The presumption of realism

Abstract

Within contemporary metaethics, it is widely held that there is a “presumption of realism” in moral thought and discourse. Anti-realist views, like error theory and expressivism, may have certain theoretical considerations speaking in their favor, but our pretheoretical stance with respect to morality clearly favors objectivist metaethical views. This article argues against this widely held view. It does so by drawing from recent discussions about so-called “subjective attitude verbs” in linguistics and philosophy of language. Unlike pretheoretically objective predicates (e.g., “is made of wood”, “is 185 cm tall”), moral predicates embed felicitously under subjective attitude verbs like the English “find”. Moreover, it is argued that the widespread notion that moral discourse bears all the marks of fact-stating discourse is rooted in a blinkered focus on examples from English. Cross-linguistic considerations suggest that subjective attitude verbs are actually the default terms by which we ascribe moral views to people. Impressions to the contrary in English have to do with some unfortunate quirks of the term “think”.

1 The presumption of realism

Within contemporary metaethics, it is widely held that moral realism enjoys the position of default view, since it accords with our pretheoretical mindset with respect to morality. Anti-realist views may have certain theoretical appeal, as, for instance, expressivism is taken to have in accounting for the presumed fact that moral thoughts are internally motivating. Realist views are, however, supported by the fact that we speak and think about moral matters as if they are fully factual, according to a very common line of thought. Drawing from Nagel (1986), Brink articulates this notion with an analogy to other realism-antirealism discussions:

In many areas of dispute between realism and antirealism, realism is the natural metaphysical position. We begin as realists about the external world or the unobservable entities mentioned in well-confirmed scientific theories. Generally, people become antirealists about these things (if they do) because they become convinced that realism is in some way naive and must be abandoned in the face of compelling metaphysical and epistemological objections. So too, I think, in ethics. We begin as (tacit) cognitivists and realists about ethics. […] We are led to some form of antirealism (if we are) only because we come to regard the moral realist’s commitments as untenable, say, because of the apparently occult nature of moral facts or because of the apparent lack of a well developed and respectable methodology in ethics. (Brink, 1989, p. 23. Cf. also McNaughton, 1988, Chap. 3)

Thus, in ethics, just like in discussions concerning the nature of reality at large, and in the philosophy of science, realism is the default view. This “presumption of realism” in ethics is often articulated as a feature of discourse. Here is Sayre-McCord:

[…]by all accounts, moral realism is, at least initially, the default position. It fits most naturally with what we seem to be doing in making moral claims, and it makes good sense of how we think through, argue about, and take stands concerning moral issues. (Sayre-McCord, 2007, p. 42)

It should be stressed that these theorists are not merely making claims about their own pretheoretical judgements. The claim is that we, all of us, experience morality in this way. Cuneo believes that this is part of how “ordinary mature human agents” work:

A realist conception of moral and epistemic features is something that, all else being equal, ordinary mature human agents whose cognitive faculties are functioning adequately in a world such as ours take for granted in their everyday doings and believings. (Cuneo, 2007, p. 11)

While the theorists quoted so far subscribe to some or other brand of moral realism, similar ideas are prevalent among anti-realists. For instance, J.L. Mackie famously held that there is a “claim to objectivity” built into moral judgements, and made this assumption a cornerstone in his argument for the error theory (Mackie, 1990/1977, p. 35). Contemporary expressivists nearly all subscribe to some version of Blackburn’s quasi-realism, i.e. the project to account for seemingly objectivist features of moral discourse within an expressivist framework (Blackburn, 1993). The presumption of realism hypothesis, or at least something in its vicinity, is widespread among realists and anti-realists alike within contemporary metaethics.

Moreover, the presumption hypothesis occupies an important role in how the dialectic between realism and anti-realism within metaethics is currently conceived of. It is taken to show that anti-realists have the burden of proof. It is for the anti-realist to present arguments to show that morality, contrary to appearance, either rests on a mistake (error theory) or was never in the business of describing “the fabric of the world” to begin with (expressivism and related views). Indeed, Dancy even goes so far as to suggest that the presumption of realism is “the first and perhaps only argument” in favour of moral realism (1986, p. 174). I suppose that most moral realists would disagree with this pessimistic assessment of the number of arguments available in favour of their view, but Dancy’s suggestion is nevertheless indicative of the importance ascribed to the presumption of realism.

The purpose of this paper is to argue against this widely held view. Specifically, I argue against its semantic facet. It is a very common trope in metaethics to point to an alleged perfect similarity between moral discourse and paradigmatic objective modes of discourse; and from that argue that views which attribute a special semantic status to moral talk and thought, like expressivism, are far-fetched and unmotivated in light of this (in addition to the references above, see e.g. Fine, 2001, p. 4; Schroeder, 2008, p. 5; Williamson, 2023). It is this alleged similarity between moral discourse and paradigmatic objective discourse which is the topic of this paper.

I argue against the similarity claim, and thereby, against the presumption hypothesis, by pointing out that there is crucial difference between moral discourse and paradigmatic fact-stating discourse, namely that moral predicates embed felicitously under subjective attitude verbs like the English “find”. I then point out that moral views, in contrast to beliefs about objective matters, are naturally designated as “opinions”. I go on to argue that the widespread notion that moral discourse bears all the marks of fact-stating discourse is rooted in a blinkered focus on examples from English. Cross-linguistic considerations suggest that subjective attitude verbs actually are the default terms by which we ascribe moral views to people.

The structure of the paper is as follows: § 2 deals with some conceptual preliminaries, § 3 outlines the argument against the hypothesis and § 4 expands on it and consider some objections. § 5 discusses some further objections, and § 6 compares the argument to the semantic arguments in favour of the presumption hypothesis. § 7 proposes some take-aways of the discussion.

2 Objective and subjective

Let me begin with some stipulations. I, like most of my interlocutors, use “moral realism” to denote the family of views according to which there are mind-independent moral properties, which are, as Dancy puts it in the quotation above, parts of the “fabric of the world”. For a particular kind of discourse to be “objective” in the sense intended here, is for that kind of discourse to purport to represent some mind-independent part of reality. To use a well-worn example, talk and thought about witches and magicians are objective in this sense, since they purport to represent some part of mind-independent reality, even though there are no witches and magicians. For moral talk and thought to be “subjective,” in the way I will understand the term here, is for some semantic theory along the lines of naïve subjectivism (“wrong” means “wrong for me”), expressivism (moral discourse does not serve to describe the world but instead to express non-cognitive states of mind) or semantic relativism (the truth-values of moral statements are relative to points assessments) to be true. On all these views, moral discourse is in one way or another relative to the “perspective” of the speaker or assessor, rather than referring to the world as it is independently of any perspective.

Another way of getting at the intended meaning of “subjective” is through ostension. It seems very natural to believe that thought and talk about what is tasty and what is funny is subjective in the relevant sense. Whereas the nature of this subjectivity is a contested matter within the literature of predicates of personal taste (see, e.g., MacFarlane, 2014, for an overview), everyone should agree that there is some natural, pretheoretical sense in which talk and thought about the tasty and the funny is subjective. It is in this sense, whatever philosophers might ultimately decide that it amounts to, that I use the term here.

Having made these distinctions, we are now in a position to distinguish two facets of the presumption to realism. One of these is that moral discourse is objective in nature (have “objective purport”), the other is that moral realism is true, i.e., that there really are mind-independent moral facts. In this text I will be concerned with the former facet. It is notion that moral discourse attempts to describe mind-independent reality in the first place, that is at issue here.

3 Subjective attitude verbs and opinions

The argument against the presumption of realism hypothesis that I will employ in the following is drawn from the recently emerging literature on so-called “subjective attitude verbs” within linguistics and philosophy of language (Bouchard, 2012; Bylinina, 2017; Coppock, 2018; Fleisher, 2013; Franzén, 2020; Hirvonen, 2014; Kennedy, 2013; Kennedy & Willer, 2016, 2022; McNally & Stojanovic, 2014; Silk, 2021; Stephenson, 2007). It is, at bottom, quite simple: moral predicates are embeddable under subjective attitude verbs, and this is, as argued in § 4, in fact the default way of referring to moral views. Since subjective attitude verbs elicit infelicity when combined with pretheoretically objective matters, we should conclude that, as far as the semantics of moral predicates is concerned, morality does not belong to this domain. I make a similar point concerning the term “opinion”.

Let us start with the notion of a “subjective attitude verb”. Coppock (2018) gives, as examples of subjective attitude verbs, Swedish “tycka”, Norwegian (bokmål) and Danish “synes”, French “trouver”, Mandarin Chinese “jué dé” and English “find”. To these examples we can also add the German “finden” and the Spanish “resultar”. Since English is the lingua franca of academic discourse, I will here focus on “find”, even though English exhibits some idiosyncrasies, to be discussed in § 4, that make it less than ideal for articulating the point.

The defining feature of subjective attitude verbs are the following. They designate a kind of attitude that an individual can have only towards a class of contents which are intuitively subjective. For instance, Kennedy (2013, p. 261) offers the following example. It makes perfect sense to ascribe to an individual the attitude of finding the Italian dish trippa alla romana tasty:

• (1) Karl finds trippa alla romana tasty.

By contrast, it does not make sense for Karl to find trippa alla romana to be vegetarian:

• (2) # Karl finds trippa alla romana vegetarian.

Similarly, one may find a philosophy paper interesting, but it does not make sense to find it to be 8,000 words long:

• (3) Karl finds the paper interesting.

• (4) # Karl finds the paper 8,000 words long. (Cf. Kennedy, 2013, p. 265)

Also, Karl might find Tuesday nights boring, but he certainly cannot find it to be Tuesday:

• (5) Karl finds Tuesday nights boring.

• (6) # Karl finds it Tuesday. (Cf. Coppock, 2018, p.126).

As noted, there is thriving literature on subjective attitude verbs (see references above). In the literature, this contrast in acceptability is the basic, but only one of several, interesting properties of this class of verbs which theorist seeks to account for. However, for the present context, I will to the furthest extent possible stay clear of linguistic subtleties and focus on something on which there is near uniform agreement in the literature.¹ This is the fact that this basic property, the contrast in acceptability in examples like the above, track pretheoretical subjectivity. Whether a dish is tasty or a philosophy paper interesting is intuitively a subjective matter, whereas what a dish is made of or how many words a philosophy paper contains is a pretheoretically objective matter. As Sæbø puts it:

[…]there is an intuition that it does not make sense to entertain a subjective attitude to something which is either a fact or not a fact, regardless of the subject of the attitude […]. (Sæbø, 2009, p. 329)

Coppock, in discussing why the Swedish subjective attitude verb “tycka” does not accept “It’s Tuesday” as its complement, puts the matter in a similar way:

If you ask a Swedish speaker why one can’t embed “It’s Tuesday” under “tycka”, the answer is typically, ‘because it’s a matter of fact’. (Coppock, 2018, p. 129)

In these passages, Sæbø and Coppock cash out the distinction between subjective and objective in terms of their being a fact of the matter regarding the issue.² I will occasionally speak in this way myself, but it’s important to note that one might think that even subjective matters are matters of fact. Nothing much hangs on the particular phrasing here. What is important is that, in virtue of this contrast in acceptability, subjective attitude verbs can be used to distinguish matters that we pretheoretically consider to be objective from those that we do not. If the matter is subjective, like whether a paper is interesting or a dish tasty, the relevant predicate embeds under subjective attitude verbs. If it is a pretheoretically objective matter, like how many words there are in a paper or what a dish is made of, the corresponding predicates are unacceptable with a subjective attitude verb. This is an interesting and, from a philosophical point of view, exciting piece of data. It gives us access to some of the ontological commitments that are built into our ordinary thought and discourse, prior to philosophical reflection. It should be clear why it is relevant in assessing the presumption of realism hypothesis.

The observation speaks firmly against the hypothesis, since moral predicates, unlike objective predicates, interact felicitously with subjective attitudes. That is, one can felicitously assert sentences like:

• (7) Karl finds it wrong to eat meat.

• (8) Karl finds the government’s stance on climate change morally reprehensible.

• (9) Karl finds the EU’s new immigration laws cruel.

The fact that moral predicates, just like intuitively subjective predicates, interact felicitously with subjective attitude verbs stands in stark contrast to the widespread view among metaethicists in recent decades that subjectivist theories are in prima facie disaccord with how we speak and think about morality.

This is not the only linguistic observation that speaks against the presumption of realism hypothesis. A similar point can be made with corresponding noun, “opinion”, which has received considerably less attention. Andrés Soria-Ruiz (pc) has dubbed the similar contrast exhibited by “opinion” “The Big Lebowski test”, referring to the Cohen brothers’ movie from 1998. In one scene of the movie, Jeffrey “The Dude” Lebowski’s bowling team is approached by Jesús, an opponent for an upcoming game. The following exchange takes place:

• Jesús: Liam and me, we are going to fuck you up [in Saturday’s game].

• The Dude: Well, that’s just, like, your opinion, man.

The comic effect created by the scene is partly due to the fact that The Dude’s comeback is linguistically strange. Who is going to win the bowling game is not a matter of opinion since, presumably, there is or will be a fact of the matter to the question.

Consider also beliefs about seemingly objective matters, as, for instance, the beliefs that Lionel Messi is 170 cm tall (which he is), that iron reacts with oxygen to form iron oxide (which it does) and that the battle of Blenheim took place in 1702 (which it did not). In all these cases, we would find it strange if someone reported their views of these matters as “opinions”:

• (10) # In my opinion, Lionel Messi is 170 cm tall.

• (11) # In my opinion, iron reacts with oxygen to form iron oxide.

• (12) # In my opinion, the battle of Blenheim took place in 1702.

By contrast, compare how natural it is to refer to seemingly subjective views as opinions:

• (13) In my opinion, this is one of the best restaurants in town.

• (14) In my opinion, he is wasting his time studying philosophy.

• (15) In my opinion, the exam was too easy.

Again, we see that moral statements do not behave like statements about seemingly objective affairs. It is indeed very natural to speak of “moral opinions”, and to refer to one’s moral views as opinions:

• (16) In my opinion, eating meat is morally wrong.

• (17) In my opinion, everyone should have access to free healthcare.

With these two points – the felicitous interaction of moral terms with subjective attitude verbs and “opinion” – in mind, let us return to the similarity claim invoked by Brink and others, that was quoted at the outset of this paper. Just as there is a default presumption of realism about the external world, and about the microparticles postulated by modern physics, so there is a presumption of realism in morality, these theorists claim. The previous considerations allow us to see wherein the analogy falters. Moral terms are embeddable under subjective attitude verbs and are naturally described as opinions. By contrast, it is not at all natural to think of the beliefs about the existence of things and relations external to our minds as opinions, or to say things like:

• (18) # I find electrons much smaller than protons.

• (19) # I find this this stone mental in nature.

• (20) # In my opinion, electrons are much smaller than protons.

• (21) # In my opinion, this stone is mental in nature.

To be clear, this does not go to show that anti-realist views in these areas, such as instrumentalism in the philosophy of science, or, for that matter, idealism, cannot be supported on philosophical grounds. And of course, one cannot directly infer something about the existence of mind-independent moral properties from observation about how we speak and think about morality. It does however undermine the analogy. In so far as there is a presumption of realism in the way we speak of microparticles in physics, it does not extend to the moral domain.

While I have argued that these data points track subjectivity, I have not said anything about what kind of subjective view they favour. The reason for this is that this is very much a live issue within the semantic discussion about subjective attitude verbs (very little has been said about “opinion” so far). For instance, Sæbø (2009) takes them to favour a form of a semantic subjectivism for evaluative terms, while Coppock (2018), focusing on the Swedish “tycka”, argues for what she calls an “outlook based semantics” (similar in spirit to semantic relativism). Kennedy and Willer (2016, 2022) think they track a specific kind of underdetermination of the meaning of the relevant terms, which is “resolved” by the speaker or someone else in the context. Franzén (2020) argues that “find” designates non-cognitive attitudes, and that the data is favourable to a form of expressivism. My purpose here is not to adjudicate between these views, but to focus on what seems very plausible and with which they all agree (with the exception of Silk (2021), discussed below), namely that felicitous interaction with these verbs is a mark of subjectivity. It is this notion, and its disaccord with the widely held view within metaethics that there is a presumption of realism in moral thought and discourse, that is the central concern of this paper.

4 The ubiquity of subjective attitude verbs

As we have seen, the notion that felicitous embedding under subjective attitude verbs tracks a kind of subjectivity is a commonplace within linguistics and the philosophy of language. My only claim to originality in what’s been argued this far is in bringing forth the relevance of these observations for debates about moral realism within metaethics. However, the considerations of the present section will, in addition to defending the argument from a certain line of possible response, make an original contribution to the discussion of the nature of subjective attitude verbs, by arguing that these are actually much more common than is standardly thought.

To get us off the ground, consider a certain kind of objection one might want to raise against the argument against the presumption hypothesis. It has been contended that moral predicates in fact do not embed felicitously under “find” (e.g. Hirvonen, 2014, Chap. 4). If this was the case, it would undermine the whole line of reasoning of the present article.

One way to dispatch with this idea is to add some authentic uses of “find” within ethical contexts to our stock of examples. Here is one from Weinberg in her book about the ethics of procreation:

• (22) Since future people are not capable of consenting to anything, some may find it unproblematic that we create them without their consent. (Weinberg, 2016, p. 26)

And here is Railton, in the context of discussing naturalistic moral realism:

• (23) Depending upon the nature and circumstances of given individuals, they might have objective interests in things we find wrong or repulsive, and that do not seem to us part of a good life. (Railton, 1986, p. 177, n. 20)

And here is Rosen, in a critical discussion of Blackburn’s quasi-realism:

• (24) At this stage there is no concept of wrongness in view, and so no such things as finding something wrong. (Rosen, 1998, p. 389)

These examples clearly show that moral predicates do interact with subjective attitude verbs in English. But there is another worry in the vicinity. Focusing exclusively on “find” when discussing subjective attitude verbs and their interaction with moral predicates may give the impression that we are dealing with a fringe phenomenon. A defender of the presumption hypothesis might perhaps be inclined to concede that moral predicates interact felicitously with “find” in some circumstances, but still maintain that it is not very common and therefore of lesser importance to moral semantics (cf. Stojanovic, 2019).

Some potential support for such reasoning can be found in a recent corpus study conducted by McNally and Stojanovic (2023) in which they investigated the prevalence, respectively, of some moral and taste terms embedded under subjective attitude verbs. What they found was that whereas the moral predicates they investigated indeed occurred embedded under “find”, they were less likely to do so than the predicates of personal taste included in the study. The moral terms they investigated were instead about three times as likely to occur with “consider” than with “find” as in:

• (25) Karl considers meat-eating immoral.

Notably, “consider” can also occur with (arguably) objective complements as in:

• (26) Trump supporters consider the election stolen (see Kennedy & Willer, 2016, 2022, for discussion).

One might take moral predicates’ relative preference for “consider” over “find” as evidence that they are somehow relevantly different from predicates of personal taste, especially regarding the extent to which they are pretheoretically subjective or objective.

Instead of addressing this alternative hypothesis head-on, let me instead make a more general point about subjective attitude verbs in English. The relative infrequency of moral predicates with subjective attitude verbs like “find” in English is, I think, rooted in the fact that it’s not the standard way by which we attribute moral views in this language. Instead, we tend to use “think” when attributing moral views, as in:

• (27) I think it is morally wrong to eat meat.

One might think that this observation spells even more trouble for our argument against the presumption hypothesis since “think”, importantly, does not exhibit the identifying trait of subjective attitude verbs. It also accepts intuitively objective complements as in:

• (28) I think that the table is made of wood.

Whether something is made of wood or not is, supposedly, a pretheoretically objective matter.

But this is an instance where I think the dominance of English has been to the detriment of metaethical theorizing. “Pure” subjective attitude verbs are much more common in some other European languages.³ For instance, in Scandinavian languages like Danish (“synes”) and Swedish (“tycker”), subjective attitude verbs are, as I will show below, the default terms for attributing moral views. This observation might at first instance invite the hypothesis that Scandinavians are less objectivistically inclined with respect to morality than people from English speaking countries. I do not think that is the right conclusion to draw. Rather, it indicates that “think” in English is ambiguous between a sense closer to that of “believe” and one close to that of subjective attitude verbs like “find”.⁴ Call a word that exhibits such an ambiguity an “impure” subjective attitude verb.

That “think” is an impure subjective attitude verb can be argued for on the basis of translation facts.⁵ Take “immoral”, which is one of the terms included in McNally and Stojanovic’s study. Consider the following ascription of a moral view with “immoral” (to himself) by the character Peter Griffin in the tv-series Family Guy:

• (29) As a matter of fact, Joe, yes. I think it’s immoral.

In the Danish, Swedish, French and German subtitles to this episode, “think” is translated with “synes”, “tycker”, “trouver” and “finden” respectively.⁶ These are all subjective attitude verbs in the sense outlined above.⁷

To provide some quantitative support for the same point, I searched a multilingual corpus of subtitles with translation into the French for sentences where both “think” and “immoral” occur.⁸ The search yielded 47 hits, out of which 40 were deemed to have the relevant kind of construction.⁹ Of these 40 sentences, “think” was translated with “trouver” (to find) 14 times, with “penser” (to think) 12 times, and 14 times with some other construction, including omittance of the attitude verb.¹⁰ When translating attitude attributions with “think” and “immoral” into French, the subjective attitude verb “trouver”, literally meaning “find”, was thus the most common choice. Swedish, which like other Scandinavian languages lacks a clear counterpart to “to think” (as an attitude verb), yields an even clearer picture. A search in the same multilingual corpus of subtitles for sentences with both “think” and “immoral”, generated 21 results that had been translated into Swedish. Of these, 15 were of the relevant construction. Of these, 11 translated “thinks” with the subjective attitude verb “tycker” and 4 with other constructions, including omitting the attitude verb.

It should be recognized that in the French, the results were somewhat less clear-cut with respect to the other three moral terms Stojanovic and McNally investigated. With “unethical”, of 21 relevant constructions, only 3 were translated with “trouver” and 4 with “penser”, whereas the other 14 translations went with some other construction, including, most commonly, omitting the attitude verb. With “ethical”, of 20 relevant constructions, 6 were translated with “penser” and 3 with “trouver”. Since “moral” is also a noun, it yielded considerably more results (80), with an adjoined difficulty of isolating the relevant constructions.

In the Swedish translation of “think” with “unethical”, out of 9 examples with the relevant construction, 4 were translated with “tycker” and 4 omitted the attitude verb. With “think” and “ethical”, out of 5 examples with the relevant construction, 3 were translated” with “tycker”.

Overall, then, we see that English “think” with a moral predicate is very often translated with the subjective attitude verb “trouver” in French, despite the fact that French has a direct counterpart to “to think” (as an attitude verb) in “penser”. Both of the latter are impure subjective attitude verbs on the hypothesis we are exploring. Moreover, in Swedish, lacking a direct counterpart to “to think”, the translation with a pure subjective attitude verb, “tycka”, is in a very clear majority. Indeed, it is the only frequent choice, if one discounts the instances where the translator omits the attitude verb.

We can conclude then, that subjective attitude verbs are actually very common in combination with moral terms. Moreover, the translation data also indicates that when “thinks” occurs with a moral predicate, it is most often with its subjective attitude meaning (i.e., “find”), rather than with its meaning that also allows for objective predicates like “being made of wood”. By virtue of being a default way of attributing moral views, subjective attitude verbs are of central importance to the semantics of moral terms. The appearance to the contrary in English has to do with the, from a theoretical point of view, unfortunate fact that English employs an impure subjective attitude verb as its standard way of attributing moral views.

McNally and Stojanovic, remember, argue that the subjectivity involved with moral terms is of a different kind from predicates of personal taste, manifested in moral terms’ preference of “consider” over “find”. As far as I can see, what I argue here is compatible with this hypothesis, since nothing I’ve said undermines the claim that moral terms prefer “consider” over “find” in English. But I don’t think there is any evidence that this potential difference has bearing on the presumption of realism, since the translation data shows that moral terms are actually very common with subjective attitude terms which, unlike “considers”, are infelicitous with objective complements. Again, as Coppock puts in the quotation above, intuitively, what matters for felicitous embedding under “tycka” is whether there is a fact of the matter regarding the issue or not.¹¹

Having said this, there is of course an interesting question to be pursued regarding why moral terms are relatively infrequent with “find” as compared to “consider” and, for that matter, why English speakers seem to prefer “think” over “find” in most cases when attributing moral views. If I might offer a conjecture, I think it has to do with the fact that “find” prefers to take a small clause rather than a full propositional clause (“I find it tasty” rather than “I find that it’s tasty”). French “trouver” is less restrictive in this matter, to the point that Ducrot (1975) seem to consider the propositional construction the default one, and accordingly, moral terms seem to be more common with “trouver” in French than with “find” in English.¹² With Swedish “tycka”, a full propositional complement is mandatory, and is accordingly very common with moral terms, as evinced by the translation data. I’m not sure why this would matter in terms of frequency, but one might speculate that the construction with a full propositional clause is more flexible, and, accordingly, more likely to be used in the various different circumstances in which we might want to attribute moral views.¹³

5 More challenges

One way of challenging the argument here is to deny that these terms actually track pretheoretical subjectivity. While I think such views face an uphill battle against common sense – there is to my mind obviously “something subjective” going on with such terms – we should try to give alternative views a run for their money.

First, let’s consider some possible alternative uses of the term “opinion”. It should be recognized that Cambridge dictionary does offer “a judgement made by an expert” as one meaning of the term.¹⁴ And indeed, it seems right that one can refer to this specific genre of judgements as “opinions”, even if the matter at hand is fully objective. For instance, an expert on the post-Roman history of Britain might felicitously assert:

• (30) In my opinion, King Arthur has never existed.

Whether King Arthur existed or not is presumably a completely objective matter. On the basis of this, a defender of the presumption might retort that the fact that we designate moral views as opinions has no bearing on whether they are factual or not, since we sometimes designate judgements, the contents of which are fully objective, as “opinions”.

But such examples are, I think, of limited help to the defender of the presumption. To be relevant, it would have to be argued that the sense of “opinion” in which the term designates an expert’s view on some contested matter is the same as the one in which it designates ethical views. That does not seem very likely. Clearly, ethical views, like other normative and evaluative views, are naturally described as opinions quite independently of whether we consider the person an expert on the matter or not. And it seems to be precisely in the sense of “opinion” in which it is not synonymous with “expert’s belief” that the term allows for embedding of ethical, aesthetical and taste terms and moreover does not allow for embedding of paradigmatic objective matters.

Could a defender of the presumption disregard the dictionary’s reference to experts, and instead argue that opinion just means a belief with some degree of uncertainty? Merriam-Webster lists such a meaning to the term whereas Cambridge dictionary does not.¹⁵ “Opinion” would, on this view, mean something like “uncertain belief”, or at least the word would have one such meaning, and it would be in that sense alone that moral views would be opinions.

But this would not resonate with how we willingly ascribe opinions even to people who are highly confident about some moral issue, like, say, someone being of the opinion that death-penalty is barbaric. Ascribing such a view does not in any way suggest that the individual is less than certain about their judgement. To be opinionated is not necessarily to exhibit epistemic humility.

In light of this, one could instead be tempted to locate the uncertainty, tracked by the term opinion on this hypothesis, at the societal level. One might think that “opinion” means something like “a belief about a contested matter”, that is, a matter on which people have diverging views, like that regarding King Arthur’s historical existence. On this view, the belief-holder could be highly confident, while the belief would still merit the name “opinion” due to it being about a contested matter. But to see that this cannot be the only meaning of the term, just consider a case where I prepare a dish, taste it myself, and find it tasty. In such a scenario, I need not be the slightest uncertain about the judgement, nor is the matter contested, since no one else has tasted the dish. But it would still be natural to ascribe to me an opinion about the dish I’ve cooked. And it seems to me that moral views work similarly here. I can be of the opinion that something that I just did, like lying to a friend, was wrong, without being the slightest bit uncertain about the matter and without it being contested (since no one else knows that I lied). Here again, moral terms behave like predicates of personal taste and unlike paradigmatic objective terms, which can only interact with “opinion” in the term’s epistemic reading.¹⁶

Having said all this, I am willing to concede that the use of “opinion” as a subjectivity tracker is complicated by the existence of an epistemic meaning, the nature of which is quite difficult to get at (witness the discrepancy between the dictionaries concerning this matter, referenced above). The issue is subtle. For the reasons stated above, I doubt that that a convincing case can be made for the view that all felicitous interactions of moral terms with “opinion” is of the epistemic kind. However, the case is more clear-cut with respect to subjective attitude verbs, which do not have a similar epistemic meaning complicating the matter.

Is there anyway of pushing back on “find” and other subjective attitude verbs being subjectivity trackers? Silk (2021) can be interpreted as doing so. Silk offers a compositional semantics for evaluative terms which, he argues, explains their felicitous interaction with “find” without necessarily conferring subjective status to such terms. If Silk is right, it would undermine our argument against the presumption hypothesis. The essence of Silk’s semantics for evaluative terms is the idea that the truth and falsity of the sentences in which they figure are sensitive to a “perspective” variable and that perspectives are dependent on context. This is similar to how terms like “tall” are often thought to be context-sensitive: the contexts in which “tall” is used are thought to provide thresholds, “standards”, for what counts as tall in them (e.g. Kennedy, 2007). For instance, in the context of a basketball team, the threshold might be set at 205 cm, while it in the context of a class of 10-year-olds might be set at 145 cm. Speakers can agree about the length of someone, while still disagreeing on whether someone should count as tall in the relevant context. In such scenarios, the following assertion would be “discourse-oriented”:

• (31) Karl is tall.

That (30) is discoursed-oriented means that it does not concern how tall Karl is (suppose that its already common ground that he is 190 cm) but rather whether that is tall enough to count as tall in the context. The conversational contribution concerns the contextual threshold, rather than absolute length.

On Silk’s view, “find” is discourse-oriented in the same way. When using “find”, speakers “distinguish possible ways the contexts might be”. To take a simple example, suppose that perspectives might differ with respect to whether they condone lying or not. By asserting “Karl finds lying wrong”, a speaker has attributed to Karl a state of mind which eliminates from the open possibilities of the way the context might be all ways in which the perspective supplied by the context condones of lying.

As Silk explains, the perspectival relativity of evaluative terms cannot derive from the threshold dependency of gradable terms. This is made evident by the fact that whereas “tall” can arguably be felicitous under some circumstances with “find”, it paradigmatically is not in the comparative form, where the threshold is fixed, so to speak, to a comparison with some other individual:

• (32) # I find Karl taller than Olle.

By contrast, evaluative terms are felicitous with “find” even in the comparative form:

• (33) I find meat-eating crueler than child abuse.

So whatever the nature of perspectival-sensitivity, it cannot be reduced to threshold-sensitivity.

At first glance, the talk of perspective-sensitivity as a distinguishing feature of evaluative terms, among them morals terms, which Silk justifies partly with reference to the felicitous embedding of such terms under “find”, seems to be very much in line with the notion that such terms are subjective in nature. However, Silk explicitly disavows such implications. There might very well be a fact of the matter as to which perspective is “the right one”, i.e. the one that matters for the truth or falsity of an utterance:

[…]a defender of the objectivity of morality—or at least the objective purport of moral language—might identify the operative moral perspective with moral norms determined independently of particular speaker attitudes. If a universal body of moral norms was correct, the same moral perspective would be supplied across contexts. This would be a substantive normative matter rather than something built into the conventional meaning and representation of semantic competence (Silk, 2021, p. 149).

“The operative perspective” is similar to the standard supplied by a given conversational context for what counts as tall in that context. The difference here is that, on the objectivist meta-normative view which Silk thinks can be coupled with his semantics, it would be the same “perspective” supplied in all conversational contexts for determining what is, for instance, morally right and wrong.

It’s important to distinguish two different points that Silk might be making here. The first is that the data surrounding “find” is in principle compatible with moral realism. Despite us using subjective attitude verbs when attributing moral views, there might very well be an objective morality which settles questions about, for instance, what is morally right and wrong, in all contexts. No semantic view can by itself settle facts about extra-linguistic reality. If this is Silk’s point, I do not contest it.

However, this notion is importantly different from saying that, for all we know, it might very well be that speakers pretheoretically conceive of morality as objective in the relevant sense. Speakers would then assume that there is one and the same attitude independent body of norms settling the truth of moral utterances. Silk’s talk of “objective purport” in the quotation above is naturally interpreted in this way.

But if speakers tacitly assumed that there is only one moral perspective which determines the truth of moral statements in all contexts, why would the perspective be treated as a separate parameter, rather than just a feature of the world in general? Think of the corresponding case with “tall”. If we all agreed that the standard for tallness is the same in all contexts, there would, obviously, be no sense in calling the term context-sensitive. Similarly with perspectival relativity.¹⁷

Perhaps Silk point should instead be taken to be that we should take speakers’ implicit meta-normative views to pattern with their use of “find”. Only speakers who are committed to some form of non-objectivism use moral predicates with “find”.¹⁸ Here the cross-linguistic considerations put forth in § 4 are relevant again. As noted, looking only at English might give the impression that we are dealing with a relatively rare phenomenon, since “find” is not that common with moral terms. But as we have seen, “trouver” and “tycker” are the most common translation of “thinks” with a moral complement in French and Swedish, indicating that subjective attitude verbs are very standardly used when attributing moral views. Even if it were true that a substantive number of speakers refrain from using moral terms with “find” because of an implicit commitment to objectivism (I doubt it), the fact that the use is so widespread would undermine the claim that there is a presumption of realism in moral thought and discourse.

In sum, then, stressing the compatibility of “find”-data and moral realism does not undermine the point we are making against the existence of a presumption of realism in moral thought and discourse. The claim that would undermine this argument, that the “find”-data is compatible with speakers tacitly assuming that there is an objective moral perspective, is not supported by Silk’ considerations.

6 The semantic argument in favour of the presumption hypothesis

This section compares our argument against the presumption of realism with the arguments that have been raised in its favour. First, theorists have pointed to a family of features that are taken to speak against a specific form of subjectivism: non-cognitivism. These include, first, the declarative form of moral sentences. Paradigm cases of non-descriptive discourse like commands and questions have their own grammatical clause-types. Since moral utterances do not, it seems reasonable to assume that they are involved in the fact-stating business, like other declarative sentences. In addition, talk and thought about morality involve notions such as “moral beliefs”, “moral truths” and “moral knowledge”, which again seem to indicate that such thought and talk are in the business of stating facts (see Ewing, 1959, for an early articulation of this point.) Secondly, theorists point to the fact that moral predicates allow for interpersonal and even intracultural disagreement. This speaks against semantic subjectivist hypotheses, according to which moral predicates have hidden argument places for speakers (“wrong for me”) or cultures (“wrong in my society”). The fact that moral sentences have declarative form and felicitously interact with terms like “knows”, “is true” and “believes” (speaking against non-cognitivism) and that they warrant the relevant form of disagreement (speaking against semantic subjectivism in its variations) are together taken to indicate that, at least on the face of it, moral terms are non-subjective in nature.

These considerations might invite the hypothesis that the evidence regarding moral predicates points in different directions.¹⁹ Perhaps we sometimes use moral terms subjectively and sometimes objectively. However, this is not the conclusion that I think we should draw.

As a first push-back, let me first make a restricted remark about the use of “beliefs” with moral predicates in English. One can indeed refer to moral views as beliefs, but it seems to me that the most salient situations where it is natural to do so in English are cases where you want to attribute a certain level of uncertainty:

• (34) Karl believes that the government’s policy on climate change is morally reprehensible.

I think that a natural interpretation of a moral belief attribution such as the one above is that Karl is somewhat hesitant about his judgement, perhaps because he does not really know what the government’s climate policy is (“but given their track record on these things, it’s probably as reprehensible as their other policies”). When we really want to attribute a fully-fledged moral conviction, it seems to me that “think” is preferable (even though “think” can also mean “believe”), and as argued in § 4, the default interpretation of “think” in such circumstances is as carrying its subjective attitude meaning, rather than its belief meaning.

Future research will have to settle whether this hypothesis about the fine-grained differences between using “believes” and “think” with moral predicates is correct. The general problem for the above argument for there being a presumption of realism in moral thought and discourse that I want to stress is that it over-generates. Taste discourse, with terms like “tasty” and “funny”, also has a declarative form, felicitous interaction with cognitive discourse and the capability to elicit disagreement. When we say that something is funny, we do so in the declarative mood. Likewise, it makes perfect sense to say that one believes and that one knows that something is funny (see, e.g., Kneer, 2021; Lasersohn, 2009; Loeb, 2003; MacFarlane, 2014, p. 158; Pearson, 2022):

• (35) I don’t know where it comes from or who first made it, but I do know it is tasty.²⁰

• (36) You know it’s tasty when the outside of the dish looks like that!!²¹

Moreover, one can certainly disagree, in the relevant sense, on whether something is funny or not. And whereas there is, of course, the odd realist about taste issues, it is not by any plausibility the default view. Declarative form, felicitous interaction with cognitive vocabulary and the capability to elicit the relevant form of disagreement can therefore not be convincingly invoked as objectivity markers of moral thought and discourse (see Loeb, 2003, for a similar line of thought). The semantic test proposed above, by contrast, gets the intuitively right result in these cases since it separates paradigm cases of objective predicates (“being made of wood”) from paradigm cases of subjective predicates (“tasty”). Members of the latter class embed under subjective attitude verbs and interact naturally with “opinion” in the non-epistemic sense.

All this is not to say that the general challenge incurred by non-cognitivists and other subjectivist views to account for belief embeddings and similar sentential contexts is nonexistent. On the contrary, it is a very real challenge, on which much ink has been shed ever since Geach (1960) first pointed out that non-cognitivist views struggle to account for embedded uses of moral terms. However, it has not been recognized that moral objectivists face a very similar challenge: to provide a semantic theory, objectivist in spirit, which makes the correct predictions regarding the embedding under subjective attitude verbs and “opinion”. Virtually no ink has been shed on this challenge.

7 Conclusion

Let’s take a step back and consider some general take-aways from all this. In his book-length defence of moral realism, David Enoch writes:

I suspect that as a psychological matter, I hold the metaethical and metanormative view I in fact hold not because of highly abstract arguments in the philosophy of language, say, or in the philosophy of action, or because of some general ontological commitments. […]Like many other realists (I suspect), I pre-theoretically feel that nothing short of a fairly strong metaethical realism will vindicate our taking morality seriously. (Enoch, 2011, p. 8)

When Enoch speaks of “philosophy of language” in the quotation above, he plausibly has something like some or all of the features of the semantic presumption of realism in mind. I share Enoch’s suspicion that for many theorists, the motivating reason behind adherence to objectivism and realism comes from normative considerations concerning the importance of morality rather than from the arguments that are more often explicitly presented in its favor. A take-away from our current discussion is that realists might be better served by bringing these normative considerations to the forefront of their argument, instead of relying on questionable arguments from the philosophy of language. The latter are, as I’ve argued, not convincing.²²

A second take-away concerns the status of subjectivism. As mentioned, most contemporary subjectivists in metaethics today subscribe to Blackburn’s quasi-realism, i.e., roughly, the program of developing a subjectivist theory which also accounts for objectivist-seeming features of moral thought and discourse. There is a risk of excess here. Often, quasi-realism is articulated as the program of accounting for the alleged fact that moral discourse bears all the marks of fact-stating discourse. This has given rise to the so-called “Problem of creeping minimalism” i.e. the problem of distinguishing quasi-realism from objectivism and realism proper (e.g. Dreier, 2004; Fine, 2001). The argument of this paper calls for moderation. While it is right that subjectivists need to formulate their theories so as to be compatible with, for instance, the fact that moral terms are embeddable under terms like “knows”, it is not the case that moral discourse bears all the marks of objectivist discourse. If quasi-realism is understood as the project of accounting for the latter alleged fact, it is misguided. If it is instead understood as the project of accounting for why moral discourse, while being subjective, shares only some important features of objectivist discourse, there is no problem of distinguishing it from objectivism proper. This is not to say, of course, that there isn’t still an important question to be pursued concerning which subjectivist theory best accounts for all the data in this regard.

Let me make a final point. There is a tendency in our field to dismiss students who come to their first metaethics classes with strong subjectivist inclinations as “sophomore relativists” and explain away these inclinations as due to “post-modernist” cultural influence.²³ Even these students, it is commonly thought, are at a “deeper level” committed to realism, and their reported views therefore do not, on this line of thought, speak against the existence of a presumption of realism (Enoch, 2017, pp. 31–32). The argument presented here speaks against this tendency to dismiss these student sentiments as being triggered solely by some contemporary intellectual fashion. There are deeply rooted subjectivist aspects of our moral thought and discourse. It is likely that this is one source from which many “sophomore relativists” draw when they try to articulate their views. Coarse-grained as such views may be, these students would then manifest a better awareness of how moral thought and discourse actually work than the teachers who explain to them that moral discourse bears all the marks of paradigmatic fact-stating discourse.