2.a Fusionism

Fusionism is the following package of claims:

The first claim, Fusion Identity, identifies groups with fusions of individual group members, where these members are taken to be concrete particulars. It also specifies a necessary condition on group membership: x is a member of a group g only if x is part g. Footnote ³

The second claim, Singular Reference, tells us about the behavior of ordinary language group terms. A common assumption in the metaphysics of social groups literature is that we are giving an account of what ordinary language speakers and thinkers refer to when they use social group terms. Specifically, fusionists take group terms to refer to individuals; the terms are nonetheless group terms because they refer to fusions of a certain kind.

The last claim, Fusion Extensionality, puts a constraint on the nature of fusions, namely, that they are individuated by their proper parts. One motivation for extensionality comes from supplementation principles; such principles clarify the sense in which a composite object cannot have only one proper part.Footnote ⁴ Common versions of these principles entail extensionality. Another motivation is more philosophical: a fusion is the minimal object composed by two or more things.

To see how the coincidence objection arises, we consider the standard example by Uzquiano (Reference Uzquiano2004). Imagine that all and only the members of the United States Supreme Court (SC) were assigned to a Special Committee on Judicial Ethics (JC). And let us assume these groups have the same members until they both fail to exist. A natural interpretation of this scenario is that SC and JC share the same proper parts. Given Fusion Extensionality, SC = JC.Footnote ⁵ Suppose the Supreme Court meets to discuss a standard court case, where none of their business concerns judicial ethics. Now consider the following.

Singular Reference tells us that ‘the Supreme Court’ and ‘the Special Committee’ refer to SC and JC. But of course, SC = JC, so the two expressions are co-referring. Leibniz’s Law (the Indiscernability of Identicals) says that identical objects have the same properties, so it follows that (1) and (2) must either both be true or both be false.

But the intuition is that (1) is true while (2) is false. There are various ways to justify this type of intuition. One line of reasoning is that the members of the Supreme Court, on that occasion, are not acting as a Special Committee on Judicial Ethics. Another line of reasoning appeals to the differences in deontic and origin facts between the two groups (Uzquiano Reference Uzquiano2004, 144–45). Deontically, the Supreme Court has powers that the Special Committee does not. A slightly different example will illustrate this.

The intuition is that (3) is true while (4) is false. Finally, you could imagine that the Special Committee was created or appointed solely by the Senate.

The intuition is that (5) is false while (6) is true. But if Fusionism is true, members of each pair—(1) and (2), (3) and (4), (5) and (6)—cannot differ in truth-value.

One response to this objection is to keep Fusion Extensionality but insist that social groups have social structures as parts. However, this view would be better classified as a structuralist view in the spirit of Ritchie’s (Reference Ritchie2015, Reference Ritchie2020) “structural wholes” account of social groups. Another response to the objection is to drop Fusion Extensionality in favor of a nonextensional mereological theory (Cotnoir Reference Cotnoir2010; Cotnoir and Bacon Reference Cotnoir and Bacon2012; Cotnoir Reference Cotnoir2013). However, the relevant metaphysical interpretation of nonextensional mereology generally takes a structuralism form; the view is that two objects have the same proper parts but they are nonidentical because they realize different social structures.

2.b Pluralism

Pluralism is the following package of claims.

A plurality differs from a fusion in that a plurality is many while a fusion is one. And instead of a parthood relation on fusions, the pluralist is committed to an is one of (or is among) relation on pluralities, where pluralities are not distinct from the things that are among them. So Plural Identity simply identifies the Supreme Court with its members.

The pluralist is committed to plural reference. We use ‘Bertrand Russell’ to refer to a single individual: namely, Russell. Similarly, we use ‘Russell and Whitehead’ to refer to multiple individuals: namely, Russell and Whitehead. Much has been written to clarify and defend the notion of plural reference and its usefulness in understanding natural language meaning (McKay and MacKay Reference McKay2006; Oliver and Smiley Reference Oliver and Smiley2016; Linnebo Reference Linnebo and Zalta2017; Ben-Yami Reference Ben-Yami2019; Florio Reference Florio2021). I will assume the notion of plural reference is sound. Though I should note that Horden and López de Sa (Reference Horden and de Sa2020) claim that group terms are plural but nonrigid; so ‘the Supreme Court’ does not refer to the same individuals at each time. In some circumstances, we refer to the actual Supreme Court; in others, we refer to a past Supreme Court. The appeal to flexible reference is intended to account for the temporal and modal flexibility of group-talk. For the sake of simplicity, I will assume the fusionist takes group terms to be similarly flexible.

Lastly, pluralists are committed to extensionality. This is intuitive. If a plurality is identical to the individuals among it, it is hard to see how pluralities could fail to be extensional. Speaking of a plurality is just a convenient way of speaking of the individuals among it; the term ‘plurality’ should not be ontologically inflated. I use xx, as opposed to x, to indicate a plural variable.

The coincidence objection applies to Pluralism in the general way it applies to Fusionism. The main difference is that we take expressions to plurally refer and therefore we must appeal to a notion of plural identity, and plural identity seems to be governed by Leibniz’s Law, just like singular identity. So ‘Supreme Court’ and ‘Special Committee’ plurally refer to the plurality of members of the Supreme Court and the plurality of members of the Special Committee, respectively, but those are identical, given Plural Extensionality. We get the coincidence objection because this rules out the intuitive possibility that (1) “The Supreme Court is in session” could be true even if (2) “The Special Committee is in session” were false.

3.a Semantic shifting

Where does this linguistic shifting occur? The natural culprit is to take the semantic content (or the literal meaning) of sentences about groups to shift what they say about group roles; one meaning of “g is F” concerns g satisfying group role F ₁ while another meaning concerns g satisfying group role F ₂.

One way to do this is to endorse what Fine (Reference Fine2003) calls the predicational-shifting strategy, in which our sentences shift between ascribing different role-properties to the same group. Recall (1) and (2), reprinted as (7) and (8).

To be in the session as the Supreme Court is to play the Supreme Court role with respect to being in session. If (7a) is the meaning of (7), then (7) is true. And if (8a) is the meaning of (8), then (8) is false. Moreover, the falsity of (8) does not commit us to coincident groups because (8) is false in virtue of the fact that the group does not play the role of the Supreme Court at the time.Footnote ⁶ (I omit time indices, for simplicity.)

As Uzquiano (Reference Uzquiano2004) and Horden and López de Sa (Reference Horden and de Sa2020) point out, the predicational-shifting approach introduces semantic complexity. To start, it introduces the possibility of predicate shifting every time we have a group term. So we cannot simply say that the group predicates ‘is in session,’ ‘wins,’ or ‘decided’ express properties like is-in-session, wins, or decided. Rather, they must be relativized either by introducing complex monadic properties—e.g., is-in-session-as-X, wins-as-X—or relations that take groups and group roles as relata—e.g., is-in-session-as, wins-as. Since the same story applies to singular social roles, many predicates applying to individuals will also have relativized values. One wonders if this added complexity is good semantic theory.

Additionally, the predicational view has to account for the felt falsity of:

Uzquiano (Reference Uzquiano2004, 143) makes this point. Hawley (Reference Hawley2017, 405) suggests that (9) is true but pragmatically infelicitous. If we accept Hawley’s suggestion, however, the theory becomes complicated enough that a purely pragmatic explanation—which I will discuss soon—may be just as good if not better.

Setting aside predicational shifting, there is another kind of semantic shifting that materialists sometimes appeal to. Perhaps the meaning of the noun phrase shifts from a group interpretation to a group role interpretation (Hawley Reference Hawley2017; Horden and López de Sa Reference Horden and de Sa2020). Call this phenomenon nominal shifting. Horden and López de Sa (Reference Horden and de Sa2020) give an example of a seemingly sound argument:

Of course, if group terms refer to particular groups, materialists must deny that the argument is sound. But Horden and López de Sa (Reference Horden and de Sa2020) claim that, in such cases, the terms ‘Supreme Court’ and ‘Special Committee’ refer to their respective group roles, not particular groups. They add that this role use of group terms is secondary because group roles cannot do things we standardly attribute to groups, like walking, singing, or voting. Horden and López de Sa (Reference Horden and de Sa2020) do not think nominal shifting completely explains the coincidence intuitions, but they think it is an important part of the story.

I am skeptical. Nominal shifting is not straightforward. Suppose ‘The Supreme Court’ in (10) refers to the abstract role of the Supreme Court. In that case, we get:

But that isn’t right. The abstract object doesn’t have the power to interpret the Constitution. Perhaps anticipating this, Horden and López de Sa (Reference Horden and de Sa2020) do not give this straightforward translation of (10); instead they propose the following interpretation.

Now (14) seems true, but it doesn’t say the same thing as (10). There is a difference between having a power—being permitted or licensed to do something—and involving a power. This suggests that nominal shifting will inevitably involve more than shifting the meaning of the noun phrase from a group meaning to a group-role meaning; rather, the relevant properties being attributed must change, as well. But it is unclear what the relevant shift will look like.

Here is another case that highlights the problems with the nominal-shifting strategy.

I believe (15) sounds true, but how should it be interpreted? Suppose we think ‘The Supreme Court’ refers to the abstract role of the Supreme Court. The problem is that it is clear that the abstract role did not exercise its power by making the Dred Scott decision.

There are other cases like this. Consider the following.

The particular group can meet on Monday but not the social role. But if (16) is true, it seems like the particular group also has the power to interpret the Constitution. For one more example, imagine the following dialogue.

One cannot point at the group role of the Supreme Court, but you can point at the particular group.

Now, we can certainly cook up an interpretation of these sentences in which the group role is part of their interpreted logical forms. But natural language semantics is not a matter of giving sentence-by-sentence interpretations. We need a compositional semantic theory, which, in this case, requires specifying the meanings of group terms and how they compose with the meanings of other subsentential expressions. But this is not what we get from the nominal-shifting strategy.

3.b Pragmatic explanations

The semantic-shifting strategy attempts to vindicate the idea that (7) “The Supreme Court is in session” is true while (8) “The Special Committee is in session” is false. But an alternative strategy is to insist that, strictly speaking, (7) and (8) are both true, but (8) is misleading or inappropriate to assert. This is the pragmatic strategy which has most recently been pursued in detail by Horden and López de Sa (Reference Horden and de Sa2020).Footnote ⁷

On this view, (8) suggests that the Special Committee is in session because they are, or plan to, play the Special Committee role. However, they are not, so you ought not assert (8). Horden and López de Sa (Reference Horden and de Sa2020) think the inappropriateness of (8) can be expressed via “metalinguistic negation.”Footnote ⁸ For example, consider:

Given the pragmatic explanation, (19) is literally false, but it is appropriate because it expresses the following idea: you should not say (8) “The Special Committee is in session” because it misleads people; (8) suggests that the Special Committee is in session because it plays the Special Committee social role. Horden and López de Sa (Reference Horden and de Sa2020), following work by Schnieder (Reference Schnieder2006) and Almotahari (Reference Almotahari2014), argue that metalinguistic negation is the reason why arguments that purport to prove the existence of coincident nonidentical objects fail. Recall (10)–(12), reprinted below as (20)–(22).

The claim is that (21) is literally false. Consequently, the argument is unsound. Nonetheless, (21) is an appropriate thing to say, given its metalinguistic interpretation.

One immediate problem with the pragmatic approach is that it does not give a clear verdict in the cases involving asymmetric relations. Here, I give a version of an example discussed in the propositional attitude reports literature (Richard Reference Richard1983; Salmon Reference Salmon1986; McKay Reference McKay1991; Salmon Reference Salmon1992).

Insofar as (21) sounds true, (23) sounds true. Having-more-powers-than is an asymmetric relation. But if it is, then the truth of (23) requires the Supreme Court to be distinct from the Special Committee.

More generally, I worry that the pragmatic strategy lacks a properly linguistic justification. I take it that metalinguistic negation is not always in effect in nonidentity arguments that rely on Leibniz’s Law; otherwise, the apparent validity of all such arguments for nonidentity could be explained away. We also cannot simply say that metalinguistic negation exists in these cases because it would support the plausibility of materialism; such an explanation gives a metaphysical justification for a linguistic hypothesis, but we need at least some independent linguistic support for metalinguistic negation in these particular cases. I assume there must be something about social groups, or social-group terms, that trigger the metalinguistic interpretation. But what? The materialist who appeals to pragmatics needs a principled answer to this question. So while I acknowledge that the pragmatic explanation could be correct, and that language users could be mistaken about the content of their utterances, I see no obvious linguistic basis for rejecting the hypothesis that (8) is literally false and its falsity is transparent to language users.

To be fair, neither Hawley (Reference Hawley2017) nor Horden and López de Sa (Reference Horden and de Sa2020) purport to give a full account of the semantics or pragmatics of group-talk. Their primary tasks are to defend the relative plausibility of fusionism and pluralism, respectively. My critiques are not intended to show that their initial proposals are completely off base, only that they are incomplete in crucial ways. We need an account that can vindicate materialism while being linguistically principled, and I offer such an account in what follows.

4.a The psychological theory explains linguistic intuitions

I believe materialists have erred in trying to explain the illusion of coincident but nonidentical groups by purely linguistic means. The deeper explanation is psychological. Specifically, I explain intuitions of coincidence by claiming (a) that the same social group can sometimes fall under conflicting social norms and (b) one set of social norms tends to dominate others, from the perspective of agents.

The relevant notion of a social norm is specified by Bicchieri:

On this view, social norms have two components: empirical expectation and normative expectation. For example, there are norms that tells us how to form a queue; we expect people to take their rightful place in the queue, and when they skip ahead of others in the queue, we view this behavior unfavorably. Empirical expectation says that most people in my city do not expect others to skip their place in a queue. (Notice that it does not have to be the case that the rule is actually followed by most people.) Normative expectation says that most people in my city believe that you should not skip your place in a queue.

Let us distinguish between global and local conformity to social norms; we may obey a norm globally (in most circumstances) while disobeying it locally (in some particular circumstance). For example, I can globally conform to the norm of taking my rightful place in a queue even if I violate that norm by skipping the queue on some particular occasion; in the case of a violation, I fail to locally conform to the social norm.

Some care must be taken to specify what it takes for a group to conform to a social norm. The distributive reading is simple: a group conforms to a norm just in case most or all of its members do (most or all of the time). But there may be cases where a group must collectively conform to a norm. The baseball team is expected to practice and play together; this norm cannot be met by each individual team member. There are subtleties here, but I will assume there is a coherent notion of collective group conformity to a social norm. Thomasson (Reference Thomasson2019) goes into more detail on the relationship between social groups and social norms, but I only need these basic facts for my account.

So far, I have described ways in which groups conform or fail to conform to social norms. But it would also be useful to describe how social norms are mentally represented. To this end, I propose we model the empirical and normative components of social norms as sets of worldly states. By a state, I am referring to states of affairs, worldly facts, or parts of possible worlds.Footnote ⁹ For example, I might believe that Clay satisfies the athlete norms, in the actual world, by playing college basketball on a regular basis. We can name the temporally extended state of playing college basketball regularly basketball. The state basketball has various parts: basketball _t , which concerns Clay playing basketball at some particular time t; basketball _l , which concerns Clay playing basketball at some particular location; and so on. I then represent others in my reference network who I take to be satisfying the athlete norms, so Jill playing tennis regularly will be designated tennis. The point is that we can collect the sets of states—call it E—that actually satisfy the athlete norms. This accounts for the representation of empirical expectation. For normative expectation, we make a similar move. The difference is that the set of states consist of all the possible states that satisfy the athlete norms. We can imagine Clay playing tennis regularly and Jill playing basketball regularly. We include these additional states in the set that represents normative expectation. We name this set of states N.

Once we have a representation of both empirical and normative expectation, we can represent them collectively using an ordered pair <E, N>. To be clear: <E, N> is not the social norm itself, nor does it represent the social norm. Rather, <E, N> is the representation that specifies the empirical and normative content of the rule that individuals may prefer (conditional upon whether other people actually conform to the rule and believe they ought to conform to it). Call pairs like <E, N> norm representations. I do not take these representations to be propositions, even though there are natural ways to construct propositions from them.Footnote ¹⁰ My claim is that there is some mental representation—propositionally structured or not—that captures both empirical and normative expectation.

Now consider two different norm representations <E ₁, N ₁> and <E ₂, N ₂> that involve exactly the same individuals. Suppose we are comparing the same people as athletes and students. We know these norm representations will be different because, for example, N ₁ will include a state basketball in which Clay plays basketball regularly, while N ₂ will not; instead, it will include a state study, in which Clay studies regularly. We also know these norm representations contain the same individuals because basketball and study share a common part—namely, Clay. It is clear that norm representations can differ while having some of the same objects in common.

I will now discuss how social norms and norm representations can help with the coincidence objection. In cases of seemingly coincident but nonidentical groups, we have three features: (i) the same group conforms to multiple, conflicting social norms; (ii) the same group will activate different norm representations in different agents; (iii) the difference in norm representations affects judgments about the acceptability and truth-values of sentences involving group terms.

The same individual can conform to multiple, conflicting social norms. This is clearest in cases of social roles. If you are a student athlete, you are governed by norms concerning students and norms concerning athletes. These norms will inevitably conflict, as when Clay must decide between studying for a big exam or preparing for a big game. Similarly, the same group can conform to conflicting social norms. Maria and Jack are a married couple, but they are also business partners. The norms of marriage and the norms of business partnerships will sometimes conflict in a particular circumstance. Nonetheless, a group can globally conform to both sets of norms. Or consider the Supreme Court and the Special Committee on Judicial Ethics; one set of norms concerns directly deciding court cases, while another concerns ethical conduct.

The possibilities for thought and deliberation can substantially differ depending on what social norms one takes an individual or group to follow. Even if you know you are both a student and athlete, the relevant norms present different possibilities for action and thought. Thinking like an athlete will lead to one path; thinking like a student will lead to another. At any given time, one set of social norms may be dominant (to use terminology from Burke [Reference Burke1994, Reference Burke1997]) from the perspective of observers or the one following the norm.Footnote ¹¹ More precisely, when we take the athlete norms to be dominant, we activate norm representations that represent Clay as conforming, and aspiring to conform to, the athlete norms rather than the student norms.Footnote ¹² To represent an object o as conforming to a social norm is to take o to be part of an e in E for some norm representation <E, N>. To represent an object o as aspiring to conform to a social norm is to take o to be part of an n in N for some norm representation <E, N>.

The normative expectations of these norm representations will shape our sense of what is practically possible. Here is a group example: if I take Maria and Jack to be business partners, I expect different behavior from them than I would if I expected them to be a married couple. Even if I know that they are both, I can nonetheless hold different expectations of them, depending on which social norms I take to dominate. The difference in perspective also holds internal to the group; the space of possibilities will look different for Maria and Jack depending on what social norms they take to be dominant at a given moment.

These representational differences creep into our judgments about the acceptability and felt truth-value of sentences. Recall López de Sa’s (Reference López de Sa2007) case in which John Roberts is both the Chief Justice of the Supreme Court and the Head of the Special Committee. When we use the terms ‘Chief Justice’ and ‘Head of the Special Committee,’ we think of the relevant social norms. Because the norms are different and can conflict, we sometimes have conflicting judgments of acceptability and truth about the sentences that contain these terms. Or suppose you are faced with utterances of (1) “The Supreme Court is in session” and (2) “The Special Committee is in session.” As the case is described, you take the social norms associated with ‘Supreme Court’—the Supreme Court norms—to be dominant. As a consequence, you take (2) to be defective because you do not expect the group to locally conform to the Special Committee norms. The sense of defectiveness may be semantic (falsehood) or pragmatic (inappropriate), but my point is that our judgments are affected by the presence of conflicting social norms and the dominance of one set of norms over another. We can explain the acceptability of (19) “The Special Committee isn’t in session” in the same way. We do not take the Special Committee norms to be dominant, so (19) seems right. Finally, consider the asymmetric case: (23) “The Supreme Court has more powers than the Special Committee.” This is a case where we contrast different norms applying to the same group, but we nonetheless take the Supreme Court norms as dominant. The asymmetry will consist in an asymmetry of normative expectations.

Now, there is no guarantee that we will always have the intuition that some particular set of norms is dominant. For example, suppose the Supreme Court plays a football match against some members of Congress and wins.

Both (24) and (25) seem equally appropriate or true. Why? Because the Supreme Court and Special Committee norms are not in conflict; rather, they are irrelevant to the subject matter at hand. We see the social group independently of the norms that are associated with the relevant group terms. The group did not win as the Supreme Court, nor did they win as the Special Committee. They won simpliciter.

This brings me to an important metaphysical difference between the current materialist theory and its structuralist rivals. Consider Thomasson’s (Reference Thomasson2019) norm-theoretic view of social groups. She says: different people can constitute the same group, where groups are defined by their social norms. I say: the same group can conform to different sets of social norms. Both views have a place for social norms, but the structuralist takes social norms to be definitive of social groups while the materialist does not. More generally, structuralists take social groups to be defined partly by social structures—norms (Thomasson Reference Thomasson2019), networks (Ritchie Reference Ritchie2013, Reference Ritchie2015, Reference Ritchie2020), profiles (Epstein Reference Epstein2019), etc. The materialist merely takes social groups to realize, instantiate, or conform to social structures.Footnote ¹³

My hypothesis is that the intuition of coincident but nonidentical social groups stems from the fact that the same group falls under social norms that conflict in some particular circumstance. The conflict drives us to take one set of norms to be dominant over the other. But when there is no conflict, we have no need for taking one set of norms as dominant. Furthermore, we should expect these psychological shifts to percolate up to our evaluation of the acceptability or truth of sentences containing group terms.

4.b The psychological theory explains nonlinguistic intuitions

You might be tempted to view the psychological theory as simply a norm-theoretic version of existing linguistic approaches. You might think: the utterance of (1) “The Supreme Court is in session” expresses or conveys the proposition that the referent of ‘The Supreme Court’ satisfies the social norms associated with the phrase ‘The Supreme Court.’ A similar story goes for (2) and ‘Special Committee.’ Such an account will not fundamentally differ from existing semantic and pragmatic accounts. So what makes the psychological theory fundamentally different from existing linguistic ones?

The psychological theory is not simply a version of existing linguistic theories because the psychological model helps explain the psychology of group representation more generally; representations triggered by language are only a special case. To show this, I will describe five ways in which the psychology of group representation is independent of the semantics and pragmatics of group terms.

One: some group sentences will either fail to activate a norm representation or the norm representations will fail to make a difference to the acceptability of group sentences. Recall (24) “The Supreme Court won the football match” and (25) “The Special Committee won the football match.” Given the subject matter of the case, the Special Committee and Supreme Court norm representations do not impact our acceptability judgments. So you cannot straightforwardly associate each sentence with a norm representation.

Two: the same group sentences will activate, or fail to activate, different norm representations in a way that has no clear linguistic basis. Another way to put it: intuitions about these coincidence cases are highly unstable. If someone says (1) “The Supreme Court is in session” when the court is considering matters of judicial ethics, (1) feels unacceptable. But consider the following discourse:

To my ear, “The Supreme Court was in session for the entire day” sounds acceptable when considered in light of all of the information provided by (26). But this contradicts the intuition that we can only properly describe the Supreme Court as being in session when they are conforming to the Supreme Court norms.

Three: there may not be a determinate norm representation or social-role property triggered by particular linguistic expressions. On standard semantic and pragmatic shifting accounts, there is an expression, like ‘Supreme Court,’ that semantically expresses or conveys a specific role property: playing the Supreme Court role. So far, I have described norm representations as if they are triggered in a similar, one-to-one manner; so ‘Supreme Court’ triggers a particular representation <E, N>. However, it may be more psychologically realistic to assume that (a) the relevant norm representations are fuzzy, and (b) the norm representations do not perfectly match these role properties (if they exist). I will start with the first point. Instead of taking ‘Supreme Court’ to activate a single representation <E, N>, we can take it to activate a set of related representations {<E1, N1>, <E2, N2>, …}. Call such sets fuzzy norm representations. Agents need not track a single set of Supreme Court norms or Special Committee norms; they only need to distinguish between fuzzy-norm representations. This brings me to the second point. Agents do not need to represent social role properties (if they exist) like playing the Supreme Court role. For example, playing the Supreme Court role might essentially require having been appointed by the US President. However, this information may not be relevant to someone who simply wants to distinguish between the different actions of the group called ‘Supreme Court’ and ‘Special Committee’; the fact that the group does not decide court cases when acting under ‘Special Committee’ may be enough. So there may be essential properties of group roles that do not go into the content of the relevant norm representations. Additionally, there may be inessential properties of group roles that do go into the content of norm representations. For example, suppose the group is consistently more jovial and humorous when acting under the name ‘Special Committee.’ If this association is robust, then joviality is enough to distinguish between when the group is conducting Supreme Court business and when it is conducting Special Committee business; joviality is part of the norm representation but clearly not part of the social-role property playing the Special Committee role. In either case, it is important to notice that, even though norm representations help explain thoughts activated by language, there is no systematic linguistic mechanism being described, here.

Four: norm representations can be activated in the absence of particular linguistic expressions. Imagine a different version of the Supreme Court case. On this version, there is no announcement that the Supreme Court or Special Committee is in session. The group members simply assemble themselves and begin doing the business associated with the Supreme Court norms or Special Committee norms. Now imagine an assistant whose job consists in keeping separate the two distinct businesses of the group. How does the assistant do this? The assistant assesses the actual business of the group for any given time period and determines the closest norm representation of the group; the assistant compares the actual behavior E _@ to either N ₁ or N ₂, where the latter represent compliance to Supreme Court and Special Committee norms. When the group is making judgments about ethics, they are interpreted as conforming to the Special Committee norms; when they are deciding court cases, they are interpreted as conforming to the Supreme Court norms. The assistant can keep track of the same group behaving in very different ways, even in the absence of special linguistic expressions that trigger norm representations.

Five: norm representations can be activated in the absence of language entirely. Suppose the same group regularly performs two different types of activities: hiking and studying. There is never a time when it is publicly announced that there are two different clubs, the hiking club and the studying club. Rather, the group members settle into performing reliably different activities together. Hiking and studying are considered separately; they do not hike immediately after studying or vice versa. And the expectations for hiking and studying are clearly different; you bring trail shoes for hiking, you bring a laptop and reading materials for studying, etc. It seems plausible that one can have different norm representations activated solely by observing the (mainly nonverbal) activities of the group. To explain how people represent the same group in different ways, we do not have to talk about language or idealized social-role properties; instead, we can consult the literature on social norms and group identity. Bicchieri writes:

Linguistic approaches do not say anything concrete about how the same group is mentally represented in different ways. In contrast, the psychological approach regards mental representation as the central phenomenon to be explained.

In general, it is unclear when and why a norm representation fails to be associated with the relevant linguistic expressions. Our judgments appear to vary, but there is no clear linguistic reason for this variation. If this is true, I would hesitate to call the resulting account a linguistic theory. The account is simply not systematic enough, or sufficiently tied to distinctively linguistic mechanisms, to merit the title. Nonetheless, it does appear that the account explains the underlying psychological attitudes that people have when representing groups.

I have described five ways in which the psychological account is detached from, and goes beyond, explaining language. These differences in the target of explanation lead to broader differences in methodology.

To start, much of the literature around group coincidence treats the issue as a special case of well-known puzzles around semantic opacity. Opacity is familiar from the philosophy of language.Footnote ¹⁴ Consider propositional-attitude reports. Lois works with Clark Kent, who she sees as a bad reporter. But she doesn’t realize that Clark Kent is the superhero Superman.

The argument from (27) and (28) to (29) feels invalid. And it would be invalid if propositional attitude terms—like ‘believes’—generated genuinely opaque contexts, contexts in which solely substituting co-referring expressions within a sentence will change the truth-value of that sentence. You may think group terms also generate opaque contexts. This may be so, but I think the core phenomenon consists of the fact that we represent groups in different ways, not that our language generates a puzzle. The representation of groups requires an additional or different kind of explanation. Humans are particularly sensitive to social norms in a way that is not captured by arbitrary instances of opacity.

Another methodological difference between the psychological approach and the linguistic one is that the psychological approach is relatively neutral about the exact semantics and pragmatics of group terms. But one might wonder: How can neutrality be possible? To answer this question, it is important to distinguish between metaphysics and natural language semantics. If you ask a natural language semanticist about the meaning of ‘table,’ they will say: the property is a table or the set of all tables. For the purposes of compositional semantics, this simple answer is good enough, but it will clearly fail to satisfy most metaphysicians, who participate in lively debates about the nature of ordinary objects like tables. So there is a sense in which natural language semantics does not say enough about the underlying metaphysical questions. On the other hand, natural language semantic theories may say too much, or the wrong things, about the underlying metaphysical questions. For example, event semanticists believe that semantic theories works best when they take linguistic meanings to contain events in their logical forms (Davidson Reference Davidson and Rescher1967; Parsons Reference Parsons1990; Champollion Reference Champollion2015, Reference Champollion2016). You may accept such views, for linguistic reasons, while rejecting the the underlying ontology of events (or at least some of its implications). With few exceptions, metaphysicians generally distinguish between the semantics of ordinary language and metaphysical inquiry.Footnote ¹⁵ There are various ways to cash out this distinction, but each account posits some separation between the metaphysics of X and the semantics and pragmatics of ‘X.’Footnote ¹⁶ The distinction matters because it shows that it is possible and coherent to not take a language-first approach to metaphysics; investigating the nature of X does not require answering all of the questions about how the term ‘X’ functions. My view is that, metaphysically speaking, social groups are concrete material particulars. This is because I take social groups to be mainly associated with the facts that other materialists like Hawley (Reference Hawley2017) and Horden and López de Sa (Reference Horden and de Sa2020) point to: the fact that groups participate in causal relations; they act and are acted upon; they persist through time; they are created and destroyed. One may object to such a view, but my current point is that the plausibility of this view does not depend on providing a semantic theory.

While I do not have a semantic or pragmatic theory to offer, my account does make minimal linguistic predictions. The first prediction is that we mostly use group terms in ways that suggest groups participate in causal relations. The second prediction is that the dominance of certain norm representations may affect our judgments of the appropriateness or truth of sentences involving group terms. In this way, my theory pays adequate attention to language without making strong commitments about natural language semantics.Footnote ¹⁷