A Naturalistic Argument for the Irreducibility of Collective Intentionality

1. Introduction

Several philosophers and scientists have argued that human sociality is explained, at least in part, by the capacity to share attitudes of various kinds with others. “Sharing” can be given several meanings, and it serves a variety of functions in everyday life. But it is clear that what is needed for any two persons to interact successfully is that they understand and experience things in the same way. Such behavior is underlain by the kind of “meeting of minds” that people establish when they share a piece of information or an emotional state about the action scene and about their personal history. How do people attain this form of mutual understanding that is necessary for basic interactions like communication and cooperation? Social theorists and philosophers have typically confronted this question within the social sciences and humanities, but few are aware of the advances that have recently been made in the subject across the social and cognitive sciences.

Analyses of “collective intentions” in philosophy provide the most refined toolkit of conceptual resources to investigate the conditions for sharing minds. ¹ The basic idea of collective intentionality is that whenever people engage in a common activity, each person comes to an understanding of the goal of their actions as something that “we are to pursue together.” For example, imagine that you and I come across a friend in difficulty and we offer help. There is a sense in which collective action is predicated over the individuals, so the “we” in the expression, of course, refers to you and me acting on our own. But this is not the sense in which we mean to bring about joint assistance: what makes our effort joint is that we intend to do it together. It is because you and I see each other as part of the same group, that we understand you doing your part and I doing mine, only as part of our doing it together. The sharing of minds essential to joint action is thus created by individual minds that represent themselves as first-person plural “persons” relative to each other.

Despite its intuitive strength, this idea clashes against the implicit assumption of much social and psychological research that a general theory of intentionality is always given from the perspective of the first-person thinking and experiencing subject I. The response of collective intentionality theorists has been to frame this problem as a demand for the conditions of reduction of collective to individual intentional states. And answers, by and large, fall in two camps. For nonreductionists, there are plenty of arguments supporting the idea that the sense of “we-ness” experienced by the participants in a shared activity can only be captured by some primitive sense of sociality. Reductionists, on the other side, hold that explanations of collective attitudes can be cashed out in terms of facts about individual action and thought. In spite of differences between the two approaches, the opposition between and within the two camps hides a remarkable consensus on how best to interpret the irreducibility question.

Everyone agrees that if no noncircular account succeeds in decomposing the concept of collective intentional behavior into purely individualistic concepts, we will have reasons for believing in the “naturalness” of collective intentionality. If any account succeeds, the opposite conclusion is warranted. So, in John Searle’s (1990) primitivist account of collective intentionality, for example, you find a number of counterexamples to the idea that it is impossible to articulate noncircular accounts of shared intentionality by reference to individual intentionality alone. Searle then goes on to argue that collective intentionality is a “biological primitive phenomenon that cannot be reduced to or eliminated in favor of something else” (1995, 24). Yet, what justifies talk of collective intentionality as a biological phenomenon? What would constitute a truly naturalistic explanation of collective intentionality? Although Searle as well as the other “founders” of collective intentionality theory are all dedicated naturalists in principle, they have rarely pursued this line of research in practice. Their accounts have contributed remarkable insights into the logical conditions under which mental states count as shared, or collective, but none has proved decisive in settling the question of the nature of collective intentional states.

In this article, I argue that the question of the irreducibility of collective intentionality cannot be settled on the basis of ordinary language analysis and intuition alone. To begin with, there is a sense of “irreducibility” by which the reduction of collective to individual intentionality is an epistemic issue. It is the question of whether collective intentional behavior can be explained using the concepts that we deploy in understanding individual thought and action. If this cannot be done, in the sense that statements about the former are not “translatable” in statements about the latter without loss of meaning, the analysis asks for some irreducible concept of we-ness. Questions concerning the conditions for having knowledge of collective intentional behavior, however, are different from questions about the existence and identity of shared mental states. Evidently, this formulation calls into question ontological, rather than epistemic, ² claims as to whether collective intentionality is real (i.e., true of the reality we inhabit). For example, one such claim is that group thinking is not exhausted by the parts and operations that underpin individual intentionality, and it involves some dedicated mechanism of social cognition. Failure to acknowledge the distinction between epistemic and ontological questions has slowed on progress in the naturalization of collective intentionality.

The appropriate method of investigation depends on the kind of question at stake, in fact. So, if the question is whether collective intentionality is an empty referent, or not, from a biological point of view is not to be determined by producing novel and stronger conceptual arguments; it is an issue that calls for scientific scrutiny instead. On this side, while social scientists have recently engaged analyses of collective intentions with theories of “team reasoning” in economics (Bacharach 2006; Gold and Sugden 2007; Bardsley 2007; Hakli, Miller, and Tuomela 2010), considerably less attention has been paid to the experimental work done in the cognitive sciences. In discussing Michael Tomasello’s influential program of research in the development and evolution of the human mind in the second part of the article, it will become clear that the ontological question of irreducibility is an empirical question in character and therefore calls for a methodologically pluralist and integrated approach to be addressed.

In a number of contributions, Tomasello and his fellow researchers try to apply theories of group thinking suggesting that the capacity to engage in meaningful episodes of interaction early in human development calls for a dedicated—cognitive and motivational—mechanism for sharing mental states (Tomasello and Rakoczy 2003; Tomasello et al. 2005; Rakoczy and Tomasello 2007; Tomasello 2008, 2009). This working hypothesis—the “Shared Intentionality Hypothesis” (SIH) hereafter—applies paradigmatically to instances of joint attention, for example, where interaction is made possible by the agents representing the referred-to object as something they attend to together. The prosocial ability to understand and enact things as a “collective” would then be essential for young infants to share attention, setting the stage for later and more complex forms of social cognition to develop. While developmental scientists have not started yet to explore the hypothesis at length, neither have collective intentionality philosophers worked out its consequences enough to realize that it has the potential to shed significant light on the oft-discussed question of the naturalness of collective intentionality.

There are of course many open questions about the scope of Tomasello’s data and, more generally, about whether scientific theories can establish anything conclusive about the place of collective intentionality in the nature of things, but I will not engage myself with any such discussions. Instead, I propose an account of what it would take to give a naturalistic account of collective intentionality given the current state of the art in the discipline; to this goal, I draw on the SIH to show what is left out of approaches to the ontological problem that employ conceptual analysis alone.

The article is organized as follows. In section 2, I outline the problem of collective intentionality with a focus on nonreductionist approaches. In section 3, I reconstruct the irreducibility thesis along the lines pursued by collective intentionality philosophers, while distinguishing the epistemic from the more implicit, though widespread, ontological interpretation of the problem. If the overall goal is to explore the ontology of collective intentionality in the natural realm, the question is empirical and falls in the domain of natural science. ³ Tomasello’s account of the role of shared intentionality in grounding early phenomena of sociality in human development offers a good example of how to turn collective intentionality into a phenomenon testable in the laboratory. The SIH is presented in section 4; in the final section, I argue that claims concerning the biological reality of collective intentionality should follow from the integrated—conceptual and empirical—task of demonstrating its putative causal role in bringing about physical effects in the world.

2. The Irreducibility Thesis

The central problem of collective intentionality in philosophy is whether collective intentional states are irreducible to individual intentional states. One of the central concepts in contemporary philosophy and science, reduction indicates the process by which entities of any kind (facts, properties, theories, states, and so forth) are redefined in terms of other entities. In this section I first illustrate the theoretical framework in which the problem of collective intentionality arises so as to examine the status of the claims made by philosophers when they advocate or reject the irreducibility thesis. For clarity, I take intentional action in the philosophical sense of intentionality to mean a form of behavior performed with a certain “reaching out” attitude of the mind toward a state of affair or goal in the world (Searle 1983). What makes it collective is the fact that two people might act in the pursuit of the same goal, which thus presupposes their intentional attitudes to be shared. ⁴ A classic statement of collective intentionality is exemplified along the following lines. ⁵ Suppose that you and I go running together every now and then, until we decide to step up and register for the next London Marathon. We show up on the race day, run the entire marathon, and reach the final line exhausted. At the end, you shout at me, “We did it! You and I ran the marathon!” There is of course a distributive understanding of the expression in terms of individual intentions (Ludwig 2007): action is predicated over the individuals, so the “we” in the expression refers to you and me running the race individually. Indeed, on this reading, it would not have made any difference if we had not planned to embark on the marathon in advance and just met by chance on the race day. We would still have run it in the (distributive) sense that you did it, and I did it. But in our example, when we set to run the marathon, we meant it together.

Something is clearly left out of the distributive reading, that is, a form of understanding that represents our action as something that we not only happened to do side by side but jointly. If we ran the marathon with the intention to do it together, then of course it is true to say that each of us ran the marathon; but the opposite is not. The claim that you and I ran the marathon does not imply that we did it together and we did it intentionally. What is implied is a form of collective understanding, or intentionality, according to which when two or more agents intend and enact things to achieve a common goal—the fact that they do something together does not mean that any member of the group does it on her own (Bardsley 2007). It takes two to tango, on a recurrent example of the collective intentionality literature—in the sense that the individual contributions to the joint performance are not to be partitioned over the single dancers if the “togetherness” of the collective action is to be captured. Hence, there is more to the understanding of people in joint action than is rendered by a purely individualistic or distributive reading.

In spite of being a pervasive and often implicit assumption of social science research and philosophy, this idea has proved challenging to characterize beyond intuition. In the early days of collective intentionality theory, philosophers made a start by investigating the intentional attitude of the agents. Back to the marathon story, the difference between running the race on one’s own and as a group might be a feature of the way in which each of us understands the action’s goal. Instead of having a thought expressed by the words “I intend to run the marathon with you,” you and I might mean something like “We intend to run the marathon together.” Depending on how each understands the “we” of the relevant intentional state, we would intend to engage in the marathon as a collective or simply as individual runners. Accordingly, collective intentionality has become shorthand for the ability of people to think of individuals, including themselves and others, as the bearers of we or group thinking.

One difficulty with group thought, however, is that accounts of social phenomena are traditionally built upon a fundamentally individualistic approach to agency and cognition. “Individualistic” means that it is implicitly assumed that people represent the goal of their actions in the first-person singular: states of the world are thinkable insofar as they are given to minds in an individual perspective. By allowing individual-level mental representations to be accessed in the first-person plural, so to speak, philosophers are now faced with the problem of explaining what makes this form of thought and action possible for individual agents. One option is to say that for an action to be intentional, there must be somebody bearing the relevant attitude. The we understood collectively would then point to a specific ontological referent, a plural subject lying over and above individuals. Yet, this conclusion is evidently at odds with the tenets of ontological individualism, something collective intentionality philosophers are unwilling to give up. ⁶ The alternative option is that the mark of collective intentionality consists in the kind of intentional attitude exhibited by its bearers.

Among the first philosophers to systematize this idea, Raimo Tuomela and Margaret Gilbert famously argued that we (qua theorists) fail to grasp the we-ness of collective intentional behavior unless we ascribe an irreducible attitude to the agents. “We-mode mental states and actions typically are joint states and actions in a strong sense involving an irreducible, thick ‘we’ (that is, a ‘we-together’), and this makes the ontic ‘jointness’ level central for the construction of the social world” (Tuomela 2007, 10). Analogously, Gilbert postulates a form of understanding consisting in “a grasp of a subtle conceptual scheme, the conceptual scheme of plural subjects” (Gilbert 1989, 416), which treats the relevant grasp as an irreducible notion, thus leaving it unspecified (Tollefsen 2002). Finally, the irreducibility of the we-mode finds its clearest formulation in John Searle’s account of social ontology (1990, 1995, 2010). Searle is motivated to believe in the irreducibility of collective intentionality by failure of traditional individualistic accounts to make sense of the intersubjectivity of human cognition in a way that does not resort to the very concept in need of explanation (Searle 1990). A dedicated individualist himself, he concludes that the analysis of joint behavior cannot be assembled out of more elementary units, notably individual states even supplemented with mutual knowledge. The only way to explain the specificity of collective intentional behavior in noncircular fashion is to acknowledge that individual agents can represent aspects of the world also from an irreducibly collective perspective.

In the past two decades, Searle’s theory of collective intentionality has become a salient point of reference and a critical target in the philosophy of society. His stance on the irreducibility problem has sparked a number of reactions, which have received wide coverage in the literature. ⁷ Searle often remarks that the distinctive feature of collective mental representations does not consist in the fact that they have the same content: people intending to play a duet together may well represent the goal of their actions in different ways and yet act upon the overall goal of playing a duet together (Searle 2010). Neither does collective intentionality consist in a distinct representation of we-ness or group thinking such that whoever intends and enacts things together with somebody else can only do it if they possess the relevant concept of collective intentionality. Instead, the fact that distinct mental representations might be shared is a matter of the distinctive psychological mode in which they are accessed (Searle 1997). For individuals to think and act as a group is for them to see together, think together, perceive together, believe together, and so on, the goal of the shared activity.

In the 20 years since the founding accounts were proposed, the theory of collective intentionality has generated an ongoing and prolific dialectic between advocates of the irreducibility thesis and those who defend reductionist accounts. ⁸ While commentators have mostly focused on the differences among and within the two camps, my main concern is with one common feature of nonreductionist theories that has surprisingly gone unnoticed.

3. Epistemic and Ontological Irreducibility

For advocates of the irreducibility thesis, there is an explanatory “gap” between collective and individual intentionality. How to analyze it has marked the main divide within the nonreductionist camp. While this concern has dominated the field until now, little if no attention has thus far been paid to what actually unifies the irreducibility claims of collective intentionality theorists. There is in fact a remarkable consensus behind the scenes, which has produced profound consequences on the debate. To unearth it, let us consider the way in which nonreductionists construe the very question of collective intentionality.

When philosophers dispute whether collective intentional states can be reduced to their individual components and their interrelations, what they mean is that we need to postulate an irreducibly collective concept to capture the sense of we-ness enjoyed by the agents in a shared activity. The “classic” construal of the irreducibility thesis is epistemic in other words: it is the question of how we describe, conceptualize, infer, or know about collective intentionality. And it is a question that collective intentionality philosophers have mostly faced with mixed results by means of ordinary language analysis that aims at setting out a priori, exceptionless and intuitively acceptable conditions for the reduction of concepts of sociality to their individual components. If this is a correct picture, it is hard to predict whether one account will ever gain enough currency to adjudicate the debate over the conditions for mental states to be shared, as far as philosophers of opposing schools will nurture genuinely contrasting intuitions about the concept of collective intentionality.

It might be argued that this is the reason why the philosophical debate on the structure of collective intentional states has stalled. This argument comes from those who usually look with suspicion at theories that are explicitly built upon irreducible or primitive new notions. If it is accepted that collective intentionality figures at the foundation of social reality—so the critique goes—it is an essential part of any theory of society that the concept of collective intentionality be further explained. If, on the contrary, the primitive notion constitutes the very explanandum of the theory, there are good reasons for discarding the irreducibility thesis as unsatisfactory. Although this is an argument that points to the limits of primitivism in general, it is widely received in the collective intentionality theory under the form of a challenge to Searle’s theory of social ontology, in which collective intentionality features prominently as a building block (Meijers 2003; Vromen 2003; Pacherie 2007).

An advocate of Searle’s, as well as of any primitivist account of collective intentionality, might counter this line of criticism with a consideration about the meaning of “primitive.” In general, the claim that a feature is a theoretical primitive in a domain means that it cannot be explained away in terms of simpler constituents of the same definitional domain. Primitive does not mean, however, that an intentional predicate cannot be given a reductive explanation in another theoretical framework. ⁹ So, the point that collective intentionality is primitive is not equivalent to saying that it is not reducible at any level of explanation. Accounts that postulate a primitive sense of we-ness in the mind of the subjects involved in a joint action are relevant for analyses of everyday social concepts that are traditionally couched in the folk-psychological scheme of intentional agency. Outside of this framework, it might well be the case that the sense of collective intentionality admits of a reductive explanation.

There is, however, a better-suited reply to skeptics of the irreducibility thesis. I argue that the challenge to Searle is wrongheaded because it trades on a misleading interpretation of the explanandum of the irreducibility thesis. While concentrating primarily on the logical structure of collective intentional states—what it means for people to think and act in we-mode—philosophical arguments often slip into the description of the nature of collective intentionality. But if the target of philosophers is the nature of intentional predicates, the key question is what conditions make those predicates true of the world we live in. Analyses of collective intentional predicates are no exception. In describing we-mode thinking as (epistemically) primitive, theorists tend quite naturally to suggest that it might be underpinned by a capacity that is not reducible to individual intentionality.

What does it mean for a capacity, or mechanism, to be irreducible? To clarify the question, let us go back to Searle’s claim of irreducibility. Famously, Searle does not refer just to an irreducible form of collective understanding; he also defines collective intentionality as a biologically primitive form of mental life. But this presupposes a different sense of irreducibility: if “primitive” and “irreducible” have thus far been used with the same epistemic meaning that understanding of collective intentionality cannot be decomposed in the concepts we use to understand individual intentionality, now “primitive” is presented as the attribute of a putative biological capacity. ¹⁰

On this construal, the irreducibility thesis is an ontological and not an epistemic question, one that concerns the conditions of existence and identity, rather than of knowledge, of collective states. So, the argument that there is a difference in mode between first-person plural and first-person singular intentional predicates comes down to the idea that such a difference might have its roots in physical structures. ¹¹

Failure to make the distinction between epistemic and ontological irreducibility explicit in the literature has brought about considerable confusion about the best method for settling either question. This confusion is even clearer in assessing both the weaknesses of Searle’s ontological account and the inadequacy of standard critiques to his case for the ontological irreducibility of collective intentionality. In fact, as a staunch defender of “biological naturalism” (Searle 1983, 2007), the view that it is only scientific ¹² theory and practice that provide a reliable answer to what there is in nature, one might expect Searle’s argument to be informed by scientific language and research. But he opts surprisingly for another argumentative strategy. The biological reality of collective intentionality is rather presented as a self-evident “fact,” which finds its place in Searle’s writings next to the (epistemic) argument that collective intentional states cannot be reduced to, or eliminated, in favor of something else (1995, 24).

However, the fact that collective intentionality is irreducible to individual intentionality in epistemic terms is no proof that it is a primitive, i.e. brute, fact of the mind. In other words, when we say that one entity is reduced to another, the former can be explained in terms of the latter also, precisely because the two are ontologically the same. We still refer to the two entities as if they were distinct, when in fact they are one and the same. But to make one thing more intelligible by showing that it actually is another thing is an epistemic quality of reduction, not a new fact in ontological terms. To be ascertained, such a fact would require more than an analysis of the uses of collectivity concepts in everyday discourse and social science research. It would rather ask for an inquiry into what there is, which implies a willingness to draw on the theory and practice of natural science for addressing philosophical puzzles (Papineau 2007). Since this inquiry is lacking in Searle’s biological claim, it then looks as if conceptual, instead of factual, evidence were given a prominent role in settling the empirical question of the place of collective intentionality in the natural realm.

Indeed, this is the route that most commentators have taken in pointing out that any hypothesis that places collective intentionality among the brute facts of the brain would be “magical” in the absence of scientific scrutiny (Hornsby 1997, 432). This critique is not totally unjustified. The question of how to capture the conditions under which intentional attitudes count as irreducibly collective is a logical question. Hence, it is separate ¹³ from the question whether there actually is any viable scientific theory that succeeds to meet the criteria for the existence of collective intentional states. Searle’s account contributes remarkable insights into the former question, but it tells very little about the brute facts, if any, that enable thinking and acting in we-mode. Thus, while he correctly emphasizes the role of science in helping enlighten philosophical puzzles (and vice versa), he does not actually make a step forward in the naturalization of collective intentionality. However, if the route taken by critics to attack the very idea that there might be a fact of the matter for the capacity of collective intentionality is methodological, then it is fair to conclude that no convincing argument has been provided yet that discards Searle’s proposal on a truly naturalistic basis.

The reason is that the question of the biological reality of collective intentionality is a scientific question. Accordingly, it suggests a set of issues and considerations that coalesce in another debate in the collective intentionality field that bears terminological resemblance with, though it is substantively distinct from, discussions of epistemic irreducibility. If the two debates are conflated into a unique interpretation of irreducibility, then it is impossible to realize that each question constitutes a problem area on its own and calls for certain methods to be addressed. The point is that logical analysis is insufficient to deal with fact-of-the-matter questions about the alleged place of collective intentionality in the natural order. If the question is whether there is a specified mechanism of social cognition at the biological level, we need to look outside the social sciences and humanities where the problem of collective intentionality is subjected to laboratory tests and therefore confronted as an empirical rather than logical question. In the next section, I illustrate one such approach in science that gives us a solid naturalistic framework for illustrating the empirical character of the ontological question in section 5.

4. Shared Intentionality Hypothesis

In the cognitive sciences, the concept of “we-intention” figures in a diverse body of knowledge ¹⁴ that finds a powerful synthesis in the work of psychologist Michael Tomasello. Drawing on the resources of collective intentionality theory, Tomasello has articulated a theory of the foundations of sociality—the SIH ¹⁵ —according to which certain human-like social-cognitive skills have evolved thanks to a unique mechanism for sharing intentional mental states (Tomasello and Carpenter 2007). In this section I examine one skill in particular, joint attention, which has recently attracted much interest in philosophy and psychology for its role in grounding the development of human social cognition. Through the lens of the joint attention debate, I present the SIH as belonging to the family of nonreductionist accounts of collective intentionality along the same lines suggested by Searle.

Joint attention develops out of gaze following, the capacity of one person to look reliably in the same direction as another’s gaze. Forms of protocommunication involving gaze following are observed very early in life, before they reach some constancy at about 6 months of age. You can easily imagine a scene where an adult introduces a third object close to the mutual line of regard, which prompts the child to alternatively look either to her or to the object. Dyadic interactions become more complex until a suite of new behaviors are on display by the age of 9 to 12 months, approximately. Infants now engage themselves in a triadic relation with both inanimate and animate beings, which outgrows ritualization and involves a broader range of coordination-attention abilities than those observed in dyadic relations. Typically, 1-year-olds hold up objects for adults to share attention to and check back and forth the others’ body reactions and their focus of attention.

For developmental scientists, it is relatively uncontroversial that joint attention behaviors are episodes of intersubjective engagement, where the outside entity must not only be actively selected by the subjects but also be experienced as an entity that is present to both (Moore and Dunham 1995). Disagreements arise when it comes to describing the nature of the special bond whereby the participants in the perceptual triangle of joint attention realize that they attend to the same object together (Tomasello 1995). In fact, questions like how to capture the full sense of “jointness” typical of triadic relations and what sort of mechanisms bring it about are the subject of a heated controversy between two classes of explanation, so-called rich and lean. Since Tomasello’s theory features prominently in the field of rich accounts, we might begin with asking why he advocates such approach. But this question is only tangential with the present discussion. ¹⁶

The interesting question for our purposes is why, in subscribing to the rich approach, Tomasello characterizes joint attention as one possible manifestation of collective intentionality (the first one in ontogeny, in fact). What particular account of collective intentionality would best suit joint attention behaviors, if any? And what does this strategy tell us about the irreducibility of collective intentionality in the context of scientific research?

Despite the similarity between the joint attention and the collective intentionality debate, there is one major difficulty in answering these questions. Philosophical analyses of shared intentions focus on cases of collective intentional behavior where individuals deliberately engage in a planned course of action to achieve a specified goal. On this conception of joint action, not only do people coordinate to attain the goal but, for the goal to be shared, each subject must be aware of it as mutually present to all (Bratman 1992). As we have seen, deep disagreements have arisen among collective intentionality philosophers as to how best to capture the mutuality of awareness—the oft-mentioned sense of we-ness—that seems necessary for a shared activity to come about. Most render the state in which two subjects understand each other as attending to the same entity and display awareness of this very fact as one of “mutual knowledge” ¹⁷ (Schiffer 1972). The upshot is a set of highly sophisticated theories of joint action that postulate detailed representations in the mind of the single agents specifying shared tasks and goals. These representations are presented as the result of conscious and reflective processes of social cognition, which are usually described in the context of paradigmatic interactions, such as an orchestra performing a piece or sport teams playing together.

While this approach has undoubtedly furthered our understanding of the structure of planned joint action, it leaves uncovered a good deal of interactions, such as cases of “emergent” (spontaneous, unplanned) coordination, where interaction does not require detailed representations in the agents’ mind. Such cases of sociality are easily observable in infant and primate social behavior, where sophisticated representational capacities of the sort postulated in philosophical accounts of shared intentions are not yet fully developed or are lacking. Evidently, joint attention appears as one such form of face-to-face interaction. So, as Tollefsen (2005) has remarked at length, it is very problematic to make sense of the evidence that children are capable of engaging in joint attention behaviors so early in infancy in terms of irreducible collective states that are postulated under conditions of mutual knowledge. But, then, since Tomasello explicitly draws on the resources of collective intentionality theory to account for joint attention, the question becomes where he stands relative to the problem of mutual knowledge and to the irreducibility question more broadly.

In its first formulation, the SIH was preceded by the observation that joint attention behaviors underlie a natural and species-unique tendency to grasp others’ reasons for action as a prerequisite for sharing attention (Tomasello 1999). The idea is that the kind of perceptual copresence experienced by the subjects in joint attention stems from an active exercise of intention understanding, with the child being in the position to process information about the others’ thoughts and motives, as well as to represent the focus of attention as mutually present to her and the caregiver. ¹⁸ Some wording thus suggests that Tomasello leans toward a conception of the mutuality of joint attention in terms of mutual knowledge ¹⁹ that is compatible with reductionist accounts of collective intentionality (Tomasello 1995). “Reductionist” here means that if all that matters for the subjects to achieve the we-ness characteristic of joint attention is their capacity to read into the others’ minds, then no irreducible we-mode needs being postulated for explanatory reasons.

However, Tomasello’s phrasing does not reflect any particular preference for the mutual knowledge solution; it is rather driven by considerations that point to significant uncertainty on how best to specify the exact meaning of concepts such as “mutuality” and “sharing” (Tomasello 2009, 69). Confusion arises especially from the idea that if mutualism is enabled by the subjects’ recognition that their actions are directed to a shared goal, which is achieved via mindreading, then the question is what cognitive and conceptual resources are needed for one to realize that they have a joint goal to start with. At first glance, this account carries the same danger of explanatory circularity that affects all accounts of collective intentionality invoking mutual knowledge. ²⁰

The decisive argument against this characterization comes from an influential discovery in primate cognition. Tomasello and his collaborators have run a set of comparative studies testing the capacity for gaze following in juvenile chimps, and, contrary to the initial hypothesis that postulates a theory of mind as the distinctive trait of human cognition, they have provided evidence that chimps are highly social creatures in the specific sense that they display some rudimentary propensity for intentional understanding (Tomasello et al. 2005; Tomasello 2008, 44-49). The evidence, however, also shows that great apes do not enjoy the characteristic sense of jointness observed in humans when everything is out in the open and mutually known, so to speak. In the “role reversal” task, for example—one that tests the ability of the players to change their roles for the sake of achieving a specified goal in coordination—it is proven that, unlike human infants, juvenile chimps do not reverse roles and perform their action without reference to the others, that is, in an instrumental rather than mutualistic way (Tomasello and Carpenter 2005).

Primates know what their conspecifics see, though from a purely individualistic perspective. In other words, their understanding of the intentional structure of action would not be shared in the way in which children’s understanding of the focus of attention is said to be shared in joint attention. But, then, if recognition of the mutuality of joint attention is species specific in that it enables human-like forms of social interaction unknown in the animal kingdom, it must ask for more than intention understanding since this capacity is displayed by chimps too. The lesson of these experiments is that any theory of social cognition that aims at capturing the special bond of joint attention should strike a balance between two desiderata that pull in opposite directions. One is to not overintellectualize the child’s mind despite her alleged ability to engage in joint actions that invoke some rather complex abilities for sharing intentionality. The other is to allow infant cognition to be rich enough to make sense of the differences observed in humans and chimps when it comes to representing action as shared.

This tension finds a solution in the latest version of the SIH: not only are goals and intentions to be grasped interpersonally, but they also must be shared in the full sense of joint attention. ²¹ Chimps do not share mental states, because they are incapable of representing the intentional structure of action from a “bird’s-eye view”—one that represents their and the others’ goals in a single format (Tomasello 2008, 179; 2009, 68). Humans, instead, frame the interaction scene in a single representation: each individual comes to an understanding of the goal of her actions as one that “we are to pursue together.” It might be argued that for the agents to represent the “we” on a collective instead of distributive reading, they must be in the position to know what it means to be thinking and acting together rather than on their own. And this might be interpreted as asking that the participants in a joint activity must possess a distinctive concept of we-ness that does not come down to individual-level representations of the goal as shared. But this conclusion should be resisted on grounds that it implies conceptual and metarepresentational abilities that can hardly be attributed to one-year-old infants.

One formulation that suits better the bird’s-eye-view metaphor, while meeting questions about early cognitive and conceptual development, is to say that the SIH postulates not a specific representation of “togetherness” but rather distinct attitudes of seeing together, perceiving together, thinking together, and so forth (Gross 2010). The sense of jointness of joint attention would then be achieved by the subjects representing the relevant goal in different ways (i.e., contents), though in the same psychological we-mode—which Tomasello identifies as the “uniquely human sense of ‘we,’ a sense of shared intentionality” (Tomasello 2009, 57-59). Several passages from Tomasello’s latest writings suggest that this might be his considered view. If so, there are good reasons for arguing that Tomasello leans toward an irreducible definition of shared intentionality following closely in the steps of Searle’s view of collective intentionality.

However, the natural-scientific approach of Tomasello’s research yields one conception of the irreducibility of collective intentionality that hardly finds its place in the philosophical-conceptual theories on offer. In the next section, I contend that this methodological difference is fundamental to understanding why issues of ontological irreducibility are empirical and call for a mixture of conceptual as well as experimental methods to be tackled.

5. Joint Attention, Reference, and the Ontological Question

Tomasello (2009) speculates that the biologically inherited “infrastructure” of social proclivities and mind-reading skills that enable the meeting of minds necessary for joint action is a species-specific capacity. This claim has sparked an intense cross-disciplinary discussion on the alleged uniqueness of human cognition. ²² What seems to be particularly problematic from a philosophical point of view is that in spite of being backed up by a robust body of data, the SIH does not encompass a structured or even partially worked-out response to the problem of the biological reality of collective intentionality. So, the claim that the capacity for group thinking might stem from a unique evolutionary trajectory and follow a distinctive developmental path early in life can only take us thus far, and it remains a speculation in need of further elaboration. However, based on the results of well-controlled laboratory experiments, Tomasello and his collaborators do postulate a biological adaptation for sharing intentional states in the framework of a natural-scientific explanation of human development. So, in light of the distinction between epistemic and ontological interpretations of irreducibility, where does this conclusion take us with regard to the question of the nature of collective intentionality?

One might, of course, read this question as suggesting that the scientific evidence from Tomasello’s experiments would be somehow decisive to vindicate Searle’s claim that there is a foundation for the capacity of collective intentionality in the biology of organisms. ²³ But, then, this reading would raise an even more controversial set of questions about the legitimacy of some version of “ontological naturalism”—the view that what science at its best tells us is approximately true of the constituents of the natural realm, no matter whether animate or inanimate. To be sure, whether the SIH can contribute to the solution of problems of social ontology and, if so, how to spell out the naturalization process in detail are issues of great importance in the collective intentionality debate and would deserve much more attention than it is paid here. Though, they largely fall outside of the scope of this article. In fact, before we embark on the project of evaluating the potential of the SIH to meet questions about the biological reality of collective intentionality in the future, we need to explain why ontological claims can only be approached with an integrated methodology of empirical and conceptual tools. This is a direct consequence, perhaps the most important implication, of distinguishing between the epistemic and ontological sense of the irreducibility thesis. While aspects of the SIH have already been discussed in detail in relation to theories of shared intentions (Tollefsen 2005), in this final section I refer to the naturalistic approach underlain by the SIH to argue that knowledge about the existence of collective intentionality can only be ascertained through factual, rather than merely conceptual, evidence.

In section 3, we identified an ontological sense of “primitive” according to which collective intentional predicates cannot be reduced to the parts and operations that underpin individual states. In social theory and philosophy, this claim is likely to be drawn on folk-psychological expertise and commonsense intuitions, and the result is a growing proliferation of views that, despite the commitment of their authors to naturalism, make hasty existential conclusions about the alleged natural conditions under which mental states count as shared. Searle (1990, 1995), for example, resorts to the concept of a primitive capacity for group thinking both as an explanatory tool that helps describe the sense of acting together experienced by the participants in joint activities and as a dedicated process of social cognition rooted in the biology of human brains. What justifies talk of irreducibility in the latter sense?

Searle offers one important reason for believing in the ontological irreducibility of collective intentionality: the causal continuity between the we-mode and the occurrence of individual action. He observes that people engaging in joint action have their understanding of the shared goal in the first-person singular (I-mode) settled by collective, we-mode representations. “Intuitively, in the collective case the individual intentionality, expressed by ‘I am doing act A,’ is derivative from the collective intentionality ‘We are doing act A’” (Searle 1990, 92). The intuition of “derivation,” however, is barely developed in Searle’s work. A better formulation would be to say that when one refers to a goal to be pursued with somebody else, framing it in the first-person plural enables the agent to “see” the action scene in a way that would simply be precluded had it been represented in the first-person singular. The upshot would be a novel understanding of the goal, which would then cause a certain course of action to be selected at the individual level. Hence, collective intentionality is primitive and individual intentionality derivative in the sense that the former is prior to the latter in situations of joint action.

Yet, in what sense of causal priority would collective intentionality exercise any power in bringing about individual thought and action? This is a question that is frequently, though implicitly, posed by discussions of joint action combining insights from the collective intentionality theory with the research in developmental psychology. But it has no clear answer in the philosophical literature. The reason is that it is a question that the analysis of collectivity concepts is unsuited to meet on purely aprioristic grounds, unless it is supplemented by the research that investigates the extent to which (collective) intentional phenomena yield behavioral effects in the world. The study of causation typically involves comparison and controlled variation to ensure that the allegedly independent variable is separated in ways that would be impossible to achieve in nonexperimental circumstances. In the laboratory, in fact, everyday explanations invoking correlations between, say, the capacity of collective intentionality and individual action, are subjected to tests that purport to isolate the relevant causal factors, turning those associations into reliable causal generalizations.

Scientific research is thus needed to spell out what it means for a mechanism to be primary in the sense of bringing about a certain effect. In developmental social cognition in particular, it is by designing the best possible conditions for testing forms of protocommunication that it has become possible to “observe” the emergence of joint attention behaviors at a certain age, isolating their causes from those that bring about earlier (dyadic) relations of interaction. And it is in the context of empirical research, rather than in abstraction of it, that it has turned out that the capacity for sharing mental states is indeed primary in joint attention situations in one particular sense of “priority”, that is, ontogenetic. This is a crucial finding in order to tackle the question of the ontological irreducibility of collective intentionality. So, to illustrate the point in detail, let us analyze how treating this question in empirical terms—against the theoretical background sketched by Searle—allows us to deal with the problem of the ontology (ontogeny) of collective intentionality in a way that logical analysis alone does not.

Recall that Searle’s philosophical concern is with the formation of shared intentional states, that is, the conditions under which two people come to realize that the goal of their actions is shared for the sake of joint action. By claiming that the content of individual intentional states can derive from collective intentionality, Searle suggests that group thinking would be causally efficacious to the extent that it contributes to settle the problem of reference—the problem of how any two persons understand on their own that they mean the same thing in interaction. The philosophical literature on reference has for a long time acknowledged that mutual understanding requires some sharing at the mental level. But this problem has been cast, unsurprisingly, as a logical problem, with little work being done on what constitutes the relevant sharing that brings about knowledge of reference. A decisive step in this direction was made by broadening the investigation to the psychological conditions that settle understanding of reference in ontogeny, when infants start communicating in a proficient way before language is acquired in earnest. What does it mean to be “proficient communicators” early in development? Prelinguistic children do not make utterances, yet they display basic skills for engaging meaningfully in referential acts. Although this observation is interpreted in a variety of different and often competing ways in the literature, there is wide consensus in describing the sociality of primitive forms of communication as the result of an active “negotiation” of attention between infants and caretakers.

Interestingly for our purposes, this argumentative line in psychology has been brought forward by Tomasello, who points to the triangle of joint attention as the first manifestation of understanding of reference in ontogeny (Tomasello 1998). Theories of joint attention, including the SIH, can thus be understood as theories of reference broadly conceived. Specifically, the SIH has its theoretical roots in the research paradigm established by Jerome Bruner in the 1970s, when questions about joint attention entered the agenda of psychologists in the guise of questions about language development and the evolution of intentional communication, such as the transition from preverbal to verbal reference (Scaife and Bruner 1975; Bates, Camaioni, and Volterra 1975; for an overview, see Moore and Dunham 1995). The intuition of Bruner is that, given its relevance as a scaffold for later forms of social cognition, joint attention must provide the kind of format that allows individuals to establish mutual understanding of reference. In brief, reference is “a form of social interaction having to do with the management of joint attention” (Bruner 1983, 68; emphasis in original).

One example of joint attention behavior in prelinguistic communication—and, presumably, in early humans before language—is the vast repertoire of attention-coordination gestures to direct and follow someone’s gaze to a given target. Among these, deictic gestures are responsible for transferring knowledge of reference from the caretaker to the child. In the typical joint attention scene, for instance, infants and caregivers look at something together and exchange eye-to-eye contact in a way that proves that they share the same attitude toward it. Joint attention procedures thus shape the field of reference by imposing the relevant constraints for what is to be attended jointly, getting the infant started “in the business of figuring out what was being meant by what was being said—what interpretants were needed to form a bridge between a sign and its significate(s)” (Bruner 1998, 220-21; emphasis in original). However, reference needs more than some contextualized clue to be fixed. Young infants are so proficient in comprehending and producing acts of reference that they seem to know what specific object is mutually present to them and their caretakers in the joint attention triangle. Though, how do they come to this form of understanding if they do not know that they share enough in their attentional focus (Bruner 1995, 2)?

As remarked in the previous section, Tomasello argues that representing their and the others’ goals as something that “we want to pursue together” allows infants to grasp the specific aspect of the goal to be attended jointly. This is the sense in which shared intentionality ultimately is a “shared space of common psychological ground” (Tomasello and Carpenter 2007, 121). So, the we-mode is primary in bringing up certain representations of the referred-to object rather than others—precisely those against which communicator and recipient understand each other’s referential acts based on experiences and activities done together. ²⁴ Collective intentionality is thus developmentally prior to knowledge of reference in the I-mode in situations where goals are shared. That is, people’s recognition that they attend to something together in communication is the precondition, rather than the consequence, of their reading into one another’s mind in the first-person singular to “search” the exact reference of thoughts and gestures.

It is worth stressing that while the implications of this theory are still under debate and point to future directions of research, the emphasis of my argument is on the way in which the particular sense of ontogenetic priority is brought about by integrated conceptual and empirical methods. This is to say that when the investigation of ontological issues is carried out in a truly naturalistic way, the inquiry does not only consist in the kind of reflective thinking that characterizes more traditional philosophical training and practice. Conceptual clarification emphatically follows from the experimental task of testing associations between putative causes and effects, while keeping background factors under control. The morale is that when naturalism is achieved in practice, the outcome is an interdisciplinary account where conceptual and empirical insights are intertwined in a way that realizes the contiguity of philosophy and science in a fruitful way. The SIH can thus be seen as a naturalistic account of the mind, or at least of part of it, because it exploits the tools and techniques of scientific psychology to tackle problems of social ontology.

6. Concluding Remarks on Empirical Social Ontology

The work of Tomasello and colleagues has attracted a great deal of interest among those who want to give a naturalistic account of the roots of sociality based on the capacity for collective intentionality. But the hype surrounding this research program seems premature; more needs to be done to elucidate the assumptions of what I called the SIH, before we evaluate its implications for the naturalization of collective intentionality.

In this article I have tackled this question by presenting the SIH in the context of the central problem of the irreducibility of collective to individual intentionality. Central to my argument is the often-neglected distinction between two issues of “irreducibility”—epistemic and ontological—which suggests an equally important distinction between methods of investigation to tackle either. Searle’s argument for the biological reality of collective intentionality offers a paradigmatic example of one version of the irreducibility thesis that can only be assessed in its own merits, along with critiques to it, with an integrated and interdisciplinary methodology. Failure to acknowledge that questions about the existence and identity of shared mental state cannot be fixed by analysis of everyday social concepts alone has resulted in unfortunate consequences for the debate on the naturalization of collective intentionality.

My argument builds on Tomasello’s experimental research in developmental social cognition, showing that group thinking, or thinking in we-mode, enters the conditions for understanding reference in I-mode. This interpretation brings to fruition the work of pragmatist accounts of reference, from Bruner (1983) to Tomasello (1998), according to which knowledge of reference relies on forms of shared intentionality, such as joint attention, to the extent that it requires that the subjects attune one another’s minds via a joint construal of the referential field. By integrating recent achievements in the study of the development and evolution of the “social mind” into the philosophy of collective intentionality, I have argued that this strand of scientific research in psychology helps us understand the empirical nature of the question of the ontological irreducibility of collective intentionality.

However, the argument that shared intentionality is a precondition of reference acquisition exploits only part of the potential of the SIH in tackling philosophical issues. There are still plenty of questions about the nature of sociality that the SIH could fruitfully address that nonetheless prove recalcitrant to the methods of purely conceptual analysis. The current trend toward an interdisciplinary approach to the philosophy of society goes precisely in the direction of strengthening the role of empirical evidence in philosophical quarrels, to be coupled with an increased attention to the conceptual foundations of experimental research.