Ross (2002: 16) wondered “whether there is any general character which makes right acts right, and if so, what it is.” In other words: what makes right acts right, in general?
Here’s a familiar answer.
|(1)||An act is right in virtue of maximizing happiness.|
This is a grounding claim because, roughly, it expresses non-causal explanation or metaphysical dependence. This claim is generic because it isn’t about any particular right act.
In contrast, consider:
|(2)||Gwen’s act of charity is right because it maximizes happiness.|
This claim is particular because it’s about Gwen’s act and what makes it right, not what makes right acts right, in general.
One natural idea is that generic grounding claims state mere regularities of ground. For example, if an act is right in virtue of maximizing happiness, then every possible right act is right in virtue of maximizing happiness. The generic claim generalizes over particular grounding relations.
In this essay, I argue that this simple story is wrong. Generic grounding claims are not merely quantificational; rather, they express real definitions, where real definitions are (in part) claims about essence.
The resulting view has two major upshots.
First, it makes better sense of the semantics of generic grounding claims. If generic grounding claims merely express regularities, then we cannot make sense of debates where generic grounding claims are at issue. I focus on the debate over moral laws.
Second, it clarifies the distinction between reductive and non-reductive metaphysical theories. The difference between, say, reductive and non-reductive naturalism is the difference between there being a regularity of ground between moral and physical facts and there being a definition of moral facts in terms of physical facts.
For the sake of simplicity, I make three assumptions. First: veridical grounding claims are explanatory truths, where an explanatory truth is a true sentence of the form: ⌜P metaphysically explains Q⌝. Second: grounding claims correspond to grounding relations in the sense that sentences of the form ⌜P metaphysically explains Q⌝ are true just in case there is a grounding relation that holds between the facts expressed by P and Q. Third: grounding is a many-one relation between facts.
What are generic (as opposed to particular) grounding claims? To answer this question, I identify four characteristic features of generic grounding claims. (For the sake of brevity, I will often call generic grounding claims generics.)
I start with the most obvious feature.
Generics are not explicitly about particular objects (with respect to grounds, grounded, or both).
A grounding claim is explicitly about a particular object just in case it features a phrase that signals that one is talking about a particular object. Here are a few paradigm generics.
|(3)||An act is right because it maximizes happiness.|
|(4)||An agent knows a proposition is true, in part, because they believe the proposition is true.|
|(5)||An object is a table because it is composed of atoms arranged table-wise.|
On their surface, these claims are not about particular right acts, propositions, tables, and so on.
In contrast, consider the following particular grounding claims.
|(6)||Gwen’s act of charity is right because it maximizes happiness.|
|(7)||Luke knows that Gwen is philanthropic, in part, because he believes that Gwen is philanthropic.|
|(8)||The fact that Jack’s table is a table is grounded in the fact that Jack’s table is composed of atoms arranged table-wise.|
(6) explicitly references Gwen’s act of charity. (7) references Luke and the proposition that Gwen is philanthropic. (8) references Jack’s table. Meanwhile, (3), (4), and (5) do not play favorites.
My paradigm examples of generic and particular grounding claims are cases where each claim is fully generic or fully particular.
But there are mixed cases. For example:
(9) An act is right because God commands it to be.1
This claim is not fully generic because it references a particular individual, God, in its grounds. This claim is not fully particular because the grounded is not about any particular right act. I nonetheless count (9) as a generic grounding claim because it is generic with respect to the grounded.
The second striking feature of generics concerns the ways they can be formulated.
Generics typically take indefinite singular, bare plural, and bare singular forms.
So far, I have formulated generics using the indefinite singular form. For example, (3) describes an act being right. The indefinite article “an” prefaces the singular noun “act.” However, generics can also take the bare plural form, as in:
|(10)||Acts are right because they promote happiness.|
They can also take the bare singular form, as in:
|(11)||Knowledge is grounded in true belief.|
It’s important to know what syntactic forms generics may take, but we don’t want to become overly sensitive to their syntax. For this reason, we should represent such claims using a canonical logical form.
We can represent generics using a two-place operator, Gen, which binds variables in a generic claim. This operator is used, in philosophy of language, to represent ordinary (non-grounding) generics like “Birds fly” (Carlson 1989; Lewis 1975; Krika et al. 1995).
|(12)||Gen x.(x is a bird)(x flies)|
This notation can also help us represent generic grounding claims.
|(13)||Gen x.(x is a right act)(x is a right act because it maximizes happiness)|
This notation is supposed to represent the generic claim, the claim that we often express using indefinite singular, bare plural, and bare singular forms. By abstracting away from these specific forms, we can focus on the content of the Gen operator.
Moving away from the syntactic markers of generics, we come to the semantic markers of generics. We will examine the entailments of generics, starting with their existence entailments.
Generics fail to entail the existence of (at least some of) the objects in their grounds and grounded.
For example, the truth of (3) is compatible with the non-existence of right acts. The resulting world would be unfortunate, but not contradictory. More generally, (3) is compatible with the non-existence of acts, things that are right, and things that promote happiness! A similar story goes for (4) and (5).
In contrast, if Gwen’s act of charity doesn’t exist, then (6) isn’t true. To explain what makes Gwen’s act right, Gwen’s act must exist! Similar stories can be told about the (7) and (8).
Unlike particular grounding claims, generic grounding claims don’t depend on the existence of particular objects for their truth. Here’s a more precise statement of existence-independence.
Let ϕ be a grounding claim, where F and G are the relevant properties in ϕ. Then ϕ is existence-independent iff ◊[ϕ ∧ ¬∃x.(x is F ∧ x is G)].
(I assume, as is standard, that the logic of metaphysical modality is S5 and that the ☐/◊ operators represent metaphysical necessity/possibility.)
In other words: a grounding claim is existence-independent iff its truth is metaphysically compossible with the non-existence of an individual of the kind featured in the grounds and grounded. It follows that (3) is existence-independent.
Notice that this definition of existence-independence also covers impure cases, like (9). In such cases, a thing might possess the property F but not G. The generic still counts as existence-independent because it can be true in cases where no object has both properties F and G.
Say that ϕ is existence-dependent iff it isn’t existence-independent. Then it follows that (6) is existence-dependent. It’s existence-dependent because it’s impossible for it and the negative existential to be true.
Now there is a question about what happens when an apparently generic grounding claim has necessary existents in both grounds and grounded. Consider:
|(14)||Numbers are grounded in mathematical structures.|
If you think both numbers and mathematical structures are metaphysically necessary, then (14) fails to count as a generic grounding claim, given EXISTENCE-INDEPENDENCE. However, it certainly seems generic.
In this case, (14) entails numbers and mathematical structures, thereby violating the letter of existence-independence, but the entailment is vacuous. Numbers and mathematical structures are always on the scene, so (14) fails to relevantly entail numbers and mathematical structures.
In the ultimate account of generics, EXISTENCE-INDEPENDENCE should be modified to reflect relevant entailment rather than classical entailment. However, such a modification would muddy the waters at the current stage of inquiry. Instead of constructing a complicated definition that accounts for all cases, I will start with a simple definition that covers the vast majority of cases. This means I will retain the classical notion of entailment and ignore generics with necessary existents. The hope is that this simple account can be extended at a later point.
Setting aside necessary existents, paradigm generics are existence-independent and paradigm particular grounding claims are existence-dependent. Existence-independence seems to be a reliable guide to generic grounding claims.
Lastly, we can identify generics by their modal properties.
Generics are downwardly necessary.
There are different notions of necessity. The standard one is as follows.
If ϕ fully grounds ψ, then it’s metaphysically necessary that: if ϕ, then ψ.
This notion is widely accepted among grounding theorists (Audi 2012; deRosset 2013a, 2013b; Trogdon 2013; Rosen 2010). But because it is accepted as a principle of grounding, in general, it does not help single out generic grounding claims. However, there is another modal principle that does single out generics.
If ϕ fully grounds ψ, then it’s metaphysically necessary that: if ψ, then ϕ.
The first formulation is upwardly necessitating: grounds necessitate groundeds. The second formulation of necessity is downwardly necessitating: groundeds necessitate grounds. (Terminology due to deRosset (2013a).) Although downward necessity may not be true of grounding in general, it certainly seems true of generics.
More precisely, the following seems true.
If ϕ is a true claim of the form ⌜An x is G in virtue of being F⌝, then for every possible x, if x is G, then x is F.
Intuitively, when we say what makes an act right, in general, we are specifying the kind of ground that all right acts share. For example, if right acts are grounded in the fact that they maximize happiness, then, for any possible x, if x is a right act, then the fact that x is right is fully grounded in the fact that x maximizes happiness.
Here is another way of seeing the same point: it seems contradictory to assert a generic while denying downward necessity. Suppose I say: an act is right in virtue of maximizing happiness, but there are some right acts that don’t maximize happiness. This seems contradictory. If there were right acts that didn’t maximize happiness, why would you say that acts are right in virtue of maximizing happiness? It appears that generics are downwardly necessary.
In contrast, it’s immediately implausible to think that every particular grounding claim is downwardly necessary. Consider the fact that the ball in my closet is colored. The ball is colored in virtue of being red, but it doesn’t follow that every colored object is colored because it’s red.
Although generics are downwardly necessary, they do allow for multiple realizability. For example, you might say that an object is colored in virtue of possessing some specific color. In this case, the fact that x is colored can have different grounds with respect to color, but there will always be a ground that concerns x being some particular color or other. So the generic will be
|(15)||An object is colored in virtue of possessing some specific color.|
as opposed to
|(16)||An object is colored in virtue of being red.|
Read as generics, (15) is plausibly true, while (16) is certainly false. However, the pragmatics of language might give (16) a different interpretation, like the one where “an object” refers to a particular object.
To summarize, generic grounding claims have four characteristic features: EXPLICIT, FORMS, EXISTENCE-INDEPENDENCE, and NECESSITY. These properties, taken collectively, constitute necessary and sufficient conditions for being a generic grounding claim.
Note that my account of generic grounding claims does not account for every sense of “generic.” For example, there are physicalists who are willing to say
|(17)||Pains are grounded in physical states.|
while believing that pain can exist in worlds where there are no physical states. In such worlds, perhaps pain is just a primitive property.
There is a sense in which (17) is generic. The claim seems to be: pains are usually grounded in physical states. The cases of interest, for me, are ones where the sense of genericness is stronger.
My account also does not address the pragmatics of generics. For example, in a context where we know we aren’t talking about fundamental pain states, we might use (17) to convey a generic (in my sense) grounding claim, such as:
|(18)||Non-fundamental pains are grounded in physical states.|
Pragmatics is messy. Although I will have some things to say about the default and deviant interpretations of generics, I will not provide a model of the pragmatics of metaphysical language.
Now that we have a preliminary characterization of generics, we can see how generics are important. Many of the paradigm examples of grounding claims are generics. Metaphysicians are not simply interested in what grounds what. They are interested in what, in general, grounds what. They want to know what, in general, grounds the moral, mental, modal, and mathematical facts. Of course, particular grounding claims are often used as a proxy for generic ones. For example: pain being grounded in neurological states is an instance of mental states being grounded in physical states. But in such cases, the generic thesis is of primary interest.
Generic grounding claims are important because of their generality. But are these claims importantly different from particular grounding claims? The standard answer appears to be: no.
You might think that the difference between general and particular grounding claims is simply the difference between quantified and unquantified formulas. The corresponding proposal identifies the content of a claim like (3)—An act is right because it maximizes happiness—with
|(19)||Necessarily, for every x, if x is a right act, then the fact that x is right is fully grounded by the fact that x maximizes happiness.|
If this view is correct, the relationship between (3) and (6)—Gwen’s act of charity is right because it maximizes happiness—is quite simple. The latter is an instance of the former.
In full generality, the regularity theory of generic grounding is as follows.
Gen x.(x is G)(x is G because x is F) is true iff ☐∀x.(if x is G, the fact that x is G is fully grounded in the fact that x is F).
Applying REGULARITY to (4) and (5) is straightforward. It should also be clear how these generalizations take (7) and (8) as instances.
If REGULARITY is true, then there is nothing particularly interesting about generics. They express necessary generalizations. The story ends here.
The story continues. Why? Because the regularity theory doesn’t make sense of metaphysical debates where generics are front and center.
I focus on the debate between moral generalists and moral particularists. This debate concerns the existence and explanatory role of moral laws (or principles). By a “moral law,” I do not simply mean a rule of thumb, but the kind of principle that normative ethicists use to characterize and explain the moral domain. Here are a few examples.
These laws (purportedly) characterize consequentialist, virtue-theoretic, and contractualist ethical theories, respectively. Though these formulations make explicit reference to biconditionals and equalities, they are intended to be asymmetric explanatory generalizations.
A simple inspection shows that these statements of moral laws have all the marks of generics. The presence of these features gives us good reason to think that moral generalists and particularists—at least the more metaphysically-minded among them—are debating about the truth of generic grounding claims.
Note: I have not assumed that there are moral laws. My claim is that metaethicists make statements of moral laws, and those statements correspond to generics. (For more on moral laws and grounding, see Berker (2019) and Rosen (2017a, 2017b).)
Suppose you are a regularity theorist. Then you interpret (19) as the moral law expressed by (3). The generalist/particularism debate concerns whether (19) is true.
The truth of these claims, however, is not all that is being debated. We should distinguish between two versions of moral generalism. On one view, moral laws are part of the grounds of particular acts. Call this the governing conception of generalism.
Suppose Gwen’s act of charity is right. According to the governing theory, this fact is grounded by the fact that (a) Gwen’s act maximizes happiness, and (b) acts are right because they maximize happiness.
The moral law governs the particular moral facts in the sense that it serves as a partial ground of such facts. The governing conception captures the intuition that laws are prior to the instances that they subsume.
On another view, moral laws do not ground particular facts; their existence is enough for generalism to be true. Call this the non-governing conception of moral laws. (Beebee (2000) distinguishes between governing and non-governing natural laws.)
In what follows, I argue that, however you understand moral laws—governing or not—the regularity theory fails to capture their content.
Suppose that generics express moral laws and these laws are captured by regularities of ground. I start by arguing that this position attributes controversial views to moral generalists; specifically, it commits them to the Humean view that moral laws are grounded in their instances. I then argue that the Humean view is an implausible interpretation of moral generalism.
Given REGULARITY, it follows that moral laws are regularities of ground. For example, consider (19), which is reprinted here as (20).
|(20)||Necessarily, for every x, if x is a right act, then the fact that x is right is fully grounded by the fact that x maximizes happiness.|
Now we ask the question: what grounds (20)?
On the standard interpretation, (20) is quantifying over possible worlds (or possibilities); it tells us that, in every possible world, a certain conditional holds. Notice that (20) is a universally quantified statement, and it’s natural to think that such statements are partially (though perhaps not fully) grounded in their instances. (See Fine (2012) and Skiles (2015) for discussion of the grounds of universal generalizations.)
The instances of (20) will be quantificational facts like:
|(21)||For every x in world w1, if x is a right act, then x is right in virtue of maximizing happiness.|
What grounds (21)? Since it’s a quantificational fact, we take it to also be partially grounded in its instances. Here is such an instance:
|(22)||[Gwen’s act maximizes happiness (in w1)] ≺ [Gwen’s act is right (in w1)]|
By the transitivity of grounding, it follows that (20) is partially grounded in (22).
Notice that if (20) is the genuine content of a moral law, and (22) grounds (20), then it follows that moral laws, if they exist, are (at least partially) grounded in their instances. I call this the Humean conception of laws.
At this point, two questions arise. The first question: is the Humean conception of laws plausible? The second question: is it plausible that moral generalists are universally committed to Humean laws? I am concerned with the second question, not the first. The current question is not whether Humeanism is the right view, but whether every generalist is a Humean. I will argue that it’s implausible that all (or even most) generalists are Humean.
To start, the Humean view rules out the possibility of primitive moral laws. So a moral generalist could never consistently say:
|(23)||Acts are right in virtue of maximizing happiness, and I take this to be a primitive moral law.|
If the fact that laws are grounded in their instances were part of the content of (23), we would expect (23) to sound inconsistent or at least slightly odd. But there is nothing amiss when uttering (23). This should give us pause about taking moral generalists to always be committed to Humean laws.
In addition, the Humean view is in tension with the governing conception of laws. If laws govern via grounding their instances, then the Humean view is clearly circular. If Humeanism is correct, then (22) partially grounds (20). And if the governing conception of laws is correct, then (20) partially grounds (22). So (22) partially grounds itself, on this view. But this clearly violates the (widely accepted) asymmetry of grounding.
Now the question here is not: is the Humean view of laws correct? The question is: are moral generalists committed to the Humean view? I think the answer to the latter question is no. My reasoning: if you think laws are in the business of grounding instances, you generally don’t think laws are also grounded in their instances.
Finally, the regularity theory is wrong because it seems possible for someone to think (a) moral laws do not ground their instances, and (b) moral laws are grounded but not in their instances. Part (a) seems to follow from the fact that we can make a coherent distinction between governing and non-governing laws. To defend part (b), I make the following observation: in saying that an act is right in virtue of maximizing happiness, I am not thereby committing myself to the view that this fact is grounded in its instances. Instead, I might think the moral laws are all grounded in God’s infinite wisdom. Or perhaps they are grounded in non-moral, natural facts. Or there could be an infinitely descending chain of moral laws, each law grounded in the one below it. If the regularity theory is correct, all of these possibilities (and more) are ruled out by semantic fiat. But this seems implausible.
Let us review. I have pointed out that REGULARITY will commit generalists to Humean moral laws—laws that are grounded in their instances. Then I argued that this interpretation of generalism is implausible, presenting three cases where the interpretation is implausible: first, the moral generalist who thinks laws are primitive; second, the moral generalist who thinks laws ground their instances; third, the moral generalist who thinks laws do not ground instances and are themselves grounded in something other than their instances.
In none of these cases do I take a stand on the plausibility of moral generalism or the Humean conception of laws. The discussion is about what we are committed to when we make generic grounding claims. I am interested in the truth-conditions of these statements rather than the truth of any particular claim.
This distinction is clearer when we consider routine philosophy of language. Consider the phenomenon of quantifier domain restriction (Stanley and Szabó 2000). Imagine Jose asks for a bottle of beer and Tasha replies:
|(24)||Every bottle is empty.|
What is the content of (24)? There are two natural interpretations.
|(25)||Every bottle in the universe is empty.|
|(26)||Every bottle in Tasha’s house is empty.|
Most people don’t think (24) literally means (25). Why not? Because (25) does not plausibly reflect the commitments of the parties involved. Tasha isn’t trying to communicate (25). It clearly isn’t what Tasha is talking about. The more plausible interpretation is (26).
Of course, there are semanticists like Borg (2004) and Cappelen and Lepore (2005) who think (25) (or something like it) is the correct interpretation of (24), but my point is methodological. The goal is not to interpret Tasha so that she has plausible beliefs; the goal is to accurately capture the content of Tasha’s claims. If you look at what Tasha believes at the context, those beliefs suggest that Tasha did not mean (25).
Similarly, my claim is not: generics have a different content because, if they didn’t, they would be implausible. Rather, the claim is that generics have a different content because the standard story does not reflect the commitments of those who sincerely assert them.
Lastly, I should note that nothing in my discussion depends on grounding claims having normative subject matters. Recall my two non-moral paradigm cases of generics, (4) and (5), reprinted here as (27) and (28).
|(27)||An agent knows a proposition is true, in part, because he or she believes the proposition is true.|
|(28)||An object is a table because it is composed of atoms arranged table-wise.|
We cannot refer to an explicit generalist/particularist debate in the case of (27) and (28), but we can construct analogous debates. We can imagine epistemic generalists and particularists as well as ordinary object generalists and particularists.
Even though the subject matters differ, we have the same theoretical choice points as we do for moral generalism and particularism. The current discussion, then, is not solely one about normative grounding.
If the regularity theory is false, how do we interpret the content of generic grounding claims? It appears that these claims must be generic in some more fundamental sense.
For example, Fine (2016) thinks that some generics (specifically, statements of identity criteria) are deeply generic: that is, they cannot be understood purely in terms of quantification and grounding facts. For example, he wants to know what, in general, grounds the identity of any two sets. One natural view is that the identity of any two sets is grounded in the fact that the sets have the same members. The regularity theorist represents the claim as follows:
|(29)||For any possible sets x and y, the fact that x and y have the same members grounds the fact that they are identical.|
Fine (2016: 13–14) points out that, on this interpretation, “there is not a single explanandum but a bunch of explananda, one for each pair of sets; and what we want to know, in the case of each pair, is what it is about the pair that would account for their being the same.” On his view, the regularity theory reduces a generic statement of ground to several statements of particular ground. But the intuition is that we are giving one general explanation, not several particular explanations.
Consider the moral generalist case. (20) tell us the distribution of ways in which right acts are right. Every right action appears to be grounded in the same way, but the existence of a common way is simply another fact about the distribution of ways in which right acts are right. There is no fundamental difference between (20) and a disjunctive statement of what makes each particular right act right. But we need an explanation of what makes an right, in general, and such an explanation makes no reference to particular acts at all.
If my previous arguments were correct, generics are deeply generic in the way that Fine indicates. However, it is unclear what this kind of genericity amounts to. The regularity theorist will ask, “What’s missing from my explanation? What undergirds the intuition that it fails to be sufficiently generic?” This is a fair question. There are several possible answers, here.
The most obvious proposal is that the genericity stems from the genericness of properties. We posit grounding between properties as opposed to facts. On this view, to say that an act is right because it maximizes happiness is to say that the property being a right act is grounded by the property being a happiness-maximizing act. A similar story goes for relations. The relation being identical to another set is grounded in the relation having the same members as.
Here is the proposal, generalized.
Gen x.(x is G)(x is G because x is F) is true if and only if G is fully grounded in F.
PROPERTY makes sense of the intuition that there is a single explanandum here—the property of being a right act—as opposed to many explananda. This solution is simple and elegant.
While PROPERTY captures the intuition that there is a single explanandum involved, it requires us to answer questions about the relationship between fact and property grounding that we do not have good answers to.
For example, consider the following question: if G is fully grounded in F, does it follow that every G is G in virtue of being F? If the answer is no, then PROPERTY fails to capture the ground-theoretic regularities associated with generics.
If the answer is yes, this problem can be avoided, but the answer is not obviously yes. Suppose there is a single fundamental property M from which all other (qualitative) properties derive. So the property of being a dog is fully grounded in M. But it’s unclear whether the fact that Sparky is a dog must be grounded in the fact that Sparky is M. My intuition is that Sparky does not have the property M at all.
But suppose my intuition is wrong, and that full property grounding does entail the relevant ground-theoretic generalizations. Another question arises: if F1 and F2 jointly fully ground G, does it follow that every G is G in virtue of possessing both F1 and F2? It is unclear what happens in these cases. It does not seem like every G must possess both properties on the assumption that neither property individually fully grounds G. One possibility is that every G possesses the fusion of F1 and F2. The plausibility of this idea depends on your metaphysics of properties.
To answer the questions I have raised, we need an account of property grounding. Specifically, we need an account that clarifies the relationship between property and fact grounding. At the moment, this relationship is poorly understood, and I see no easy way of revising PROPERTY to account for the content of generics.
This is not to say that properties cannot be a part of an account of generic grounding. (As I will later show, my account of generics does have a place for properties.) It is only to say that we need more than simple property grounding.
If property grounding does not make sense of “deep generality,” what does? In what follows, I give my account of the content of generics.
I propose we understand generic grounding claims as statements of real definitions.
Real definitions are not semantic or conceptual definitions; rather, they are metaphysical definitions—definitions of worldly objects, facts, or properties. To give a real definition of courage is to “say what it is for a person to be courageous—to identify that in which the courage of the courageous person consists” (Rosen 2015: 189). Real definition is said in many ways: to be G is to be F; being G consists in being F; being G reduces to being F; what it is to be G is to be F.
Consider (Rosen 2015: 200)’s account of real definition.
F is a real definition of G iff ☐G∀x. (if x is G or F, the fact that x is G is fully grounded in the fact that x is F)
Read this as: F is a real definition of G if and only if it lies in the nature of G (☐G) that: if something is either G or F, it is G because it is F.
Suppose we wanted a real definition of right action. One might look like this.
|(30)||It lies in the nature of what it is for an act to be right that: for every x, if x is a right act or x maximizes happiness, then the fact that x is right is fully grounded by the fact that x maximizes happiness.|
Simply put, (30) tells us that right acts and happiness-maximizing acts are coextensive, and that every right act is right in virtue of maximizing happiness. (For more discussion of the coextension claim, see Rosen (2015).)
Claims like (3) express contents like (30). In full generality, the view is as follows.
Gen x.(x is G)(x is G is grounded in x is F) is true iff ☐G∀x. (if x is G or F, the fact that x is G is fully grounded in the fact that x is F).
In what follows, I argue that DEFINITION is the best proposal for three broad reasons: (i) it makes sense of the content of generics, (ii) it makes sense of the reductive/non-reductive distinction in metaphysics, and (iii) it accounts for generics in a way that is both ideologically conservative and relatively neutral about the nature of grounding.
To start, DEFINITION captures the semantic properties of generics. (30) allows generics to be existence-independent. It also accounts for the downward necessity of generic grounding. Overall, it accounts for the regularities involved in generics.
What is more impressive: DEFINITION makes sense of deep generality. It does this by appealing to essences.
Let me explain. The regularity theory identifies patterns of grounding, but the patterns themselves do not entail anything about the nature of what it is for an act to be right. This is what prompts the intuition that the notion of genericness is “deeper” than a modal regularity.
To dig deeper is to note that such regularities of grounding stem from facts about essence. If we encounter a necessary generalization characterizing the grounding of right acts, the natural conclusion is that this generalization holds in virtue of the nature of right action.
Here is a less abstract way of seeing this. Consider what moral generalists sometimes say about moral laws. Bakhurst (2008: 201) writes:
When someone has a concept, ϕ, their knowledge is general in kind; it is knowledge they bring to particular cases and which enables them to identify whether these cases fall under the concept. We want to say that when someone has the concept of cruelty, they know something general about the nature of cruel actions. They know what characteristics an action has to have to be cruel, and they know what kind of disvalue cruelty lends the action.
Jackson et al. (2000: 88) writes:
Our language for talking about them—the predicates ‘is something that there is good reason to do’ and ‘is generous’—must apply because of the nature of what they apply to.
These passages point toward the natural view that, if generalism is true, then our practice of giving moral explanations is guided by our grasp of the natures of normative properties.
More generally, we get the intuitive view that, in many cases, knowledge of grounds is guided by knowledge of essence. If you miss the essentialist element, you miss part of what is at stake.
On this picture, particular statements of ground are insufficient explanations of what, in general, grounds what. Those statements must be unified by an essentialist fact. The general essentialist idea is familar from Fine (1994). On his view, whenever you have a necessity, that necessity holds in virtue of the nature of the objects or relations involved.
Though I am appealing to machinery developed by Fine and other essentialists, I do not think—as they do—that every modal claim is grounded in (or is otherwise explained by) the natures of the objects, properties, or relations involved. (See Berker (2019) for a critique of the essentialist-grounding view.) Moreover, I’m not committed to the view that every grounding claim (generic or otherwise) is underwritten by an essence. I’m only committed to the view that there are some essences that figure into some grounding claims.
So DEFINITION explains generality in terms of essence. But where do properties (or universals) fit into this picture? It is natural to think of a real definition as one property, G, being defined by another property, F. The result is a property-theoretic proposal that trades property-grounding for property-definition.
Another possibility is that you simply stipulate that sentences like ⌜F is a real definition of G⌝ hold just in case ☐G∀x. (if x is G or F, the fact that x is G is fully grounded in the fact that x is F). This view does not posit a relation between properties.
On either view, we have enough to explain the deep generality of generics. (Or at least, if we do not, then we have an in-house dispute between two theorists of real definition. This is progress.)
Before moving on to another advantage of DEFINITION, it is worth noting that it is a generalized account of generics. If we look at the closest rival views to mine (aside from the regularity theory), we see that they are all concerned exclusively with the moral case.
It is not a problem, in itself, to focus on the moral case. (That is what I have done!) The problem comes when one constructs a solution that only fits the moral case.
For example, Rosen (2017b) considers various accounts of moral laws. If we are to transpose his discussion into the current one, where we consider generics, we would get the following proposal.
⌜An act is right because it maximizes happiness⌝ is true iff it is normatively necessary that: if an act is right or maximizes happiness, it is right in virtue of maximizing happiness.
On a slightly different version of this proposal, one might appeal to the existence of a distinctively normative grounding relation (following Fine (2012)).
In either case, the proposal is limited to the case of moral laws. We could not handle claims like “Knowledge is grounded in true belief” with this framework. But, qua generic grounding claim, there does not appear to be a fundamental difference between claims with normative subject matters and those with non-normative subject matters.
It is nice to be able to make sense of generics. It would be really nice if DEFINITION illuminated other things. As it happens, it does.
DEFINITION makes sense of the distinction between reductive and non-reductive theories of a domain. Rosen (2010: 132) presents two cases of this:
Consider a version of non-reductive materialism in the philosophy of mind according to which every fact about phenomenal consciousness is grounded in facts about the material organ of consciousness (in our case, the brain) even though no phenomenal property is reducible to any neurophysiological property or to any functional property that might be realized by a brain state. On this sort of view, I might be in pain in virtue of the fact that my c-fibers are firing, even though my being in pain would not consist in the firing or my c-fibers, nor in any disjunctive state of which c-fiber firing was a disjunct, nor in some existentially general state of which c-fiber firing was an instance… As another example, consider a Moorean position in metaethics according to which moral properties like right and good are indefinable, and yet every right act is right in virtue of possessing some right-making feature.
In both cases, some set of facts metaphysically depend on others, but the former facts are nonetheless not definable/reductive in terms of the latter. How do we make sense of this idea?
Here is my proposal. We understand definition/reduction in terms of real definition. Then we say that it is possible for some regularity of ground to hold even if there is no corresponding real definition. For example, it may be possible that every right act is right in virtue of maximizing happiness even if the nature of right action is silent about what grounds right acts. Similarly, the phenomenal facts may always be grounded in the material facts, even if there is nothing in the nature of the phenomenal that says so. (Rosen 2010: 133) has a helpful metaphor: “To put the point in epistemic terms, we are imagining views on which one might know everything there is to know about the nature of pain or rightness without knowing the first thing about c-fibers or happiness.”
Here is how the distinction plays out, more precisely. A physicalist takes the mental facts to be fully grounded in the mental facts. Whether one is a reductive or non-reductive physicalist hinges on whether the mental facts are defined by the physical facts. A similar story goes for moral naturalism. Reductive naturalists think the moral is defined by the natural; non-reductive naturalists do not. For more on this proposal see Rosen (2017a).
The intelligibility of non-reductive theories, here, is somewhat controversial. If non-reductive physicalism is possible, then we must give up on a certain kind of essence-grounding link. Specifically, we cannot believe that the nature of the groundeds will always specify what grounds them. The resulting view puts pressure on the Finean view of modality, whereby truths about necessity always hold in virtue of the essences of things. I say “puts pressure on,” as opposed to “rules out,” because there are nuances involved. Strictly speaking, the non-reductive theorist can believe that every grounding fact follows from the essence of something or other. Consider non-reductive moral naturalism. It may be true that it lies in the nature of something that, if an act is right or maximizes happiness, it is right because it maximizes happiness. But that something may not be rightness (or the property of rightness).
So the resulting framework will be controversial. But this is not a problem because the distinction between non-reductive and reductive theories is itself controversial. Framing the issue in terms of DEFINITION at least identifies, exactly, where the controversy lies.
In light of the previous discussion, we can now see how DEFINITION makes sense of the non-default, or deviant interpretations of generics. Consider some of the theses I have discussed.
My theory predicts that, on their default interpretations, these claims express real definitions, not mere regularities of ground. Moreover, my theory can make sense of cases where these claims take a deviant interpretation. A deviant interpretation arises when either (a) a sentence takes on a context-sensitive meaning that differs from its default one or (b) the speaker uses a sentence to convey a meaning that differs from its default one.
For example, there may be a case where two metaphysicians utter “Knowledge is grounded in true belief,” but one is making a real definition while the other is not (or at least not intending to). What is happening in such a case? Most likely, the person who is not expressing a real definition is someone who wants to merely assert the regularity because she does not adopt a reductive theory of knowledge. For each of the theses I have mentioned, one might plausibly be a non-reductionist about the relevant domain, while accepting the relevant generalizations.
As a contingent fact about our current linguistic practices, we do not have a way to clearly express the fact that we are speaking in a definition-theoretic way versus a regularity-theoretic way. This unclarity contributes to the confusion between the two views.
This brings us to the following question. If generics sometimes take the regularity reading, why not think of the regularity reading as the default one? The real definition reading could be considered a deviant interpretation, or merely part of what people generally presuppose. I think the real definition is the default interpretation of generics because, as far as I can tell, that’s what generics normally mean. Normality, here, is a fact about conventional uses of language.
Ideally, we would have better linguistic practices, ones that would succinctly distinguish between definitions and regularities. Here is my pitch for language reform. We should explicitly speak of real definitions when we want to make claims about real definitions. When we do not make claims about real definitions, we refer to grounding itself. Or if you fully subscribe to grounding pluralism, we should explicitly refer to the specific kinds of grounding.
DEFINITION is ideologically conservative yet theoretically substantive. In order to account for generics, we do not need to posit radically new metaphysical tools.
Real definitions are constructed from familiar materials—essence and grounding. Nonetheless, they are quite powerful when we use them to carve out a substantive distinction between generic and particular grounding claims. The current account does not posit a primitive notion of a metaphysical law, for instance.
Additionally, DEFINITION is relatively neutral about the nature of grounding. It does this in two ways.
First: the view I have sketched is compatible with a deflationist, or explanation-theoretic account of grounding. On these views, grounding should be understood in terms of metaphysical explanation, as opposed to vice versa. (See: Litland (2013), Thompson (2016, 2018), Richardson (2018: forthcoming), Litland (2013), and Dasgupta (2017).)
We start by thinking of grounding as holding between truths and propositions. We take a grounding explanation to be a metaphysically explanatory truth. On this view, we think of a real definition as a metaphysically substantive set of necessary and sufficient conditions, instead of a definition of properties.
Second: DEFINITION allows for the possibility that grounding is upwardly contingent even though many paradigm cases of grounding appear to be upwardly necessary. Generic grounding implies that many particular grounding claims will be both upwardly and downwardly necessary. However, I do not assume that grounding, in general, is upwardly necessary. For all I have said, grounding could be upwardly contingent.
This is a welcome result for those who think grounding is upwardly contingent (Leuenberger 2014; Dancy 2004; Chudnoff 2011; Schnieder 2006; Skiles 2015). To explain away the impression that grounding is upwardly necessary, you might point out that the paradigm cases of upwardly necessary grounding are, in fact, generics. If generics express real definitions, then certainly those grounding claims are upwardly necessary, but this is only a reflection of real definition, not grounding more generally.
Another way to think of it: we should endorse a kind of pluralism about grounding. On my view, real definition is just another species of grounding. Typically, we understand a species as its genus plus its differentia. For Aristotle, a human (species) is a animal (genus) that is rational (differentia). For me, a real definition (species) is grounding (genus) plus definition (differentia). The resulting pluralism is modest because it does not imply the disunity of grounding. It also does not imply the indefinability of the species of grounding. Nonetheless, it singles out a theoretically significant species of grounding—real definition. This sort of grounding may be necessary while other kinds of grounding may be contingent. (For more details on this view, see Richardson (forthcoming).)
I have said this view should be appealing to contingentists. But it should also be appealing to necessitarians. Why? Because the notion of a real definition provides some shared notion between necessitarians and contingentists. Even if one disagrees about the modal properties of grounding, by itself, there is less disagreement about what definitions require. And since in many cases, metaphysicians appear to be interested in definitional claims, the dispute about the modal force of grounding may not be as divisive as once thought.
My goal has been to show that DEFINITION is a promising way to deal with the inadequacies of the regularity theory. However, my goal has not been to show that DEFINITION is the only or best way to deal with those inadequacies. The latter appraisals can only be made after comparing DEFINITION with additional rival theories. It is unclear what such rival theories would be, but even if it were, such a discussion is better suited for a future paper.
In the current paper, I’ve argued that generic grounding claims are distinctively generic, and I’ve accounted for this genericness using real definitions. There is more to be said about the differences between generic and particular grounding claims. I have constructed an initial framework for this discussion.
I want to thank audiences at MIT, Yale, and the 2018 Canadian Metaphysics Collaborative Meeting. I am particularly grateful for the feedback given by Stephen Yablo, Sally Haslanger, Brad Skow, Jack Spencer, and the anonymous referees.
The author has no competing interests to declare.
Audi, P. 2012. Grounding: Toward a Theory of the In-virtue-of Relation. Journal of Philosophy, 109(12): 685–711. DOI: https://doi.org/10.5840/jphil20121091232
Bakhurst, D. 2008. Laughter and moral ambiguity: particularist reflections on the ethical dimensions of humour. In: Strahovnik, V, Potrc, M and Lance, MN (eds.), Challenging Moral Particularism, 53–74. New York, NY: Routledge.
Beebee, H. 2000. The non-governing conception of laws of nature. Philosophy and Phenomenological Research, 61(3): 571–594. DOI: https://doi.org/10.2307/2653613
Berker, S. 2019. The explanatory ambitions of moral principles. Noûs, 53(4): 904–936. DOI: https://doi.org/10.1111/nous.12246
Borg, E. 2004. Minimal Semantics. New York, NY: Oxford University Press. DOI: https://doi.org/10.1093/0199270252.001.0001
Cappelen, H and Lepore, E. 2005. Insensitive Semantics: A Defense of Semantic Minimalism and Speech Act Pluralism. Malden, MA: Wiley-Blackwell. DOI: https://doi.org/10.1002/9780470755792
Carlson, GN. 1989. On the semantic composition of english generic sentences. In: Chierchia, G, Partee, BH and Turner, R (eds.), Properties, Types and Meaning: Volume II: Semantic Issues, 167–192. Dordrecht, Netherlands: Springer. DOI: https://doi.org/10.1007/978-94-009-2723-0_5
Chudnoff, E. 2011. What Should a Theory of Knowledge Do? Dialectica, 65(4): 561–579. DOI: https://doi.org/10.1111/j.1746-8361.2011.01285.x
Dancy, J. 2004. Ethics Without Principles. Oxford, UK: Oxford University Press. DOI: https://doi.org/10.1093/0199270023.001.0001
Dasgupta, S. 2017. Constitutive Explanation. Philosophical Issues, 27(1): 74–97. DOI: https://doi.org/10.1111/phis.12102
deRosset, L. 2013a. Grounding Explanations. Philosophers’ Imprint, 13(7). URL: http://hdl.handle.net/2027/spo.3521354.0013.007.
deRosset, L. 2013b. No Free Lunch. In: Schnieder, B, Hoeltje, M and Steinberg, A (eds.), Varieties of Dependence: Ontological Dependence, Grounding, Supervenience, Response-Dependence (Basic Philosophical Concepts). Munich, Germany: Philosophia Verlag.
Fine, K. 1994. Essence and modality. Philosophical Perspectives, 8: 1–16. DOI: https://doi.org/10.2307/2214160
Fine, K. 2012. Guide to Ground. In: Correia, F and Schnieder, B (eds.), Metaphysical Grounding, 37–80. Cambridge University Press. DOI: https://doi.org/10.1017/CBO9781139149136.002
Fine, K. 2016. Identity criteria and ground. Philosophical Studies, 173(1): 1–19. DOI: https://doi.org/10.1007/s11098-014-0440-7
Krika, M, Pelletier, FJ, Carlson, GN, ter Meulen, A, Chierchia, G and Link, G. 1995. Genericity: An Introduction. In: Carlson, GN and Pelletier, FJ (eds.), The Generic Book, 1–124. Chicago: University of Chicago Press.
Leuenberger, S. 2014. Grounding and Necessity. Inquiry, 57(2): 151–174. DOI: https://doi.org/10.1080/0020174X.2013.855654
Lewis, D. 1975. Adverbs of quantification. In: Keenan, EL (ed.), Formal Semantics of Natural Language, 178–188. Cambridge University Press. DOI: https://doi.org/10.1002/9780470758335.ch7
Litland, JE. 2013. On Some Counterexamples to the Transitivity of Grounding. Essays in Philosophy, 14(1): 3. DOI: https://doi.org/10.7710/1526-0569.1453
Richardson, K. 2018. Grounding Pluralism: Why and How. Erkenntnis. DOI: https://doi.org/10.1007/s10670-018-0083-8
Richardson, K. Forthcoming. Grounding is necessary and contingent. Inquiry: An Interdisciplinary Journal of Philosophy, 1–28. DOI: https://doi.org/10.1080/0020174X.2019.1612777
Rosen, G. 2010. Metaphysical Dependence: Grounding and Reduction. In: Hale, B and Hoffmann, A (eds.), Modality: Metaphysics, Logic, and Epistemology, 109–36. Oxford, UK: Oxford University Press. DOI: https://doi.org/10.1093/acprof:oso/9780199565818.003.0007
Rosen, G. 2015. Real Definition. Analytic Philosophy, 56(3): 189–209. DOI: https://doi.org/10.1111/phib.12067
Rosen, G. 2017a. Metaphysical relations in metaethics. In: McPherson, T and Plunkett, D (eds.), The Routledge Handbook of Metaethics, 151–169. New York, NY: Routledge. DOI: https://doi.org/10.4324/9781315213217-10
Rosen, G. 2017b. What is a moral law? In: Shafer-Landau, R (ed.), Oxford Studies in Metaethics, 135–159. Oxford University Press. DOI: https://doi.org/10.1093/oso/9780198805076.003.0006
Schnieder, B. 2006. Truth-Making without Truth-Makers. Synthese, 152(1): 21–46. DOI: https://doi.org/10.1007/s11229-004-7905-8
Skiles, A. 2015. Against Grounding Necessitarianism. Erkenntnis, 80(4): 717–751. DOI: https://doi.org/10.1007/s10670-014-9669-y
Stanley, J and Szabó, ZG. 2000. On quantifier domain restriction. Mind and Language, 15(2–3): 219–61. DOI: https://doi.org/10.1111/1468-0017.00130
Thompson, N. 2016. Grounding and Metaphysical Explanation. Proceedings of the Aristotelian Society, 116(3): 395–402. DOI: https://doi.org/10.1093/arisoc/aow012
Thompson, N. 2018. Irrealism about Grounding. Royal Institute of Philosophy Supplement, 82: 23–44. DOI: https://doi.org/10.1017/S1358246118000206
Trogdon, K. 2013. Grounding: Necessary or Contingent? Pacific Philosophical Quarterly, 94(4): 465–485. DOI: https://doi.org/10.1111/papq.2013.94.issue-4
Zagzebski, LT. 2004. Divine Motivation Theory. Cambridge, UK: Cambridge University Press. DOI: https://doi.org/10.1017/CBO9780511606823