The first desideratum for a good characterization of non-naturalism is that it be

What are these pre-theoretical claims? The first claim is that normative properties and facts are entirely of their own kind (sui generis). Shafer-Landau (2003: 55) says that his view, following Moore, is that “moral facts are different in kind from any other.” And Parfit (2011: vol. 2, pt. 6) and Scanlon (2014: ch. 2) claim that normative truths are “irreducibly normative”. So, the issue that concerns non-naturalists is not just whether normative properties and facts are natural—or as Enoch (2011: 134) says, whether they’re the same “kind of stuff our best science says (or will shortly say) that there is.” The issue is whether normative properties and facts are different in kind from any other kinds of stuff there might be. Non-naturalists are thus equally at odds with supernaturalists like Divine Command Theorists, who claim that normative properties are supernatural properties like being commanded by God (McPherson 2012: 210).

The non-naturalist’s claim that normative properties are sui generis is mysterious, though, because it’s not clear what kind-talk for properties amounts to. One might think that two properties are entirely different in kind if and only if they don’t share any second-order properties. But any two properties share some arbitrary property: being good and being an electron share the property of being a property and being normative or physical (Leary 2017: 79, fn. 12). Instead, one might appeal to the notion of elite properties to distinguish between metaphysically significant and arbitrary properties. According to Lewis (1983), some properties are more elite than others in the sense that they make for more objective similarity and are more explanatory, and so, the concepts and terms that refer to more elite properties better carve reality at its joints.¹ So, one might take the claim that normative properties are entirely different in kind to amount to the claim that natural properties and supernatural properties (and whatever other kinds of properties there might be) each have something in common that makes for a high degree of objective similarity, but which normative properties lack. But since natural properties are so diverse (e.g., having negative charge, being in pain, being a toaster, being a city, etc.), the assumption that they all have something in common that makes for a high degree of objective similarity seems highly implausible. So, what makes normative properties of the same kind or different in kind from any other sets of properties can’t be a matter of sharing or lacking objective similarities with those properties. An ILLUMINATING characterization of non-naturalism should thus elucidate the non-naturalist’s pre-theoretical claim that normative properties are of their own kind by explaining what such kind-talk amounts to in precise metaphysical terms.

A related pre-theoretical claim that non-naturalists are committed to, which aligns them with supernaturalists, is that countenancing normative properties is incompatible with an entirely scientific view of reality. Enoch (2011: 135) and Scanlon (2014: 19) point out that non-naturalism is logically consistent with our best scientific theories, since these theories do not explicitly claim that there are no non-natural normative properties or contain an “and that’s all there is” clause. But so long as non-naturalists take normative properties to be different in kind from all the “stuff of science”, they are committed to claiming that there is some sense in which science cannot provide a complete account of reality.² An ILLUMINATING characterization of non-naturalism should thus also clarify in what sense the sciences fail to provide a full account of reality.³

Moreover, a good characterization of non-naturalism should also be

I’ll explain each of these in turn.

SUBSTANTIVE & LOCAL is straightforward. Non-naturalism is not trivially true or trivially false, and it’s a view about normative properties and facts in particular. So, a good characterization of non-naturalism cannot be a trivial claim nor one that can be easily settled by general metaphysical considerations that have nothing to do with normative properties in particular.

FITTING and GENERAL require more explanation. Since non-naturalism is often stipulatively defined in the literature, a good characterization of non-naturalism need not classify all self-proclaimed non-naturalists as genuinely non-naturalist. Sometimes philosophical progress will result in a recategorization of people’s views. But a good characterization of non-naturalism should at least aim to classify paradigm non-naturalists like G. E. Moore as genuinely non-naturalist. And it should also aim to do so in a way that best captures the spirit of the view. Consider an analogy: suppose that all the vegans in my class think wearing leather is disgusting, but everyone else in my class likes to wear leather. I could correctly taxonomize my vegan and non-vegan students as those who don’t like to wear leather and those who do. But this taxonomy doesn’t capture the spirit of my vegans’ view: they don’t have a clothing preference, but instead an underlying moral view that it’s wrong to consume animal products. So, a better taxonomy is along these lines. For any taxonomy, then, we should aim to classify paradigm parties to the relevant debate in a way that best captures the spirit of those parties’ views. That’s FITTING.

But when we’re trying to specify what a philosophical view P amounts to, we’re not just interested in the contingent, historical question of what paradigm defenders of P have in common. We’re also interested in a further conceptual question: what is it that all possible views that capture the spirit of P have in common? This is what makes taxonomical questions substantive and interesting. They’re not just about categorizing existing literature, but they’re about carving up the conceptual space of views into certain families that share the same concerns, including those spaces that are not occupied by anyone thus far (perhaps even for good reason). So, a good characterization of non-naturalism should aim to not only correctly classify paradigm non-naturalists in the literature, but also to offer the broadest characterization that captures all possible views that speak to the pre-theoretical concerns of non-naturalists. That’s GENERAL.

But both FITTING and GENERAL involve “as much as possible” clauses, which serve two important functions. First, it makes these two desiderata inherently comparative: whether a characterization is FITTING or GENERAL depends on whether there are alternative characterizations that do a better job of classifying paradigm non-naturalists or providing a more general characterization of the view that captures their pre-theoretical concerns. Second, these “as much as possible” clauses signal that each desideratum needs to be weighed against the other desiderata. It may be that the only characterization of non-naturalism that is ILLUMINATING and SUBSTANTIVE & LOCAL provides a slightly less general characterization of the view than alternative characterizations or requires recategorizing some paradigm non-naturalists from the literature. If so, that would still be the best characterization of the view. So, the idea here is that the best characterization of non-naturalism is the one that strikes the best balance amongst all four of these desiderata.

Of course, it could turn out that there is no single best characterization of the view: there could be multiple characterizations that each do an equally good job of meeting the desiderata and different characterizations might better fit different paradigm non-naturalists. My methodology here does not rule out this possibility. But I will ultimately argue that there is one characterization that is best: I argue that an essence-characterization does a better job of meeting these desiderata than identity or grounding characterizations.

But before I get there, one more disclaimer: notice that these desiderata ignore how well the resulting taxonomy fits with paradigm naturalists. This is because, since non-naturalism has been poorly defined since it was first presented by Moore, I suspect the naturalist reaction to Moore may have been motivated by very different pre-theoretical concerns. So, if we try to specify the debate in a way that captures the pre-theoretical concerns of naturalists, we may end up with a different taxonomy. If so, that would show that naturalists and non-naturalists have been talking past one another all along. I remain neutral about this here and am simply making the first step in diagnosing the debate by focusing on what non-naturalism amounts to.

One might think that non-naturalism is just the following claim:

But the problem is that there’s no way of specifying the “other kinds of properties and facts” in such a way that makes Non-Identity SUBSTANTIVE & LOCAL and ILLUMINATING.

If we specify “other kinds of properties and facts” as non-normative ones, where these are properties and facts that are metaphysically distinct from normative properties and facts, Non-Identity is trivially true. A property cannot be both metaphysically normative and non-normative. This is why Jackson (1998) and Shafer-Landau (2003) instead specify the “other kinds of properties and facts” as descriptive ones, where this is metalinguistic rather than metaphysical. That is, descriptive properties and facts are just any properties and facts that are referred to by descriptive, non-normative terms and normative properties and facts are those referred to by normative terms. Non-Identity, thus understood, states that normative predicates don’t refer to the same properties as descriptive predicates.

But this metalinguistic characterization renders Non-Identity trivially false. As Jackson (1998: 119, fn. 10) himself notes, if I’m thinking about rightness, ‘the property that I’m thinking of right now’ is a descriptive predicate that may refer to the same property that ‘right’ refers to. In order to avoid Non-Identity being trivially false, then, Jackson seems to restrict the set of descriptive predicates relevant to Non-Identity to only those that don’t have that sort of form (‘the property such that … ’). But Dunaway (2015: 632) points out that the descriptive predicate ‘is actually being thought about by me’ lacks this form but may still refer to the same property as ‘right’. So, Dunaway suggests that we restrict the set of descriptive predicates relevant to Non-Identity to those that do not merely accidentally refer to some property because they involve indexicals like “I”, “actually”, and “now”.

But even with these restrictions, Non-Identity still seems false because there are other descriptive predicates that don’t involve indexicals or have the relevant form but nonetheless seem to refer to the same properties as normative terms. For example, ‘plays a central role in deliberation about what to do, praising and blaming practices, and reactive attitudes’ seems like a descriptive predicate that refers to normative properties. And we can even imagine a simple descriptive predicate that refers to a particular normative property. Consider Eklund’s (2017: 75) hypothetical aliens who, when learning English, have an “unspecific or mistaken idea of what [‘right’] might stand for” and introduce the term ‘thgir’ into their language to refer to what we refer to with our term ‘right’; but ‘thgir’ doesn’t play the sort of conceptual role for them that is characteristic of our term ‘right’ (‘thgir’ doesn’t have a strong connection to their deliberation, motivation, and reactive attitudes). Because of its different conceptual role, ‘thgir’ seems like a descriptive predicate rather than a normative one, but it plausibly refers to the same property as ‘right’. So, just by reflecting on how we can use language, we see that it’s false that normative predicates don’t refer to the same properties that descriptive predicates refer to. Non-Identity, understood in this metalinguistic way, is thus trivially false.

Moreover, it seems misguided to construe non-naturalism metalinguistically in the first place. What Eklund’s ‘thgir’ example shows is that normative and descriptive terms have different conceptual roles but may nonetheless refer to the same property. Many naturalists accept this and claim that normative terms may refer to natural properties, even if normative terms don’t have the very same meaning as natural terms. But even non-naturalists should be able to accept that normative terms and descriptive terms can co-refer—to non-natural normative properties. This is because the non-naturalist’s central pre-theoretical concerns are about metaphysics and not about language: whether normative properties and facts are sui generis and something beyond the “stuff of science”. Whether we can refer to that same sui generis stuff with descriptive language seems beside the point. Non-Identity, understood metalinguistically, is thus not only false but doesn’t capture the pre-theoretical concerns of non-naturalists.⁴

This suggests that we need a metaphysical specification of the “other kinds of properties and facts” in Non-Identity. But how do we avoid making it trivially true? We could simply list all the other kinds of properties and facts there are and specify non-naturalism as the claim that normative properties are not identical to anything on the list:

But the problem with Non-Identity_List is that it presupposes, rather than elucidates, the notion of properties being of the same kind. For example, it presupposes that certain properties are natural—that is, of the same kind as paradigmatic scientific properties—and states that normative properties are not identical to any of those properties. But Non-Identity_List does not tell us what makes having negative charge, being in pain, being a toaster, and being a city natural properties.

So, consider a view according to which every normative property is identical to some long disjunction of paradigmatic scientific properties (Jackson 1998). Is this a view according to which Non-Identity_List is true or false? To answer this, we need to know whether a long disjunction of scientific properties is a natural property. But Non-Identity_List doesn’t give us an answer. Of course, it may seem obvious that a disjunctive property is of the same kind as its disjuncts. But this simply presupposes the pre-theoretical notion that we are trying to elucidate and Non-Identity_List doesn’t help. So, for Non-Identity_List to be ILLUMINATING, we need a supplemental account of what it is for properties to be of the same kind.

One candidate account is in terms of grounding: perhaps what makes having negative charge, being in pain, being a toaster, and being a city natural properties is that all facts involving these properties are ultimately fully grounded in paradigmatic scientific facts. And since all facts involving disjunctions of those natural properties are also ultimately fully grounded in paradigmatic scientific facts, those disjunctive properties are natural too. This is a currently mainstream proposal to which we’ll now turn.

While the grounding terminology is relatively new, it’s meant to capture an old, familiar idea. We often say that certain facts obtain in virtue of other facts, where this “in virtue of” talk is not about causation, logical entailment, or semantics. For example, I might say that I’m identical to the five-year-old in that photograph in virtue of the fact that we’re psychologically continuous, or that Michelangelo’s David has a six-pack in virtue of the fact that the hunk of marble is S-shaped. I’m not saying that the fact that I’m psychologically continuous with the photographed five-year-old caused us to be identical or that it logically entails or literally means that we are. Nor am I saying that the hunk of marble’s shape caused David to have a six-pack, or that it logically entails or literally means as much. In each case, I’m saying that some fact metaphysically determines and explains another fact in a non-causal sort of way. Grounding is that non-causal metaphysical determination.

There are many controversial debates about grounding, but we can mostly ignore them here.⁵ I assume that grounding is a relation that holds between facts, but everything I say here could be translated to accommodate other views. I also assume that grounding is typically transitive: if x grounds y, and y grounds z, then x grounds z (perhaps with a few exceptions). And I assume there’s full and partial grounding: the fact that there’s stuff in my glass is fully grounded in the fact that there’s beer in it, whereas the fact that the beer is cold and hoppy is partially grounded in the fact that it’s cold and partially grounded in the fact that it’s hoppy. These minimal assumptions suffice for my purposes.

As suggested earlier, grounding seems apt to elucidate kind-talk for properties. We can think of properties being of the same kind insofar as their instantiations are all fully grounded in the same set of facts: if all facts involving F are fully grounded in facts that involve properties G, H, and I, then F is of the same kind as G, H, and I. So, we can define natural properties as follows:

For any F:

The naturalist’s claim that all normative properties are natural is thus the claim that all normative facts are either identical to or fully grounded in paradigmatic scientific facts. The supernaturalist denies this, insisting that normative facts are either identical to or at least partially grounded in facts involving supernatural entitles like God.

What unites the naturalist and the supernaturalist, though, is that they both endorse

Whereas the non-naturalist endorses

On this grounding-characterization, what it is for normative properties to be sui generis is for their instantiations to not be fully metaphysically explained by the instantiations of paradigm non-normative properties⁶. So, countenancing normative properties is incompatible with a purely scientific view of reality that says that all facts are fully metaphysically explained by scientific facts. Anti-Full-Ground thereby speaks to and elucidates the non-naturalist’s pre-theoretical commitments—it’s ILLUMINATING.

In light of the above discussion of Moore, one might think we should simply characterize non-naturalism in terms of real definition. Rosen (2017) rejects this idea because it doesn’t allow room for a non-reductive naturalist opponent. He notes that some physicalists in the philosophy of science claim that higher-level scientific properties are not definable in terms of fundamental physical properties, and that an analogous naturalist view about the normative should be intelligible. But recall that it’s an open question whether naturalists and non-naturalists share the same pre-theoretical concerns. So, if the characterization of non-naturalism that best meets our desiderata ultimately doesn’t allow room for a non-reductive naturalist opponent, that’s reason to think that so-called “non-reductive naturalists” are not engaged in a genuine disagreement with non-naturalists after all. It’s not reason to reject the characterization.

Regardless, though, my proposal here is that we characterize non-naturalism using the related but broader notion of essence, which does make room for a non-reductive naturalist opponent.¹⁵ The essence of x is simply the set of propositions that directly state the nature of x (Fine 1994)¹⁶, whereas a real definition of x is a proposition that directly states the nature of x in a particular form: “To be x is to be φ”, where φ is a complex condition not containing x (Rosen 2017). Essence is thus broader than real definition because the essence of x may contain some proposition that directly states the nature of x, but which cannot be put in the form of a real definition. For example, consider disjunction: we cannot define what it is for p or q to obtain in terms of a condition not containing “or”, but it nonetheless seems to be essential of disjunction that disjunctive facts are grounded in their disjuncts (Rosen 2010: 131). So, while x’s real definition is always at least part of x’s essence, x’s essence (or part thereof) may not be a real definition; and x may even have an essence, without having a real definition at all.

We can now distinguish between two different views on which normative properties have no real definition. On the first, while basic normative properties lack real definitions, they nonetheless have essences that include propositions that specify natural sufficient conditions for their instantiation. For example, on this view, the essence of goodness may include that, if x is pleasurable, x is good, but since goodness is multiply realizable, being pleasurable does not define what it is to be good (nor is goodness merely the disjunction of its realizers). On the second view, basic normative properties not only have no real definitions, but their essences also don’t include propositions that specify non-normative sufficient conditions for their instantiation. The former view captures the spirit of non-reductive naturalism, whereas the latter view captures the spirit of Moore’s view.

So, we can use essence to characterize the non-naturalist’s opponents as follows. First, since essences are just a set of propositions, let's say F is involved in the essence of G just in case F is a constituent of some proposition contained in the essence of G. Next, let’s say that a property F is of the same kind as the A-properties just in case the essence of F ultimately involves the A-properties, which amounts to satisfying a recursive definition in terms of the A-properties. For example, F is natural just in case its essence ultimately involves paradigmatic scientific properties like so:

The naturalist’s view that all normative properties are natural thus amounts to the claim that they satisfy this recursive definition.¹⁷ Whereas the supernaturalist claims that the essences of some normative properties involve supernatural properties, thereby denying that they are natural.

Now we can specify what the non-naturalists’ opponents have in common. First, we can define the essentially non-normative properties—those properties that ultimately involve (and are thus of the same kind as) paradigm non-normative properties—as follows:

For any property F:

Then, we can formulate the non-naturalist’s opponents’ view as simply

And so, the non-naturalist’s view is

The non-naturalist thus insists that some normative properties fail to satisfy (i)–(iv) above: they are neither identical to nor have essences that ultimately involve paradigm non-normative properties.

This characterization of non-naturalism makes the non-naturalist’s property-kind-talk precise. The basic idea behind the recursive definition of essentially non-normative properties is that normative properties would be of the same kind as paradigm non-normative properties if they were ultimately “built up” out of such properties. But if a normative property is not ultimately built up out of such properties, then it’s something “over and above” all other properties. And since paradigm non-normative properties include scientific ones, this implies that there is some feature of reality that is not ultimately built up out of scientific features. Essentially Normative thus speaks to and elucidates the non-naturalist’s pre-theoretical claims that normative properties are sui generis and incompatible with a purely scientific worldview—it’s ILLUMINATING.

Essentially Normative is also SUBSTANTIVE & LOCAL. Whether all normative properties ultimately involve paradigm non-normative properties is a non-trivial question that must be settled by theorizing about normative properties in particular, and not by general metaphysical considerations.

Moreover, this characterization allows for the sort of non-naturalist view that is arguably Moore’s view according to which basic normative properties lack a real definition, but all normative facts of the form Fa are fully grounded in natural, non-normative facts. And it offers a diagnosis of why this view counts as non-naturalist that is very similar to Moore’s own thinking, since essence is a close cognate of real definition. Essentially Normative thus classifies Moore’s view as genuinely non-naturalist for reasons that capture the spirit of his view better than Rosen and Dasgupta’s grounding-characterizations.

Essentially Normative is also compatible with my hybrid property view on which all normative facts of the form Fa are fully grounded in non-normative facts and the normative grounding-facts themselves are fully grounded in autonomous facts about essences (specifically, the essences of hybrid properties). What makes this view non-naturalist, on my characterization, is that the primitive normative properties, which are grounded by hybrid normative properties, are not essentially non-normative.¹⁸ So, Essentially Normative also provides a broader characterization of non-naturalism than both Rosen and Dasgupta’s grounding-characterizations.

But Essentially Normative fits well with other non-naturalist views too. First, consider Fine’s view according to which there are no metaphysically necessary connections (including grounding connections) between normative and non-normative properties whatsoever. Fine (2002: 271) claims that the reason why there can’t be metaphysically necessary connections between these properties is that there are no essential connections between them:

So, Essentially Normative captures the spirit of Fine’s view well too, since he takes the primary issue to be about essence.

One might worry, though, that Essentially Normative does not fit as well with Bridge Law Non-naturalists like Enoch (2011) and Scanlon (2014) who claim that all contingent, particular normative facts of the form Fa are partially grounded in some non-normative fact Ga together with a fundamental normative principle connecting F and G. These non-naturalists rarely explicitly talk about essences.

But Enoch (2019: 14) expresses some sympathy toward appealing to essences and doing so would help the Bridge Law Non-naturalist’s view. As I point out elsewhere, it’s unclear what makes the normative sui generis on this view, since some derivative natural facts plausibly have a similar grounding structure (Leary 2017: 92). For example, the fact that there’s a statue may be partially grounded in the fact that there are particles arranged in such a way that compose a statue together with the metaphysical law that if F composes G, then if F exists, G exists. So, what makes the normative facts sui generis, but not statue facts? One might think it’s simply the content of the normative laws: they involve normative properties. But this assumes that there’s something distinctive about normative properties to begin with. And in order to substantiate that assumption, it seems natural to appeal to essences. The Bridge Law Non-naturalist may claim that what makes normative properties sui generis while being a statue is not, even though normative facts and statue facts are grounded in the same way, is that normative properties are essentially normative, while being a statue is essentially non-normative (in the sense defined above).

So, Essentially Normative is both GENERAL and FITTING: it provides a characterization of non-naturalism that encompasses a wider spectrum of views that speak to the non-naturalist’s pre-theoretical commitments and it provides a better fit with paradigm non-naturalist views (including Moore’s) than the alternative characterizations. The central claim held by non-naturalists is that the essences of some normative properties do not ultimately involve paradigm non-normative properties (as defined above). But this leaves open downstream metaphysical questions about which non-naturalists may disagree, which gives rise to a variety of non-naturalist views (see Figure 2).

One objection one might have to this characterization, though, stems from the worry that debates about essences are irresolvable. Fine (2002) claims that it’s obvious that the nature of being in pain and the nature of badness don’t involve one another, but others may just as easily claim that the nature of being in pain involves badness (Leary 2017) or that badness involves being in pain. If there’s no clear method for resolving such disputes, then even if Essentially Normative provides the most fitting and general characterization of the non-naturalist’s pre-theoretical concerns, perhaps this is not what we should be concerned with. In other words, one might think that another desideratum for a good characterization of non-naturalism is tractability: it should characterize non-naturalism as a thesis for which there is some clear methodology for determining whether it’s true.

But the alternative grounding-characterizations are not more tractable. Consider Anti-Full-Ground_Fa. Most contemporary non-naturalists agree that all normative facts of the form Fa are at least partially grounded in non-normative facts. So, the main issue relevant to the truth value of Anti-Full-Ground_Fa is whether all such normative facts are also partially grounded by fundamental normative principles. But what’s the methodology for answering that question?

It’s particularly mysterious how we could answer that question without appealing to overly general metaphysical considerations, which would thereby render Anti-Full-Ground_Fa not SUBSTANTIVE & LOCAL. For example, Berker (2018a) offers a dilemma for the view that normative principles play a partial role in grounding particular normative facts, which stems from reflecting on first-order normative theories like utilitarianism. Berker claims that either the utility principle is a grounding-claim (necessarily, an action is right if and only if, and because, it maximizes happiness) or it’s just a necessary biconditional (necessarily, an action is right if and only if it maximizes happiness). If it’s a grounding-claim, Berker argues, it can’t partially ground a particular act A’s being right because this would involve redundant circular grounding, since the very principle states that an act’s maximizing happiness is by itself sufficient for its being right. But if the utility principle is a necessary biconditional, Berker argues that it’s implausible that it partially grounds a particular act’s being right because other facts are not grounded in this way: the fact that x is red is not grounded in the fact that x is red and round together with the fact that, necessarily, if an object is red and round, then it’s red (Berker 2018a: 912). But this argument generalizes: the very same argument could be used to argue that bridge laws can’t play a partial role in grounding particular mental or biological facts. So, if Berker’s argument is correct, it would actually show that the debate about Anti-Full-Ground_Fa is not sufficiently local, and thus not the right way to think of non-naturalism.

So, while Essentially Normative may be difficult to get traction on, it is not less tractable than grounding-characterizations of non-naturalism. The lesson to be learned here is simply that it’s much more difficult to answer metaphysical questions about particular kinds of properties and facts than it is to answer more general, structural metaphysical questions. So, relative intractability is an unavoidable feature of the naturalism-non-naturalism debate. I suspect that the only way to resolve the debate, then, is indirectly and in a holistic way. We need to ask which view is better at vindicating realism, explaining how we can think about and refer to normative properties, explaining supervenience, and so on.

I’ve argued that an essence-characterization of non-naturalism does a better job than identity or grounding characterizations at meeting four desiderata: it speaks to and elucidates the non-naturalist’s pre-theoretical commitments, while also rendering non-naturalism a substantive, local claim about normativity (unlike identity-characterizations) and it also provides the broadest characterization that fits best with the spirit of paradigm non-naturalists (unlike grounding-characterizations). I take this to show that, contrary to popular opinion, non-naturalism is best specified in terms of essence, rather than grounding or identity.

I’ve admitted that, even on my essence-characterization, non-naturalism is an unavoidably difficult view to evaluate directly. But this admission should not be taken to undermine the importance of the taxonomical task addressed here. This is because the essence-characterization of non-naturalism has significant implications regarding non-naturalism’s overall merits.

The most important implication concerns non-naturalists’ resources in explaining why the normative supervenes on the non-normative, which has been one of the most longstanding challenges to the view. The supervenience challenge is to explain why there are metaphysically necessary connections between the normative and the non-normative even though the normative is sui generis. There’s an apparent tension here because the non-naturalist’s claim that the normative is sui generis requires that normative properties are metaphysically distinct from non-normative properties, while explaining the metaphysically necessary connections between these properties requires that they are somehow intimately metaphysically related. But if the non-naturalist’s claim that normative properties are sui generis amounts to a claim about the essences of normative properties, which is compatible with a wide range of views about how normative facts are grounded, the non-naturalist has a wide selection of explanations for supervenience available to her.

For example, consider my hybrid property view according to which some basic normative properties like goodness and badness are essentially normative (in the sense defined in §4), but there are hybrid normative properties whose essences involve both normative and non-normative properties and thereby explain why certain non-normative facts ultimately fully ground facts about what’s good and bad. Hattiangadi (2018: 603) claims that this view is not genuinely non-naturalist because it’s a view on which all particular normative facts are fully grounded in non-normative facts. But this presupposes that non-naturalism is about grounding. If non-naturalism is instead about essence, as I’ve argued here, my explanation for supervenience is a genuinely non-naturalist one.

But the essence-characterization also supports other potential non-naturalist explanations for supervenience. In Leary (2017), I considered two alternative views that take general normative principles to play some role in explaining why the normative is grounded in, and thus supervenes on, the non-normative. The first is Enoch’s Bridge Law Non-naturalist view according to which the general normative principles are metaphysical laws that themselves partially ground particular normative facts together with the particular non-normative facts. The second view is one according to which the normative principles are metaphysical laws that don’t partially ground the particular normative facts, but instead ground the very grounding relations between non-normative facts and normative facts. I rejected both of these explanations for supervenience in favor of my hybrid property view in part because on both views it was unclear what makes the normative facts sui generis, since some derivative natural facts plausibly have the exact same grounding structure: for example, contingent facts about statues may be grounded by contingent facts about the arrangement of particles with the help of metaphysical laws about composition (Leary 2017: 92). So, my criticism of these alternative explanations for supervenience (like Hattiangadi’s criticism of mine) was that they are not obviously non-naturalist. But the essence-characterization of non-naturalism that I’ve presented here shows that my earlier criticism of these alternative explanations for supervenience is wrong: what makes the normative sui generis, while derivative natural facts that plausibly have the same grounding structure are not, is that normative properties are essentially normative, while derivative natural properties are essentially natural (in the sense defined in §4).

So, here’s the important upshot: because non-naturalism is a view about the essences of normative properties, this allows non-naturalists to appeal to metaphysical explanations for supervenience according to which the particular normative facts have the exact same grounding structure as derivative natural facts and are even fully grounded in natural facts. And there are a variety of explanations to choose from. This doesn’t completely resolve the supervenience challenge because there may still be reasonable debate about which non-naturalist explanation is better or whether these explanations are worse than naturalist ones. But it goes a long way to diffusing the apparent tension that motivates the challenge: there is no inherent incompatibility between the non-naturalist’s view (about essence) and the claim that there are metaphysically necessary connections (even grounding connections) between the normative and the non-normative.

Finally, my argument in favor of an essence-characterization of non-naturalism also has significant implications for metaphysical theorizing, more generally. Within metaphysics, it’s controversial whether we need the ideology of essence for metaphysical theorizing. But if my argument that essence is best suited for characterizing what non-naturalists take to be at stake in the debate between them and their opponents, this provides some justification for adopting the ideology of essence in the first place. After all, it shows that the ideology of essence is useful for capturing a significant first-order metaphysical debate.