# The Paraphrase Argument Against Collective Actions **Authors:** Johannes Himmelreich **Published in:** Australasian Journal of Philosophy 95(1), 2017 **DOI:** [10.1080/00048402.2016.1153685](https://doi.org/10.1080/00048402.2016.1153685) *This is the author's manuscript. Please cite the published version.* --- **Abstract:** This paper is about the status of collective actions. According to one view, collective actions metaphysically reduce to individual actions because sentences about collective actions are merely a shorthand for sentences about individual actions. I reconstruct an argument for this view and show via counterexamples that it is not sound. The argument relies on a paraphrase procedure to unpack alleged shorthand sentences about collective actions into sentences about individual actions. I argue that the best paraphrase procedure that has been put forward so far fails to produce adequate results. **Keywords:** Social Ontology, Collective Actions, Reduction, Paraphrase, Methodological Individualism. # Introduction Are there collective actions? Or are there only actions of individuals, of which we sometimes speak as if they were collective actions? Proponents of the *paraphrase view* answer the latter question in the affirmative and contend that all statements about collective actions are merely a shorthand for statements about the actions of individuals. For example, “the Supreme Court strikes down the Defense of Marriage Act” is a shorthand for “the justices strike down the Defense of Marriage Act”. This is taken to support the claim that collective actions metaphysically reduce to individual actions. In this paper I argue that this paraphrase view is untenable. I raise principled doubts about its overall strategy, reconstruct an argument for it, and show with three different counterexamples that it fails. Are statements about collective actions merely a shorthand for statements about individual actions? This paraphrase hypothesis is often expressed,[^1] but it has never been fully defended.[^2] The hypothesis is not obviously true; it demands a proof. Furthermore, the hypothesis makes a claim about “statements” (cf. Quinton (1975, 25)). But how does an observation about language bear out metaphysical facts about reduction? To answer this question, we need an argument. With this paper, I aim to rectify these shortcomings. I begin on a sceptical note in Section 2 and caution against undue optimism about the potential of paraphrase arguments in general. I then outline desiderata they should satisfy. In Section 3 I proceed more constructively and present a paraphrase procedure that systematically unpacks statements about collective actions into statements about individual actions. In Section 4 I reconstruct the argument for how this procedure vindicates the metaphysical reduction of collective actions. I then return to a sceptical outlook. I observe in Section 5 how this paraphrase procedure does not meet the desiderata discussed in Section 2. I show with three counterexamples different ways in which the paraphrase argument fails. I need to make two clarifications upfront. First, this paper focuses on the question of whether collective actions metaphysically reduce to individual actions. I take it to be a further question as to whether collective actions are eliminated by a reduction or whether they retain their status as entities \[List and Pettit 2011, 10; Sylvan 2012\]. Eliminative reductionists deny that collective actions exist. In contrast, non-eliminative reductionists accept that there are collective actions even though they metaphysically reduce to individual actions. I am interested in the common ground between these two positions, which is the reductionist claim that collective actions *just are* the actions of individuals. Second, I leave questions of constitution, grounding, contribution, and participation aside to focus solely on the reductionist claim. Constitutionists contend that collective actions are constituted by, or grounded in, the actions of individuals. This claim involves a grounding relation, but the reductionist claim involves an identity relation. The two relations differ importantly in their formal properties.[^3] In so far as the one involves an asymmetric (grounding) and the other a symmetric relation (identity), the constitutionist claim and the reductionist claim are incompatible. Similarly, we must distinguish the reductionist claim from claims about contribution. The reductionist claim is about the agency of actions. In contrast, the contribution claim is concerned with how an action comes about. An individual can contribute to a collective action without being an agent of it. It is consistent to claim that any collective action is brought about by individual contributions but that this collective action is distinct from the individual contributions. Similar considerations extend from contributions to participations. Each individual may participate in a distinct collective action. # Desiderata for a Paraphrase Procedure A paraphrase is a relation between sentences. It relates one sentence, the shorthand, to another sentence, the longhand. A *paraphrase procedure* is a method for producing longhand sentences from shorthand sentences, such that the former paraphrase the latter. The procedure is employed to determine what exists. In most cases a “\[p\]araphrase is offered as an escape route from excessive ontological commitments” (Jackson 1980, 304). The procedure needs to meet several desiderata, particularly when it is used to play this theoretical role (von Solodkoff 2014). I mention three desiderata and then focus on the last one. First, a paraphrase procedure needs to be supplemented with an explanation of why only the longhand sentence but not the shorthand sentence straightforwardly expresses ontological commitments (Alston 1958). This asymmetry, on which an eliminativist reduction rests, demands an explanation for two reasons. First, longhand sentences are not commonly used in language. Why should we rely on the longhand and not instead on the more commonly used shorthand to determine what exists? Second, a paraphrase usually merely puts in different terms what a shorthand sentence already expresses. What, in addition to restating the shorthand sentence in different terms, is a paraphrase procedure doing? This question leads to the next desideratum. Second, a paraphrase procedure should make the logical relation between shorthand and longhand sentences clear (Nolan 2014). The relation cannot be the biconditional, that is, that the two sentences have the same truth value. Having the same truth value is necessary but insufficient to relate longhand sentences to shorthand sentences. Many sentences have the same truth value but by no means stand in the paraphrase relation. A different proposal says that the relation between the sentences is a necessary conditional. That is, two sentences stand in the paraphrase relation if and only if, necessarily, when the shorthand sentence is true, the longhand sentence is true. Again, this is insufficient to relate longhand sentences to shorthand sentences. “Peter was more than 20 minutes late” necessarily implies “Peter was more than 10 minutes late”, but the former is not a paraphrase of the latter. We can strengthen this further and take the logical relation to be the necessary biconditional. Under this proposal, shorthand sentences and longhand sentences necessarily have the same truth value. But this is still insufficient to account for the logical relation between longhand and shorthand sentences, because all necessary truths have the same truth value but by no means stand in the paraphrase relation. Third, the paraphrase procedure must be universal. It must produce an adequate longhand sentence for any given shorthand sentence. What makes a longhand sentence adequate? I focus on the following necessary condition for adequacy. *Minimal Adequacy*.  \ A longhand (individual action) sentence is adequate only if it has the same truth value as its shorthand (collective action) sentence. If a paraphrase procedure fails to give an adequate longhand paraphrase for an alleged shorthand sentence, then the shorthand sentence might not actually be shorthand at all. A paraphrase procedure that fails to be universal also fails to support claims about metaphysical reduction. Reductionist claims need a universal paraphrase procedure because they are universally quantified, as, for example, the claim that *all* collective actions just are the actions of individuals. Given these desiderata, the prospect of formulating a paraphrase procedure to support a metaphysical reduction seems unlikely (Nolan 2014). Indeed, I argue that in the case of collective actions, the best paraphrase procedure known so far is not universal. It therefore fails to support the reductionist claim. # Paraphrase Procedure The paraphrase procedure I detail here rests on a regimentation of action sentences (Davidson 1967; Ludwig 2007). Let an *action sentence* be any sentence in a natural language that refers to an action. Consider the following examples. “I build a boat.” “We build a boat.” “The judge finds the defendant to be guilty.” “The judges find the defendant to be guilty.” In sentences 1 and 3 the subject term is singular, and in sentences 2 and 4 it is plural. But the form of the subject does not determine whether a sentence refers to individual or collective actions. Action sentences with a singular subject term can also refer to collective actions. An example is the action sentence “the Security Council condemned the violence in Syria”. Despite the singular subject term, this is a collective action in so far as the Security Council consists of several members. To avoid this complication, I focus on action sentences that, when they have singular subject terms, are about individual actions. These are *individual action sentences*. Analogously, all sentences with plural subject terms are *collective action sentences*.[^4] Apart from differing in the form of their subject term, the sentences share a similar surface structure. ## Individual Actions Before developing a paraphrase procedure to unpack collective action sentences, it would be useful to understand how to regiment individual action sentences. Davidson (1967) developed the standard proposal to regiment action sentences in first-order logic. I indicate a regimented sentence with an $R$, such as $R(1)$ for the regimented form of sentence 1. $\textbf{\emph{R}(1)}$. $(\exists e)[\textrm{agent}(a,e) \:\&\: \textrm{building}(e) \:\&\: \textrm{of}(p,e) ]$ The formula $R(1)$ reads: There exists an $e$ such that $a$ is an agent of $e$, this $e$ is of the type building, and this $e$ involves $p$. The letter $e$ is a variable denoting an event. The letter $a$ is an individual constant that refers to the person to which “I” refers in sentence 1. This person is an agent of the event $e$ because any $a$ that occurs in the first place of the *agent–event relation* “$\textrm{agent}(a,e)$” is an *agent* of the event $e$. Any event $e$ is a *collective action* if and only if its agent is a group. The expression “$\textrm{building}(e)$” is an *action predicate*, which says that this event is a certain action type. The individual constant “$p$” refers to the boat that is built. The boat is the patient of the action. The expression “$\textrm{of}(p,e)$” is a *patient–event relation*. Consider sentence 3 as a further example: “The judge finds the defendant to be guilty”. Let there be an event $e$ where a judge $a$ finds a defendant $p$ guilty. The action predicate $\textrm{finding-guilty}(e)$ expresses this action type. The action sentence 3 is then regimented as follows. $\textbf{\emph{R}(3)}$. $(\exists e)[\textrm{agent}(a,e) \:\&\: \textrm{finding-guilty}(e) \:\&\: \textrm{defendant}(p,e) ]$ These examples illustrate a simple way of regimenting action sentences that can be generalised with a schema. I abbreviate this schema as SI for *individual* action sentences. Sentences that are instances of this schema are *non-reductive regimentations*. SI. $(\exists e)[\textrm{agent}(a,e) \:\&\: \emph{action-predicate}(e) \:\&\: \emph{patient--event relation}(p,e)]$ Regimenting action sentences in this way has limitations. It leaves aside aspects of a sentence that are not relevant to our discussion. For example, to keep the formal language simple we ignore the time at which an action occurs. Furthermore, we do not allow for different agent–event relations (ignoring so-called thematic relations or *Theta roles*). We admit different action predicates for various types of actions, such as $\textrm{building}(e)$ or $\textrm{finding-guilty}(e)$, but we hold fixed the relation $\textrm{agent}(a,e)$. It is true that different action predicates may require particular senses of agency. But this distinction is only relevant for comparisons across sentences with different action types. In contrast, we investigate how individual action sentences relate to collective action sentences within one action type. So nothing speaks against holding the agent–event relation fixed. In fact, there are two important reasons in favour of doing so. First, in philosophy of action there is only one concept of being an agent of an action, that is, what is required of $a$ to be an agent of $e$.[^5] The agent–event relation encodes this concept of agency. Second, when attempting to reduce collective actions to individual actions we must keep fixed what we mean by “actions”. Otherwise, we risk moving the bar of what is necessary for being an agent of an action depending on whether we talk about a collective or an individual agent. We keep the meaning of “actions” fixed across individual and collective actions, by allowing only one agent–event relation. What does it take to be the agent of an action? Theories of intentional agency differ on this. But there are two widely accepted necessary conditions underlying the agent–event relation (Davidson 1963). The first is an intentional condition, the second a causal condition. The intentional condition says that an $a$ is an agent of $e$ only if there is a description of $e$ under which $a$ intends $e$. This condition is not too demanding. It allows that I am an agent of actions that turn out differently to how I intended. For example, when I try to wriggle a block out of a rickety tower in the *Jenga* game and the tower collapses, I am an agent of the collapse despite intending the tower not to collapse. The collapse can be described as an attempt to win the game, which is what I intended to do. The collapse is a side effect of my attempt, that is, it is something that I foresee might happen but that I intend not to happen. The second condition for being an agent of an action is a causal condition. An $a$ is an agent of $e$ only if $a$ stands in a causal relation to $e$. Different theories of action and causation give different interpretations of what this means. But we can bracket this issue for our present purposes. I will return to the conditions of the agent–event relation after presenting the paraphrase procedure for collective actions. ## Collective Actions How should we regiment collective action sentences? Our aim is to find a regimentation that can be used as a paraphrase procedure. The schema SI cannot be used for this purpose because it treats individual and collective action sentences alike. For example, the formula that results from applying schema SI to “we build a boat” is orthographically identical that of applying SI to “I build a boat”. The only difference is that the agent constant $a$ refers not to an individual “I”, but to a collective “we” suggesting that the event is a collective action. Hence, SI fails to unpack collective action sentences into individual action sentences. Ludwig (2007) develops a proposal that can be employed to regiment collective action sentences.[^6] The idea is to replace the plural subject term “we” with something like “each of us”. We amend the earlier schema SI by adding “$(\textrm{Each } x \textrm{ of } a)$” after the quantifier “$(\exists e)$”.[^7] Furthermore, we replace occurrences of $a$ with the new variable $x$. This gives us a new schema. I abbreviate this schema as SC because it will be used to regiment *collective* action sentences resulting in *reductive regimentations*. SC. $(\exists e)\,(\textrm{Each } x \textrm{ of } a)[\textrm{agent}(x,e) \:\&\: \emph{action-predicate}(e) \:\&\: \emph{patient--event-relation}(p,e)]$ In English, individual and collective action sentences may have the same surface structure. But proponents of the paraphrase argument contend that this similarity is misleading. Introducing the restricted quantifier $(\textrm{Each } x \textrm{ of } a)$ brings out an important difference between collective and individual action sentences. When we speak about a group doing something, we are in fact saying that individuals are doing something.[^8] The new schema formalises what we really mean with a collective action sentence. It will help to illustrate this proposal with some examples. Consider the sentence “we build a boat” and its regimentation $R(2)$. $\textbf{\emph{R}(2)}$. $(\exists e)\,(\textrm{Each } x \textrm{ of } a)[\textrm{agent}(x,e) \:\&\: \textrm{building}(e) \:\&\: \textrm{of}(p,e) ]$ The collective action sentence is regimented into a formula about individual actions that in English could be read as: “Each of us is an agent of an event that is a building of the boat”. To many, this sounds wrong. It seems to say that each of us builds the boat. But none of us builds the boat alone. Is there a flaw in the regimentation? No. We need to look at what $R(2)$ says precisely. It says that each of us *stands in the agent–event relation* to the construction of the boat. Standing in the agent–event relation means that the intentional and the causal condition for agency are met. In a joint action, there are several individuals that stand in the agent–event relation to the same action. Does each of us individually stand in the agent–event relation to the construction of the boat? We can consider the two necessary conditions for the agent–event relation. First, it seems plausible that each of us intends to build a boat. For each of us there is some description under which he or she intends to build the boat. Some of us want to hone their carpentry skills, while others simply want to get out of the house. Second, it seems plausible that each of us stands in a causal relation to the construction of the boat. If one of us slacks, then his or her part will be left undone or will have to be built by others. So it is plausible to assume that each of us meets the intentional and causal conditions necessary to stand in the agent–event relation to the action in question. Hence, the regimentation $R(2)$ gives a minimally adequate paraphrase of what seems to be a sentence about a collective action. Consider as a further example the case of a democratic election. Some argue that it presents a counterexample against the paraphrase procedure. I argue that the paraphrase procedure handles the case adequately. 5\. “We elect Anne as the mayor.” $\textbf{\emph{R}(5)}$. $(\exists e)\,(\textrm{Each } x \textrm{ of } a)[\textrm{agent}(x,e) \:\&\: \textrm{electing}(e) \:\&\: \textrm{of}(p,e) ]$ According to $R(5)$, each of us stands in the agent–event relation to Anne’s election. But does each of us meet the two necessary conditions of the agent–event relation? Consider first the intentional condition. For each of us there is a description of the election under which he or she intends it. Some intend to elect Anne and some others intend to elect her opponent, while others only intend to fulfil their civic duties and vote. Consider now the causal condition. Whether one stands in a causal relation to an event depends on the theory of causation. In some theories of causation, none of us individually stands in a causal relation to the event (cf. Lewis (1973)). But in other theories of causation, each of us stands in a causal relation to the event (cf. Chockler and Halpern (2004)). As such, the regimentation $R(5)$ may be a minimally adequate paraphrase, depending on the theory of causation. Hence, action sentences about democratic elections do not immediately make a decisive case against this paraphrase procedure. In summary, the regimentation schema SC defines a paraphrase procedure for simple collective action sentences. It produces longhand sentences that are minimally adequate. It provides a *linguistic reduction* of collective actions by giving an account of how sentences that seem to be about collective actions are in fact just sentences about individuals doing something. However, further assumptions are needed to get from the existence of a paraphrase procedure to metaphysical conclusions. Presumably, such an argument follows Quine (1960). But I share the observation of von Solodkoff (2014, 571) that “whilst it’s clear what role paraphrasing is supposed to play according to Quine, it is remarkably unclear how exactly this strategy works”. Let us see if we can reconstruct how it works in the next section. # Reduction Argument The paraphrase argument aims to show that collective actions metaphysically reduce to individual actions. It needs to show how the linguistic reduction of a paraphrase procedure supports a metaphysical reduction. I reconstruct this argument with two assumptions. The first assumption is that there is a universal paraphrase procedure that meets the minimal adequacy constraint as defined above. *Existence of a Paraphrase Procedure*.  \ F or each collective action sentence there is a paraphrase that has the same truth value as the corresponding collective action sentence. In particular the assumption requires that the reductive regimentation of any collective action sentence has the same truth value as its non-reductive regimentation. It does not make a difference to a sentence’s truth value whether it is regimented using schema SI or SC. But the reductionist claim that all collective actions just are actions of individuals does not yet follow. We require an additional assumption such as the following. *Paraphrase Implies Identity*.  \ A collective action *just is* an individual action if and only if, given a true collective action sentence about this action, there is an adequate paraphrase of that action sentence into an individual action sentence. This principle is the missing link between the existence of a paraphrase procedure and a metaphysical reduction. It states that the existence of a paraphrase shows that collective actions just are the actions of individuals.[^9] This assumption is reminiscent of Quine (1960, 241), who writes: “To paraphrase a sentence into the canonical notation of quantification is, first and foremost, to make its ontic content explicit.” With this additional assumption, the reductionist conclusion follows.[^10] The two assumptions form a valid argument for the conclusion that collective actions metaphysically reduce to individual actions. Is the argument sound? In the next section I argue that the first assumption is false. The paraphrase procedure fails to be universal and adequate. # Counterexamples Even if the paraphrase procedure works for *some* collective action sentences, it does not work for *all*. In this section, I discuss three counterexamples in which the paraphrase procedure gives inadequate results. In these examples, the procedure leads to individual action sentences with a different truth value than the corresponding collective action sentences. Each counterexample takes a different line of attack. The first is based on a central result of judgment aggregation. In this case, the collective but not the individuals meet the intentional condition. The second is based on a counterexample against functionalism. In this case, the collective but not the individuals meet the causal condition. The third counterexample explores the possibility that intentionality might be fundamental on the collective level. Since actions presuppose intentionality, reducing collective actions to individual actions is then a non-starter. Formally, we can trace the failure in each of these cases to a common source: the collective meets the conditions to stand in the agent–event relation but the individuals do not. Leaving the other conjuncts aside, a reductive regimentation (using schema SC) is equivalent to a non-reductive regimentation (using schema SI) if and only if the expression “$\textrm{agent}(x,e)$” in a reductive regimentation has the same truth value for every $x$ in the domain of the $(\textrm{Each } x \textrm{ of } a)$ quantifier as the expression “$\textrm{agent}(a,e)$” in the non-reductive regimentation. *But it is not a theorem that* $(\textrm{Each } x \textrm{ of } a)\,\textrm{agent}(x,e)$ *is a semantic consequence of* $\textrm{agent}(a,e)$. ## Discursive Dilemma Consider again action sentence 4: “The judges find the defendant to be guilty”. I first employ the paraphrase procedure to reductively regiment this sentence as an instance of schema SC. Compare this with a non-reductive regimentation under schema SI, with the constant $a$ for “the judges”. $\textbf{\emph{R}}_{SC}\textbf{(4)}$. $(\exists e)\,(\textrm{Each } x \textrm{ of } a)[\textrm{agent}(x,e) \:\&\: \textrm{finding-guilty}(e) \:\&\: \textrm{defendant}(p,e) ]$ $\textbf{\emph{R}}_{SI}\textbf{(4)}$. $(\exists e)[\textrm{agent}(a,e) \:\&\: \textrm{finding-guilty}(e) \:\&\: \textrm{defendant}(p,e)]$ For an adequate paraphrase, these two regimentations must be equivalent. But there exist situations where the reductive regimentation (SC) is false despite the non-reductive regimentation (SI) being true. Consider the example of a discursive dilemma (Kornhauser and Sager 1993; List and Pettit 2002; List and Pettit 2011; Tollefsen 2002, 36). Three judges have to decide on a civil case. The question before them is that of whether or not a defendant is guilty of a breach of contract. Assume that there are three conditions underlying the verdict. The defendant is guilty if and only if (1) there is a valid contract, (2) the defendant committed a breach of this contract, and (3) court papers were filed in accordance with all procedural requirements. Each judge intends to convict the defendant if and only if she finds the defendant guilty. |   | valid contract | breach | due process | guilty | |:-----------|:--------------:|:------:|:-----------:|:------:| | Judge 1 | ✓ | ✓ | ✗ | ✗ | | Judge 2 | ✓ | ✗ | ✓ | ✗ | | Judge 3 | ✗ | ✓ | ✓ | ✗ | | Collective | ✓ | ✓ | ✓ | ✓ | Profile of the judges’ opinions on the four propositions. As Table 1 illustrates, the judges vote on each of the three conditions and each condition achieves a majority. Two out of three judges find that there was a valid contract, a breach of contract, and a procedurally sound filing. This implies that they find the defendant to be guilty. Nevertheless, there is no judge who finds the defendant to be guilty. If the judges were to vote on whether the defendant is guilty, they would agree that she is not guilty. But the rules of the court prescribe that a collective decision is reached by voting only on the conditions. Hence, although none of the judges individually finds the defendant to be guilty, collectively they find the defendant to be guilty and the defendant is convicted. The paraphrase $\emph{R}_{SC}(4)$ says that each judge individually stands in the agent–event relation to the event of finding the defendant guilty. This sentence is false for two reasons. First, a conviction is not an unintended side effect but rather it is something that is done intentionally.[^11] Yet none of the judges intends this event to occur, so none of them is an agent of the conviction. Accordingly, the paraphrase $\emph{R}_{SC}(4)$ is false because the conjunct “$\textrm{agent}(x,e)$” is false.[^12] Second, the conjunct “$\textrm{finding-guilty}(e)$” is false as well. “Finding the defendant guilty” describes the formation of a belief. But no judge holds this belief.[^13] Of course, each judge does something. Each judge decides, votes, and hence contributes to the defendant being found guilty. Although each judge stands in the agent–event relation to these individual contributions, it does not follow that each stands in the agent–event relation to the defendant being found guilty. Now consider the sentence $\emph{R}_{SI}(4)$, on which there is a collective action. This sentence states that the judges collectively stand in the agent–event relation to the event of finding the defendant guilty. This non-reductive regimentation stays very close to the original sentence in English. So it is reasonable to suppose that if the original sentence is true, then this non-reductive regimentation should be true as well. But let me in addition sketch an elimination argument for the conclusion that the judges collectively meet the necessary conditions for the agent–event relation. Let us assume that finding the defendant guilty is an action and therefore requires an agent. Furthermore, there is at least one agent who intentionally finds the defendant guilty. None of the judges individually intends to find the defendant guilty and therefore no individual judge is an agent of intentionally finding the defendant guilty.[^14] By elimination, the judges collectively are an agent of finding the defendant guilty.[^15] Hence, the collective action sentence is true while the paraphrase into an individual action sentence is false. This violates the minimal adequacy condition and the paraphrase procedure fails. ## China’s Avatar I now advance a second counterexample against the reduction argument. This time we focus not on the intentional but on the causal condition. The group as a whole meets this condition, but the individual group members fail to do so. This counterexample adapts the China-body system of Block (2006). I call it China’s avatar. Suppose that the members of the Chinese nation are arranged so as to form a functional duplicate of a brain. The structure in which they are arranged and the dynamics in which they interact are identical to that of a brain. Each member simulates a brain neuron and uses a radio transmitter to connect her with others. All individuals have instructions on what to do in response to the signals they receive. Together they control a humanoid body, which instead of a brain has a remote control interface wired into its cranium. This body is China’s avatar. The avatar and the members of the Chinese nation together form the China-body system. I make two assumptions about the China-body system. First, the China-body system is a collective. Like many collectives, it may act as if it were one individual, but it really consists of many individuals. Second, the China-body system has intentions. The China-body system is a functional duplicate of a human individual. The two only differ in how these functions are realised. I take it for granted that differences in realisation do not make a difference concerning intentions.[^16] Now suppose that the China-body system intends to butter a slice of toast. In other words, it intends its avatar to move in such a way that its movements are an action of buttering a slice of toast. Sure enough, just as any individual would act on this intention, so does the China-body system.[^17] “The China-body system butters a slice of toast.” I contend that the paraphrase procedure fails for this sentence. The non-reductive regimentation is true but the reductive one is false. This time, the problem is not the intentional condition but rather the causal condition underlying the agent–event relation. The China-body system as a whole stands in the right causal relation to the action of buttering the toast. But its members do not. Figuratively speaking, while it is true of the system as a whole that it butters the toast, it is not true of its parts. There are at least three arguments for this conclusion. The first argument is that there are distinct causes. The collective, but not any individual, instantiates the relevant property that causes the slice of toast to be buttered. The second argument is that there are distinct effects. The event of buttering the toast is distinct from any individual action. The third argument is based on intuitions. It is intuitively false to claim that each individual member of the China-body system meets the causal condition for being an agent of buttering the toast. The first argument contends that the relevant cause of the toast being buttered is a property that is only instantiated by the collective. This argument draws an analogy to the ontological status of the mental properties of individuals over and above the neurological properties of their brains (Kincaid 1986). In the domain of individuals, we might assume that the property of having an intention is distinct from properties that realise this intention on a neural level. The same intention might be realised by different configurations of properties on the neural level. In this way, only the intention, but not any property on the neural level, stands in a robust causal relation to the action. This robustness makes the intention the relevant cause explaining the action. By analogy, the intention of the Chinese nation is distinct from properties realising this intention on the level of its members. Analogous to the view about individuals’ intentions, the same collective intention might be realised by different configurations of properties on the level of individuals. In this way, only the collective intention causes the action robustly and thereby stands in the relevant causal relation to the action of buttering the toast. This analogy is plausible because the Chinese nation forms a functional duplicate of a brain. It is one first argument to suggest that the paraphrase of sentence 6 is inadequate. Only the China-body system as a whole meets the causal condition for being an agent of buttering the toast, and hence only the collective stands in the agent–event relation to the action. The second argument for the conclusion that the reductive regimentation is inadequate contends that there are distinct actions. On the one hand there is the collective’s action of buttering the toast, while on the other hand there are the individuals’ actions of operating their transmitters. We appeal to the individuation conditions of events to show that the collective action is distinct from any individual action. There are two ways to individuate events. First, according to a spatio-temporal individuation, the collective action and the individual actions are distinct events because they do not occur at the same locations and at the same time.[^18] Second, according to a causal individuation, buttering the toast and responding to a signal are distinct because they adhere to different causal conditions. An individual intention is necessary and sufficient for the individual action to occur. However, no individual operating her transmitter is necessary for the collective action to occur. If some individuals do not operate their transmitter correctly, others will compensate for their failure and the toast will still be buttered. Furthermore, an individual responding to a signal is not sufficient for the collective action of buttering the toast to occur. Therefore, the collective action is distinct from the individuals’ actions because of differences in causal properties. The final argument arises from intuition. It contends that buttering the toast is an action only on the part of the China-body system because intuitively, the individuals do not meet the causal condition for standing in the agent–event relation to buttering the toast. The influence that each individual has on the occurrence of the action is vanishingly small. Intuitively, individual causal responsibility decreases with increasing group size (Lagnado et al. 2013). Hence, taking our intuitions about causation as a guide, no individual member of the China-body system stands in the agent–event relation to buttering the toast. The paraphrase procedure fails to give an adequate paraphrase. This second counterexample has similar ramifications to the first. The reduction argument is not sound because the paraphrase procedure fails. Some believe that there is a better paraphrase procedure; but there is insufficient evidence to support this conjecture. So far, it is reasonable to conclude that there is a collective action: the China-body system butters a slice of toast. ## Hive Minds Finally, we can question a tacit presumption of the previous cases to suggest yet another way in which the paraphrase procedure fails to be adequate and universal. The presumption on which all previous cases operate is that collective actions involve individual actions. But this is not necessarily true. There can be collective actions without individual actions. How are such cases possible? There may be a collective that has intentions but which consists of individuals who lack intentions. Actions require an intention. Without individual intentions there are no individual actions to which the collective action could be reduced. To find such cases, we need to look outside the domain of human interactions. In human groups, collective intentions may be derivative in the sense that a human group has intentions in virtue of its members’ intentions. But we can imagine a reversal of this relationship. In other species, collective intentions are the fundamental intentions from which individual intentions are derived (if there are individual intentions at all). These are cases of *hive minds*. Bees and ants seem to have fundamental collective intentions. An individual bee or ant is akin to a simple robot. Nevertheless, collective intentions may arise from the interactions of these robots. A bee hive can be thought of as “a kind of exposed brain that hangs quietly from a tree branch” (Seeley 2010, 204). Similarly, Hofstadter (2000, 328) suggests that “ant colonies are no different from brains in many respects”. The analogy is that just as intentions emerge from a brain that itself lacks intentions, intentions in a bee hive or an ant colony may emerge from individuals that themselves lack intentions. Similarly, Wilson (2001, S265) claims that “it is only groups of social insects, not individual members of those groups, that possess a mind . . . a *group-only* trait”. Consider the following example by Watkins (1952, 188–9). When a bee hive splits, we can observe that the bees in the two resulting hives will change their behaviour so that in each new hive the same proportion of bees occupies each role as in the original hive. For example, one hive will start rearing a new queen to fill the vacant position. It seems that this reorganisation is an intentional action. It can be argued that a hive as a whole intends a certain internal organisation. There are at least two ways to make this argument. First, one can appeal to an instrumentalist theory of intentions and argue that ascribing the intention to maintain a certain proportion of roles to the bee hive is useful for predicting its behaviour (Dennett 1989). Second, one can appeal to a functionalist theory of intentions and argue that the bee hive realises states that play the role of having an intention. Either way, a bee hive can be considered a collective agent that performs a collective action, even though no individual intentional actions are involved in this collective action (cf. List and Pettit (2006)). For a concrete action sentence, let us take the example of a bee hive choosing a new nest site. “The bees decide to relocate to a new nest site at $L$.” In deciding between alternative nest sites, individual bees implement cognitive functions similar to the functions implemented by human cognitive systems Seeley (2010, 204). Similarly to the earlier comparison between the China-body system and a brain, this suggests an analogy between a bee hive and a human cognitive system. One can contend that just as a human agent has intentions that are distinct from his or her brain states, the collective intention of the hive is distinct from the states of individual bees. This may sound similar to the counterexamples above, but the case of a hive mind shows something much stronger. What this counterexample suggests is that it is not possible to have a paraphrase procedure that is universal and adequate. In hive minds, collective actions cannot be reduced to individual actions because there are no individual intentions to which the collective intention for an action could be reduced. Without individual intentions there cannot be an individual action. Any adequate paraphrase procedure will have at best a restricted domain that excludes the collective actions of hive minds. # Conclusion Many contend that statements about collective actions are merely a shorthand for statements about individual actions and that, therefore, collective actions reduce to individual actions. I have argued that this paraphrase view about collective actions is untenable. I first outlined general desiderata for paraphrase procedures to then focus on the minimal requirement that a paraphrase must not change the truth value of a sentence. Second, I put forward a procedure for collective action sentences and constructed the reduction argument. Finally, I used three different counterexamples to demonstrate that the reduction argument does not succeed. This leads to two important suggestions. First, the failure of the paraphrase argument provides evidence in favour of two kinds of collective actions. There are *joint actions*, for which the paraphrase procedure works; these are just actions of individuals. But there are furthermore *corporate actions*, for which the paraphrase procedure fails; these are genuine collective actions (cf. Pettit and Schweikard (2006)). Second, the paraphrase project is far from complete. Given the popularity of the hypothesis that collective action sentences are merely a shorthand for individual action sentences, we can see relatively little momentum in developing a procedure to prove this suspicion right. Until an improved paraphrase procedure is developed, the claim that collective action sentences are merely a shorthand for individual action sentences still awaits vindication. **Acknowledgements:** I am grateful to Richard Bradley, David Chalmers, Ryan Cox, Richard Holton, Holly Lawford-Smith, Christian List, Kirk Ludwig, Marco Meyer, Daniel Nolan, Wlodek Rabinowicz, Jesse Saloom, Ying Shi, Jennifer Szende, and two anonymous referees for valuable comments and discussions at various stages in the development of this paper. The paper also benefited from comments from audiences at the PhD Seminar at the London School of Economics, the ENSO-III conference at the University of Helsinki in October 2013, and the 2014 APA Eastern Division meeting. # References Alston, William P., 1958. Ontological commitments, *Philosophical Studies* 9 1-2: 8–17. Block, Ned, 2006. Troubles with functionalism, in Eckert, Maureen, ed., *Theories of Mind: An Introductory Reader*, Rowman & Littlefield. Chockler, Hana and Halpern, Joseph Y., 2004. Responsibility and blame: A structural-model approach, *Journal of Artificial Intelligence Research* 22: 93–115. Cooper, D. E., 1968. Collective responsibility, *Philosophy* 43 165: 258–268. Copp, David, 1979. Collective actions and secondary actions, *American Philosophical Quarterly* 16 3: 177–186. Davidson, Donald, 1963. Actions, reasons, and causes, *The Journal of Philosophy* 60 23: 685–700. Davidson, Donald, 1967. The logical form of action sentences, in Rescher, Nicholas, ed., *The Logic of Decision and Action*, University of Pittsburgh Press. Dennett, Daniel Clement, 1989. *The Intentional Stance*, MIT Press. Elster, Jon, 2007. *Explaining Social Behavior: More Nuts and Bolts for the Social Sciences*, Cambridge University Press. French, Peter A., 1979. The corporation as a moral person, *American Philosophical Quarterly* 16 3: 207–215. Hofstadter, Douglas R., 2000. *Gödel, Escher, Bach: An Eternal Golden Braid*, Penguin Books. Isaacs, Tracy, 2011. *Moral Responsibility in Collective Contexts*, Oxford University Press. Jackson, Frank, 1980. Ontological commitment and paraphrase, *Philosophy* 55 213: 303–315. Kincaid, Harold, 1986. Reduction, explanation, and individualism, *Philosophy of Science* 53 4: 492–513. Kornhauser, Lewis A. and Sager, Lawrence G., 1993. The one and the many: Adjudication in collegial courts, *California Law Review* 81 1: 1–59. Lagnado, David A., Gerstenberg, Tobias, and Zultan, Ro’i, 2013. Causal responsibility and counterfactuals, *Cognitive Science* 37 6: 1036–1073. Lewis, David, 1973. Causation, *The Journal of Philosophy* 70 17: 556. List, Christian and Pettit, Philip, 2002. Aggregating sets of judgments: An impossibility result, *Economics and Philosophy* 18 01: 89–110. List, Christian and Pettit, Philip, 2006. Group agency and supervenience, *The Southern Journal of Philosophy* 44 S1: 85–105. List, Christian and Pettit, Philip, 2011. *Group Agency: The Possibility, Design, and Status of Corporate Agents.*, Oxford University Press. Ludwig, Kirk, 2007. Collective intentional behavior from the standpoint of semantics, *Noûs* 41 3: 355–393. Massey, Gerald J., 1976. Tom, Dick, and Harry, and all the king’s men, *American Philosophical Quarterly* 13 2: 89–107. Miller, Seumas, 2001. *Social Action: A Teleological Account*, Cambridge University Press. Nolan, Daniel, 2014. The question of moral ontology, *Philosophical Perspectives* 28 1: 201–221. Pettit, Philip and Schweikard, David, 2006. Joint actions and group agents, *Philosophy of the Social Sciences* 36 1: 18 –39. Quine, Willard Van Orman, 1960. *Word and Object*, MIT Press. Quinton, Anthony, 1975. The presidential address: Social objects, *Proceedings of the Aristotelian Society* 76: 1–viii. Seeley, Thomas D., 2010. *Honeybee Democracy*, Princeton University Press. von Solodkoff, Tatjana, 2014. Paraphrase strategies in metaphysics, *Philosophy Compass* 9 8: 570–582. Sylvan, Kurt L., 2012. How to be a redundant realist, *Episteme* 9 03: 271–282. Tollefsen, Deborah, 2002. Collective intentionality and the social sciences, *Philosophy of the Social Sciences* 32 1: 25–50. Tuomela, Raimo, 1989. Actions by collectives, *Philosophical Perspectives* 3: 471–496. Wall, Edmund, 2000. The problem of group agency, *The Philosophical Forum* 31: 187–197. Watkins, John, 1952. The principle of methodological individualism, *The British Journal for the Philosophy of Science* 3 10: 186–189. Watkins, John, 1973. Historical explanation in the social sciences, in O’Neill, John, ed., *Modes of Individualism and Collectivism*, Heinemann. Wilson, Robert A., 2001. Group-level cognition, *Philosophy of Science* : S262–S273. **Funding Information:** No conflicts of interest declared. **Institutional Affiliation:** Institut für Philosophie and IRI THESys, Humboldt-Universität zu Berlin, Unter den Linden 6, 10099 Berlin, Germany. [^1]: French (1979, 211), Wall (2000, 195–6), Elster (2007, 13), and List and Pettit (2011, 2–3) are examples of authors referring to this hypothesis using the term “shorthand”. Watkins (1973), Quinton (1975, 23–5), Copp (1979, 178), Tuomela (1989), and Miller (2001, 10) advance a paraphrase argument under a different heading. A similar hypothesis concerns not collective actions but collective responsibility (see Cooper (1968); Isaacs (2011, 82)). [^2]: Ludwig (2007) and Massey (1976) could be seen as aiming to vindicate the paraphrase hypothesis. [^3]: The identity relation is symmetric: when $A$ just is $B$ then it is also the case that $B$ just is $A$. The constitution relation is asymmetric: when $A$ is grounded in $B$ then it is not the case that $B$ is grounded in $A$. [^4]: Action sentences with plural subject terms could be understood either *collectively* or *distributively* (Ludwig 2007, 361–2). Because only the collective reading refers to collective actions, I henceforth assume the collective reading. [^5]: Agency in philosophy must be distinguished from the linguistic property that goes by the same name. [^6]: I do not ascribe to Ludwig (2007), this paraphrase procedure, or the subsequent reduction argument. [^7]: “$(\textrm{Each } x \textrm{ of } a)$” is in fact a shorthand for “$(\forall x : x\in G)$” where $G$ extensionally refers to a set of individuals. [^8]: More precisely, we can distinguish between pragmatic and semantic versions of the paraphrase view (von Solodkoff 2014). [^9]: The assumption above is formulated for the specific domain of collective and individual actions. It is an instance of a more general domain-neutral principle. *Paraphrase Implies Identity (Domain-neutral)*.  \ A n entity of type $X$ *just is* an entity of type $Y$ if and only if, given a true sentence about a token of type $X$, there is an adequate paraphrase of that sentence in terms of tokens of type $Y$. [^10]: Eliminativists go a step further. They argue that only individual actions exist, and that there are no collective actions. For this conclusion they need a different assumption. *Paraphrase Implies Elimination*.  \ A n action exists only if, given a true action sentence about this action, there is no adequate paraphrase of that action sentence in terms of another action. [^11]: Ludwig (2007) does not make this assumption. [^12]: This argument rests on a contradiction between four premises. (1.) Each judge is an agent of the conviction. (2.) There is no other agent of the conviction. (3.) There is at least one agent of the conviction that also intends to convict the defendant. (4.) No individual judge intends to convict the defendant. Because this is a contradiction, one premise must be given up. We could give up the second premise and accept that the court is a group agent; but opponents of group agency cannot do so. Denying either the third or the fourth premise comes at a considerable cost. When we argue that convictions are not intentional actions, we must accept that our justice system relies on events that no one intended. And when we argue that the judges individually intend to convict the defendant, we must explain how this intention coheres with their epistemic judgment that she is not guilty. [^13]: This is one important way in which this case differs from the election case above. [^14]: It might be that any judge intends to reach a verdict by means of the proper aggregation procedure. But this intention is too unspecific to say that she *intentionally* finds the defendant guilty. [^15]: We are assuming that either the judges individually or the judges collectively are an agent of the conviction. See footnote 12. [^16]: I set aside the question of whether the system is phenomenally conscious. [^17]: To avoid an exception to the assumption made earlier that all action sentences with singular subject terms are individual action sentences, sentence 6 could be rephrased as “The members of the Chinese nation butter a slice of toast.” [^18]: The events are not disjointed. The individual actions are strict spatio-temporal subsets of the collective action.