1. Introduction

Michael and MacLeod begin their paper on theories of reference for intentional concepts with the following pronouncement: “It is a curious fact that many philosophical discussions about the reference of everyday concepts of intentional states—from folk psychology in the 1980s to theory of mind and mind reading more recently—have, in recent decades, been predicated on descriptive theories of reference” (Reference Michael and MacLeod2013, 212). Although Michael and MacLeod are correct that the eliminativism debates of past decades were so predicated, it is surprising to hear that the Theory of Mind (ToM) and mindreading debates—the ones on which Michael and MacLeod focus their attention—are so predicated, because these debates have largely ignored issues about the reference of intentional concepts. As a result, there is some chance that one might be able use the insights from theorists working on reference to illuminate some of the debates over ToM development on which Michael and MacLeod focus, especially if one were able to show that an overreliance on descriptive theories of reference has hindered any fruitful interaction. Unfortunately, Michael and MacLeod’s paper neither shows the negative effects of descriptivism on theories of reference for intentional concepts nor provides an adequate picture of how the sort of theory they advocate might explain either the reference of intentional concepts or the puzzles of development on which they focus. In this article, I give reasons to think that the prospects are dim for applying causal approaches to the major topics raised by Michael and MacLeod.

2. The Michael and MacLeod Account

Since Michael and MacLeod frame their project as a flight from descriptivism, it will be important to articulate which descriptivist features they seek to shun. Michael and MacLeod’s causal alternative involves “tracking mechanisms” that, unlike mechanisms involved in descriptivist accounts,

1. 1. Can be used to fix reference in a relatively theory-free way, allowing the reference of terms/concepts to exist independently from any particular theory of mind (hence, descriptivism leads to dependence)

2. 2. Allow for a different model of reference continuity in the face of descriptive change or inadequacy (hence, descriptivism leads to discontinuity)

3. 3. Establish early causal relations that then enable the cognitive system subsequently to build on that basic causal relation in order to develop a capacity for representing the intentional state (hence, descriptivism leads to a different sort of representation-building process, one that may change the extension of the concept).

Michael and MacLeod’s alternative proposal should be judged on the extent to which these features provide an adequate account of the phenomena of reference fixing and conceptual development without involving the contrasting descriptive elements.

The impetus for this causal analysis seems to be Michael and MacLeod’s view that intentional concepts bear important similarities to certain scientific concepts. In particular, Michael and MacLeod seem impressed with causal accounts of scientific concepts because they allow scientists to misdescribe the entities they are studying, and yet still refer to the entities in question. Because causal accounts allow a scientist to lock onto the entity in a way that does not depend on having correct descriptions of the entity, she will then be in a position to clarify and develop various descriptions of that very entity that is responsible for the phenomena confronting her. According to Michael and MacLeod, the child is in a relevantly similar position when acquiring intentional concepts. He must figure out what sort of state is causing the phenomena confronting him (the observed behaviors or indicators), and once he has locked onto that cause, he will then be in a position to clarify and develop various descriptions of that very entity.

Concepts that fit this causal analysis, either in scientific practice or ontogeny, would need to emerge early enough to provide stability through the acquisition process and obviate the need for a costly conceptual overhaul (Michael and MacLeod Reference Michael and MacLeod2013, 226). Such early emerging tracking processes would provide robust and steady causal connections whose presence allows for some type of subsequent descriptive or theoretical improvement (what Michael and MacLeod dub “conceptual development”). Although Michael and MacLeod provide evidence that the circumstances required by their account are present in ontogeny, they never discuss how a descriptivist might explain this period in development.Footnote ¹ Nevertheless, it would seem to follow from Michael and MacLeod’s take on descriptivism that such theories will be ill suited to capture reference-fixing practices in science and in ontogeny because they lead to some unfortunate combination of theory dependence and referential discontinuity (1–3).

3. Descriptivism under Any Guise?

One puzzling feature of Michael and MacLeod’s account, beginning with their choice of title and continuing throughout their paper, is their suggestion that the account on offer should be described as a causal theory of reference. Although they provide the reader quite a bit of detail about the genesis of the causal theory in the philosophy of language, Michael and MacLeod are surprisingly silent about what role descriptions should actually play in their own theory (i.e., when applied to psychological states).

What Michael and MacLeod do offer about the role of descriptions in their account is a concession to worries that (pure) causal theories seem to make reference too easy (Reference Michael and MacLeod2013, 216–17). They agree that there is at least the appearance of such a problem and address it by mentioning the possibility of causal-descriptive accounts. Unfortunately, this concession, without significant elaboration, is not edifying because there has never been anything like a consensus in the philosophical literature as to how or whether these hybrids are supposed to work. Not only do the causal-descriptivists cited by Michael and MacLeod disagree with one another about how and whether the other causal-descriptivists incorporate descriptions successfully into their accounts,Footnote ² the perspective they offer is hardly a vindication of the causal approach. Take, for instance, Kroon’s way of introducing such hybrids: “Causal descriptivists thereby express large, perhaps even complete, agreement with the semantic intuitions that causal theorists of reference use to support causalism. They agree with the evidence, but still think that the best explanation for the evidence is a kind of descriptivism, and not causalism” (Reference Kroon1987, 1). Since Michael and MacLeod concede that a strict causal theory will not suffice for concepts, they must explain exactly where it fails and what resources, especially descriptive resources, are needed to supplement it.

Michael and MacLeod begin to address something like this issue for their causal-descriptive account in the following passage:

If we take this quotation as providing something like a reference criterion, we end up with an account that fixes reference by whatever descriptions are most “crucial” and “productive.” Hence, not just any old description will do, but unlike “straightforward” causal accounts, some descriptions apparently will be involved in reference fixing and must be shared across development to ensure referential continuity of the sort that interests Michael and MacLeod. This appears to be the version of the causal-descriptive account that they endorse. But, if this interpretation is correct, their account is strikingly similar to what I take to be their most likely descriptivist target (they never give an actual target)—the theory-theory of concepts.

Theory-theorists (Carey Reference Carey1985, Reference Carey, Carey and Gelman1991; Keil Reference Keil1989; Murphy Reference Murphy2002) take the basic insight that concepts are rich bodies of information and develop it into a robust picture of how such theories meet the various desiderata that a theory of concepts must meet (e.g., as articulated by Laurence and Margolis Reference Laurence, Margolis, Margolis and Laurence1999; cf. Machery Reference Machery2009). Although these little theories (concepts) and the bigger theories that contain them (the theories for some domain) work swimmingly to explain the information on which category judgments and other inferences are drawn, at least some of these theorists now acknowledge that such accounts do not work well at explaining concept stability and concept individuation. In order to deal with these problems, they have invoked a distinction between core and nonessential (although perhaps stereotypical) features for concepts (e.g., Carey Reference Carey2009), thus allowing the theory to play the role in judgment it is supposed to play (as in categorization and other judgments) while still allowing the subject to maintain a (descriptive) core of the concept that resists change and aids in concept individuation and reference determination and stability. Upon reflection, Michael and MacLeod’s account, with its core or crucial descriptive contents fixing reference and maintaining referential stability over time, seems difficult to distinguish from such forms of descriptivism. Yet, even if Michael and MacLeod have acknowledged the need for some sort of causal-descriptive account, one would assume that it would not simply collapse into a common variant of a descriptivist theory.Footnote ³ Such an account would surely not free us from viewing the ToM debates as predicated on descriptivism, and at a minimum, we need to know what role causal features actually play in their account that distinguishes it from other descriptivist accounts.

However, if we take the causal-descriptivist authors cited by Michael and MacLeod as a guide, we find a very minimal role for causal relations to play in explaining reference. For both Kroon (Reference Kroon1985, Reference Kroon1987) and Psillos (Reference Psillos1999), for example, true descriptions fix the reference of a term, whereas causal relations merely serve as a constraint to ensure that the description is not free-floating but can trace its etiology back to the phenomenon in question. Assuming that we embrace the causal approach’s standard view that such causal contact need not be direct, but that indirect causal links (e.g., those gained through other people) will meet this constraint, the causal requirements are indeed quite minimal. It is not clear when a person who possessed a true description of an entity (especially a more theoretical one; see below) would not count as so connected to the entity or magnitude in question so long as it results from some even distant contact with that entity (or the phenomena for which it is responsible) or any other person who was so related to the entity.Footnote ⁴ Would theory-theorists or other descriptivists be in trouble if they agreed to this minimal constraint, and if so, would such a brand of causal-descriptivism count as perniciously predicated on descriptivism? It is hard to square such a minimal role with the title and tenor of Michael and MacLeod’s paper.

Assuming that Michael and MacLeod would wish to resist such overt forms of descriptivism, we might seek a distinct account of reference fixing in another similarity they draw between reference fixing in scientists and in children. On this view, they would explain reference fixing causally and explain the phenomenon they describe as “conceptual development” descriptively. What follows might be seen as such an endorsement: “We would now like to look more closely at the role that reference fixing plays in conceptual development. … Children actively contribute to sustaining and refining the causal relationships between their bodily tracking devices and others’ intentional states. This makes it possible for intentional concepts to further articulate the links to causes of behavior that have been fixed by tracking systems and, thus, to track ever more sophisticated and fine-grained intentional states” (Michael and MacLeod Reference Michael and MacLeod2013, 222–23).

In the penultimate sentence in this quotation, we are told that such “conceptual development” consists at least in part in a child’s being able to “contribute to sustaining” and “refining” already established causal links. This claim makes the role of descriptions relatively clear and unproblematic, since nothing about this sustaining and refining can affect what the concept is about. The final sentence, however, is potentially more problematic for a causal account. Tracking a more “sophisticated” or a more “fine-grained” intentional state could involve at least two things. First, one might track the same state but do so in a more sophisticated or fine-grained way. This is how both Fodor (Reference Fodor1998) and Millikan (Reference Millikan2000) articulate their distinct causal accounts for natural kind and substance concepts, respectively. But, one might come to track distinct types of states that are more sophisticated or fine-grained. On the face of it, this is much more at odds with the basic causal picture because whatever the concept under scrutiny is a concept of can apparently change due to the developmental processes described above.

Before we can assess this account of conceptual development, we need to know how these resources—both causal tracking and descriptive—are brought to bear on not just cognition and theorizing about intentional phenomena, that is, how the agent comes to understand or think about the phenomena through development, but on the nature of intentional concepts and their reference. What is on offer, after all, is a theory about the reference of intentional concepts, not just a theory about how children develop their capacities to think about or understand other agents.

4. Categorizing Concepts and Conceptions

Michael and MacLeod’s perspective is influenced heavily by Millikan’s account of substance concepts. On her account, the reference of a substance concept is fixed early in ontogeny in a nondescriptive way. By bringing the basics of her account on board, Michael and MacLeod get the sort of model they sought: no theory is needed to fix or capture the reference, so reference can remain continuous despite myriad changes and differences in descriptions. Such descriptive elements are not relevant to concept individuation, on Millikan’s view. These elements are part of the conception of that substance, not the concept of it (Rey Reference Rey1983). If they hope to take a stand on the issues stemming from the concepts and reference literature, Michael and MacLeod must draw this distinction clearly and explain the extent to which any feature that they label as “conceptual development” fits on one side of the divide or the other.

According to Millikan and Michael and MacLeod, this process of conceptual development is an extended, ongoing process of augmentation and refinement through which an agent becomes better and more sophisticated at interacting with the entity in question. This augmentation process or “filling out the concept into a more and more adequate one” could take one of three forms (Millikan Reference Millikan2000, 92, see also 39, 50, 92–94): (1) the referent is fixed early on, and the concept is thereby individuated—only the conception is filled in (or out) through subsequent descriptive development; (2) equivocation occurs throughout the process of refinement, and there is no sense to be made of what the referent is or what concept the agent really possesses; or (3) the concept gets filled in in some way that is distinct from how the conception gets filled in (according to 1), and this filling in might affect the referent.

Michael and MacLeod’s suggestion of merely taking some general causal approach is powerless on its own to explain critical features in development because taking a broadly causal approach results in this full range of options regarding where to place what is acquired or developed. Is the best developmental explanation that a concept is locked into place early (i.e., nondescriptively), with its reference fixed, and that all subsequent descriptive change (no matter how dramatic) is an improving conception of that phenomenon? Or, is it better for us to look at the developmental data and acquiesce in the equivocation perspective, simply acknowledging the discriminations infants make (or fail to make) but admitting that there is nothing to say about concepts in this developmental period? On this view, it is not until the reference gets “focused” that one can explain what the agent actually understands and tracks,Footnote ⁵ and there is good reason to insist that such focusing comes only late in ontogeny.Footnote ⁶ Or, do we, in effect, give up on the causal theory and allow some descriptive elements to settle issues of reference but in a way that Michael and MacLeod need to more fully explain (would it be the suggested account in sec. 3 or some other account)? Not only do they vacillate among these options, seemingly unable to articulate the depth of the causal aspects of their account, but once the options are laid out, it is not at all clear which of the options is best supported by the empirical literature cited by Michael and MacLeod. For each of the four phenomena they consider (emotion contagion, mimicry, gaze, belief), there are good reasons to find certain features locked in early, with other features only coming later. But, this fact, on its own, does not favor any of the options more than any other. In the end, what matters most is how these subsequent developments are placed on the concept-conception divide and how that affects the features that are responsible for reference fixing.

5. Passing in the Night: Theories of Concepts and Theory of Mind

Even if Michael and MacLeod could convince us that an appropriately causal theory of reference could be given for intentional concepts, it would be a pyrrhic victory. This is because almost none of the issues about conceptual development that interest Michael and MacLeod will be affected by issues about reference fixing and concept individuation of the sort that a theory of reference could settle. Perhaps this is part of their point—if we can formulate an account of development in which intentional concepts are present and fixed early, then some of the current ToM debates about when the adult or mature concept finally emerges in ontogeny just will not arise, and we will no longer try to explain developmental changes in judgments or discriminatory capacities as the result of changes in concepts.

But, aside from quashing this particular set of debates about the presence and timing of adultlike concepts, it is not clear why the account of development that Michael and MacLeod seem most interested in needs to be couched in terms of reference and concepts at all. It is even less clear why it should be couched in terms of a causal account. Michael and MacLeod seem most interested in allowing descriptive improvement or development in the resources used by scientists or children in a way that need not lead to conceptual discontinuity. But, whether one is a causal theorist, a descriptivist, or a causal-descriptivist, one can maintain reference in the face of descriptive upheaval, so long as a reasonable concept-conception distinction can be drawn. If it is primarily referential continuity that one wants to maintain, the pure causal approach accomplishes this most effortlessly (although it comes with the side effect of “too easy” reference), but descriptivists can manage this as well.Footnote ⁷

Now, other than being able to establish that an intentional concept is present and can or will maintain its reference through subsequent maturation, it is not clear how any claim about the reference of some intentional concept (or how that was fixed) will be a primary factor in explaining ToM judgments or their development. As with most of the empirical work described as investigating “concepts,” the experimental work Michael and MacLeod describe is about the conceptions people use in making judgments about the intentional states of others.Footnote ⁸ At most, Michael and MacLeod can explain how “conceptual development may depend upon the consistency of the causal relationships between other agents’ behavior and one’s own bodily and cognitive responses to that behavior” (Reference Michael and MacLeod2013, 226). But, note that this does not involve the development of a concept, over time, but the development of its conception. In Millikan’s terminology, these “individual recognitional abilities” that an agent develops that serve as “the ‘means’ by which the thinker knows how to reidentify a substance” not only allow her to have her concept activated in the presence of its object via tracking mechanisms, but more critically, the particular means may then become integrated with other information and may thereby stand in inferential relations that make possible various categorizations and judgments (Reference Millikan2000, 174). On the sort of causal account we have been trying to piece together, it appears that researchers will be left investigating these various “means” by which the concepts are applied and integrated with one another, with many such conceptions commonly falling under one concept. Debates about which sorts of conceptions will count as really representing the intentional states of another, as such or qua intentional state, will surely have to take place by invoking the sorts of means or modes of presentation, and so on, that are the purview of the study of conceptions and will not be settled in terms of whether a state is (merely) tracked.

At bottom, such is the odd result of taking a causal approach to intentional concepts: to the extent the causal approach to concepts and their reference is correct, the fact that an agent possesses this sort of tracking device will fall far short of providing much of anything except serving as something of a sine qua non for the actual descriptive means of tracking that will be invoked in explaining development. But, I take it that this result should be of no surprise for those supporting a causal account of reference.

6. Causal Accounts and Theoretical Concepts

In bringing this discussion to a close, I want to stress some reasons to think that a causal account of reference has been, from the outset, quite ill suited for concepts of the intentional variety. First, it is surprising that Michael and MacLeod take theoretical concepts in science as a case through which to promote a causal account of reference. Although causal accounts have proven successful (but certainly not uncontroversial) in dealing with proper names and natural kind concepts, there are substantial obstacles in applying them to theoretical scientific concepts, especially concepts for unobservable entities. The causal views of the Fodor-Millikan variety have rarely, if ever, been viewed as viable by their proponents for anything other than concepts for observational entities, and as argued in section 3, the causal (descriptivist) accounts that have sought to deal with theoretical entities hardly offer a vindication of the causal theory itself for such cases.

Michael and MacLeod’s method of dealing with an entity’s unobservability appears to be to posit an unobservable causal agent responsible for the phenomenon that can be observed and to have reference fixed to that agent. However, Kroon, anticipating such a move, notes that even if such a posit is made, what fixes the reference is a substantive theory-embedded description, and not some such thin acknowledgment of the presence of some unknown entity: “the burden of reference for the [theoretical] term introduced rests broadly on the theory within which the term is embedded, and not on some cautious causal descriptions of the form ‘〈whatever ψ it is that is responsible for such and such phenomena〉’ (or less cautiously, 〈whatever ψ is responsible for such and such phenomena〉, where ψ is a sortal predicate)” (Reference Kroon1985, 150).Footnote ⁹

Second, causal theories are best equipped for concepts that have a certain independence from their conceptual neighbors (i.e., for those that may be given an atomic rendering). Yet, as Armstrong famously notes, “The logical dependence of purpose on perception and belief, and of perception and belief upon purpose, is not circularity in definition. What it shows is the corresponding concepts must be introduced together or not at all. … Correlative concepts or mutually implicated concepts … [of mental states] will be a complex and interlocking set of causal factors” explaining purposeful behavior (Reference Armstrong and Lycan1999, 23–24). Even if such causal-functional accounts of these phenomena fail to yield a complete picture of the mind, there is still a strong case to be made that these intentional concepts are logically dependent on one another. Roughly, the seeming logical or conceptual interdependency, when coupled with the unobservability of intentional states, suggests that possession of, say, Belief will not be a mere matter of a triggering of that single concept in the presence of an unobservable belief but something involving the acquisition and deployment of many related concepts whose individuation conditions will reflect these interrelations rather than merely any one of those concept’s typical causes.

7. Conclusion

In this article, I have argued that Michael and MacLeod’s account is underdescribed and that as it has described more details, it turns out to be difficult to distinguish from the descriptivism that they claim to reject. There may be ways of fleshing out the picture of reference that they suggest, but it needs to be much clearer where their account places critical features on the concept-conception scale. In supporting a causal account of reference for intentional concepts, Michael and MacLeod and other like-minded theorists must take special care in explaining the role that causal features—and descriptive ones—should play in that account.