (that are not composed, even in part, by beliefs) can explain the desire-like features, but struggle to explain the belief-like features.
Hybridists aim to have the best of both worlds by maintaining that a moral judgement is a hybrid state, composed in part by a belief (the belief-component), and in part by a desire-like attitude (the desire-component).For instance, on the toy version of the view we'll use for illustrative purposes, the judgement that stealing is morally wrong is composed of desire-like attitude (a1) and belief (b1): (a1) Disapproval of actions insofar as they have descriptive property F.
(b1) The belief that stealing is F.
The broad attraction of hybridism lies in its potential to explain the belief-like features of a moral judgement by pointing to its belief-component, and the desire-like features by pointing to its desire-component.
Hybridism itself comes in two importantly different varieties (cf.Ridge 2007: 54), depending on whether the hybridist accepts or denies Inheritance:

Inheritance
It is necessarily the case that: a moral judgement is true iff its belief-component is true.
Hybrid cognitivism accepts Inheritance: the truth-conditions of a moral judgement are identical to the truth-conditions of its belief-component (Copp 2001;Boisvert 2008).
Since this point will be important later, it's worth working through an example.
Suppose we have an agent, Connie the consequentialist, who thinks that stealing is wrong and thinks that lying is wrong.For Connie, 2 the former judgement is realised by (a2) and (b2), while the latter is realised by (a2) and (b3): 2 Hybrid expressivists usually hold that moral judgements are massively multiply realisable, e.g. on the toy version of the view, that the value of 'F' can vary interpersonally (or intrapersonally over time).What matters is that one has a desire-like attitude (of the right kind) and a belief that are related in the right kind of way, e.g. on the toy version, as (a1) is related to (b1).Hence why hybrid expressivism as we understand it is sometimes known as relational expressivism (Toppinen 2013;Schroeder 2013;Ridge 2014).We follow orthodoxy in making this assumption here, but it is strictly dispensable for our argumentative purposes.The same is true of the assumption that a moral sentence expresses a hybrid state composed of a desire-like attitude and a belief, rather than (a2) Disapproval of actions insofar as they fail to maximise utility.
(b2) The belief that stealing fails to maximise utility.
(a2) Disapproval of actions insofar as they fail to maximise utility.
(b3) The belief that lying fails to maximise utility.Now suppose that (i) stealing fails to maximise utility, but (ii) lying does not fail to maximise utility.But suppose that the correct first-order moral theory is deontological, e.g.(iii) an action is wrong iff it violates the categorical imperative.And suppose further that (iv) stealing does not violate the categorical imperative, but (v) lying does.Given (i), the belief-component of Connie's judgement that stealing is wrong (b2) is true; but from (iii) and (iv) it follows that stealing is not wrong, and thus that the moral judgement itself is false.Thus the truth of the belief-component of a moral judgement is not sufficient for the truth of the hybrid state as a whole.
Moreover, given (ii), the belief-component of Connie's judgement that lying is wrong (b3) is false.And yet it follows from (iii) and (v) that lying is wrong, and so the moral judgement itself is true.Thus the truth of the belief-component of a moral judgement is not necessary for the truth of the hybrid state as a whole either. 3xpressing both the desire-like attitude and belief (as in Ridge's (2006) earlier formulation of the view).

The Frege-Geach problem
The Frege-Geach problem, as we'll understand it here, is the difficulty of making sense of the logico-semantic properties of moral sentences, given the nature of the mental states the expressivist says these sentences express.
The expressivist is an "expressivist" because she aims to explain the meaning of a moral sentence in terms of the mental state that it is used to express. 4 The expressivist is therefore committed to making sense of the logico-semantic properties of moral sentences -such as the consistency, inconsistency, and entailment relations they stand in -by appeal to these mental states.These properties include, for example:

Sentential entailment for modus ponens (SEMP)
'p ' and 'p→q' together entail 'q'.   4 This claim is traditionally understood as a semantic thesis: the semantic expressivist identifies or reduces the meaning of a moral sentence to the mental state it expresses (Rosen 1998;Schroeder 2008Schroeder , 2010)).More recently, it has been understood as a metasemantic thesis: the metasemantic expressivist aims to explain why a moral sentence has the meaning it has in terms of the mental state it expresses, without necessarily identifying the meaning with the mental state (Pérez Carballo 2014;Ridge 2014: 8-9;Chrisman 2016).This distinction does not matter for our purposes: either way, the expressivist needs to make sense of the logico-semantic properties of moral sentences by appealing to the mental states they express (Schroeter & Schroeter 2017;Woods 2017).
For ease of presentation, in this paper we'll primarily focus on SIN.However, as we'll occasionally note, our discussion straightforwardly generalises to other logicosemantic properties like SEMP. 5 The reason that there is supposed to be a problem for the expressivist in particular here is because the cognitivist is entitled to resources that the expressivist is not automatically entitled to.Take SIN.The cognitivist can maintain that 'p' expresses the (ordinary descriptive) 6 belief that p, that '¬p' expresses the belief that not-p, and that the belief that p is inconsistent with the belief that not-p (since they have inconsistent contents, and everyone agrees that beliefs with inconsistent contents are inconsistent).This would explain why the sentences are inconsistent by appealing to the inconsistency between the beliefs they express.In general, the cognitivist can explain the logico-semantic properties of moral sentences in terms of the logico-5 SIN is the locus of discussion in the literature on what is called the 'Negation Problem' (Unwin 1999(Unwin , 2001;;Schroeder 2008); discussion often focused on SEMP in the 80s and 90s (e.g.Blackburn 1984Blackburn : ch.6, 1993: ch.10;: ch.10;van Roojen 1996).Strictly speaking, what needs explaining is logical inconsistency and logical entailment.However, to sort logical inconsistency (entailment, etc.) from semantic inconsistency (entailment, etc.) one merely needs to clarify that the relevant property holds under any substitution of the non-logical vocabulary (Baker and Woods 2015: 397).For the sake of brevity and clarity, then, we will normally leave this clarification implicit.
6 Expressivists usually allow that moral judgements are "beliefs" in some expanded sense of the term, while maintaining that they differ in kind from the representational states expressed by descriptive sentences like 'grass is green'.We generally use 'belief' in the narrower sense here, occasionally using 'ordinary descriptive' to emphasise the point.
semantic properties of the beliefs they express, which the beliefs themselves inherit from their contents.
According to pure expressivists, however, moral judgements are not beliefs, but desire-like attitudes, so this strategy is not available.Take SIN again.Suppose that the pure expressivist holds that 'stealing is wrong' expresses disapproval of stealing.
How, then, is she to explain the sentence's inconsistency with 'stealing is not wrong'?
There are two aspects to the problem.First, saying what state is expressed by the negation.Unlike the cognitivist, the pure expressivist cannot simply push the negation into the content of the state expressed by the atomic sentence: given that 'wrong' is, ex hypothesi, used to express disapproval, it is 'not stealing is wrong' that expresses disapproval of not stealing, and 'stealing is not wrong' is obviously not equivalent to 'not stealing is wrong'.Second, saying why these states are inconsistent.Whatever state is expressed by 'stealing is not wrong', the "inconsistency" that holds between that state and disapproval of stealing will not be the uncontentious inconsistency of beliefs with inconsistent contents, but will involve a different "clash" of attitudes, the cogency of which has been questioned (e.g.Hale 1993;Schroeder 2008: ch.3).The difficulty generalises: if moral judgements are desire-like attitudes rather than beliefs with moral content, then the expressivist cannot maintain that moral sentences inherit their logico-semantic properties from the content of the beliefs they express.

The simple "offloading" strategy
There are, of course, pure expressivist responses to the Frege-Geach problem, but these have proven controversial and none has gained anything like consensus.
Hybrid expressivists, however, claim to solve the problem 'on the cheap'.After all, according to hybrid expressivists moral judgements are hybrid states that are composed, in part, by beliefs; it is thus thought that the hybrid expressivist can use the very same explanatory resources that are available to the cognitivist, by "offloading" the explanation of the logico-semantic properties of moral sentences onto the belief-components of these hybrid states.
In more detail, the hybrid expressivist's proposal comes in two stages.First, they give a systematic, compositional explanation of what hybrid mental state is expressed by a logically complex moral sentence as a function of the hybrid states expressed by its atomic components.For example, if 'stealing is wrong' expresses a hybrid state composed of ( a1) and (b1), its negation 'stealing is not wrong' expresses a hybrid state composed of ( a1) and ( b4): (a1) Disapproval of actions insofar as they have descriptive property F.
(b1) The belief that stealing is F.
(a1) Disapproval of actions insofar as they have descriptive property F.
(b4) The belief that stealing is not F.
The desire-component (a1) is held constant.The belief component (b4) is the belief whose content is the negation of the content of (b1), the belief-component of the hybrid state expressed by the sentence being negated.The hybrid expressivist thus has a neat answer to what mental state is expressed by a logically complex moral sentence.We have no objection (here) to this aspect of the hybrid expressivist's proposal.
Our objection concerns the second stage of the proposal: the explanation of why moral sentences stand in the relevant logico-semantic relations.( b1) and (b4) are inconsistent, in virtue of being beliefs with inconsistent content.The hybrid expressivist thus proposes to explain why 'stealing is wrong' and 'stealing is not wrong' are inconsistent in terms of this inconsistency between the belief-components of the hybrid states they express.7 Generalising, the hybrid expressivist aims to explain the logico-semantic properties of moral sentences in terms of the logico-semantic properties of the belief-components of the hybrid states they express, thus using the very same explanatory resources that are available to the cognitivist.This is the simple "offloading" strategy.
(There are, in fact, two ways this could go.The direct approach is to explain the inconsistency of the sentences by appealing to the inconsistency of the relevant beliefcomponents directly.The indirect approach is to maintain that the hybrid states themselves are inconsistent in virtue of having inconsistent belief-components, and to explain the inconsistency of the sentences by appealing to the inconsistency of the relevant hybrid states.We focus on the direct approach, since this is the one hybrid expressivists seem to go in for (see e.g.Ridge's definition of validity, quoted in fn.10); but, as we'll note, our objection applies either way.) The problem is as follows.The hybrid expressivist denies Inheritance: the truth of the belief-component of a moral judgement is neither necessary nor sufficient for the truth of the judgement as a whole.The truth-conditions of a moral judgement, and thus those of the sentence that expresses it, float free of the truth-conditions of its belief-component.This is a constitutive commitment of hybrid expressivism as such.
But, as just discussed, the simple "offloading" strategy also seeks to explain the logicosemantic properties of moral sentences by maintaining that the relevant relations between different moral sentences perfectly track the corresponding relations between the truth-conditions of the belief-components of the hybrid states they express.For instance, 'stealing is wrong' is inconsistent with 'stealing is not wrong' because the former expresses a state composed in part by (b1), the latter expresses a state composed in part by (b4), and ( b1) and (b4) are inconsistent.But if the truth-conditions of the individual moral sentences float free from the truth-conditions of the relevant belief-components, then this perfect correlation should be the last thing we expect!If the truth-conditions of 'stealing is wrong' float free from the truth-conditions of (b1), and the truth-conditions of 'stealing is not wrong' float free from the truth-conditions of (b4), then the fact that the latter are inconsistent gives us no reason whatsoever to think that the former are inconsistent too.On the contrary, any such correlation would be surprising and would itself cry out for explanation.
The point generalises: in the hybrid expressivist's framework, we have no reason to think that any relevant relations between the truth-conditions of moral sentences will track the corresponding relations between the truth-conditions of the relevant belief-components.But the simple "offloading" strategy postulates just such a perfect correlation, across all moral sentences.Absent further explanation, then, the hybrid expressivist is simply asking us to accept a vast cosmic coincidence.(On the indirect approach, the coincidence concerns the correlation between the relations between the hybrid states and those of their belief-components.) Any explanation of the logico-semantic properties of moral sentences that appeals solely to the logico-semantic properties of the belief-components of the hybrid states they express is therefore plainly explanatorily inadequate.If we are to truly explain the logico-semantic properties of moral sentences -that is, if we are to solve the Frege-Geach problem -then something more is needed.

What went wrong
We think the following distinction is useful for clarifying where the simple "offloading" strategy goes awry.It also enables us to offer a charitable explanation of why its inadequacy has been missed, and it will be useful when presenting our positive proposal in §2.
Logico-semantic properties like inconsistency and entailment are typically defined in terms of truth: a set of sentences is inconsistent iff it is not possible for them all to be true at the same time.A set of sentences 'p1', …, 'pn' entails a sentence 'q' iff it is not possible for all of 'p1', …, 'pn' to be true and 'q' false.But the expressivist does not want to explain the meaning of a moral sentence, at least in the first instance, in terms of its truth-conditions.So, to explain properties of moral sentences like SIN, the expressivist first looks to explain psychological properties, like:

Attitudinal inconsistency for negation (TIN)
The mental state expressed by 'p' is inconsistent with the mental state expressed by '¬p'.
She then aims to use these psychological properties to explain the sentential properties, as befits a psychologistic (meta)semantics.
Now, an explanation of TIN would suffice to explain SIN.However, as argued, the simple "offloading" strategy cannot explain TIN: that the belief-components of a pair of hybrid states are inconsistent does not tell us anything whatsoever about whether the states themselves are inconsistent.
What the hybrid expressivist can explain is:

Agential inconsistency for negation (GIN)
Someone who accepts 'p' and '¬p' at the same time is thereby guaranteed to have inconsistent mental states. 8 Someone who accepts 'stealing is wrong' and 'stealing is not wrong' at the same time has two hybrid mental states.The first hybrid state has (b1) as its belief-component; the second has (b4).( b1) and (b4) are inconsistent.So, any agent who accepts both sentences is thereby guaranteed to have inconsistent mental states (namely, (b1) and (b4)).
Page 14 of 36 However, GIN is only true because (b1) and (b4) are inconsistent; and, as argued, by the hybrid expressivist's lights, the fact that (b1) and (b4) are inconsistent gives us no reason whatsoever to think that the hybrid states of which they are components, or the sentences that express them, are inconsistent.So, unlike TIN, GIN does not suffice to explain SIN. 9 To sum up: the hybrid expressivist can explain GIN but not TIN; and while TIN would suffice to explain SIN, GIN does not. 10e suspect that this distinction may help explain why the inadequacy of the simple "offloading" strategy has been missed, given that: (1) the hybrid expressivist can explain GIN; (2) in the normal run of things, GIN and TIN come along together; and (3) TIN would suffice to explain SIN.The hybrid expressivist thus has everything they would need to explain SIN, in the normal run of things.The problem is that we are not in the normal run of things.Given a commitment to hybrid states whose truthconditions float free from the truth-conditions of their belief-components, it should be 9 There's an excellent joke we could make here about GIN being the source of SIN, but we don't want to distract you.
perfectly possible for TIN (and so SIN) to be false even if GIN is true.Ironically enough, it is the constitutive commitments of hybrid expressivism that introduce mystery into what would otherwise be a satisfactory explanation.

Conclusion to Part I
We have argued that the simple "offloading" strategy employed by hybrid expressivists to solve the Frege-Geach problem is explanatorily inadequate.Contrary to its advertising, hybrid expressivism cannot solve the Frege-Geach problem 'on the cheap'.
However, we have not argued that hybrid expressivism constitutes no advance at all on pure expressivism.It's worth emphasising, for example, that hybrid expressivists have a neat, compositional story to tell about what states are expressed by logically complex sentences -that's no mean feat.We have only argued that hybrid expressivists have not solved the Frege-Geach problem, not that they cannot do so.
One option here is to endorse a weaker connection between the truthconditions of a moral sentence and those of the relevant belief-component than Inheritance, but which is still sufficiently strong that the relevant relations are guaranteed to hold between the moral sentences when they hold between the beliefcomponents. 11However, we suspect that the hybrid expressivist will be loath to endorse any such connection (particularly as they take moral judgements to be massively multiply realisable -see fn.2), and we cannot think of a principle that is both plausible and sufficiently strong.Another option is to try and give an account of 11 Thanks to an anonymous referee for raising this possibility.
what it is for two mental states to be inconsistent, such that we can infer the inconsistency of two hybrid states from the inconsistency of their belief-components. 12Again, we do not rule this out.Note, however, that this move -of expanding the notion of attitudinal inconsistency beyond merely beliefs with inconsistent contentsis precisely the kind of controversial move employed by pure expressivists.The attraction of hybrid expressivism vis-à-vis the Frege-Geach problem was supposed to be that she could solve the problem by appealing to very same resources as the cognitivist.This hope has been frustrated.
We suggest a different strategy.

II. HOW TO SIDESTEP THE FREGE-GEACH PROBLEM
Expressivists, pure and hybrid, have been going about the Frege-Geach problem in the wrong way.Their efforts have focused on explaining what it is for, say, 'p' to be inconsistent with '¬p'.Since the expressivist aims to make sense of the logico-semantic properties of moral sentences in terms of the mental states they express, the claim is that these sentences are inconsistent in virtue of expressing inconsistent mental states.
They then face the difficulty of saying what it is for the relevant mental states to be 12 For instance, Ridge might try explaining attitudinal inconsistency using his account of disagreement in prescription (Ridge 2014: ch.6).This strikes us dialectically costly, however.Hybrid expressivists are generally thought to have a harder time explaining disagreement than others (hence why Ridge has to develop a non-standard such view).On this approach, hybrid expressivism's most marketable feature over pure expressivism, its response to the Frege-Geach problem, winds up relying on one of the most controversial aspects of the view.Thanks to James Brown for raising this possibility and for discussion.
inconsistent.For the pure expressivist, it's controversial whether desire-like states "clash" in a way that counts as the right sort of inconsistency.The hybrid expressivist initially seems better off, since she can appeal to inconsistency between the beliefcomponents of the hybrid states -but we have seen how that cannot suffice to explain the inconsistency of the sentences themselves.
Our instance explain what it is to think that certain sentences are inconsistent.We will first explain how the sidestepping strategy works in general, before applying it to inconsistency in particular, and then generalising it to other logico-semantic properties.

The expressivist sidestep
The apposite term 'expressivist sidestep' was coined by James Dreier (2015: 273) to describe a particular strategy that has proved popular among expressivists ever since Simon Blackburn (1984Blackburn ( , 1993) ) began his quasi-realist project.We do have a slight disagreement with Dreier's characterisation of it, though.Where we agree with Dreier (2015: 284) is that expressivists normally 'think the illuminating way to understand our life of moral talk and thought is by stepping off to the side and looking at the phenomena from there.Instead of thinking about wrongness, we think about thoughts and talk of wrongness.Everything we wanted to explain turns out to have a much clearer explanation from this side-on vantage.'Where we disagree with him (2015: 273, emphasis added) is that this is 'a favorite maneuver of expressivists by which they avoid answering the traditional questions of metaethics and give satisfying answers to different questions-questions that arise when we look at the same phenomena "side on"-instead.' We do not think the expressivist sidestep should be used to dodge questions.
The sidestep is a way to tackle problems from a different angle.When we say we're sidestepping the Frege-Geach problem, then, we don't mean that we're dodging it or avoiding answering it; we mean we're taking a step to the side to look at the problem in a different way -with the intention of solving it.
The expressivist sidestep has been employed to show that expressivism is compatible with various prima facie plausible claims.For example, according to Blackburn (1993: 4), the conundrum is whether the expressivist 'can make sense of several ideas: that truth is the aim of judgement; that our disciplines make us better able to appreciate it, that it is, however, independent of us, and that we are fallible in our grasp of it.'Sceptics allege these ideas would not make any sense if moral judgements are desire-like attitudes rather than beliefs that robustly represent a realm of moral facts.Blackburn's quasi-realist project involves having to 'earn the right' (1993: 198) for expressivists to continue using these ideas, rather than impoverishing moral discourse by rejecting any talk of truth, fallibility, mind-independence and, as we'll be discussing, inconsistency and entailment.An example will help to illustrate the strategy.Consider the charge that expressivism cannot make sense of morality being mind-independent (Jackson and Pettit 1998;Suikkanen 2009).The first step towards answering the charge is to identify a target sentence to make sense of.In this case, let us choose 'torture would be wrong even if we thought otherwise'.Given expressivism, does this sentence make any sense or is it as nonsensical as 'adgjg jep uii'?Recall from §1.2 that expressivists explain the meanings of sentences in terms of the mental states they express.This is why the expressivist sidestep is part of answering the challenge rather than dodging it: if the expressivist can identify an expressivist-friendly mental state for the target sentence to express, then she can maintain that the target sentence is meaningful.Blackburn's (1998: 296) proposal is, roughly, that our target sentence expresses the speaker's disapproval of the torture that goes on in possible worlds in which we permit torture.Problems have been raised for this proposal (Berker 2020, cf. Baker 2021), but if it works then the expressivist has shown that talk about moral mind-independence is perfectly meaningful.Following that, we may proceed to the stronger claim that there is nothing incoherent involved in believing expressivism and having the disapproval expressed by the target sentence (Schroeder 2014).Finally, the issue becomes first-order: is there good reason to so disapprove, or is the justification for the target sentence undercut by awareness of the mental state it expresses?The answer is of course there's good reason to disapprove of torture even in worlds where it's permitted -regardless of what we think about torture, it still hurts, undermines autonomy and dignity, yields unreliable intelligence and so on.This is how quasirealists earn the right to claim the kinds of things that 'tempt people to realism' (Blackburn 1984: 171).
We believe expressivists should do the same thing with inconsistency.They should not try to widen the definition of inconsistency and argue that some moral sentences (such as contradictory ones), or the mental states they express, can meet those broader conditions.Instead, they should take a step to the side and examine what we're up to when we think a set of sentences is inconsistent.
This marks an important difference in strategy.we will have earned the right to make the claims.Just as there is 'nothing improper' (Blackburn 1984: 171) about the use of moral predicates, we aim to show there is nothing improper about thought and talk of moral inconsistency either.

Inconsistency is modal
A standard view of inconsistency is that sentences (claims, propositions, beliefs) are inconsistent when they can't all be true simultaneously.Likewise, sentences are consistent if it's possible for them all to be true at once.Consistency and inconsistency are modal notions.Fortunately for us, expressivists have already explored moral modals.In a moment we'll look at how this helps solve the inconsistency problem.
Before we do, let us make our position more explicit.
What is it for a set of sentences {p, q, …} to be inconsistent?In the first instance we give the standard answer: it cannot be the case that all members of the set are true. 13If you don't think expressivists have earned the right to truth yet, then we can disquote: a set of sentences {p, q, …} is inconsistent if and only if it is impossible that p, q, … .)The worry is that this has merely shifted the bump along the rug.Why can't certain moral sentences, such as contradictory ones, be true together?This is where the sidestep strategy comes in.We aren't able to explain at the first-order level why, for example, 'lying is wrong' and 'lying is not wrong' cannot be simultaneously true -they just can't.We've hit explanatory bedrock here, and it isn't in a place to be particularly embarrassed about.The law of non-contradiction is as fundamental as it gets.However, what we can do is sidestep it, explain what mental states are involved in accepting it, and eventually earn the right to say it.
The target sentence we will earn the right to assert in the following two subsections is 'lying cannot be both wrong and not wrong', a modal sentence that (if we're right) is equivalent to saying that 'lying is wrong' and 'lying is not wrong' are inconsistent.Different theories of moral modals will assign different mental states to be expressed by this sentence, and we do not wish to adjudicate between theories of modals in this paper.Of course, there is a possibility that the expressivist cannot make sense of moral modals at all, in which case our strategy is doomed -but then the quasirealist project is independently doomed, so let's remain optimistic for present purposes.These next two subsections are illustrations of how expressivists can earn the right to inconsistency claims, dependent on which theory of moral modals turns out to be true.14

Moral modals 1: ideal advisors
There are many possible ways to understand moral modals in terms of ideal advisors.
We will use a simple version inspired by James Lenman (2003): to think that a moral claim 'p' is possibly true is to think one's ideal advisor might think that p.This is not merely descriptive because 'ideal' is an evaluative notion, but the modal 'might' is non-moral (this theory assumes we already know what non-moral modals are) and so the result is a hybrid state.For example, 'lying might be wrong' expresses: (a5) Approval of certain qualities, e.g.'clever, fully informed and/or whatever' (Lenman 2003: §8).
(b5) A non-moral modal belief that an advisor with those qualities might think that lying is wrong.
This is essentially a generalised version of Simon Blackburn's (1998: 318;2009) account of thinking oneself might be incorrect.An important feature is that the truth of the belief-component (b5) is not necessarily linked to the truth of the moral modal claim that lying might be wrong.That is, the denial of Inheritance applies here as well.
Let us turn now to the task of earning the right to the target sentence by first assigning a coherent mental state that constitutes accepting it.'Lying cannot be both wrong and not wrong' expresses: (a5) Approval of certain qualities.
(b6) A non-moral modal belief that it is impossible that an advisor with those qualities thinks that lying is wrong and not wrong.
There is nothing incoherent in having these mental states, and so expressivists can coherently believe and assert that lying cannot be both wrong and not wrong.This is equivalent to saying that 'lying is wrong' and 'lying is not wrong' are inconsistent.
Expressivists, then, can talk of inconsistency without shame.
But hybrid expressivists can do even better.On certain dialectically plausible assumptions, hybrid expressivists can show that we are committed to accepting that such moral contradictions are inconsistent.
Consider what ideal advisors are thought to be like.Typically, one of the ideal qualities is taken to be consistency in non-moral belief.Ideally we'd want our advisors to have true descriptive beliefs rather than false ones (Firth 1951: 333;Smith 1997: 89), but at a minimum they'd better be consistent.Now, we showed in §1.4 that hybrid expressivists have fairly non-controversially secured agential inconsistency for negation (GIN) -anyone who accepts moral 'p' and '¬p' at the same time is guaranteed to have inconsistent descriptive beliefs.This means that (b6) is true: it is impossible for a descriptively consistent advisor to believe lying is wrong and not wrong.
Therefore, if hybrid expressivism is true, anyone who approves of descriptive consistency in (a5) should also believe (b6).And, if this theory of moral modals is true, this just is to believe that lying cannot be both wrong and not wrong, which is to believe that 'lying is wrong' and 'lying is not wrong' are inconsistent.
Given the denial of Inheritance, the truth of (b6) does not guarantee the truth of the overall judgement, so we have not provided any direct argument for the truth of this moral inconsistency claim.What we have shown is that we are committed to believing it.What we have, then, is a transcendental argument: given our pre-existing commitment to the law of non-contradiction for descriptive claims, we are rationally compelled to believe in the law of non-contradiction for moral claims. 15Thus, not only have we shown that it is meaningful and coherent to say that 'lying is wrong' and 'lying is not wrong' are inconsistent, but we have given a conclusive reason to believe it as well.The right to inconsistency has been earned.

Moral modals 2: acceptable standards
Echoing the previous subsection, there are many possible ways to cash out a theory of moral modals in terms of acceptable standards.We will use a simple version that Ridge (2015: 12) also uses for illustrative purposes (before he develops a more complex theory): '"must-p" corresponds to "Any acceptable epistemic standard would, given some contextually specified body of evidence, require believing that p."'This is not merely descriptive because 'acceptable' is an evaluative notion which, the expressivist will insist, involves desire-like attitudes.According to Ridge, these desire-like attitudes rule out sets of guiding standards for the formation of further attitudes.As a hybrid theorist, Ridge also includes a belief-component that refers to the standards not ruled out.For example, 'the butler must have done it' expresses: (a7) Ruling out sets of standards.
(b7) The ordinary descriptive belief that any sets of standards not ruled out would, given the evidence, require believing that the butler did it.
Page 26 of 36 Two people looking at the same evidence may come to different conclusions by having different descriptive beliefs about what the relevant standards require us to believe on the basis of that evidence, or by having ruled out different sets of standards for belief formation.Once again, expressivists must deny Inheritance -so the truth of (b7) is neither necessary nor sufficient for the truth of the hybrid state as a whole.
Let us return to our target sentence, 'lying cannot be both wrong and not wrong'.In our context of considering inconsistency, we are not relying on any particular body of evidence.Here is one possibility for what the target sentence expresses: (a7) Ruling out sets of standards.
(b8) The ordinary descriptive belief that any sets of standards not ruled out would, whatever the evidence, require not believing that lying is wrong and not wrong.
Another possibility involves strengthening (b8) so that it isn't about requiring the absence of believing the moral contradiction, but is instead about requiring believing the negation of the moral contradiction.In either case, reasoning parallel to the previous subsection will apply.
First, (a7) and (b8) are perfectly coherent mental states and so it is coherent to accept 'lying cannot be both wrong and not wrong'.Second, given that we already rule out any standards that permit us to believe descriptive contradictions, the remaining sets of standards require not believing moral contradictions due to GIN.This is to say that anyone whose standards require us to abide by the law of noncontradiction for descriptive beliefs in (a7) should also believe (b8).And, if this theory of moral modals is true, this just is to believe that lying cannot be both wrong and not wrong, which is to believe that 'lying is wrong' and 'lying is not wrong' are inconsistent.As before, this is not a direct argument for the truth of the inconsistency claim.It is a transcendental argument to believe it.

Conclusion to Part II
We have argued that, according to two prominent expressivist theories of moral modals, the mental state expressed by 'lying cannot be both wrong and not wrong' is perfectly coherent.It should be clear that there was nothing special about lying as the subject nor wrongness as the moral property, and that the same strategy can be generalised to other instances of moral inconsistency.We thus make sense of talk of moral inconsistency.Moreover, we argued that the mental state expressed by the target sentence cannot be rationally rejected.We thus have conclusive reason for accepting it.
This strategy straightforwardly generalises to other logico-semantic properties.
Consider entailment.We accept that 'all theft is wrong' and 'taxation is theft' together entail 'taxation is wrong'.To explain why, we might use the following target sentence: 'it cannot be that all theft is wrong, that taxation is theft, and that taxation is not wrong'.The strategy then proceeds as before.
So, by approaching the Frege-Geach problem from side-on, the hybrid expressivist obtains a transcendental solution to it.
Note that our justification of SIN makes no reference to TIN. 16 It is a dramatic change of strategy.Nevertheless, if anyone is committed to explaining the truth of SIN in terms of TIN, they may do so.They may employ our strategy at the level of thought to first obtain a transcendental argument for TIN: the judgements that lying is wrong and that lying is not wrong are inconsistent, for example, because those judgements cannot be simultaneously true (and to think so is to be in the kinds of states discussed in the previous two subsections, just directed at thoughts rather than sentences).They can then say that the sentences are inconsistent because they express inconsistent mental states.
We acknowledge that this strategy for sidestepping the Frege-Geach problem relies on a number of substantial assumptions that may be contested.We'll consider three of them here.To begin with, our sidestepping solution depends on there being a plausible expressivist-friendly theory of moral modals.If the reader thinks that expressivists cannot make sense of possibility or impossibility at all, then the reader should not think that expressivists can solve the inconsistency element of the Frege-Geach problem.
Relatedly, even though we have argued that two different kinds of theory would do the required work and we are open to others, it is not the case that anything goes.A silly theory of modals that holds 'lying might be wrong' expresses hostility to chimpanzees would not be amenable to our strategy.It's notable that the two theories we discussed are both hybrid theories -even when coming from Lenman, a pure expressivist -but hybridism is not the necessary feature of modals for our argument Page 29 of 36 to work.17All we need is a theory that allows us to, in some sense or other, rule out descriptive inconsistencies.For example, a pure theory may posit that believing it is impossible that p is a matter of planning never to accept p given any evidence.As long as the theory has the resources to capture our existing animosity to descriptive inconsistency, our reasoning will find a foothold.We may even consider this a condition of plausibility for future theories of moral modals, given its utility for solving the Frege-Geach problem.
Second, it is worth noting that not everyone is averse to contradiction.
Dialetheists such as Graham Priest (2006) do not rule out contradictory descriptive beliefs.Our strategy will therefore not persuade this aberrant group of philosophers.
This doesn't seem like much of a problem, though.Dialetheism implies we shouldn't automatically believe that 'lying is wrong' and 'lying is not wrong' are inconsistent on the basis of form alone.Instead, judging that lying cannot be both wrong and not wrong will be a matter of substantive moral theorising, not a foregone conclusion; the option that 'lying is wrong' and 'lying is not wrong' are simultaneously true would be open.
In the ordinary debate about the inconsistency problem, our dialectical opponents are those who already accept that contradictory descriptive sentences cannot be true, who then challenge expressivists to show why contradictory moral sentences follow suit.Our strategy involves a rejection of any head-on answer.We don't directly explain what it is for moral sentences to be inconsistent, just as expressivists don't directly explain what it is for torture to be wrong.We sidestep.To think moral sentences are inconsistent is to think they can't all be true, which is -in some sense to be specified in a theory of moral modals -to rule out ever accepting them at once.Except for dialetheists, we all rule out descriptive contradictions and, given GIN, this means we effectively rule out moral contradictions.We therefore believe moral contradictions are inconsistent.The same reasoning applies to other moral sentences we take to be inconsistent. 18 It's important for us to stress that we are not merely saying non-dialetheists are committed, on pain of inconsistency, against accepting moral contradictions.This much was already apparent in §1.4 with GIN.Our goal was to show that nondialetheists are committed to accepting that contradictory moral claims are inconsistent.
Finally, our transcendental move may be contested.The jump from 'we are committed to believing p' to 'p' is not logically valid.Couldn't the world be different to what we must believe about it?Couldn't we be tragically committed to falsehoods?
It is of course controversial whether transcendental arguments work.If they do not, we have still made progress because we have shown how expressivists can hold that moral inconsistency claims make sense, are coherent, and are rationally justified.It would be nice to make that transcendental jump, though.We can't defend the legitimacy of transcendental arguments in this paper, but note that they have been 18 Note, then, that our proposal does not leave room for the view that there are moral contradictions but no non-moral contradictions.So be it: since we see no reason to think that this is an independently attractive view, we see no reason to think this is an objection to our proposal.
used several times in the expressivist literature (Gibbard 2003: ch.5;Schroeder 2010: 160;Toppinen 2017;Bex-Priestley forthcoming).While there does seem to be something prima facie worrying about transcendental moves within domains in which we are robust realists, perhaps those worries lose their force once we abandon that realism.Perhaps if we are not 'representing a way the world might be,' we can be less anxious that the 'world might simply not cooperate' (Ridge 2018(Ridge : 2955)).We suggest that expressivists can take rational commitments as decisive.

III.CONCLUSION
The Frege-Geach problem is arguably the biggest problem for expressivism.Hybrid expressivists have gone a long way towards solving compositionality, but we argued that they have not yet solved inconsistency.This is due to two things in combination.
Firstly, they have been tackling the issue head-on, trying to explain what it is for moral sentences and attitudes to be inconsistent.These explanations rely on the inconsistency of the belief-components of moral judgements.Secondly, for hybrid expressivism to remain a form of expressivism proper, they deny any connection between the truth of those belief-components and the truth of the moral judgements themselves.Consequently, we are left wondering why the belief-components have any bearing on moral consistency if they have no bearing on moral truth: the fact that the belief-components of two moral judgements can't both be true gives us no reason to think that the two moral judgements, or the sentences that express them, can't both be true.
proposal is to do what expressivists have done with other problems: sidestep it.Rather than explaining what it is for certain sentences (or the mental states they express) to be inconsistent, the expressivist should in the first Blackburn isn't explicit about what it means to 'make sense' of ideas and claims but there are a few different ways we might cash out the notion.First, it could be a matter of showing the relevant sentences, given expressivism, are meaningful rather than nonsensical.More strongly, it could involve showing the ideas are coherent with expressivism -that is, not only are they meaningful, they might be true.Finally, and more strongly still, it could encompass showing that the claims can still be justified if expressivism turns out to be true -that is, not only are they meaningful and coherent, If we can show that there are sensible expressivist-friendly mental states that constitute accepting moral inconsistency claims, and that there is good reason to have those mental states, then Previously, expressivists have sidestepped ordinary moral claims 'p' and '¬p' to figure out what mental states are involved in accepting them, and then argued head-on (and unsuccessfully, in our view) that those mental states are inconsistent.Contrastingly, we go one level up: we will be sidestepping the claim that 'moral sentences "p" and "¬p" are inconsistent' to figure out what mental states are involved in accepting that.