Disagreement and inconsistency: a problem for orthodox expressivism

What makes two sentences inconsistent? Expressivists understand the meaning of a sentence in terms of the mental state it expresses. In order to explain the inconsistency between two sentences, the expressivist must appeal to some inconsistency feature of the mental states expressed. A simple explanation is that two sentences, e.g., “murder is wrong” and “murder is not wrong” are inconsistent by virtue of expressing mental states that disagree. Schroeder (2008) argues that the expressivist lacks a plausible explanation of the disagreement. Baker & Woods (2015) argue that Schroeder is wrong. With these authors, I agree that expressivists have an explanation of disagreement, but this does not adequately explain why two sentences are inconsistent. The reason is that two intuitively inconsistent sentences do not necessarily express mental states that disagree. Moreover, assuming that the expressivist gives a structurally identical explanation for moral and non-moral language, the problem generalizes to non-moral language. It is also argued that the problem extends to thought. How expressivists can and should conceive of inconsistency thus remains a challenge.


Introduction
What makes two sentences inconsistent? If we think of the meaning of a sentence in terms of its truth-conditions, i.e., the conditions that have to obtain in order for the sentence to be true, two sentences are inconsistent if the sentences cannot be true simultaneously. Expressivists conceive of the meaning of a sentence in a radically different way. Mark Schroeder explains the orthodox expressivist idea as follows.
[T]he expressivist holds that the meaning of 'grass is green' and of 'stealing is wrong' can both be explained by saying what mental state they express -the former expresses the ordinary, run-of-the-mill belief that grass is green, and the latter expresses a negative attitude toward stealing which I am here calling 'disapproval.' (Schroeder, 2014: 279-280) This approach to meaning rules out explaining the inconsistency in terms of truthconditions. Rather, the expressivist, Schroeder argues, must show that two sentences are inconsistent by virtue of "appealing to some 'inconsistency' feature of the attitude that each expresses" (Schroeder, 2008: 39). Schroeder claims that expressivists have "left completely unexplained and apparently inexplicable why 'murder is wrong' and 'murder is not wrong' are inconsistent" (Schroeder, 2008: 47). It may be responded that the expressivist has a simple explanation. Allan Gibbard understands the main difference between the Fregean (or truth-conditionalist) view described above, and the expressivist view as follows: "Fregeans start out with negation. … Expressivists start out correspondingly with disagreement" (Gibbard, 2012: 273-274). Similarly, Derek Baker and Jack Woods argue that the expressivist can and should say that " [s] entences are broadly inconsistent in virtue of expressing discordant attitudes" (Baker & Woods, 2015: 416). This suggests that the key to understand inconsistency is in terms of disagreement.
This paper aims to argue that the expressivist explanation of inconsistency in terms of attitudes or mental states that disagree is problematic. First, it seems that two intuitively inconsistent sentences do not necessarily express mental states that disagree. Indeed, if the expressivist gives a structurally identical story for moral and non-moral language, the problem applies to both moral and non-moral inconsistencies. Second, the problem also extends to thought. Jack believes that grass is green. Julia supposes that grass is not green. Although Jack and Julia do not have attitudes that disagree, plausibly the contents of their attitudes are inconsistent. It, therefore, does not seem as if the expressivist can explain inconsistency in thought or talk simply by virtue of mental states that disagree.
The outline of this paper is as follows. The next section provides some preliminary remarks about how to understand orthodox expressivism and Schroeder's challenge, i.e., why explaining inconsistency is a problem for expressivism. Section 2 explains the main idea behind the claim that two sentences are inconsistent because they express mental states that disagree. In Sect. 3, it is argued that the expressivist explanation of inconsistency between two sentences in terms of expressing mental states that disagree does not work. Section 4 argues that a similar problem arises at the level of thought. In the final section, I sum things up and make some concluding remarks.
the mental state(s) it expresses. Moreover, this is also what meaning consists in. 1 We can thus explain the orthodox expressivist idea as follows.
Orthodox semantic expressivism: the meaning of a sentence S consists in and is explained by the mental state(s) it expresses.
Moreover, expressivists typically give structurally identical semantic stories for moral and non-moral sentences. The difference in meaning between moral and nonmoral language is explained by the attitude that the sentences express. As Schroeder writes in the passage quoted above, "grass is green" expresses the belief that grass is green and "stealing is wrong" expresses disapproval of stealing. 2 We can now consider the following pairs of sentences.
These pairs of sentences are plausibly inconsistent. On a standard explanation of "inconsistency," two sentences are inconsistent if they express propositions that cannot be true simultaneously. Again, if we think of the meaning of (3) in terms of its truth-conditions, this is easily explained. If (3) is true, then (4) is false and vice versa. This is guaranteed by the meaning of (3) and the meaning of "not." This explanation is not available to the expressivist. Rather, if the meaning of sentences is determined by the mental states they express, the expressivist will have to rely on "some 'inconsistency' feature of the attitude that each expresses" (Schroeder, 2008: 39).
Schroeder argues that expressivists have "left completely unexplained and apparently inexplicable why" (Schroder 2008: 47) a pair of sentences like (1) and (2) are inconsistent. In order to understand why, we must, following Schroeder, distinguish between two approaches to explain the inconsistency, viz., A-type inconsistency and B-type inconsistency. An A-type inconsistency is characterized by the same attitude towards inconsistent contents, e.g., the belief that grass is green and the belief that 1 We can think of "explained by" and "consists in" as corresponding to metasemantic and semantic components of expressivism respectively. Schroeder focuses on expressivism as a semantic theory. See also, e.g., Gibbard (2003: 7), Ayer (1936: 31) and Sinclair (2011: 386). Ridge (2014) defends a version of expressivism that is supposed to be a combination of metasemantic expressivism and a more standard truth-conditional semantic story. Some philosophers use "expressivism" to refer to pragmatic theses. See e.g., Yalcin (2012) and Charlow (2015). It may also be debated whether these views are more adequately described as metasemantic. The focus here is on expressivism as a semantic theory. I will thus put the other views to the side. Moreover, the focus is on pure expressivism rather than hybrid or relational variants. 2 For the purposes of this paper, I assume that the expressivist wants to tell structurally identical stories for moral and non-moral sentences. But expressivism may also be conceived as a more local thesis. Part of the motivation for ethical expressivism is the apparent non-representational function of moral language. Against such a background, it may be argued that moral sentences express some kind of non-cognitive attitude, but that the meaning of a non-moral sentence is determined by its truth-conditions. Ayer (1936) is perhaps the most explicit proponent of a kind of local expressivism. grass is not green. Moreover, beliefs belong to the class of what Schroeder calls inconsistency transmitting attitudes. "An attitude A is inconsistency-transmitting just in case two instances of A are inconsistent just in case their contents are inconsistent" (Schroeder, 2008: 43). 3 A B-type inconsistency, by contrast, is characterized by two distinct and apparently logically unrelated attitudes. For expressivists, a pair of sentences like (1) and (2) do not express beliefs. Rather, they express non-cognitive attitudes, e.g., disapproval of murder, and, following Blackburn (1988), tolerance of murder respectively. 4 Since disapproval and tolerance are distinct attitudes it is an example of a B-type inconsistency.
A-type inconsistency is, Schroeder claims, "something that we should all recognize and be familiar with" (Schroeder, 2008: 48). To believe that grass is green and to believe that grass is not green is inconsistent. It, therefore, seems as if the expressivist has an explanation of why (3) and (4) are inconsistent. B-type inconsistency, by contrast, is more esoteric. Indeed, Schroeder claims that there are no good examples of B-type inconsistencies and "[a]ssuming that disapproval and tolerance of murder are inconsistent is taking for granted everything that expressivists need to explain" (Schroeder, 2008: 48). Schroeder's challenge to the expressivist is thus to explain B-type inconsistency, e.g., why disapproval of murder and tolerance of murder are inconsistent. 5

Disagreement is the key
It may be responded that the expressivist has a simple way of answering Schroeder's challenge, viz., that inconsistency should be explained by virtue of disagreement. A prominent idea in the expressivist tradition is that people do not merely disagree in belief, but also in attitude. Approval of murder disagrees with disapproval of murder. Hence, as Baker and Woods argue, two "[s]entences are broadly inconsistent by virtue of expressing discordant attitudes" (Baker & Woods, 2015: 416). This suggests the following general view: Expressivist explanation of inconsistency: Two sentences are inconsistent if the mental states that they express disagree. 3 Wondering, supposing and desiring, Schroeder claims, are not inconsistency-transmitting attitudes. Hence, there is nothing inconsistent in supposing that grass is green and supposing that grass is not green (see Schroeder 2008: 40). 4 Part of the problem is also to say what a sentence like (2) expresses. This is the kernel of the negation problem, i.e., saying what mental state that is paired with "murdering is not wrong." See Schroeder (2008) for discussion. 5 It is worth considering whether A-type inconsistency is consistent with expressivism. Schroeder (2008) seems to think it isn't. The reason is that it requires an explanation of content that is independent of attitudinal expression. I will not here investigate this claim any further. Rather, the issue is whether an expressivist can provide a respectable explanation of B-type inconsistency, i.e., some kind of explanation of why two apparently logically unrelated attitudes are inconsistent.
For example, if "grass is green" expresses the belief that grass is green and "grass is not green" expresses the belief that grass is not green, then the sentences express mental states that disagree. This, at least on the surface, is an example of an A-type inconsistency. However, the challenge to the expressivist is to explain B-type inconsistency. If the key to answering the challenge is disagreement, we need to investigate the nature of disagreement more thoroughly.
Following Gibbard (2012), we can start by thinking of disagreement in terms of mental states that exclude one another. On such a view, believing that grass is green excludes believing that grass is not green. "The explanation […] appeals to some initial understanding of disagreement and of one state's of mind excluding another" (Gibbard, 2012: 277). Part of Gibbard's view regarding how one state of mind excludes another relies on Thomas Scanlon's idea that certain combinations of states of mind are structurally irrational. 6 Paradigm examples of structural irrationality include failing to intend to do what one believes one ought to do, failing to intend the means to one's ends, and believing contradictory propositions. Believing that grass is green thus excludes believing that grass is not green. This would be a kind of A-type inconsistency. However, not all examples of structural irrationality are of "the kind of inconsistency Schroeder thinks it respectable to invoke" (Gibbard, 2012: 288). Some examples of structural irrationality involve apparently logically unrelated attitudes, i.e., B-type inconsistencies. For example, many philosophers think that tolerance of murder excludes disapproval of murder. Of course, Schroeder may insist that this is "taking for granted everything that expressivists need to explain" (Schroeder, 2008: 48). The expressivist must explain why tolerance of murder and disapproval of murder disagree.
This takes us to the proposal made by Baker and Woods. They argue that B-type inconsistencies are no less amenable to explanation than A-type inconsistencies. 7 Certain combinations of attitudes are what they call "discordant." "This is a term of art and is not meant to indicate anything other than that the set of attitudes have a property that is often called 'inconsistency'" (Baker & Woods, 2015: 393). They outline two different explanations of discordance. The first explanation "identifies discordance with a strong sense of interpretive incoherence" (Baker & Woods, 2015: 402). For example, it may be argued that we cannot make sense of a person who believes that p and believes that not-p or who intends inconsistent ends. The second explanation identifies discordance in terms of functional roles. For example, in relation to Schroeder's idea, the functional role of a belief may be used to offer an explanation of why belief is an attitude that is inconsistency-transmitting: "a belief is doing its job properly when the world is the way the belief says it is. … Beliefs with inconsistent contents are discordant since they are guaranteed to misrepresent the world" (Baker & Woods, 2015: 402-403).
These two explanations also promise to explain why certain combinations of attitudes are discordant even in the absence of any inconsistency in content. Baker and Woods consider a number of different examples. For example, some attitudes plausibly do not take propositions as their objects at all, but things. John likes ice-cream. John's attitude is not plausibly interpreted as directed towards any proposition. Rather, the attitude is object directed. However, to like ice-cream seems to be in discordance with disliking ice-cream. The functional role of likes and dislikes have both inter-and intrapersonal roles. The interpersonal role is to "guide our group decisions and to coordinate our reactions to various events" (Baker & Woods, 2015: 407). The intrapersonal role is to guide the individual agent, viz., "to structure decisions over time. Since liking and disliking one and the same thing contravenes this purpose, it is discordant" (Baker & Woods, 2015: 407-408). Of course, likes and dislikes also take persons and actions as their objects. The latter is perhaps more relevant and important for expressivism. Baker and Woods argue that liking dancing and disliking dancing are discordant. For similar reasons they also argue that approval of dancing and disapproval of dancing are discordant. This gives us a promising and plausible way of explaining why two mental states disagree that applies to both A-type inconsistencies and B-type inconsistencies. 8

Inconsistent sentences without disagreement
Suppose that Baker and Woods have provided a convincing explanation of why two mental states are discordant or disagree. Does it solve Schroeder's challenge? The challenge, recall, is to explain why "murder is wrong" is inconsistent with "murder is not wrong." First, Baker and Woods never provide any explanation of what it is to judge that murder is not wrong, but let us assume that they agree with Blackburn. To judge that murder is not wrong is to tolerate murder. Moreover, suppose we agree with Gibbard that "[t]olerating an alternative is treating it as eligible for picking, whereas disapproving it is treating it as ineligible for picking" (Gibbard, 2012: 287). Given these two functional roles, it seems that tolerance of murder and disapproval of murder is a structurally irrational combination of attitudes. The functional role of the former is contravened by the functional role of the latter -and vice versa. Second, what Baker and Woods have explained is why two mental states disagree, but Schroeder's challenge concerns explaining why two sentences are inconsistent. Of course, the expressivist idea is that two sentences are inconsistent by virtue of expressing mental states that disagree. It thus seems that we have a simple and elegant explanation of why "murder is wrong" and "murder is not wrong" are inconsistent. Two sentences are inconsistent by virtue of expressing mental states that disagree. But is this really how an expressivist can and should explain inconsistency?
Let us assume that moral and non-moral language works in structurally identical ways. We can thus start by considering whether.
are inconsistent by virtue of expressing beliefs that disagree. A sentence like "grass is green" is typically used to make an assertion. If Jack asserts that grass is green, it expresses the belief that grass is green. However, it also seems plausible that one and the same sentence can be used in other ways. Julia, for example, uses "grass is green" as a supposition. We often suppose things in discussions and arguments. Sometimes we aim to settle a certain question and use suppositions as a means to show that an alternative is unacceptable. This, as Mitch Green writes, involves "a practice permitting a proposition P to be put forth as a supposition rather than as an assertion" (Green, 2000: 378).
It may be responded that one cannot use "grass is green" to make a supposition. Rather, in order to make a supposition one must say "let us suppose, for sake of argument, that grass is green" or the like. One can, of course, perform the particular speech act by making it explicit. However, it also seems possible to "put forth P as a supposition without describing oneself as doing so, but instead relying on contextual factors to indicate the force of one's remark" (Green, 2000: 377-378). Let me begin by advancing some examples where it seems that the use of an indicative sentence does not intuitively express a belief and then proceed to provide some more theoretical considerations to the same effect.
First, in certain argumentative contexts, e.g., spelling out the premises and conclusions of an argument, it seems that a speaker can utter "grass is green" without expressing the belief that grass is green. Consider this fragment of discourse: "Here is a valid argument. Grass is green. All that's green is visible. Therefore, grass is visible." Arguably, the speaker is stating the premises for the sake of argument, but "grass is green" is not asserted and thus does not express the belief that grass is green. Rather, that grass is green is merely supposed for sake of argument. Second, if my son were to say "Look dad! I am an elephant," it seems unintuitive to think that the use of "I am an elephant" expresses the belief that he is an elephant. My son is not asserting that he is an elephant. Rather, he expresses a fantasy to the effect that he is an elephant for the sake of the game that he is playing. Third, and finally, consider the following example from Martin Montminy (2020: 373).
Tariq: Guess who I saw today? Mia: I don't know. You saw your sister.
Mia's utterance of "you saw your sister" does not seem to express the belief that Tariq saw his sister, but a guess. 9 These three examples purport to show that it is possible to use an indicative sentence without expressing a belief. Of course, not everyone will be convinced by these examples. Let us, therefore, turn to some more theoretical considerations. How should we conceive of the difference between a supposition and an assertion? It is often claimed that the illocutionary point of an assertion is to represent a situation as actual or commit oneself to the truth of something. This is in part why it seems plausible that an assertion expresses a belief. To suppose something in discussion, by contrast, is not to represent a situation as actual or to commit oneself to the truth of something. This is why it seems plausible that a supposition does not express a belief. In neither of the contexts above is the speaker representing a situation as actual or committing him or herself to the truth of something. This is why it is plausible that the use of the sentences does not necessarily function to express beliefs.
Moreover, it is often argued that the sincerity condition for a particular speech act is that the speaker is in the mental state that the speech act functions to express. Indeed, it is plausible to think that Jack is insincere if he asserts that grass is green, but does not believe it. However, it does not seem as if Julia is insincere if she supposes that grass is green in an argumentative context like the one above, but does not believe it. Rather, Julia is insincere if she does not suppose that grass is green. 10 Similar considerations apply to the examples above. This also suggests that the use of the sentences does not necessarily function to express beliefs. 11 The expressivist explanation of inconsistency requires that (3) and (4) express mental states that disagree, but the possibility of using a sentence as a supposition rather than an assertion shows that this explanation is problematic. Consider an example where Jack asserts that grass is green. Asserting that grass is green expresses the belief that grass is green. Julia, by contrast, merely supposes that grass is not green. Perhaps Julia wants to explain what a deductively invalid argument is and presents the following reasoning: "If grass is green, then it is raining. Grass is not green. Hence, it is not raining." Both intuitively, and based on the theoretical considerations advanced above, it seems that Julia's utterance of "grass is not green" does not express the belief that grass is not green. Rather, it expresses the supposition that grass is not green. Hence, the mental states that the sentences express do not disagree. This is easily seen if we consider the following: (5) Jack believes that grass is green. (6) Julia supposes that grass is not green. 9 Montminy claims that Mia's utterance is not an assertion and since it is not an assertion it does not express a belief. 10 See Green (2000) for a discussion of the sincerity conditions for supposing. 11 As many philosophers have observed, assertions can be challenged in distinctive ways. If one thinks that assertion is governed by the knowledge norm (see e.g., Williamson 2000), then it is appropriate to challenge what is said by asking how someone knows what they asserted. If one cannot answer the challenge, one should retract what one asserted. However, such challenges, even a weaker one like asking "why do you believe that," seem inappropriate in the contexts discussed above. In either case, the speaker could say "I do not believe it. I am merely supposing it." Both parties could adopt the other party's mental state without ending up with a structurally irrational combination of mental states. 12 For example, Jack could end up with the following combination of attitudes.
(5) Jack believes that grass is green. (7) Jack supposes that grass is not green.
Believing that grass is green does not disagree with supposing that grass is not green. If, on some occasion, (3) expresses the belief that grass is green, but (4) expresses the supposition that grass is not green, we cannot explain the inconsistency of (3) and (4) by virtue of the sentences expressing mental states that disagree. Although (3) and (4) can express mental states that do not disagree, (3) and (4) are still intuitively inconsistent. It thus seems that two sentences can be inconsistent despite not expressing mental states that disagree. The Fregean can explain this by arguing that if "grass is green" is true, then "grass is not green" is false. Jack and Julia express mental states that do not disagree towards inconsistent propositions. Such an explanation is not available to the expressivist.
Similar problems arise regarding the use of moral language. "Murder is wrong" is inconsistent with "murder is not wrong." If we use the sentences "assertively" we express disapproval of murder and tolerance of murder respectively. 13 However, it seems possible to use either sentence to express a mere supposition. If Simon uses "murder is not wrong" as a supposition to make a point in an argument, it does not seem as if it expresses tolerance of murder, but a supposition. 14 First, Simon is not purporting to tolerate murder. Second, Simon is not insincere by virtue of not tolerating murder. These considerations suggest, for reasons outlined above, that "murder is not wrong," in this context, does not express tolerance of murder. Nevertheless, it still seems as if "murder is wrong" is inconsistent with "murder is not wrong." Hence, the expressivist explanation of inconsistency of two sentences by virtue of the sentences expressing mental states that disagree isn't satisfactory.
At this point, it may be objected that the expression of a supposition that grass is green is a kind of pragmatic entailment that is parasitic on the sentence conventionally expressing the belief that grass is green. For example, it may be argued that expressing a supposition works like a conversational implicature. On this view, 12 Interestingly, it seems that interpersonal disagreement does not mirror intrapersonal disagreement. If Jack believes that grass is green, but Jill believes that grass is not green, they disagree. Neither party could adopt the other party's belief without ending up with a structurally irrational combination of mental states. But consider a scenario where Jack hopes that grass is green and Julia hopes that grass is not green. It is not obvious that Jack and Julia disagree in any interesting sense, but it seems that neither party could adopt the other party's attitude without ending up with a structurally irrational combination of mental states. I will not here try to explain why -although I think that the explanation probably has something to do with hope having one kind of interpersonal function and quite a different intrapersonal function. At least certain desires seem similar. If Jack desires to have only one car that is red and Julia desires to have only one blue car, they do not disagree, but neither party could adopt the other party's desire without ending up with a structurally irrational combination of attitudes. See also MacFarlane (2014) for discussion. 13 Given orthodox expressivism, moral assertions, of course, do not express beliefs.
14 Part of the problem here is that it is not obvious how to understand what it is to suppose that murder is wrong on an expressivist view. "grass is green" conventionally expresses the belief that grass is green, but given the circumstances of Julia's use of "grass is green", we realize that this is not the mental state that she intends to convey. Instead, we search for a different interpretation. In this context, we realize that what she intends to convey is a mere supposition that grass is green. 15 But why think that "grass is green" conventionally expresses the belief that grass is green in the first place?
First, it may be argued that indicative form is enough for a sentence to express a belief. As Timothy Williamson writes, "the default use of declarative sentences is to make assertions" (Williamson, 2000: 258). However, the examples advanced above seem to show that indicative form is not sufficient. Green argues that uttering "You will be more punctual in the future" can be "used as a vehicle of prediction and in another utterance as a vehicle of a command or even a threat" (Green, 2020: 351) without any change in content. Montminy claims that "[t]he existence of this type of use [i.e., using indicative sentences without making assertions] should be beyond dispute" (Montminy 2020: 373). Hence, "indicative grammatical mood … does not determine its force" (Green, 2021). 16 Second, it may be argued that considerations having to do with Moore's paradox show that an indicative sentence like "grass is green" conventionally expresses the belief that grass is green. According to Moore's paradox, uttering a sentence like the following is paradoxical or absurd: (8) Grass is green, but I don't believe it is.
Uttering (8) does indeed sound paradoxical or absurd. One explanation of the absurdity of uttering (8) is that "grass is green" conventionally functions to express the belief that grass is green. Saying "grass is green" expresses the belief that grass is green. The speaker is thus implying that s/he has the belief. The speaker is also saying that s/he does not have the belief that s/he implies having. Although not a contradiction, it is linguistically odd to imply that one believes that p, but then say that one does not believe that p.
But does this show that "grass is green" conventionally expresses the belief that grass is green? If we interpret the expression of belief as conventional, at least given a standard Gricean test, it should not be cancellable. However, it does not seem to be linguistically infelicitous to utter something like the following: (9) Okay, let's see. Grass is green. I don't really think that grass is green. Remember I am only going through the premises of this argument. Everything that is green is visible. Therefore, grass is visible. 15 See below for reasons to be skeptical about this view. 16 Moreover, if the use of an indicative sentence were sufficient for making an assertion, the expressivist idea that "murder is wrong" is not an assertion would also be jeopardized. In relation to this, consider the possibility that "Bob is hiyo" functions to accost Bob. See Dreier (1996) for discussion. See Montminy (2020) and Green (2020) for considerations that purport to show that indicative form is neither necessary nor sufficient for assertion.
Given the argumentative context, there does not seem to be anything absurd in uttering (9). Similarly, my son may say "Look dad! I am an elephant. But don't worry. I don't believe that I am an elephant." There is nothing absurd in this context either. The explanation of why there is no absurdity is because "grass is green" and "I am an elephant" are not used to make assertions in these contexts. Hence, they do not express any beliefs. Rather, the sentences are used to express other mental states.
The reason why we find (8) absurd is because the default interpretation of the use of "grass is green" is as an assertion. However, as the examples above illustrate, "grass is green," and other indicative sentences, do not always express beliefs. 17 This is enough to cause problems for the expressivist explanation of inconsistency. At this point it may be argued that I have relied too much on considerations based on speech act theory (and speech act theorists usually endorse truth-conditional semantic theories). But how do expressivists conceive of the expression relation? Gibbard, for example, provides the following characterization: "[w]hen I say he "expressed" a belief, I don't mean that he has the belief. To express a state of mind, as I use the term, is to purport to have it, whether or not one does" (Gibbard, 2003: 77). When Jack asserts that grass is green, he purports to have a particular belief. However, when Julia supposes that grass is not green, she is not purporting to believe that grass is not green. She is purporting to suppose it for sake of argument. Hence, Gibbard's view does not seem to help the expressivist out of the predicament.
Of course, it may be argued that Gibbard's characterization of expression is problematic. 18 Instead, consider what Schroeder calls assertability expressivism. On this view, "to say that a normative utterance expresses a non-cognitive state of mind is to say that it is semantically correct to assert that sentence only if you are in that state of mind" (Schroeder, 2015: 53). The idea in the background is that "[e]very sentence in a language is associated with conditions in which it is semantically correct to use that sentence assertorically" (Schroeder, 2015: 49) where these conditions are useconditions rather than truth-conditions. "The rule governing 'grass is green' is […] to assert it when you believe that grass is green (Schroeder, 2015: 51). Similarly, the rule governing "murder is wrong" is to assert it only if you disapprove of murder.
I do not think this view solves the problem either. Even if the rule governing "grass is green" is to assert it only if you believe that grass is green, it seems plausible that different speech acts come with different rules. 19 For example, the rule governing "grass is green" is to suppose it for sake of argument in discussion only if you suppose that grass is green. Voicing the supposition that grass is green is not semantically incorrect if the speaker does not believe that grass is green. It is semantically incorrect if the speaker does not suppose it. When Julia puts "grass is not green" forward as a supposition for the sake of argument, she is not expressing the belief that grass is not green. Her utterance expresses the supposition that grass is not green. However, the literal meaning of "grass is not green" does not thereby seem to change. What changes is the force it is presented with. This also suggests that the expression of a belief is not explained by the conventional meaning of "grass is green." If one can use "grass is green" to express a supposition, then it seems that "grass is green" does not express the belief that grass is green by virtue of its conventional meaning.
Finally, it may be argued that it is enough that a sentence typically expresses a particular mental state. "Grass is green" and "grass is not green" are inconsistent because the sentences are typically used to express beliefs that disagree. However, it is far from obvious that this will suffice. If the expressivist wants to explain what it is for two sentences to be inconsistent, we want an explanation of why two sentences are inconsistent, not just typically, but always. "Grass is green" and "grass is not green" are always inconsistent, but as argued above, the sentences do not always express mental states that disagree. Hence, that two sentences typically express mental states that disagree is not enough to explain why two sentences are always inconsistent.
Given the observation that inconsistent sentences do not always express mental states that disagree, the expressivist may weaken his or her view. First, the expressivist may weaken the scope of the proposal. If "grass is green" can be used to express the supposition that grass is green, rather than the belief that grass is green, the expressivist may claim that they are proposing a theory of inconsistency not between sentence types, but between sentence + force types. The problem with this proposal is that "grass is green" and "grass is not green" are not merely inconsistent when asserted. Rather, the sentences are inconsistent regardless of whether the sentences are asserted or not. In other words, inconsistency is a feature of sentence types.
Second, the expressivist may weaken the force of their view. Instead of arguing that two sentences are inconsistent if they express mental states that disagree, they may opt for a modal thesis: Modal thesis: Two sentences are inconsistent if they can express mental states that disagree.
It may be argued that this thesis makes the right prediction for "grass is green" and "grass is not green." Although the sentences do not always express beliefs that disagree, they can express such beliefs. However, the modal thesis seems to yield the wrong predictions. Consider the following two sentences.
(10) I believe that grass is green. (11) Grass is not green.
(10) and (11) are not inconsistent. However, (10) and (11) can be used to express beliefs that disagree. Although (10) reports that the speaker believes that grass is green, it can also be used to express the belief that grass is green, but this does not make the sentences inconsistent. For this reason, it seems that the modal thesis does not provide a simple explanation of why two sentences are inconsistent. 4: Inconsistent contents without disagreement. Although the considerations above are not exhaustive, the expressivist idea that two sentences are inconsistent by virtue of expressing mental states that disagree is problematic. This problem, moreover, does not merely pertain to moral sentences, but also to non-moral sentences. However, the problem arguably runs even deeper.
Not only can we not adequately explain why two sentences are inconsistent by virtue of expressing mental states that disagree, but this suggestion also fails to explain why two attitudes that do not disagree can have inconsistent contents. Consider, again, the following example.
(5) Jack believes that grass is green. (6) Julia supposes that grass is not green.
Both parties could adopt the other party's mental state without it giving rise to any structural incoherence. However, the mental states have inconsistent contents. This is no less true for judging that murder is wrong and supposing that murder is not wrong. If the expressivist idea is that inconsistency in thought (and talk) is explained in terms of mental states that disagree, then the explanation seems problematic. Jack and Julia's mental states do not have any inconsistency feature, but there is nothing else that the expressivist can appeal to. As Gibbard writes, an expressivist explanation cannot "rest on a prior understanding of consistency in thoughts, propositions, or other items of mental content" (Gibbard, 2012: 275).
The Fregean, by contrast, explains inconsistency quite independently of disagreement. Jack and Julia do not disagree, but they nevertheless have mental states directed towards propositions that cannot be true simultaneously. As Schroeder claims, "grass cannot be green and not green" (Schroeder, 2008: 39). This explains why Jack and Julia have attitudes towards inconsistent contents. Gibbard agrees that it is obvious that grass cannot be green and not green, but he also thinks that "the behaviors of disagreement and exclusion are obvious. Our question is which obvious things to explain by which other obvious things." (Gibbard, 2012: 277). Even if we suppose that (Fregean) inconsistency and disagreement are equally obvious, I do not think that Gibbard's diagnosis is quite right.
If the expressivist idea is that inconsistency (in thought or talk) can and should be explained by mental states that disagree, then it does seem as if one obvious thing would have to explain another obvious thing, but this does not seem to be the case. Jack believes that grass is green. Julia supposes that grass is not green. These mental states do not disagree, but Jack and Julia nevertheless have attitudes towards inconsistent contents. It, therefore, does not seem as if one obvious thing (disagreement between mental states) explains another obvious thing (inconsistent contents).
The Fregean does not purport to explain disagreement in terms of inconsistency. Two sentences are inconsistent if they express propositions that cannot be true simultaneously, but this does not explain disagreement. In order to explain why two mental states disagree, the Fregean needs an independent understanding of why two mental states (with inconsistent contents) disagree. For example, to believe that grass is green and believe that grass is not green is a structurally irrational combination of mental states. Believing that grass is green and supposing that grass is not green is not. The contents are nevertheless the same in the two examples. Even if inconsistency is obvious, this does not, by itself, explain why two mental states disagree. Hence, one obvious thing does not explain another obvious thing.
Gibbard also claims that the expressivist can offer an explanation that is structurally identical to the Fregean one.
Instead of a function that takes, say, pieces of content to states of hope -taking, for instance, the content IT WON'T RAIN TODAY to the hope that it won't rain today -we have a function that takes states of belief to states of hope. It takes the state of believing that it won't rain today to the state of hoping that it won't rain today. (Gibbard, 2012: 290) The same kind of explanation is supposed to apply to other contentful states -they are all "treated in relation to belief states" (Gibbard, 2012: 290).
Thus, like the Fregean, the expressivist calls on a single function to schematize all hopes. Each kind of theorist, of course, owes an explanation of how her favored function works, but in using a single function for each type of "propositional attitude" such as hope, the two approaches with each other tie in their simplicity and fecundity. (Gibbard, 2012: 290) Again, I do not think that Gibbard is right. Consider how Peter Pagin and Neri Marsili describe the formation of a belief (or judgment) on Frege's view. "A judgment … is a step from entertaining a Thought to acknowledging its truth. A subject first merely thinks the Thought that p, and then, at the judgment stage, moves on to acknowledge it as true" (Pagin and Marsili 2020). This explains three things. First, it explains how a thought, or proposition, comes to be believed. Second, it explains how a proposition can be the object of different attitudes. Instead of acknowledging the truth of the proposition, one may desire it, e.g., if one finds it attractive, hate it, if one fits it repulsive, or merely suppose it for the sake of argument. Third, it explains the sense in which two attitudes that do not disagree can have inconsistent contents. To believe that p is to acknowledge the truth of p. To suppose that not-p is not an attitude that disagrees with believing that p, but the proposition that p is still inconsistent with the proposition that not-p. Even if Gibbard's suggestion does not concern how a belief or a supposition is formed, it is nevertheless a question that an expressivist should have an answer to. However, since Gibbard does not examine the function from beliefs to other contentful mental states, it is not at all obvious how different attitudes can have the same content or why two attitudes that do not disagree nevertheless can be inconsistent. 20 If the expressivist explanation is supposed to be similar to the Fregean explanation, one interpretation is the following. On the Fregean view, we can think of the function as follows: different contentful mental states, e.g., believing that p, hoping that p, supposing that p, and so on, have the proposition that p as a formal part. Similarly, on the expressivist view, we can think of the function as follows: different contentful mental states, e.g., hoping that p, supposing that p, and so on, have the belief that p as a formal part. Hence, hoping that p contains, as a formal part, the belief that p in the following sense: "S hopes that p" is equivalent to "S hopes that S believes 20 Gibbard spends roughly half a page examining preference and indifference. Preference is a function from "two belief states, self-attributing going left and self-attributing standing still. A single function then takes pairs of such self-attributions to states of mind. Preference is […] an attitude towards two conceptually possible self-attributions" (Gibbard, 2012: 291). It is not obvious how this explains how different attitudes can have the same content or why two attitudes that do not disagree can have inconsistent contents. that p." To suppose that not-p, by contrast, contains, as a formal part, the belief that not-p in the following sense: "S supposes that not-p" is equivalent to "S supposes that S believes that not-p." Hoping that grass is green and supposing grass is not green do not disagree, but their formal parts do. This, it may be argued, explains why the two mental states do not disagree, but nevertheless, in virtue of their formal parts, are inconsistent. This would make beliefs play roughly the same role as Fregean propositions. 21 However, the idea has some rather implausible upshots. It would, for example, imply that x's hope that p can be fulfilled simply by acquiring the belief that p. If I hope that my team wins the game, my hope won't be satisfied simply by forming the belief that my team won.
On Gibbard's view, it is not obvious in what sense an attitude, e.g., a hope or a supposition that p, is a function from the belief that p, or in what sense a belief that p and a supposition that not-p have inconsistent attitude without any attitudes that disagree. On the Fregean view, by contrast, this is relatively easy to understand. Note that I am not saying that the expressivist explanation does not work. However, until the expressivist offers a more thorough explanation of the function and how this explains inconsistency in content, the Fregean and the expressivist explanations are not on par regarding simplicity and fecundity. Until such an explanation is advanced, explaining inconsistency, on expressivist premises, remains a challenge.

Concluding remarks
The problem for the expressivist is not, pace Schroeder, that they lack an explanation of why disapproval of murder and tolerance of murder disagree. I think that the expressivist can explain disagreement in belief and disagreement in attitude in a uniform matter, e.g., in terms of functional roles. The problem, pace Gibbard and Baker and Woods, is that inconsistency in thought and talk is not easily explained by disagreement. Two sentences, e.g., "grass is green" and "grass is not green" are intuitively inconsistent, but they do not necessarily express mental states that disagree. Similarly, believing that grass is green and supposing that grass is green are not mental states that disagree, but they nevertheless have inconsistent contents. Explaining inconsistency in terms of disagreement therefore seems problematic.
A simple diagnosis of the problem has to do with the orthodox expressivist theory of meaning. The meaning of a sentence is determined and consists in the mental state that the sentence functions to express, but one and the same sentence can be used in different ways. Indeed, the problem of distinguishing disagreement and inconsistency can be interpreted as another facet of the Frege-Geach point. Peter Geach famously argued that "a proposition may occur in discourse now asserted, now unasserted, and yet be recognizably the same proposition" (Geach, 1965: 449). Much focus has been on contexts where a sentence is embedded, but it seems that any indicative atomic sentence, e.g., "grass is green" or "murder is always right" can be used to express a supposition rather than a belief or some kind of approval, but the sentences are nevertheless inconsistent with "grass is not green" and "murder is always wrong" despite not expressing mental states that disagree. Again, I am not claiming that the expressivist cannot solve this problem. 22 Rather, I am merely suggesting that it is not obvious that the inconsistency between two sentences or two contents can and should be explained by virtue of (two sentences expressing) mental states that disagree. This it is an issue that the expressivist needs to address more explicitly. How expressivists can and should conceive of inconsistency in thought and talk therefore remains a challenge.