Is relativism a coherent thesis? Donald Davidson’s classic ‘On the Very Idea of a Conceptual Scheme’ (1974), gives a negative answer. Davidson sought to undermine the coherence of the thesis of relativism, which Davidson takes to be thesis that alternative conceptual schemes are possible. Despite much confusion about Davidson’s argument, the master argument of the paper is surprisingly straightforward, with a radical impossibility result following from two easily specifiable, albeit controversial, premises. The first premise interprets relativism and alternative conceptual schemes in terms of the notion of an untranslatable truth. The second premise charges the notion of an untranslatable truth with incoherence. The radical impossibility result is that the thesis of relativism is not only false, but incoherent.

I argue against Davidson and for the view that relativism is a coherent thesis, as follows. (1) Davidson’s master argument against the coherence of relativism is invalid but can be patched to be valid. (2) Formulating the thesis of relativism requires being careful about alternative conceptual schemes and their attribution. Relativism should not be understood as the thesis that alternative conceptual schemes are possible, but instead should be understood as a thesis about how certain attributions are possible, attributions according to which thinkers interpret other thinkers as saying and thinking things that make use of alternative conceptual schemes. (3) The patched argument is shown to be unsound through a consideration of different kinds of semantics for saying and attitude attributions. (4) The overall result is that relativism is a coherent thesis even if the “very idea” of an alternative conceptual scheme is not.

Before getting started, let me clarify the relation between the present discussion and the recent wave of work Kölbel (2002, 2004, 2015a, 2015b), MacFarlane (2005, 2014), and Lasersohn (2005, 2016), amongst others on relativism. The views of these philosophers are strictly speaking independent of the view being defended here, according to which relativism is a coherent thesis even if the very idea of an alternative conceptual scheme is not. Broadly speaking, these philosophers challenge Davidson’s view that the very idea of an alternative conceptual scheme is incoherent by developing what are supposed to be formally and philosophically coherent accounts of perspectival content and relative truth. Although it is no part of the argument of this paper, I believe that Davidson is correct that the notion of an alternative conceptual scheme is incoherent, and so I believe that some incoherence attends the views of contemporary relativists. However, this claim is separate from and independent of the argument of the current paper, the conclusion of which is conditional: relativism is a coherent thesis even if Davidson is right that the “very idea” of an alternative conceptual scheme is incoherent.¹ I assume the antecedent of this conditional throughout.

Finally: Davidson, again? Yes. Davidson’s ideas about relativism are, if the argument of this paper is correct, still not well understood. Further, understanding them better serves to situate Davidson and relativism in a new setting, one that highlights new ideas: about the explanatory ambitions of relativism, about the role of attitude attributions and their semantics in defending the coherence of relativism, and about a connection, glimpsed but not properly deployed by Davidson, between relativism and nonsense.² Although Davidson’s paper is not new, the analysis of it and its significance as articulated here, is.

1 Davidson’s Master Argument, Patched

Davidson’s master argument runs as follows:

P1 The thesis of relativism invokes the notion of an untranslatable truth.

P2 The notion of untranslatable truth is incoherent.

C The thesis of relativism is incoherent.³

Let me briefly indicate some textual basis for this way of construing Davidson’s master argument.

Premise 1

In his presentation of the issues in his paper, Davidson takes the thesis of relativism to be the thesis that alternative conceptual schemes are possible.⁴ Davidson considers a number of proposals to make sense of the idea of an alternative conceptual scheme and finds them either to be dead ends or to fail to “improve the intelligibility while retaining the excitement” (Davidson 1974, 183).⁵ After these false starts, Davidson concludes that the most promising (but ultimately not promising enough) way of understanding conceptual schemes runs via the proposal, floated earlier in the paper, to construe conceptual schemes as sets of intertranslatable languages. He writes:

[A] We may identify conceptual schemes with languages, then, or better, allowing for the possibility that more than one language may express the same scheme, sets of intertranslatable languages. [B] Languages we will not think of as separable from souls; speaking a language is not a trait a man can lose while retaining the power of thought. So there is no chance that someone can take up a vantage point for comparing conceptual schemes by temporarily shedding his own. [C] Can we then say that two people have different conceptual schemes if they speak languages that fail of intertranslatability? (Davidson 1974, 185)

The initial thought ([A]) is to construe conceptual schemes as languages; however, given that different languages can express the same conceptual scheme, conceptual schemes should be distinguished from languages, and identified with sets of intertranslatable languages. A consequence of this is that ([C]) that when languages fail of intertranslatability, they express distinct conceptual schemes. This proposal is almost the view that is the target of Davidson’s critical argument, but it fails to make one significant point explicit, viz., about the connection to the concept of truth. But this emerges naturally enough from situating the proposal about alternative conceptual schemes in the setting of Davidson’s larger philosophy.

Before getting to that, though, note that it is not clear in this passage what the relevance is of ([B]), of the claims that languages are not “separable from souls” and of there being no chance of taking up a vantage point for comparing conceptual schemes by shedding one’s own. I suggest an interpretation of this idea below when I suggest an account of relativism that does not involve commitment to alternative conceptual schemes.

The proposal that conceptual schemes are different iff languages that express them fail of intertranslatability has the merit, for Davidson, of connecting up via the notion of translation with his interpretivist apparatus.⁶ The radical interpretation of a speaker/thinker for Davidson is part of a constitutive account of mind and meaning and culminates in attributions of saying and belief (and other attitudes, although only saying and belief will be relevant here) to the speaker/thinker, as well as in a theory of meaning for the speaker/thinker’s language. Interpretation proceeds in accordance with charity principles that broadly speaking require imposing a rational pattern on sentences held-true, under the conditions under which they are held-true, by speakers. Together the charity principles and the sentences held-true form the constitutive basis for attributions of mind and meaning.⁷

The interpretivist apparatus overall links language, mind, rationality, and truth. Davidson does not fully exploit the connections to his interpretivist apparatus in his argument against relativism, but crucial for his argument is the connection between translation and truth. Davidson makes this connection explicit by adding to the basic idea of identifying conceptual schemes with sets of intertranslatable languages the requirement that the different languages should be capable of expressing truths. The net result is that, on Davidson’s proposal, there are alternative conceptual schemes when there are possible languages L and L* capable of expressing truths, but which are not intertranslatable. Conceptual relativism thus invokes the notion of an untranslatable truth in its understanding of the notion of an alternative conceptual scheme, which is what P1 of the master argument claims.

Premise 2

However, Davidson thinks, and this is P2 of the master argument, that the notion of an untranslatable truth, and thus also that of an alternative conceptual scheme, is incoherent. The argument that Davidson deploys at this point derives from the link in Davidson’s interpretivist account between the theory of meaning and mental content and the theory of truth.⁸ The argument is that on the correct account of truth—in which the concept of truth is a semantic concept to be understood through its role in the theory of meaning and for which Tarski’s (1944) Convention T plays a central defining role—the notion of an untranslatable truth is incoherent. Here is how Davidson puts his argument:

[Alfred Tarski’s] Convention T suggests, though it cannot state, an important feature common to all the specialized [language-specific] concepts of truth. It succeeds in doing this by making essential use of the notion of translation into a language we know. Since Convention T embodies our best intuition as to how the concept of truth is used, there does not seem to be much hope for a test that a conceptual scheme is radically different from ours if that test depends on the assumption that we can divorce the notion of truth from that of translation. (Davidson 1974, 194–95)

Davidson’s point is that understanding the attribution of truth to a sentence requires being able to translate the sentence into a language one oneself understands, and so there is no prospect for understanding relativism through the notion of a truth, but where it applies to sentences that are not translatable into one’s own language.

Whether this is a good argument or not for P2 is not our main concern.⁹ The more important point for now is that, whether or not Davidson’s P2 is objectionable, I will not be objecting to it now. Even granting both P1 and P2, Davidson’s master argument is problematic, logically: it is invalid. That’s not the end of the story though, because invalid arguments can be made valid by adding premises—like P3.

Premise 3

I claim that Davidson’s argument is invalid. How could this be? How could it be that a thesis invokes a notion, that the notion is an incoherent notion, yet the thesis is not incoherent? This is a great question that we will address at length. But showing that an argument is invalid does not require answering such a question; it requires only that there be some such question to ask. But now that we have the question, we can make the argument valid by closing off the possibility that the question identifies, by adding a premise. The additional premise is:

P3 If a thesis invokes an incoherent notion, the thesis is incoherent.

This is progress. The argument as it stood was, it is now clear, invalid. Adding this premise patches the hole in the argument and makes it valid.

But is the argument also sound? This is a harder question. As we shall see, answering it will require going into issues about the semantics of saying and attitude attributions, a discussion that itself will be occasioned by a connection between relativism and attitude attributions that it is the concern of the next section to establish. More specifically, there are issues about alternative schemes and their attribution that, I will argue, need to be considered before we can endorse, or not, P3. Once we consider these issues, we will be in much better position to assess whether the patched master argument, with P3, is sound.

2 Alternative Schemes and Their Attribution

What do issues about saying and attitude attributions have to do with relativism? The answer begins with recognizing that alternative conceptual schemes are, in the first instance, explanantia and not explananda.

Alternative conceptual schemes are an explanatory posit, invoked to explain a certain intersubjective epistemic phenomenon. Call this explanandum

The intersubjective situation in relativism

The situation in which parties to an apparent disagreement who possess no epistemic advantage over each other, respond with indifference and detachment to the difficulties encountered in trying to understand and resolve their apparent disagreement.¹⁰

The intersubjective situation in relativism describes how minds are configured and related (or not) to each other in the kinds of situations that philosophers and others are tempted to invoke relativism—when a thinker is trying to make sense of “the world-views of different cultures, shifts in scientific paradigms, or differences of ethical outlook” as Bernard Williams (1974, 215) puts it. The idea is that in these kinds of cases, thinkers take themselves to disagree, recognize that their disagreement is intractable, and then respond with indifference and detachment instead of skepticism and further engagement. It is important that this indifference and detachment is not the result of posturing, or recalcitrance to engage, for, we assume in the case envisioned, both parties are trying to determine the truth for shared reasons. But they come to view their disagreement as rationally exhausted, with no plausible lines of resolution, and this is what leads to indifference and detachment.

Enter relativism. Alternative conceptual schemes are supposed to function as part of an explanation of the intersubjective situation in relativism. In this explanation, the theorist posits the existence of alternative conceptual schemes and parties to the apparent disagreement are attributed saying and mental attitudes that make use of these alternative conceptual schemes. According to this view, the relativist thesis is:

Attributed Relativism

There are alternative conceptual schemes and parties to certain apparent disagreements are saying and thinking things that makes use of these alternative conceptual schemes.

In Attributed Relativism, the theorist holds that there are alternative conceptual schemes and attributes to parties of certain apparent disagreements saying and thinking that make use of these alternative conceptual schemes. This, if coherent, is a possible explanation of the intersubjective situation in relativism. According to Davidson, however, this is incoherent.

But this is not the only possible explanation, and not even the only possible explanation that invokes the notion of an alternative conceptual scheme. A less committing explanation does not posit the existence of alternative schemes, and attribute them to thinkers, but instead recognizes the existence of attributions of saying and thinking by parties to the apparent disagreement that make use of alternative conceptual schemes. This kind of relativism is not attributed by the theorist, but is still attributional, in that the theorist attributes or reports attributions of saying and thinking that make use of alternative conceptual schemes by and to parties to an apparent disagreement. According to this view the relativist says:

Attributional Relativism

There are attributions by parties to certain apparent disagreements in which the parties attribute to each other saying and thinking that involves the use of alternative conceptual schemes.

Attributional Relativism says that there are attributions by parties to certain apparent disagreements in which they attribute to each other saying and thinking that uses alternative conceptual schemes.

Let me make some points about Attributional Relativism and its relation to Attributed Relativism.

The first point is that Attributional Relativism makes sense of what Davidson said in the passage [B] above—about how there is “no chance” that one can take a vantage point on another conceptual schemes by shedding one’s own. Davidson seems to be saying that there is no neutral point from which a thinker can conceptualize and evaluate an alternative conceptual scheme. However, this is a point that would seem to apply to parties to an apparent disagreement who are participating in the intersubjective situation in relativism and not to the theorist of relativism. The theorist is not trying to achieve any special vantage point on other conceptual schemes, but only to describe and attribute them; moreover, there is no reason to think that the theorist should have a problem accessing alternative conceptual schemes. However, the parties to the apparent disagreement do have this problem, since they are participants in the intersubjective situation in relativism. So, I submit that Davidson was confused about whether the theorist attributes alternative conceptual schemes or instead reports or attributes attributions of saying and thinking that make use of alternative conceptual schemes by and to parties to an apparent disagreement.

Next note that Attributional Relativism does not commit the theorist to the existence of alternative conceptual schemes; nor is there any attribution of saying and thinking that make use of alternative conceptual schemes; nor is the coherence of the very idea of a conceptual scheme assumed. There are, however, reportings or attributions of attributions of saying and thinking that make use of alternative conceptual schemes by and to parties to apparent disagreements.

I will say more on the contrast between Attributed and Attributional Relativism, but since this is all so terribly abstract, it will be useful to have an example to work with.

Consider, then, Liberal and Libertarian (these are people) who have what initially looks like a disagreement about whether justice permits governments to impose substantial restrictions on individual liberty to secure a certain extent of some public good. Liberal and Libertarian are interested in the general question and how it applies in different cases—to lockdowns, gun control, drug prohibition, reduced speed limits, and elsewhere. Liberal thinks that even severe restrictions, though unfortunate, are permissible; Libertarian disagrees. Suppose further their disagreement is in good faith—both Liberal and Libertarian are trying to get at the right answer for the right reasons and are not at all concerned in digging in or grinding axes because some view is their own. They are prepared to learn from, or instruct, the other, if they can.

Suppose, though, that on further reflection both Liberal and Libertarian come to worry that theirs is no ordinary disagreement and wonder whether they really understand what the other is saying and thinking. Maybe they assumed at the outset that they are speaking a common language, but at this stage, they leave this assumption behind and move into an interpretive mode in which certain assumptions about shared meanings and beliefs must be suspended for Liberal and Libertarian to make sense of each other, and so in which interpretation is a process of reconstructing what the other means and believes.¹¹

Liberal and Libertarian can be in such a state without their situation inviting or requiring relativism; at this stage, they could be engaged in a deep disagreement, the depth of which required them to move into this interpretive mode.¹² However, imagine further that interpretation proves impossible; they can assign no coherent sense to each other’s words or minds on this point. As they might put it, they just can’t understand what the other is thinking and how they have come to their view, despite lengthy and good faith discussion aimed at getting at the truth. So, they come to the following intersubjective situation in their disagreement:

Liberal believes that Libertarian said that . . . . . . and expresses her belief that . . . . . . with ‘Justice does not permit the severe restriction of individual liberties for achieving significant measures of a public good’,

where the ‘. . . . . .’ indicates that Liberal has no sentence of her language which she can use to make sense of what Libertarian says or believes. Mutatis mutandis for Libertarian.

It is here that the relativist thought comes in: despite (we assume) this principled and not merely contingent failure of interpretation, Liberal and Libertarian accept that the other is saying something and believing something, capable of truth, but by making use of an alternative conceptual scheme. So, the situation is as follows:

Relativist Attribution on Attributional Relativism

Liberal believes that Libertarian said that [justice does not permit the severe restriction of individual liberties for achieving significant measures of a public good]_Libertarian and expresses her belief that [justice does not permit the severe restriction of individual liberties for achieving significant measures of a public good]_Libertarian with ‘Justice does not permit the severe restriction of individual liberties for achieving significant measures of a public good’.

Here, ‘[justice does not permit the severe restriction of individual liberties for achieving significant measures of a public good]_Libertarian’ (and the corresponding ‘[justice permits the severe restriction of individual liberties for achieving significant measures of a public good]_Liberal’) are not only unintelligible for Liberal and Libertarian, respectively, but, prefixed with ‘that’, purport to refer, in the attribution, to elements of the alternative conceptual schemes that Libertarian and Liberal, respectively, use. However, because the idea of an alternative conceptual scheme is incoherent, and hence any concepts that are supposed to be part of such schemes non-existent, the parties to the apparent disagreement are referring to nothing, and attributing nonsense to each other.

The key here is that what the relativist is claiming is that this kind of scenario is coherent. The relevant interpretive situation involves this kind of double embedding and does not take the form of attributions by the theorist of saying and thinking that makes use of alternative conceptual schemes. The attributions do not have the theorist attributing alternative conceptual schemes, like this:

Relativist Attribution on Attributed Relativism

Liberal said that [justice permits the severe restriction of individual liberties for achieving significant measures of a public good]_Liberal and expresses her belief that that [justice permits the severe restriction of individual liberties for achieving significant measures of a public good]_Liberal with ‘Justice permits the severe restriction of individual liberties for achieving significant measures of a public good’.

The attributions made by the relativist are at one level remove from this.

Let me return to the comparison between Attributed and Attributional Relativism. I argue that if Attributed Relativism explains the intersubjective situation in relativism, so will Attributional Relativism, and that Attributional Relativism will explain the intersubjective situation even where Attributed Relativism does not. Let’s take these points in reverse order.

First, Attributional Relativism is not committed to the coherence of the idea of an alternative conceptual scheme, and so not committed to the existence of alternative conceptual schemes, and so not committed to the existence of the concepts that are supposed to make up alternative conceptual schemes. These concepts may be putative and non-existent—literally non-senses—as far as Attributional Relativism is concerned. Attributed Relativism, on the other hand, is committed to the existence of alternative conceptual schemes and so to the coherence of the very idea of a conceptual scheme. So Attributional Relativism has this advantage. However, vindicating Attributional Relativism requires explaining whether and how the reporting by the relativist of attributions of the (under these assumptions) nonsensical thinking by and to parties to an apparent disagreement is not itself a kind of nonsense (Section 3 below).

However, it might be wondered: how can it be correct that parties to the apparent disagreement are attributing nonsensical thinking to each other? My point is not that this or other examples should be analyzed like this, in relativist terms. We are trying to make sense of what relativism is if the very idea of an alternative conceptual scheme is incoherent. It is no part of the present argument to argue for or advocate for a relativist treatment in this or any specific case. I use the case of Liberal and Libertarian as representative of the kind of case philosophers and others are tempted to invoke relativism, not as a case in which relativism actually applies. My point is rather that, if a relativist treatment of this case is appropriate, and if there are no alternative conceptual schemes because the very idea of them is incoherent, then this is indeed what the parties to the apparent disagreement are doing: attributing nonsense to each other. Note, though, that this is not how parties to the apparent disagreement would understand what they are doing. As I have emphasized, parties to an apparent disagreement who transition into interpretive mode but fail to turn up any stable and comprehensible sense of what the other is up to will automatically assume that there is some thinking going on—thinking with an alternative conceptual scheme. But if alternative conceptual schemes do not exist, then the attributions are attributions of nonsense.

Second, as I have described it, what needs explaining is the intersubjective relation that holds between the parties to the apparent disagreement, and this intersubjective relation is both reflective and metarepresentational. It is the result of parties to the apparent disagreement trying to make sense of each other for purposes of learning from each other, if possible. And now the point is that Attributed Relativism will not explain the intersubjective situation in relativism in the circumstances in which the parties to the apparent disagreement are saying and thinking things that make use of an alternative conceptual scheme unless the parties to the apparent disagreement in some sense take this to be the case. For if they took things as though it were cognitive business as usual and had no sense that anything like alternative conceptual schemes were operative in the disagreement, it would be hard to understand how Attributed Relativism and there actually being alternative conceptual schemes could function to explain the intersubjective situation in relativism, since it is hard to see how there would be any such situation. But if the intersubjective situation in relativism were to be explained by Attributed Relativism, it would also be explained by Attributional Relativism, which cleaves closer to the reflective and metarepresentational nature of the situation. The thinkers in that case could report themselves as being indifferent or detached from the judgment of the other party to the disagreement because they take them to be saying and thinking things using alternative conceptual schemes.

The possibility of thinking about relativism as Attributional Relativism shows not only that there is a different explanation of the intersubjective situation in relativism from the one that Davidson and almost everybody else who takes relativism seriously considers, but that, while still invoking the notion of an alternative conceptual scheme, provides an explanation that is more conservative in its commitments, while being at least as explanatory. This suggests that relativism could be understood this way.

So let’s take stock. We found earlier that Davidson’s master argument against the coherence of relativism is invalid, but that it could be patched by adding an additional premise, P3, to render it valid. This then allowed us to attend to the question of whether the master argument is sound, by attending to the question of whether the additional premise is true. Now we have a much better handle on how to address this question. We now understand in detail the kinds of claims that the theorist of relativism wants to make. These claims invoke incoherent notions on the assumption that the idea of an alternative conceptual scheme is incoherent. Whether P3 is true or not will thus depend on whether the kinds of claims that the theorist of relativism wants to make, are, by dint of the fact that they invoke incoherent notions, also rendered incoherent. If so, the master argument is sound; if not, the master argument is unsound.

3 On the Soundness of the Master Argument

I want now to explain why the patched Davidsonian master argument is unsound—why P3 is false. Although something like P3 may be true for theses that do not introduce propositional attitude contexts, I argue on semantic grounds that it is false for theses that do introduce propositional attitude contexts. If alternative conceptual schemes figure as they do in Attributional Relativism, then relativism can be a coherent thesis even if the idea of an alternative conceptual scheme is not.

My argument begins by considering three positions on embedding nonsense in attitude reports. Our overarching question is: are attributions of nonsensical thinking themselves sensical?

According to Cora Diamond (2000), the answer is ‘no’. Diamond’s discussion is set in the context of explaining the semantic background for her “resolute” reading of Wittgenstein’s pronouncements at the end of the Tractatus according to which its propositions are nonsensical. She writes:

if you cannot make sense of the sentence “God is three persons” then you can say that Smith uttered the words “God is three persons” and you can say that he uttered them with the intonation of asserting something, but you cannot say of him that he said that God is three persons. “Smith said that p” is itself nonsense unless what we put for “p” makes sense. (Diamond 2000, 156)

According to Diamond, nonsense embedded in an attribution makes the whole attribution nonsense. Diamond’s view makes use of the intuitive idea that the meaningfulness of the attitude attribution requires the meaningfulness of the complement clause of the attitude attribution; or put the other way, if the complement clause is meaningless, then the attitude attribution itself is meaningless.

Ofra Magidor’s (2013) view in effect modus tollenses Diamond’s modus ponens. Magidor argues that propositional attitude attributions that embed putatively meaningless category-error involving sentences can themselves be meaningful;¹³ and that this shows that, in these cases, the embedded putative nonsense is not nonsense but actually possessed of genuine sense or meaningfulness. While Magidor’s view opposes Diamond’s, it does not support the point I am trying to make against Davidson, namely that complexes that embed genuine nonsense need not themselves be incoherent or nonsensical. Indeed, Magidor gives arguments against this possibility. We will return to those arguments below.

A third position, intermediate, between Diamond and Magidor, holds that nonsense can be embedded without making the larger embedding whole itself nonsense. Sorensen (2002) argues that Mates’s (1950) puzzle can be used to show that meaningless beliefs are possible.¹⁴ On Sorensen’s view, complexes that embed genuine nonsense need not themselves be nonsense: meaningful complexes can embed genuine nonsense. This intermediate position is the position I want to defend.¹⁵

I will now argue against Magidor and Diamond through a consideration of the semantics of attributions of nonsensical thinking. I will consider three different accounts of attitude attributions.¹⁶ I will argue that the Magidor and Diamond views are on safe ground on propositionalist accounts of attitude attributions, but that their views are less plausible on sententialist, and, especially, on Fregean accounts of attitude attributions.

On a propositional account, an attitude attribution is true iff the attributee stands in the relevant propositional attitudinal relation to a proposition, namely the proposition expressed by the complement clause. Suppose now, as Magidor holds, such attitude attributions are meaningful and potentially true. When they are, they are true in virtue of an attributee standing in a relation to a proposition. But if the complement clause in some such attribution is meaningless it will not express a proposition, in which case the attribution cannot be true, since it is true iff the attributee stands in the attitudinal relation to the proposition and there is no proposition. So if the attribution is true, the complement must be meaningful. Further, assuming the meaning of the whole is composed of the meanings of its constituents, if the attribution is meaningful the complement clause must too be meaningful (compare Magidor 2013, 63). Suppose now, as Diamond holds, the relevant complement clauses are meaningless. They, then, fail to express propositions, in which case, using the same assumption about the meaning of the whole being composed of the meanings of its parts, the attribution itself will be meaningless.

So, the propositionalist view supports both Magidor’s and Diamond’s view. Diamond’s modus ponens is Magidor’s modus tollens. Neither view, though, supports the intermediate position.

Things are, however, more complicated on the sententialist account. In her argument against a Davidsonian-styled sententialist defence of the intermediate position, Magidor notes that any such account “must somehow account for the fact that we can truly ascribe a propositional attitude to an individual even if that individual uses different words to express their own attitude.” To do this,

such views of propositional attitudes must assume some kind of relation between the utterances of the ascriber and the utterances or thoughts of the ascribee: a relation such as synonymy or having the same content or, following Davidson, of ‘being such as to make the ascriber and ascribee same-sayers’. But suppose that my utterance of ‘The theory of relativity is eating breakfast’ … merely has the same meaning as or the same content as (rather than being an exact repetition of) John’s original utterance. If my utterance and John’s have the same meaning or the same content, it suggests that both utterances are meaningful. (Magidor 2013, 65)

So Magidor’s argument is that meaningful and potentially true attributions require samesaying, and samesaying requires saying, and saying, Magidor suggests, requires the meaningfulness of the utterances that samesay, and so requires the meaningfulness of the sentence in the complement clause the utterance tokens.

Although Magidor’s suggestion is natural, it is not inevitable, and moreover a different conclusion is possible on grounds that emphasize Davidsonian (Davidson 1968) and more broadly sententialist antipathy towards primitive synonymy relations (Quine 1960; Field 2017).

First, note that what Magidor suggests is not inevitable because two utterances could match in content by both being meaningless or nonsense. For example, although Heidegger said ‘Das Nichts nichtet’ and not ‘the nothing nothings’, we can report this by saying ‘Heidegger believes that the nothing nothings’, and we can do this even though we may think that both ‘Das Nichts nichtet’ and ‘the nothing nothings’ are meaningless.

It might be objected that on this view the samesaying relation for attributions of meaningless or nonsensical sayings or attitudes is trivialized. Heidegger’s belief could just as well be reported as ‘Heidegger believes that the truth truths’, assuming that ‘the truth truths’ is also meaningless, and so, also samesays ‘Das Nichts nichtet’. But this is not right. The problem with this objection is that it smuggles in a prior notion of synonymy or sameness of content and constrains samesaying to respect antecedently given synonymy relations. Whether or not Magidor is correct that her view does not commit her to the existence of intensional entities like propositions (2013, Ch. 3, n. 39), it does seem to presuppose a prior notion of synonymy, a notion that sententialists are likely to think to emerge from rather than to be presupposed by an account of samesaying.¹⁷

The idea is that samesaying comes first and synonymy second (if at all), with the samesaying relation deriving from reporting practices, not from a prior notion of synonymy.¹⁸ This idea can be used to defend a non-trivial samesaying relation for meaningless or nonsensical utterances. The sententialist account can point to the fact that only ‘Heidegger believes that the nothing nothings’ and not ‘Heidegger believes that the truth truths’ is acceptable to show that the samesaying relation for meaningless or nonsensical utterances can be non-trivial. Just as we find in our reporting practices that not every meaningful sentence can figure in a complement clause to attribute some meaningful attitude correctly, so we find in our reporting practices that not every meaningless sentence can be used to attribute every nonsensical attitude correctly.

Does the sententialist need to give a deeper explanation? I don’t think so. The sententialist really need insist only that samesaying is non-trivial because this is what is determined by our reporting practices. However, if pressed, the sententialist could add that our reporting practices do not in general require synonymy. Instead, for example, when it comes to reporting nonsense saying or attitudes, it looks as though our reporting practice operate under a constraint that samesaying nonsense requires preserving what is determined elsewhere in interpretation as the meaning of the individual meaningful elements in the original utterance (compare n. 3 above).

All of this suggests that sententialist accounts are less hospitable to views like Magidor and Diamond’s and instead provide potential lines of support for the intermediate position.

Let me turn, finally, to the prospects for the intermediate position on the Fregean account. Magidor considers this possibility as well: “is [there] a feasible version of the Fregean view where the embedded category mistakes are meaningless, but the propositional attitudes containing them are nevertheless meaningful” (2013, 64)? According to Magidor, the prospects are dim, but she recognizes immediately that the argument to explain why is more complicated on the Fregean account than on the propositionalist account.

The strategy used for the propositionalist account, according to which “we … rule out this possibility by appealing to the compositionality of meaning”, will not work because on the Fregean account, complement clauses, “do not contribute to the meaning of the propositional attitude ascriptions their regular meaning, but rather a second-order meaning” (2013, 64). Here Magidor is taking into account Frege’s idea about sense and reference shift in attitude attributions, according to which expressions embedded in the complement clauses of attitude attributions do not refer to their customary referents and express their customary senses, but refer to their customary senses and express indirect senses.¹⁹ So as long as the expression has a “second-order meaning”, even if it does not have a “regular meaning”, the attitude attribution can be meaningful, and the intermediate position vindicated.

Accordingly, Magidor argues that “an expression can only have a second-order meaning if it has a first-order meaning”:

…from the perspective of the Fregean framework we are considering this principle should indeed be accepted. Plausibly, the theory should take second-order senses as computable from the first-order senses: otherwise speakers would need to learn two primitive senses for each embedded sentence, and this point is strengthened if one accepts that to accommodate sentences with multiple embeddings, one will need a distinct nth-order sense for any natural number n. (Magidor 2013, 64–65)

This is a reasonable perspective. However, the idea that the theory should take the second-order sense to be computable from the first-order sense entails that necessarily, the second-order sense presents a first-order sense, which entails that necessarily, a second-order sense is not an empty sense, a sense without a referent. But in the Fregean system, this is a possibility (Frege 1892, 152, 156–57). Why is it ruled out?

First let me spell out the possibility. Assume, using an example from Sorensen (2002), that ‘The nothing nothings’ is meaningless. This means, on the Fregean account, that it expresses no thought. Consider now ‘Heidegger believes that the nothing nothings’. Since ‘The nothing nothings’ is meaningless, the ‘that’-clause in the attribution, ‘that the nothing nothings’, refers to nothing—it is an empty name of a proposition. This entails that ‘Heidegger believes that the nothing nothings’, does not have a truth-value, which would be its Fregean referent. However, the Fregean view allows sense without reference, so although ‘The nothing nothings’ is meaningless and does not express a sense, and ‘that the nothing nothings’ is a singular term with no referent, ‘that the nothing nothings’ can have a sense. And this entails that, ‘Heidegger believes that the nothing nothings’ can have a sense. The result is sense without reference for the ‘that’-clauses that figure in attitude attributions. I say more immediately below, but this vindicates, in outline, the possibility of the intermediate position in which the complement clause is meaningless, but the attitude attribution is meaningful.

Is this view ruled out by the idea that “the theory should take second-order senses as computable from the first-order senses”? It is not clear how. Suppose it is the case that where there is a first-order sense, it determines the second-order and other higher-order senses.²⁰ This view is silent on second-order senses when there is no first-order sense. It is not clear why this view is not combinable with the view that multiple modes of presentation can designate an absence of sense. Indeed, it should not come as a surprise, for a view that allows that there can be non-trivial instances of nonsense, that nonsense could be thought of (not itself thought, since there is no sense to think) under different modes of presentations or guises. And this idea survives even if we accept that when there is a first-order sense, it determines higher-order sense.

But maybe the question is: how we are supposed to determine what the second-order sense is when there is no customary sense? How would this work? Optimism about a solution comes from the idea that there could be distinct and multiple modes of presentation expressed by different empty names. For example, it seems very plausible that although the name ‘Zeus’ and the name ‘Hera’ are both empty, they differ in sense. How is this possible? A Fregean answer would be that this is because of the possibility of rationally conflicting attitudes. The thought that Zeus throws thunderbolts is different from the thought that Hera throws thunderbolts because a thinker can rationally take conflicting attitudes to those thoughts without irrationality or any deficit of understanding.

This explanation applies here, too. Let ‘the nothing nothings’ and ‘the truth truths’ express nonsense. Nevertheless, since Carnap believed that Heidegger believed that the nothing nothings but Carnap doubted that Heidegger believed that the truth truths, the thought that Heidegger believed that the nothing nothings, is distinct from the thought that Heidegger believed that the truth truths. Although there are no propositions or thoughts *<the nothing nothings> or *<the truth truths>, there are singular, albeit empty, propositional senses <that the nothing nothings> and <that the truth truths>. What Carnap’s belief and doubt tell us is that these singular propositional senses are distinct. i.e., that <that the nothing nothings> ≠ <that the truth truths>.

We have here indicated how distinct modes of presentation could be determined along familiar Fregean lines, even though these modes of presentation fail to present any thought. So, it is not clear how the possibility of this kind of position can be ruled out unless the general idea of sense without reference can be ruled out.²¹

Let’s summarize. Although Diamond and Magidor’s views are supported by the propositionalist account of attitude attributions, they are not, or not clearly, supported by either the sententialist or Fregean accounts. The intermediate position, according to which sense can embed meaninglessness or nonsense thus remains viable, at least on most standard approaches to saying and attitude attributions.

I close this section by returning to Davidson’s master argument against relativism. I come at the issue by considering some further details regarding the Fregean account as it applies to Liberal and Libertarian. This will also allow us to see the difference that a shift from Attributed Relativism to Attributional Relativism makes.

If our discussion of the semantic considerations is correct, then, when Attributional Relativism makes its claim about the possibility of a certain kind of attributional scenario, for example in Relativist Attribution on Attributional Relativism, what the relativist is saying is coherent. If relativism is understood along the lines of Attributional Relativism, then it looks like the claims of relativism in their application to particular scenarios, are, pace Davidson, coherent, and this can be the case even if the notion of an untranslatable truth or alternative conceptual scheme is incoherent.

It might be objected that the argument actually shows that Attributed Relativism too is coherent, because it involves embedding, and the considerations here suggest that even singly-embedded attributions are coherent and sensical. This last bit is correct, but the relativist who accepts Attributed Relativism commits to the existence of alternative conceptual schemes and this part of the view does trade, outside of attributional contexts, in the notions of an untranslatable truth and alternative conceptual schemes. So Attributed Relativism remains incoherent.

But let’s put this objection to Attributed Relativism aside for the moment. Another objection runs as follows. Although the Fregean account serves to make saying or attitude attributions sensical, it does not allow them to have a truth-value, since the report will be true just in case the attributee stands in the saying or attitude relation to a thought designated by the ‘that’-clause, but there is no such thought since the ‘that’-clause is empty. This time, this last bit is incorrect. Understanding why shows further the difference between Attributed Relativism and Attributional Relativism.

To see this, let’s return to Liberal (mutatis mutandis for Libertarian). Attributed Relativism makes claims like Relativist Attribution on Attributed Relativism in its explanation of the intersubjective situation. However, although the Fregean account allows that such claims may be sensical in themselves, the Fregean account also entails that they do not have truth values. So, claims attributing attitudes that make use of alternative conceptual schemes made by Attributed Relativism to explain the intersubjective situation in relativism are neither true nor false, and so are incapable of being true.

But this means that the Attributed Relativism is incoherent in a different sense: in the sense of making theoretical claims—claims that are supposed to be true and explanatory—that are incapable of truth. The concept of an alternative conceptual scheme is thus the exact equivalent for philosophy of what the concept of phlogiston is for physics. Let me explain.

There are two differences between the way that the concept of an alternative conceptual scheme figures in Attributed Relativism and the way that the concept of phlogiston figures in physics; however, assuming, as we are, that the very idea of a conceptual scheme is incoherent, these two differences cancel out. The two differences are: first that the concept of phlogiston is sensical but empty whereas the concept of an alternative conceptual scheme is incoherent and nonsensical, and the concepts that are supposed to make up alternative conceptual schemes only purported and non-existent; but, second, that the physicist uses the concept of phlogiston in a theory that does not make use of saying and propositional attitudinal contexts whereas the relativist does make use of such contexts. The two differences cancel out because, on the Fregean account, Attributed Relativism makes use not of the purported but non-existent concepts of an alternative conceptual scheme, but of concepts of those purported but non-existent concepts of alternative conceptual schemes. These concepts of purported but non-existent concepts of alternative conceptual schemes are like the concept of phlogiston in being sensical but empty and they are unlike the concepts that are supposed to make up an alternative conceptual scheme, which are merely purported and non-existent. The net result, for both physics and philosophy, are theoretical claims that are incapable of truth.

Attributional Relativism is not subject to this problem faced by Attributed Relativism. Crucially, the claim made by the theorist who accepts Attributional Relativism involves an extra level of embedding, and indeed the kinds of attributions made by Attributional Relativism embed the kinds of attributions made by Attributed Relativism—attributions that are sensical but without truth-value. But this implies that the kinds of claims that the Attributional Relativist makes are true or false. Their truth or falsity is determined by whether the attributee stands in the relation to a thought, but because the attribution is second-order, the attributee does stand in a relation to a thought, namely that expressed by the sensical but truth-valueless first-order attribution. And this means that Attributional Relativism is not subject to the incoherence of making theoretical claims—claims that are supposed to be true and explanatory—that are neither true nor false.

4 Conclusion

Davidson’s master argument slides from the correct view that the notions of an untranslatable truth and of an alternative conceptual scheme are incoherent to the incorrect view that relativism is incoherent. Relativism understood along the lines of Attributional Relativism, which characterizes relativism as a view not about the existence of alternative conceptual schemes but about the existence of certain kinds of attributions between thinkers wherein they attribute to each other saying and attitudes that make use of alternative conceptual schemes, is not incoherent. This is not a view in which the notions of an untranslatable truth or alternative conceptual scheme are coherent—I agree with Davidson that they are not—but is a view in which attributions of attributions of nonsensical thinking are coherent. Incoherent theory is bad theory, but theory that involves the attribution of the attribution of incoherent thinking need not be incoherent. Relativism can be a coherent thesis about the explanation of the intersubjective situation in relativism even if the notions that are often invoked in connection with it, such as that of an alternative conceptual scheme, of an untranslatable truth, of relative truth, etc., are all incoherent.

Acknowledgements

This paper has been in preparation for a long time, and the ideas in it have received feedback from many people. Some, but not all, of those people are: Marion Durand, Paul Horwich, Hanti Lin, Kirk Ludwig, Mark McCullagh, Stephen Neale, Alex Rausch, Roy Sorensen, and Adam Sennet. Thanks to them all, and to others who discussed these ideas with me over the years.