Ground, Essence, and the Metaphysics of Metanormative Non-Naturalism

1. The Task of Formulating Non-Naturalism

This section introduces and motivates the task of formulating the non-naturalist’s distinctive metaphysical commitments.

As we understand it, “non-naturalism” is a family of views about certain sorts of normativity. There is not clear agreement about which sorts of normativity are included, but the target usually includes ethical normativity, and excludes the normativity of the rules of games. For simplicity, in this paper, we will use ‘normative’ and its cognates to pick out whatever sort of normativity a non-naturalist might have in mind.³ Here we set aside non-naturalists’ characteristic commitments concerning (e.g.) normative thought and talk and the epistemology of the normative, to focus on their distinctive metaphysical commitments. This is important in part because we take “non-naturalism” to pick out two different kinds of view with quite different metaphysical commitments. In particular, according to some philosophers, we can develop the core metaphysical commitments of non-naturalism in a relatively “lightweight”, “relaxed”, “quietist”, or “non-ontological” way (e.g., Dworkin 1996; Parfit 2011; and Scanlon 2014). By contrast, many other contemporary non-naturalists take their view to involve substantial metaphysical commitments (e.g., Enoch 2011; Fitzpatrick 2008; Leary 2017; and Shafer-Landau 2003). Here we are focused solely on how to understand the latter, substantial sorts of metaphysical commitments.⁴

The task of formulating non-naturalism’s core metaphysical commitments might seem straightforward. In particular, one might think these commitments should be understood as follows: non-naturalism is just the thesis that normative facts and properties are not natural. There are of course live controversies concerning what makes a fact or property “natural”. For example, if one defines the “natural” in terms of specific features of properties countenanced by our best current sciences, there is the question of which features those are (e.g., causal efficacy, empirical tractability, etc.). And one might want to resist associating the natural too closely with the current (or even an idealized) state of science. These are philosophically delicate issues. But one might be tempted to reason as follows: once those issues are settled, the formulation of non-naturalism is settled too. This thought correctly identifies a contrast with naturalism as important to understanding the non-naturalist’s metaphysical commitments. But it is otherwise insufficient, for (at least) the following two reasons.

First, ever since Moore, proponents of the non-naturalist research program characteristically distinguish the ontological commitments of their view not just from those of “naturalistic realism”, but also from those of “supernaturalist” views about normativity, such as divine command theories (see Moore 1903/1993; Shafer-Landau 2003; and Enoch 2011).⁵ The non-naturalist’s characteristic thought is that (the relevant kind of) normative properties are sui generis, or, put more colloquially, “of their own kind” (see Moore 1903/1993; Shafer-Landau 2003; and Enoch 2011). This means that normative properties are not merely different from natural properties, but also from supernatural properties, as well as other non-normative properties whose naturalistic credentials might be in doubt (such as phenomenal properties, mathematical properties, etc.).

The second reason why the tempting reasoning above is hasty is that many non-naturalists accept that normative properties bear apparently intimate metaphysical relations to natural properties. For example, it is commonly claimed that the normative properties supervene on the natural properties. On one formulation of this claim, there can be no normative difference between metaphysically possible worlds without there being some corresponding naturalistic difference.⁶ More recently, some non-naturalists—or those arguing on their behalf—have gone further than this and held that facts about the instantiation of normative properties are fully grounded in facts about the instantiation of natural properties (see Leary 2017 and Bader 2017).⁷

These facts are relevant because supervenience and grounding relations are commonly used to formulate monistic hypotheses in various subareas of philosophy. For example, within late twentieth-century metaphysics and the philosophy of mind, the dominant approach formulated physicalism as a supervenience thesis. And the most prominent contemporary alternative to this approach formulates physicalism in terms of grounding.⁸ But non-naturalism is supposed to be inconsistent with an analogous kind of metaphysical monism: naturalism about the normative. Thus, insofar as non-naturalists endorse either the supervenience of the normative on the natural (as most do) or similar claims about grounding (as some do), it’s not clear why their view even counts as genuinely “non-naturalist”.

We take these facts to put substantial pressure on non-naturalists to provide an informative characterization of their metaphysical commitments. The costs of failing to do so can be illustrated by David Enoch’s discussion in Taking Morality Seriously, which we take to be one of the most important recent defenses of non-naturalism. Enoch writes:

I am going to make use of the nothing-over-and-above relation [to understand the contrast between naturalism and non-naturalism]. I take the more philosophically-sounding relations of identity, reduction, constitution, and grounding to be attempts at precisifying this more intuitive relation. . . . And the reason I will not have to worry about the precisification suitable for our context is that I want to deny the nothing-over-and-above claim, however precisified. (Enoch 2011: 101–2, emphasis in original)

There are at least two problems with this approach. First, elsewhere in the book, Enoch embraces the thesis that the normative supervenes on the non-normative (see Enoch 2011: ch. 6). And many philosophers understand the “nothing-over-and-above” metaphor in terms of supervenience, at least in certain contexts.⁹ So it is not true that Enoch denies the metaphor however it is (reasonably) made more precise.

Second, other philosophers understand the “nothing-over-and-above” metaphor as specifically concerning reduction.¹⁰ To illustrate this line of thinking, consider the following idea: that the normative fails to reduce to X shows that it is something “over and above” X. This means that some non-reductive naturalists may also deny the slogan with which Enoch formulates the “naturalism” he denies. The non-naturalist thus needs to say something more to explain what metaphysically distinguishes non-naturalism from non-reductive naturalism.

These points interact with one of the most common and powerful motivations for accepting non-naturalism. This is the idea, neatly formulated by Enoch, that normative properties are “just too different” from natural properties for naturalistic realism to be a plausible view (see Enoch 2011: ch. 3).¹¹

A natural worry is that the apparently intimate metaphysical relations between the normative and the natural put pressure on this intuition. More precisely, if one grants that (for example) the normative facts are all fully grounded in the natural facts, one might start to suspect that the “just too different” intuition is not tracking a metaphysical contrast, but perhaps a semantic, conceptual, or epistemological one.¹² One reason this would not be surprising is that “just too different” intuitions can be understood as attempting to extract what is plausible in Moore’s famous “open question” argument against normative naturalism. And a standard diagnosis of the latter argument is that, at best, it shows something distinctive about normative thought and talk, rather than normative metaphysics (e.g., see Darwall, Gibbard, & Railton 1997: 4–5). In light of these (and other) worries, we do not endorse the “just too different” intuition, or its probative force.¹³ Our point here is that this intuition and its variants are central to the contemporary case for non-naturalism. This entails strong pressure for non-naturalists to provide a formulation of the metaphysics of their view that squares this intuition with the sorts of intimate metaphysical connections that supervenience and grounding relations are used to explicate.

In light of this, the non-naturalist faces a delicate dialectical situation. The apparent supervenience and/or grounding relations between the normative and the natural might seem to raise the question of whether there could be a plausible metaphysical view that vindicates the “just too different” intuition. If such a view cannot be formulated, one might worry that the non-naturalist’s commitments are a metanormative wish list, rather than the characterization of a viable thesis. Because of this, we think that providing an informative formulation of the non-naturalist’s characteristic metaphysical commitment is a pressing task for anyone who wants to understand whether non-naturalism is a viable research program.

This dialectical situation also suggests some important desiderata for the task of formulating non-naturalism’s metaphysical commitments. First, the formulation should provide an informative metaphysical characterization of the thesis: “over-and-above” metaphors won’t cut it. Second, the formulation should fit with the core “just too different” motivation for the view. A crucial corollary of this second point is the main theme in what follows. This is that the formulation of non-naturalism should not overgeneralize. For example, if one is arguing for non-naturalism about ethical facts, with (purported) support from the “just too different” intuition, the formulation of non-naturalism that is used should not entail that non-naturalism also holds for the rules of chess, or about sociological facts, which many people (including many non-naturalists) do not have the same “just too different” intuitions about. With these desiderata in hand, we now turn to our first class of formulations of non-naturalism.

2. Grounding Accounts of Non-Naturalism

In this section, we introduce and evaluate the most prominent grounding formulation of non-naturalism, due to Gideon Rosen (2017). We argue that it overgeneralizes. We then briefly introduce our general diagnosis of why grounding is an unpromising resource for formulating non-naturalism.

Rosen’s formulation deploys the most influential contemporary way of regimenting ideas of metaphysical dependence and explanation. This approach appeals to the idea of grounding, an allegedly unified class of asymmetric metaphysical determination relations.¹⁴ To get a feel for the idea of grounding, consider a prosaic fact: that Roberto is smiling. Contrast two sorts of questions about explanation that we might be interested in with respect to this fact. First, we might be interested in explaining what caused Roberto to be smiling. A (partial) causal explanation of this fact might be that Roberto just heard a joke that he found funny. Second, we might note that it seems unlikely that Roberto’s smiling is a fundamental fact about reality. So we might want to know what more fundamental facts explain what his smiling consists in.

A simple proposal in response to the second question is that Roberto is smiling in virtue of the muscles in his face being configured in a certain way. This configuration did not cause his smiling (his hearing the joke did that). Rather, the configuration grounds it. More generally, grounding relations are supposed by proponents of grounding to be a (perhaps the) general form of asymmetric metaphysical dependence relations. We can distinguish full from partial grounds. If a certain configuration of Roberto’s facial muscles fully grounds the fact that he is smiling, then he is smiling wholly in virtue of these configuration facts. By contrast, a specific fact among those configuration facts—say, the shape of a certain muscle in Roberto’s cheek—partly grounds the fact that Roberto is smiling, if it is among the facts that fully grounds the latter fact. Grounding relations are generally assumed to entail metaphysical necessities. For example, if a certain configuration of Roberto’s facial muscles fully grounds the fact that he is smiling, then it is metaphysically necessary that: if his face is thus configured, then Roberto is smiling.¹⁵

With these clarifications in mind, we can consider Rosen’s formulation of non-naturalism:

We can better understand the motivation for Full Grounding if we accept a further assumption linking metaphysical grounding to another metaphysical idea: essence (and, in particular, what Kit Fine calls “constitutive essence”).¹⁷ On Fine’s understanding, the constitutive essence of a thing specifies its nature or identity. One of Fine’s motivating examples can be used to illustrate this idea. In every possible world, Socrates is the only member of the singleton set that contains only Socrates. But being a part of this set is not part of what it is to be Socrates, and hence not part of Socrates’s essence (Fine 1994).

With the idea of essence in hand, we can now ask: why does the configuration of Roberto’s face ground his smiling? A plausible answer is that this grounding fact follows from the essence of smiling: what it is for Roberto to be smiling just is for his face to have this configuration. According to an ambitious kind of essentialist about grounding, every grounding fact follows from some fact about essence in this way.¹⁸

If we assume the above ideas about ground and essence are on track, we can motivate the Full Grounding formulation as follows. Suppose that normative facts were fully grounded in non-normative facts and properties. Given the ambitious essentialist thesis we introduced above, this fact would need to follow from facts about the essences of either the relevant normative or non-normative properties. But you might think that this sort of essential entanglement between normative and non-normative properties is simply inconsistent with the guiding non-naturalist idea that such classes of properties are “just too different” from each other.

One worry about Rosen’s formulation and its motivation is that it involves a controversial collection of metaphysical commitments. For example, one might be suspicious of the idea that reality has the sort of grounding and essence structure that Fine and Rosen suppose.¹⁹ We set this sort of worry aside, for the following two reasons. First, we are seeking a formulation of non-naturalism that is congenial to robust non-naturalists, who are generally comfortable with more expansive metaphysical commitments. Many reasons to doubt the grounding and essentialism framework are motivated by a general approach to metaphysics that is antithetical to this expansiveness. Second, we are here seeking an adequate formulation of the core metaphysical commitment of non-naturalism, rather than a defense of that commitment. If grounding and essence are the best ways of formulating the non-naturalist’s thesis, we take it that the philosopher suspicious of the metaphysical or epistemic credentials of these notions should thereby be suspicious of the non-naturalist’s thesis, not its formulation. An analogy may be helpful. Many contemporary metaphysicians are suspicious of the idea that reality contains a necessarily existing perfect being. But that suspicion should not prevent us from formulating the metaphysical commitments of prominent versions of theism in terms of such a being.

Despite all of this, we now argue that we should reject Rosen’s formulation, because it overgeneralizes. To see the problem, consider what we will call the Painful Hypothesis, according to which (i) painfulness is a metaphysically fundamental property (that is, facts of the form painful(α) are not grounded in anything), (ii) painfulness is a natural property, and (iii) the normative property of badness is identical to painfulness. We want to emphasize that we intend condition (iii) about identity literally. The idea is not that the properties of painfulness and badness covary, or that instantiation of one of them grounds instantiation of the other. Rather, the idea is that there is just one property here, which we pick out both with the word ‘painfulness’ and with the word ‘badness’. To make this salient below, we will sometimes call this property ‘painfulness/badness’.

Each part of this hypothesis has philosophical defenders. On (i): painfulness is plausibly a phenomenal property, and one might think that some phenomenal properties are fundamental. This fundamentality claim, while obviously controversial, fits naturally with a Chalmers-style property dualism in the philosophy of mind (Chalmers 1996). On (ii): it is controversial what makes a property a “natural” one, but on some views, fundamental phenomenal properties can be natural. For example, the epistemology of phenomenal properties is experiential, such properties can seemingly be causally efficacious, and (as Chalmers argues), property dualism is compatible with the potential for developing a natural science investigating the nomic relations linking phenomenal and (e.g.) neurological properties (see Chalmers 1996: ch. 6 for a defense of this final claim). Finally, the identification of badness with painfulness in (iii) could be part of an ambitious form of reductive naturalism based on hedonic consequentialism.²⁰

Full Grounding classifies the Painful Hypothesis as “non-naturalist”. Here is why. First, condition (iii) of the hypothesis says that there is a single property: badness/painfulness, which is normative. And condition (i) says that this property is fundamental. Because this property is fundamental, no fact of the form bad(α)—for example, Omar’s leg wound is bad—will be grounded in other facts. On Rosen’s formulation, this entails non-naturalism about badness. This is extremely implausible, because condition (ii) of the Painful Hypothesis entails that painfulness/badness is a natural property.

We can offer a general diagnosis of what has gone wrong with Full Grounding. Grounding relations mark relations of metaphysical determination, and hence of relative fundamentality. The case we have discussed causes trouble because fundamentality is just not the same thing as non-naturalness, as many natural properties are presumably perfectly fundamental.²¹ (The Painful Hypothesis makes the problem vivid precisely by considering the hypothesis that one of these fundamental natural properties is also normative.) Because of this, we should be suspicious quite generally of attempts to formulate non-naturalism in terms of grounding.

One might be tempted to defend Full Grounding by denying that the Painful Hypothesis is an apt counterexample. Perhaps, for example, this hypothesis is in some non-obvious way incoherent. We do not find this line of resistance promising. Notice that the general form of the challenge posed by the Painful Hypothesis is simply that Full Grounding will misclassify any hypothesis on which there is a property that is (i) fundamental, (ii) natural, and (iii) normative. Since it is highly plausible that a property can be both natural and fundamental, the general worry would have to be either that the very idea of a natural normative property is incoherent, or that there is some special problem with the idea of fundamental natural properties. We want to grant that one of these ideas may turn out to be metaphysically impossible. But in general we do not want our taxonomies to simply disregard controversial metaphysical impossibilities. Consider an analogy. Many philosophers working in metaethics would agree that—even if it turns out that a form of utilitarianism is correct on which the right thing to do is to maximize happiness—it is false that rightness is identical to happiness maximization, and that this is so as a matter of metaphysical necessity. But, even if this widespread thought turned out to be correct, it would be a great embarrassment if a taxonomy classified the hypothesis that rightness is identical to happiness maximization as a form of non-naturalism.

Alternatively, one might adapt Rosen’s reply to a closely related taxonomic worry: that this sort of hypothesis is “too far-fetched to bother with” (2020: 211, note 11). We disagree. The core reason is this: we think that, in this context, as with many other contexts in philosophy, a taxonomy of metaphysical positions should attempt to capture what certain sorts of views consist in, rather than simply providing helpful heuristics for identifying instances of certain sorts of views. If that is correct, then even “far-fetched” counterexamples to taxonomic principles should worry us. In certain cases, if we can’t make sufficient progress on this goal, we might of course have to settle for less, and simply stick with rough heuristics that help us distinguish different kinds of views. However, in the case at hand, we are hardly forced into such a retreat. In fact, our counterexample helps to illustrate an attractive diagnosis of what has gone wrong with Full Grounding: that the latter is using the wrong sort of metaphysical equipment to capture the intuitive idea of non-naturalness. Moreover, as we argue at the end of this paper, we have equipment that can allow us to do better, and thereby help us meet the more ambitious kind of aim for the sort of taxonomical project we have just sketched.

3. Essentialist Formulations of Non-Naturalism

As we noted above, contemporary discussions of grounding often link the idea of grounding to the idea of essence. Given the diagnosis that we have just offered, it appears more promising to formulate non-naturalism in terms of essence than in terms of grounding, because essences concern the natures of things, as opposed to dependence. In this section, we consider two essentialist formulations of non-naturalism put forward by Stephanie Leary and Gideon Rosen, and show that these formulations also overgeneralize in similar ways to the grounding formulations.

We begin with the Leary’s formulation:

It is worth clarifying the relevant notion of specification here. It is not enough for specification that we have some way of picking out the essence of a property. If we could do that, we could offer “specifications” like the essence of G. E. Moore’s favorite property, and Leary’s Specification Essentialism would be useless as a way of formulating non-naturalism. We should not read Leary this way. Rather, a specification must be understood as something like a constitutive characterization of the property’s essence.

Specification Essentialism faces the same problem as Full Grounding. Again, consider the Painful Hypothesis: that badness is identical to phenomenal badness (i.e., painfulness), which is a fundamental natural property. Because the property is supposed to be fundamental, there is no reason to suspect that its essence can be specified in non-normative terms, or in any non-circular terms. The same goes for its instantiation conditions: if this property is fundamental, facts about its instantiation are ungrounded. And so there will be no distinct metaphysically sufficient conditions for its instantiation encoded in its essence. So, on the Painful Hypothesis, Leary’s Formulation classifies phenomenal painfulness/badness as a “non-natural” property. And, as we explained in the previous section, this result runs counter to deep non-naturalist commitments.

Rosen has recently offered his own essentialist formulation of non-naturalism:

Rosen’s formulation appeals to the idea of real definition. The “real” in “real definition” signals that a real definition is meant to define an object or property, rather than the concepts or words we use to think or talk about it. So, for example, one might attempt to offer a real definition of justice, rather than of the word ‘justice’. Rosen’s preferred account of real definition is this: for Φ to be the real definition of F is for it to follow from the essence of F that, if Fx or Φx, Φx fully grounds Fx (Rosen 2015: 200). Thus, according to Rosen, real definition can be understood in terms of a combination of essence and grounding.

With Rosen’s account of real definition in hand, we can see that Definitional Essentialism is also going to misclassify the Painful Hypothesis. Consider the property of painfulness/badness, posited by this hypothesis. Because this property is fundamental, instantiations of this property are not fully grounded in anything. So, by Rosen’s account of real definition, this property lacks a real definition. Trivially, it thus lacks a real definition in non-normative terms, and hence it is classified as non-natural by Definitional Essentialism.

To be clear, we take Rosen’s formulation of real definition to be important and well-motivated (while, inevitably, controversial). Our objection is not to his formulation of this idea, but to the idea that we can use the idea, so defined, to characterize non-naturalism.

We can again offer a general explanation of why these formulations fail: like Full Grounding, they are each best understood as securing the fundamentality of the normative, as opposed to any deep contrast with the natural. Thus, they will overgeneralize on hypotheses like the Painful Hypothesis, according to which normative properties are both fundamental and natural.

4. Making Non-Naturalism Compatible with Grounded Normativity

As we have mentioned, one central challenge facing the non-naturalist is that there appear to be extremely intimate metaphysical connections between normative and natural properties, and it is not clear whether non-naturalism is compatible with the existence of such connections. One candidate for such a connection is the idea that normative facts are fully grounded in the natural facts. In this section, we briefly explain why this idea is credible, and why it appears to pose a challenge to the non-naturalist. We then consider two recent proposals, due to Stephanie Leary and Selim Berker, that aim to explain how certain full grounding theses can be compatible with non-naturalism. We argue that these attempts themselves overgeneralize in objectionable ways.

To begin, consider an initial case for the idea that every normative fact is fully grounded. This starts from the observation that whenever there is a normative fact, that fact seemingly holds because another fact holds. For example, Sadat’s headache is bad because it hurts; a paternalistic act is wrong because it fails to respect the agency of its target. Notice next that the “because” used in these explanations appears to signal a grounding explanation, and not (e.g.) a causal link.²⁴ Notice finally how bizarre it would be to say “such-and-such state is bad, but not in virtue of anything”. On such a view, the state would be bad not because it involves pain, or the frustration of desires, or ignorance, or loneliness, etc.; it’s just bad all by itself. These observations suggest, at least at first blush, that every normative fact is grounded.

How seriously should we take the apparent grounding connections between normative and natural properties? For much of the last fifty years, it has been common to formulate the core “intimate metaphysical relation” between normative and natural properties in terms of supervenience. Supervenience, at least as standardly understood by contemporary philosophers, is a purely modal notion. Put roughly, to say that the X properties supervene on the Y properties is to say that there can’t be a difference in the distribution of the X properties without a difference in the distribution of the Y properties. The core claim about the supervenience of the normative on the natural that has driven metanormative debate can be put crudely as follows: there can be no normative difference between naturalistically identical possible worlds.²⁵ One might be tempted to think that we could just use such a supervenience thesis to fully account for the intimate metaphysical relation between normative and natural properties, and leave issues about grounding aside.

There are two important reasons to reject this idea. First, as Michael DePaul (1987) and Selim Berker (2018: §3) have shown, many of the canonical discussions of supervenience in ethics clearly treat “supervenience” either as a metaphysical dependence relation, or as a proxy for one.²⁶ Thus, many defenses of supervenience in ethics might already best be understood as defenses of full grounding in ethics. Second, supervenience is a modal relation. Accepting supervenience thus leaves entirely open what explains why the relevant modal pattern obtains. It is common to argue that the necessary connections encoded in a supervenience thesis “call out” for explanation (see McPherson 2012). As Ralf Bader has pointed out, the claim that the natural fully grounds the normative seems tailor-made to do this explanatory work (2017: 116).

We can see the apparent problem for non-naturalism immediately when we consider Full Grounding. For on this account of non-naturalism, if every normative fact is fully grounded in the non-normative facts, non-naturalism is simply false. Setting this aside, we have already mentioned the general form of the challenge now at hand: full grounding is a very intimate metaphysical relation. So the non-naturalist seems to face the burden of explaining how it could be compatible with their view, according to which normative and natural facts are, metaphysically, very different.

One option for the non-naturalist that we will not discuss at length in this paper is to simply deny that the normative either metaphysically supervenes on, or is fully metaphysically grounded in, the natural.²⁷ Perhaps the non-naturalist is ultimately forced to embrace this option (for an argument that this is so, see Rosen 2020). We return to this idea briefly later on in this paper. For now, we note that in our view, rejecting these commitments is both substantively implausible, and runs counter to the clear-eyed commitments of many non-naturalists.

5. A Diagnosis and a Promising Alternative

In this paper, we have discussed several of the most prominent recent discussions of what the core metaphysical commitments of non-naturalistic realism might amount to, and how to explain the intimate metaphysical connections between the normative and the natural in a way that doesn’t conflict with non-naturalism. We have argued that each of these proposals overgeneralizes in implausible ways.

These formulations rely on the notions of metaphysical grounding and essence. This reliance raises a question: are the failures of these formulations due to idiosyncratic weaknesses, or do they arise from a deeper mismatch between the aforementioned notions and the task of formulating non-naturalism? In this section, we argue that there is a deep mismatch in the case of grounding, but not in the case of essence. We then sketch what we take to be a promising alternative framework for formulating non-naturalism’s metaphysical commitments.

We can understand why grounding is an unpromising tool for formulating non-naturalism by relating the canonical “just too different” motivation for non-naturalism to two sorts of metaphysical relations. The first sort of metaphysical relations we have in mind are metaphysical dependence relations between the more and less fundamental. We assume here that these relations are aptly regimented by the theory of grounding.³³ The second sort of metaphysical relations are objective similarities and dissimilarities between classes of properties. Consider an example where these two sorts of metaphysical relations come apart. Suppose that a strong form of phenomenal dualism is true, such that phenomenal redness, phenomenal orangeness, and negative charge are all perfectly fundamental properties. Ipso facto, there are no grounding relations among facts about the instantiation of these properties. However, plausibly, phenomenal redness is more objectively similar to phenomenal orangeness than it is to negative charge. This example shows that at least in some cases, there is no prospect of explaining objective similarity facts by appealing to grounding relations.

Now consider the canonical motivation for contemporary non-naturalism, the so-called “just too different” intuition that we introduced at the start of the paper. We can ask: what is the most plausible way of directly reflecting this motivation in metaphysical terms? On its face, if the “just too different” intuition is about something metaphysical, it is about objective dissimilarity. This suggests a deep diagnosis for why we should expect grounding formulations of non-naturalism to fail: as the red/orange/negative charge example showed, objective (dis-)similarity cannot be understood simply in terms of grounding. And so we can predict that the implausibility of the specific grounding formulations of non-naturalism we have considered is not a consequence of their idiosyncrasies. Rather, we should expect that any formulation that understands non-naturalism purely (or even just mostly) in terms of grounding is going to fail to adequately capture the “just too different” motivation for non-naturalism.

We think the prospects are different for essentialist accounts of non-naturalism. If we accept an essentialist framework, it appears plausible that objective (dis-)similarity will be a matter of essence. For example, the fact that a lot of people have thought about both normative and physical properties is not an essential fact about both properties, and thereby plausibly fails to show that they are objectively similar. By contrast, if, as some believe, the funny and the disgusting are each essentially response-dependent, one might think this constitutes a dimension of objective similarity between them. There are reasons to be optimistic that an essentialist could formulate non-naturalism precisely in terms of the objective (dis-)similarity between properties, where that in turn is understood as a relationship among essences.

The failure of the essentialist formulations we have considered in this paper is thus more plausibly a matter of the specific content of those formulations, and not of the fact that they are essentialist. For example, recall that Rosen’s Definitional Essentialism invoked the idea of real definition, which Rosen in turn understands in terms that combine essence with grounding. This means that Definitional Essentialism faces the same diagnosis as the grounding formulations: objective similarity cannot be captured in terms of grounding. But the problem arises, as it were, because this formulation is definitional (using Rosen’s account of real definition), not because it is essentialist.³⁴

We have suggested that the idea of essence is a more promising way to formulate non-naturalism’s distinctive metaphysical commitments in terms of objective similarity. We cannot fully defend this idea here, but we will seek to briefly substantiate it. For concreteness, consider one such formulation:

We unpack this in two steps.

First, Similarity Non-naturalism says that the (relevantly) normative is an objective similarity class.³⁶ This means that (i) the essences of normative properties are all similar in being normative, and (ii) this is an objective similarity (like that shared by all negatively charged things), not merely a way that we happen to group properties together.

Second, Similarity Non-naturalism says that this similarity is sui generis. Here is a first pass gloss on this idea: for a similarity to be sui generis is for it not to constitute a species or subclass of some broader objective similarity. For example, the biological properties might well form an objective similarity class. But it is plausible that this is a subclass of the natural properties: that is, the essence of every biological property also shares the objective similarity that constitutes the natural (supposing there is one). In light of this, the biological is not (on this picture) sui generis. Here is a worry about the first pass gloss: it might be that being a property is itself a dimension of objective similarity (assuming, for example, that the property/object distinction is a dimension of objective similarity). If so, every class of properties will be a subclass of the class properties. Here is one way that the similarity non-naturalist might seek to finesse this issue. The core idea is that there is a unified kind of metaphysical question that is being asked when we think about classifications like [natural, physical, biological, phenomenal, mathematical supernatural . . .], which is different from the sort of metaphysical question being asked when we (e.g.) distinguish properties from objects. Of course, it is controversial what classes of properties belong on this list. Our proposal is that for a class of properties to be sui generis is for it to belong on that classificatory list, and for it not to be a subclass of another class of properties on that list.

So: for the normative to be sui generis is for being normative to be a dimension of objective similarity, and for this not to be a subclass of another dimension of the relevant kind (i.e., [natural, phenomenal, etc.]). This characterization is, we submit, tailor-made to vindicate the “just too different” intuition: on this view, that intuition functions precisely to alert us to the status of the normative as a sui generis objective similarity class.

The next point we want to emphasize is that Similarity Non-naturalism does not face the sorts of problem cases that plagued the accounts discussed thus far. For example, recall the Painful Hypothesis, which identifies badness with phenomenal painfulness. If we suppose that phenomenal painfulness is a naturalistic property, then this property is a member of the classes natural and phenomenal. If all normative properties are like this, then the normative will in fact be a subclass of the natural, and hence not sui generis, and hence not non-natural, given Similarity Non-naturalism. If we suppose instead that the phenomenal is itself a sui generis objective similarity class, then the Painful Hypothesis entails that, while naturalism in the ordinary sense is not true of the normative (for some normative properties are not members of the naturalistic similarity class), Similarity Non-naturalism is not true either. This is because, on this hypothesis, the normative is not sui generis in the objective similarity sense we have just spelled out.³⁷

Next, consider the hypotheses we considered in the preceding section. For example, suppose that humorousness facts are all fully grounded in “hybrid humor facts”. If we suppose that the humorous is part of the naturalistic similarity class, nothing about this grounding story entails non-naturalism about humor, given Similarity Non-naturalism. And there is no danger on this account that views that identify normative properties with natural properties will count as a version of non-naturalism (as they would on Berker’s higher-order grounding proposal), even if we take the alleged identity to itself be a normative fact.

Of course, none of this shows that Similarity Non-naturalism is correct, or free of problems. But it does suggest what a more promising strategy for formulating non-naturalism might be, in light of the diagnosis we have offered of the failures of the accounts discussed thus far.

6. Conclusion

We began this paper by arguing that non-naturalists face a substantial burden to offer a clear formulation of their central metaphysical commitments. The main task of the paper has been to show that the most influential contemporary framework for discharging this burden—one which appeals centrally to grounding relations—is ill-suited to the task. We have argued that the most prominent efforts to formulate the metaphysical commitments of non-naturalism in terms of grounding and essence each overgeneralize in implausible ways by classifying certain plausibly naturalistic metaphysical hypotheses as non-naturalistic. And we have suggested that an alternative strategy, which involves formulating non-naturalism in terms of objective similarity, is more promising. As we argued, this strategy can be naturally implemented within an essentialist metaphysical framework. What is crucial is that such an essentialist account would focus on objective similarity rather than dependence.

We conclude the paper by sketching three ways in which the arguments just summarized are illuminating. We first explain the importance of our discussion for understanding our metanormative options. We then explain its significance for the core metaphysical puzzles facing non-naturalism, before drawing some general metaphysical lessons.

We have argued in this paper that grounding formulations of non-naturalism fail to reflect what we take to be central commitments motivating contemporary robust non-naturalists. Based on this, we have argued that such formulations fail to provide a helpful way of regimenting the terminology of ‘non-naturalism’ and should not be used for contemporary metanormative discussion.³⁸ Note, however, that our argument is fully compatible with the thought that the condition mentioned (e.g.) in Full Grounding might be satisfied. That is, there might be normative facts which are metaphysically fundamental. This view might aptly be called metanormative fundamentalism. The point we want to emphasize here is that metanormative fundamentalism should be distinguished from non-naturalism. Indeed, we think the fundamental/non-fundamental distinction cross-cuts the natural/non-natural distinction. We think that distinguishing these views helps us to better understand the space of possible metanormative views. To drive this point home, consider a view that combines Similarity Non-naturalism with the idea that every normative fact is fully grounded in some collection of non-normative facts. This view is very different from fundamentalism. Indeed, it suggests a view that we might aptly call emergentist non-naturalism.³⁹ This is because this view suggests that if we “put together” certain natural conditions in the right way, something totally different from anything natural emerges.⁴⁰

Our second theme is the significance of our discussion for some of the central metaphysical debates concerning non-naturalism. As we have noted in §4, many non-naturalists find it highly plausible that the normative metaphysically supervenes on the natural, and there also appear to be strong prima facie reasons to accept that the normative is fully grounded. Formulating non-naturalism in terms of objective similarity rather than grounding arguably improves the non-naturalist’s prospects for embracing the idea that the normative is fully grounded. (Indeed, this is just the combination that we dubbed “emergentism” above.) Even if mistaken, it is not obviously incoherent to think that properties in one objective similarity class are fully grounded in properties in another objective similarity class, or supervene on them as a matter of metaphysical necessity. This is not to say the problems for non-naturalists tied to endorsing strong claims about supervenience and grounding disappear given this formulation. Indeed, we think that explaining the supervenience of the normative on the natural remains one of the deepest problems facing non-naturalistic realism, and that the existing proposals for how to address it are all unsatisfying.⁴¹ We think that the parallel challenge involving the grounding of the normative in the natural is a deep one as well. However, if our discussion in this paper is correct, the upshot of both of these challenges may be a matter of “plausibility points”, as opposed to outright inconsistency, as it is on (e.g.) Rosen’s formulation of non-naturalism (see Rosen 2020).⁴²

Our final point concerns the broader significance of our discussion for metaphysics. As we have noted, it has been very common to formulate monistic theses in metaphysics (most prominently, physicalism) in terms of supervenience, and, more recently, grounding (see Rosen 2010 and Dasgupta 2014 for discussion). Our discussion in this paper suggests that this strategy should be met with suspicion. Consider a variant of the sort of non-naturalist “emergentism” we sketched above. On this view, Similarity Non-naturalism is true, but the normative facts are fully grounded in the physical. Now, it may well be that such views are impossible. But consider straightforward grounding formulations of physicalism, according to which, put roughly, physicalism is the thesis that the relevant target set of facts (e.g., facts about the instantiation of qualitative mental properties) are either grounded in (or identical to) facts about the instantiation of some specific set of privileged “physical” facts. Such views entail that if the kind of “emergentist” views we just sketched above are possible, then they are “physicalist” hypotheses. We think this is an implausible result. Rather, we think that the relevant sort of “emergentist” view is a paradigm model of how physicalism might conceivably be false. We suspect that grounding formulations of physicalism implicitly assume the impossibility of such emergentist views. This may be right, but in our view, such impossibilities should be argued for, rather than obscured by tendentious taxonomy, such as grounding and supervenience formulations of physicalism.