§1: The Quinean Account

What does it mean for a sentence or theory to commit us to the existence of a thing? The ontological commitments of a claim are what it asks of our ontology, what must exist if the statement is to be true. So understood, the ontological commitments of a claim are part of what it demands of the world. Footnote ⁵

An account of ontological commitment should be a kind of test—it should deliver an impartial verdict about what must exist according to a sentence or theory. Quine’s quantifier account is still popular:

Quine’s view is more complex than it appears, but its precisifications will not be relevant here.Footnote ⁶ According to all versions of the Quinean orthodoxy, it is the subject and not the predicate that does the work: ‘elephants are gray’ commits us to gray elephants but not to grayness. This is because it is only elephants—not their features—that must be among the values of the first-order bound variables in our translation of ‘elephants are gray’.

Thus standard moral predications will not carry commitment to moral entities. ‘Sally’s act is right’ commits us to acts, not to rightness. And much the same will be true of claims about what is good, who is virtuous, and so on.

Yet in doing moral philosophy, we say not only that Sally’s act is right, but that rightness is derivative and goodness fundamental; we wonder not only whether Lucy is virtuous but how virtues are tied to dispositions. In answering these questions, we must quantify over moral properties and so seemingly commit ourselves to such entities.

Still, we also quantify over holes, cracks, average students, and sakes, without accepting the existence of such entities. On the Quinean view, our ontological commitments arise not from our uninterpreted claims but from the translations of our claims into canonical first-order logic. As a result, it is possible to paraphrase away commitment to entities.Footnote ⁷ If we do not believe in things like holes, our aim is to be able to reconstruct our assertions to avoid quantification over such entities. The paraphrases we accept may be cumbersome (it is easier to say that the cheese has three holes than it is to say that the cheese is shaped such that there are three non-overlapping space-time regions entirely surrounded by cheese). But the loss in eloquence is worth the gain in parsimony.

Of course, holes and cracks are clearly eliminable; rightness and goodness are not. But consider austere property nominalists, who think properties of any sort are as worrisome as holes and cracks. They trade ‘orange is a color’ for ‘everything orange is colored’; they trade ‘intelligence is a feature Linus and Lucy share’ for ‘both Linus and Lucy are intelligent.’ In so doing, they hope to avoid quantification over anything that is not concrete and particular. Can the relaxed realist do the same?

I suggest that, in general, claims that quantify over moral properties may be replaced with claims about the inferential and explanatory relations between moral sentences. To say that rightness is derivative and goodness fundamental is to say that sentences of the form ‘x is right’ are grounded in sentences of the form ‘y is good’. (I hope for my position to be ontologically modest, and so I assume only that grounding is a sentential connective that obeys certain inference rules. See Bliss and Trogdon (Reference Bliss, Trogdon and Zalta2021)). To say that virtues are settled dispositions is to say that every sentence of the form ‘x is a virtue’ implies ‘x is a settled disposition’.

Of course, some first-order moral sentences can appear to be ontological but non-inferential; consider ‘Sally has a reason to help Lucy’, or ‘Linus has a right to his blanket.’ These sentences are often understood as existential statements about rights or reasons. But we can reject this interpretation and instead appeal to locutions such as ‘a is a reason for B to c’ or ‘a has a right to B’ and paraphrase accordingly.

Other examples are admittedly more worrisome. Following Shafer-Landau (Reference Shafer-Landau2003), Nolan (Reference Nolan2014) focuses on:

The most immediate paraphrase is:

Nolan objects that this paraphrase fails: perhaps Tim’s virtue merits reward, but there are some other, countervailing truths that count against his being rewarded. Yet it is unclear what this shows. To paraphrase a sentence is not to endorse a general translation scheme. Quine writes:

If paraphrase can be contextual in the way that Quine suggests, then our task is not to find the paraphrase for these sentences, but to be able to provide a possible replacement for each way we might fill in the relevant contextual parameters. If, as Nolan imagines, we think Tim’s deservingness is outweighed by some other consideration, then we can say:

And if we instead spoke without intending to commit one way or the other, we might simply understand our utterance as:

§2: Moral Things and Moral Qualities

The paraphrase strategy may seem ad hoc. We wish to eliminate commitment to moral properties, so we reinterpret seemingly innocent sentences to avoid commitment. How can such reinterpretation be defended?

This is an important challenge. The justification for our paraphrases cannot, I think, be only to avoid commitment. Rather the challenge calls for a positive, independently plausible vision of reality, such that the paraphrases proposed better capture this reality.

Begin with the general case: consider how we might justify the paraphrases that global property nominalists provide. Property nominalists think that everything that exists is concrete and particular; they posit a fundamental ontology of things. To allow properties is to divide a thing into its particularity and its qualities, making both mysterious, they claim. Thus it may be true that some elephant is gray, massive, and muddy. But these are mere descriptions of the elephant. The elephant and its environs are sufficient to satisfy these predicates; properties—whether universal or particular—are not needed. The property nominalist may therefore say that they provide paraphrases not merely to avoid commitment but to represent things as they are.

This global strategy can be applied locally. We say that though there are acts, situations, and persons, there are no moral qualities of these things. Again, to separate things from their moral qualities is to make both mysterious and thus invite complaints about the ‘queerness’ of moral properties. In the same way that the elephant can satisfy the predicate ‘is gray’ without ontological assistance, some acts—such as saving a drowning child, or calling your mother on her birthday—can satisfy the predicate ‘is right’ without ontological assistance.

If we are moral particularists, this may be a brute fact; if we are moral generalists, a theory of right action may tell us what must be true for ‘is right’ to apply. Regardless, our only concern is the conditions under which moral predicates are true of things; there is no ontological issue to engage with.

We can, then, provide a vision of morality that justifies the paraphrases I have proposed. But it proceeds from a general property nominalism. Can we provide a similar explanation without rejecting properties wholesale?

This question is merely an instance of a broader puzzle: what might justify us in accepting some properties but not others? Perhaps the most common answer is that we should endorse a causal restriction on our ontology and accept only those properties that are needed to explain the causal powers of things (Armstrong Reference Armstrong1978: 45-46). This would leave us with a broadly scientific conception of properties and so eliminate non-natural moral properties immediately.

As stated, the causal restriction will be unattractive to relaxed realists; they are hostile to a broader philosophical naturalism. But the insight behind the causal restriction can be generalized. What is important is not whether a property helps to explain causal powers per se but whether it helps to explain what needs to be explained. We may disagree about what needs to be explained or how this explanation is to proceed, of course. But we need not settle this debate here; our interest is in the relaxed realist. And if we should posit only properties that are explanatory—and we allow, as relaxed realists do, that moral properties are not explanatory—then we will have sufficient justification for our paraphrases.

§3: Second-order Quantifier and Entailment Criteria

Some claim that Quine’s focus on first-order quantification is myopic: we should instead say that a sentence is committed to the entities that its first-order and second-order quantifiers must range over. This criterion takes predication seriously; it claims that ‘elephants are gray’ commits us to elephants and to grayness. So, too, ‘some acts are right’ should commit us both to acts and to their rightness.

Yet this conclusion does not quite follow, even if we accept the second-order account of commitment. This is because the values of our second-order bound variables need not be properties. The value of the second-order variable in the translation of ‘elephants are gray’ can be understood as the set of gray things. Similarly, the second-order account of commitment need not imply that ‘some acts are right’ is committed to the property of rightness; this statement may commit us only to the set of right things.

Some may object that the set of right acts just is the familiar sui generis property of rightness that non-naturalists posit. But a non-natural moral property—whether lightweight or otherwise—cannot be identified with the set of its instances. As Jackson (Reference Jackson1998: 118-125) shows, for every moral property F, there is some disjunctive natural property G such that for any possible entity x, x is F if and only if x is G. Sets are identical when they are coextensive, however. So if properties are sets, then every moral property must be identical with some natural property. But non-natural properties cannot, of course, be identical with natural properties.

If this defense fails to convince, a more radical reply is available: we can reinterpret second-order quantifiers as plural quantifiers. Plural quantification has traditionally been eliminated in terms of first-order quantification over sets; we trade ‘there are some books on the shelf’ for ‘there is a set of books on the shelf.’ But, following Boolos (Reference Boolos1984), we may reject this reduction and take plural quantification to be legitimate: we can allow that “to be is to be the value of some variable or the values of some variables.” If we accept this reinterpretation of the second-order quantifiers, we can say that the values of the second-order bound variables in our translation of ‘some acts are right’ are simply the right things. To say that ‘some acts are right’ is only to say that there are right things, and some acts are among them.

I conclude that the second-order quantifier account does not require ontological commitment to moral properties. So let us focus instead on the entailment criterion. This account maintains that the ontological commitments of a sentence are the things that the sentence entails exist. The rationale is clear: if a theory entails that a thing exists, surely it is committed to its existence. And if a theory does not entail that a thing exists, in what sense could the theory require it?

The entailment criterion has advantages over both first-order and second-order quantifier views. ‘Elephants are gray’ commits us to elephants, but it also seems to commit us to trunks and tusks, ears and eyes. Yet these things need not be among the values of our first-order or second-order bound variables in the canonical translation of ‘elephants are gray.’ Consequently, as Bricker (Reference Bricker and Zalta2016) and Rayo (Reference Rayo2007) note, quantifier accounts under-generate our commitments. But the entailment criterion avoids this concern: ‘elephants are gray’ entails that there are elephants, and the existence of elephants entails the existence of trunks and tusks as well.

However, the entailment view is of little use as an independent test for commitment. The question of whether ‘elephants are gray’ commits us to grayness is simply converted into the question of whether ‘elephants are gray’ entails that grayness exists. And the property nominalist and the property realist will disagree, with little recourse to a resolution.

Given this, a relaxed realist who wishes to do without moral properties can simply insist that ‘some acts are right’ entails that right acts exist but does not entail that rightness exists. This is no more ad hoc or worrisome than the property nominalist’s equally unsurprising claim that ‘elephants are gray’ entails that gray elephants exist but does not entail that grayness exists.

(Some, such as Michael (Reference Michael2008), suggest that we should focus not on modal entailment but on a priori entailment. I remain neutral on this issue, but it will not affect our argument: it is not clear that e.g. ‘elephants are gray’ modally entails that there is grayness nor is it clear that ‘elephants are gray’ a priori entails that there is grayness. And mutatis mutandis for ‘some acts are right’ and rightness.)

§4: Truthmaking

Turn finally to the truthmaker account. Every truth requires a truthmaker: for every true claim, there must be something in the world that guarantees the truth of that claim. Thus, again, if we think of the ontological commitments of a claim as (part of) what it asks of the world, and we believe that all truths require a truthmaker, then it is natural to say that a sentence ontologically commits us to the things that must exist for it to be made true (Armstrong Reference Armstrong2004: 23-24; Cameron Reference Cameron2008).

The truthmaking view looks dangerous for nominalists of any stripe. Essential predications can be made true by particular things: if our elephant is necessarily gray, then he can make ‘this elephant is gray’ true. But if our elephant might be plaid or polka dotted, then he cannot necessitate this truth: he might exist and yet fail to be gray. We need something whose mere existence can ensure that it is true that the elephant is gray. The most familiar option is the (worldly, concrete) fact <this elephant is gray>—which has the universal being gray as a constituent.

Similarly, every contingent moral predication seems to require a moral fact to make it true. If Linus is contingently good, then the claim ‘Linus is good’ cannot be made true by Linus; we must appeal instead to (something like) the fact <Linus is good>. But moral facts have moral properties as constituents. Traditional realists may therefore conclude that the truthmaker account makes it impossible to avoid commitment.

Yet this conclusion can be resisted; it assumes that if A makes B true, then the existence of A necessitates the truth of B. Call this view truthmaker essentialism. It is no part of the truthmaker account of ontological commitment that truthmakers must necessitate. We can instead allow that A can make B true even if A could exist though B is false. Parsons (Reference Parsons1999: 329) writes:

Truthmaker essentialism suggests that the relation between existence and truth is necessitation: every truth must be necessitated by something (or things). Parsons, like Lewis (Reference Lewis1992) and Bricker (Reference Bricker2006), recommends only supervenience: what is true must supervene upon what there is and what it is like. This weaker relation between truths and truthmakers allows that a red rose can make it true that ‘this is a red rose’—and that a good person can make it true that ‘this is a good person.’

I cannot defend the views of Parsons and his allies without taking us too far afield. But their position is plausible and fits neatly with the nominalist form of relaxed realism we have considered.

Yet even if contingent moral predications can be made true by the particular things they are about, what could make fundamental moral claims, such as e.g. utilitarianism, true? These claims are purely moral and so do not depend on anything non-moral. Thus it seems that the only candidate for a truthmaker for utilitarianism would be some complicated fact tying the properties of rightness and maximizing utility together.

Relaxed realists deny that fundamental moral claims substantively depend on reality. So if there is any dependence between these fundamental moral truths and reality, it is trivial. I therefore suggest that everything makes fundamental moral truths true. Fundamental moral truths do not depend on contingencies and so cannot vary from world to world. Such truths are necessary—and everything guarantees the truth of a necessary truth. My existence may guarantee that 2+2=4; the existence of this sentence may guarantee that utilitarianism is true.

The claim that necessary truths possess only trivial truthmakers is not uncommon in the broader debate over truthmaking. Asay (Reference Asay2020: 58) argues:

Some, like Rodriguez-Pereyra (Reference Rodriguez-Pereyra, Beebee and Dodd2005), reject this picture: they insist that truthmaking is explanatory and that trivial explanations are impossible. But it is unclear that truthmaking is an explanatory relation. As Fine (Reference Fine, Correia and Schnieder2012) argues, explanatory structures often have levels: the biological depends on the chemical, the chemical depends on the atomic, and so on. Thus an explanatory relation should allow that A might be explained by B, which might be explained by C, and so on. But truthmaking has no such structure; it is a direct connection between what exists and what is true. So if B makes A true, then there is no further explanatory question we can ask about B using truthmaking. Further, explanation seems to be hyperintensional: explanations for necessarily equivalent truths need not be identical. But we have understood truthmaking in terms of either necessitation or supervenience—and neither is hyperintensional.

I therefore suggest that we think of truthmaking primarily as a modal tool for investigating what truths require of our ontology, and so allow trivial truthmakers for fundamental moral truths. But if moral principles may have trivial truthmakers, then moral properties are not needed to make them true. Thus moral principles do not carry ontological commitment to moral properties.

However, note that even if we insist that truthmaking is explanatory, this need not undermine our attempt to avoid commitment to moral properties. Akhlaghi (Reference Akhlaghi2022) assumes that truthmaking is explanatory but shows that relaxed realists like Parfit can nonetheless avoid commitment. While I say that everything makes fundamental moral claims true, Akhlaghi suggests that nothing makes fundamental moral claims true. Fundamental moral truths are not explained by non-moral truths; to deny this would be to lapse into a kind of naturalism. Nor are fundamental moral truths explained by other moral truths; if they were, they would not be fundamental. So fundamental moral truths are not explained by anything, and so have no truthmakers if truthmaking is explanatory.

There are costs to Akhlaghi’s approach (as he recognizes): it implies the controversial claim that some truths lack truthmakers. I would prefer to avoid this implication if possible. But Akhlaghi’s result is nonetheless important for my purposes. For, regardless of whether we see truthmaking as a kind of modal tool for examining what a theory asks of our ontology or an explanatory tool that highlights substantive dependence relations, the relaxed realist can still avoid commitment.

§5: Classifying the Theory

I have argued that moral truths need not commit us to moral properties, no matter our views on ontological commitment. This allows a form of relaxed realism that does without moral properties. Such a view may be attractive; at the very least, it is both simpler and more ecumenical than extant formulations of relaxed realism.

But is this nominalist revision truly a form of moral realism if it denies that there are moral properties? And if it is not a form of moral realism, does this matter?

Answering either question is difficult. There is persistent disagreement about the meaning of ‘moral realism’. Indeed, some, such as Foster and Schroeder (Reference Foster, Schroeder, Bloomfield and Copp2023), have become skeptical that we can provide an informative, general account of the term. Still, we may fruitfully distinguish two broad views: the first sees moral realism as an ontological doctrine about moral properties; the second sees moral realism as an attitude toward morality.

Begin with the ontological interpretation. To be a moral realist is to believe that morality is part of reality—and so moral realism requires, of course, the existence of (instantiated) moral properties. As a result, describing any view as a nominalist form of realism is, at best, misleading.

However, this ontological account of moral realism seems anti-nominalist and so question begging. We might grant that if moral realism is true, then something moral must be part of reality. But must it be moral properties? Why not instead moral objects—right acts, virtuous persons, valuable outcomes?

Perhaps, however, ontological moral realism is not about the existence of moral properties but about their relative standing. Bloomfield and Copp (Reference Bloomfield, Copp, Bloomfield and Copp2023) argue that to be a realist is simply to allow moral properties the same ontological status as other properties. So if we are global property nominalists, then the nominalist metaethics I have described is a form of realism; otherwise, it is not.

Yet the question of parity strikes me as a red herring. Suppose I am an ontological nihilist and so believe in nothing. I give moral properties (and moral things, such as right acts) the same status as protons, buildings, and persons: all are unreal. I have then met the standards required by parity. But I am, I think, no moral realist.

Still, suppose I am mistaken and moral realism should indeed be understood as a thesis about the existence or status of moral properties. There remains the more pressing question of why it matters whether we are moral realists in this ontological sense.

Many relaxed realists reply that it does not matter—indeed, Parfit (Reference Parfit2017: 59) simply abandoned the label of realism, instead opting for “non-realist cognitivism.” But Parfit never abandoned the attitude toward morality that is characteristic of moral realism. This realist attitude is difficult to capture succinctly but, following Clarke-Doane (Reference Clarke-Doane2020: 15-20), we can identify its markers: the realist attitude requires that moral claims have truth values, that some atomic moral claims are true, that moral judgments are beliefs, that moral truth is independent of our attitudes, and that moral claims may, in general, be interpreted at face value (they are not e.g. covert ways of expressing attitudes or describing what would be true according to some fiction).

Ontological moral realism can, of course, underwrite this realist attitude toward morality—and some moral realists, such as Shafer-Landau (Reference Shafer-Landau2003: 13-18) and Enoch (Reference Enoch, McPherson and Plunkett2019), seem happy to conflate realism as an ontological thesis with realism as an attitude about morality. But if the nominalist view I have proposed is plausible, we should be wary of this conflation; the realist attitude toward morality can be justified by those who reject moral properties entirely. Indeed, just as global property nominalists may insist that judgments are beliefs, that truth is independent of our attitudes, and so on, local nominalists about moral properties may say that moral judgments are beliefs, that moral truth is independent of our attitudes, and so on.

Others have, of course, advanced theories which they claim can capture the realist’s stance without their ontology. Quasi-realists (in)famously argue that we can begin from an expressivist metaethical program but earn the right to speak as the realist does. What difference is there between quasi-realism and the form of relaxed realism I have proposed?

The quasi-realist program remains controversial; stating the view succinctly is difficult. Yet, in any theory that begins with expressivism but aims to simulate realism, the expressivism must be retained somehow. Some, like Blackburn (Reference Blackburn, Haldane and Wright1993), say that moral judgments, though truth-apt, are equivalent with or at least “begin life as” wishes, interjections, or other desire-like states. Dreier (Reference Dreier2004) instead suggests that moral judgments are not desire-like states but we nonetheless possess moral judgments because of our desire-like states. Thus my belief that it is wrong to steal is not equivalent with any desire, plan, or command. But I can hold this belief only by disapproving of stealing, or planning not to steal, or accepting the command that none shall steal.

I take no stance on how the expressivist component of quasi-realism must be understood. But if there is any expressivism remaining in the quasi-realist program, this will be enough to allow the distinction I desire. The nominalist view I have proposed needs no expressivist component, any more than global property nominalists need be global expressivists. As a result, relaxed realism is distinct from both quasi-realism and traditional realism—and represents an important alternative to both.