Philosophical Studies

, Volume 165, Issue 3, pp 921–937

Conceptual analysis as armchair psychology: in defense of methodological naturalism


    • Department of Humanities and Social SciencesRose-Hulman Institute of Technology

DOI: 10.1007/s11098-012-9981-9

Cite this article as:
Hartner, D.F. Philos Stud (2013) 165: 921. doi:10.1007/s11098-012-9981-9


Three proponents of the Canberra Plan, namely Jackson, Pettit, and Smith, have developed a collective functionalist program—Canberra Functionalism—spanning from philosophical psychology to ethics. They argue that conceptual analysis is an indispensible tool for research on cognitive processes since it reveals that there are some folk concepts, like belief and desire, whose functional roles must be preserved rather than eliminated by future scientific explanations. Some naturalists have recently challenged this indispensability argument, though the point of that challenge has been blunted by a mutual conflation of metaphysical and methodological strands of naturalism. I argue that the naturalist’s challenge to the indispensability argument, like naturalism itself, ought to be reformulated as a strictly methodological thesis. So understood, the challenge succeeds by showing (1) that we cannot know a priori on the basis of conceptual analysis of folk platitudes that something must occupy the functional roles specified for beliefs and desires, and (2) that proponents of Canberra Functionalism sometimes tacitly concede this point by treating substantive psychological theories as the deliverances of a priori platitudes analysis.


Canberra PlanNaturalismConceptual analysisPhilosophical psychologyPhilosophical methodologyFolk psychologyBeliefs and desiresPlatitudes analysisNeurosciencePsychology

1 Introduction

The Canberra Plan is a philosophical method that began with the collaborative work of David Lewis, Frank Jackson, and many others who passed through the Australian National University in Canberra during the 1990s. It is perhaps most widely discussed in analytic metaphysics and the metaphysics of mind in particular. But three core proponents of the Canberra Plan, namely Frank Jackson, Philip Pettit, and Michael Smith (henceforth “JPS” when I refer to the authors collectively), have recently compiled a number of co-authored articles (Mind, Morality, and Explanation 2004) in an effort to use the Canberra Plan to connect the various subfields of philosophical inquiry, most notably philosophy of mind, moral psychology, and metaethics. I will refer to JPS’s collective project as Canberra Functionalism (henceforth CF).

At the heart of CF (like the Canberra Plan more generally) is the idea that philosophy aims to provide an a priori analysis of everyday, or folk, concepts. Using conceptual analysis (CA), philosophers determine what is required for something to satisfy a commonplace concept like ‘belief,’ ‘free will,’ or ‘hope.’ The kind of CA in question is not just any old CA however but rather the distinctive two-step procedure sometimes called the “Canberra two-step,” which I outline in the next section.

CF is analytic functionalism writ large. Like most forms of functionalism, it begins with the idea that theoretical terms, in this case folk psychological (FP)1 terms like ‘belief’ and ‘desire,’ are to be defined2 functionally, i.e. by what they do rather than by what composes them. For the analytic functionalist in particular, those functional definitions are to be determined by folk theory, the commonsense theory consisting of all the various platitudes about mental states that we all take to be true, even if only implicitly, as evidenced by our everyday practices. Analytic functionalism is commonly contrasted with psychofunctionalism, roughly the view that theoretical mental state terms are to be defined by their roles in scientific psychology. It is important to keep analytic functionalism and psychofunctionalism distinct here since the latter, which looks to empirical psychology rather than folk theory to characterize mental state terms, is at odds with the core commitments of CF. Like analytic functionalism more generally, CF sees a priori analysis of FP theory as having a sort of privileged or primary status in elucidating the nature of cognition.

One of the main sources of opposition to CF is so-called philosophical naturalism.3 Unfortunately, naturalism is a rather ambiguous term in contemporary philosophy, and it is not exactly clear how JPS understand it here. In the most general terms, naturalism may be either a metaphysical or methodological thesis. The metaphysical thesis says, roughly, that the universe is exhausted by the natural stuff studied by the empirical sciences, or that all facts are natural facts, or perhaps even that “spooky” entities are to be rejected from one’s ontology. By contrast the methodological version says, roughly, that the scientific method sets the standard for acquiring knowledge of our universe (cf. Casebeer 2005). It is often far from clear in the literature which of these two distinct ideas the label “philosophical naturalism” is supposed to pick out, though it is worth noting that in some philosophical subfields such as ethics it refers primarily to a metaphysical position (see Darwall 1998, p. 27). For this reason, and others to be discussed below, it seems to me that it has become standard practice in philosophy to treat the metaphysical thesis as primary.

JPS themselves of course share the basic metaphysical commitments of philosophical naturalism—like most self-described naturalists they reject spooky entities from their ontology—so it is the methodological position that gives rise to their conflict with naturalism. The naturalists’ characteristic antipathy toward armchair methodology puts them at odds with JPS, since CA forms the backbone of CF methodology. In this paper I aim to support the foundational commitments of methodological naturalism by way of a bipartite critique of CF. As a corollary, I intend the argument to show that the tendency to read naturalism as a thesis containing foundational metaphysical commitments, whether in whole or in part, is problematic. Naturalism is properly a methodological rather than metaphysical position despite the current convention (or lack thereof). Naturalism, as I see it, is not a metaphysical position concerned with the metaphysical status of morality, or psychological states, or scientific postulates, and so on, but rather the idea that, one’s metaphysical commitments notwithstanding, it is the methodology of the empirical sciences that sets the standard for knowledge acquisition. This is the fundamental idea that the label ought to convey.

The argument I present here against CF is two-parted in that it begins with a critique of that approach in theory before turning to a critique of that approach in practice. In §2 I briefly review and elaborate the basic commitments of the Canberra Plan and CF in particular. In §3 I outline a version of the naturalist’s objection to CF methodology and then the reply from JPS. I then raise two theoretical problems for JPS’s reply in §4, which I think cast some doubt on the indispensability of CA in elucidating the nature of cognitive processes. These arguments are intended to motivate a turn toward methodological naturalism (and not metaphysical or “philosophical” naturalism) in principle. While philosophical naturalists are right to point out the limits of CA, those limits reveal more about the methods required to adequately explain the nature of cognition than they do about metaphysics. Finally in §5 I present the second part of the critique. I argue that, principled claims about methodology aside, a careful look at some recent Canberra-style functionalism in practice suggests that at times the proponents of CF themselves tacitly concede the theoretical point by treating substantive psychological theories as the deliverances of platitudes analysis, particularly in advancing theories of moral motivation.

2 Canberra functionalism

JPS and other Canberra Planners are united, primarily, by their commitment to the so-called Canberra two-step. This is the philosophical method whereby we investigate everyday concepts through a two-step procedure. First, we choose the term or concept that we want to understand. In the moral case it will be a moral concept, like ‘right’ or ‘fair.’ In the psychological case it will be a concept like ‘belief’ or ‘desire.’ We round up all of the platitudes about that concept, that is, the claims about that concept that reflect the way we (everyone, or “the folk”) use the term. These platitudes are then conjoined in order to determine the theoretical role of the concept in question. Famously, these theoretical roles are represented using Ramsey sentences, existentially quantified sentences that contain the information gathered from the platitudes. Then, in the second step, we go out into the world to determine what if anything plays the role specified.

Canberra Planners, JPS included, generally agree that an analysis of a concept is successful iff it gives us knowledge of all and only the platitudes that—in virtue of being platitudes—endow us with mastery of the concept (Smith 1994, p. 31). According to Smith’s version, for example, in acquiring a concept we acquire a set of inferential and judgmental dispositions, which connect facts expressed in the terms of that concept to other kinds of facts (37–38). The platitudes about that concept are then a set of statements of all these various dispositions. CA is just the attempt to articulate all and only these platitudes. To analyze moral rightness is to specify which property is the property of being right by referencing descriptions of the inferential and judgmental dispositions—platitudes about rightness—given by those who have mastered the concept (39).

Because it is just analytic functionalism writ large, this approach aims to define mental states not in terms of observable features of the world but rather implicitly in terms of their role in a broader network of theoretical and observable terms in a theory. Analytic functionalists claim that the source of that theory is common, folk opinion, i.e., FP, rather than empirical science as in psychofunctionalism. JPS, however, make at least one modification worth noting. As they see it, the source of the theory in question is FP minus anything in FP that we have any reason to seriously doubt. They call this modified theory “commonplace psychology.” Like FP, commonplace psychology is given by a Ramsey sentence, though it is a “cautious Ramsey sentence” with “anything open to serious doubt deleted or modified to make it pretty much a truism” (1993, p. 302). I use the label CF to pick out this set of ideas.

3 Naturalism versus the Canberra Plan

There is a dispute between CF and so-called naturalists about the role that CA plays in generating explanations of cognition and behavior.4 Naturalists have regarded CA with suspicion for a variety of reasons, but one common thread seems to be this: armchair reflection on concepts does not deliver as many interesting truths about human cognition and behavior as many philosophers seem to think it does because it is often far from obvious which concepts stand in need of analysis. It is quite often the sciences that reveal the concepts relevant to explaining the phenomenon of interest.

This point is of course controversial, since it is not just obvious that it is primarily the sciences that tell us which concepts need analysis. I think the naturalist has something like the following argument in mind.

It is often in the course of scientific practice, for example the formulation of hypotheses, the formulation of operational definitions in accordance with existing research frameworks, literature review, data analysis, and so on, that we come to understand what phenomena—usually some pattern of data or systematically collected observations—are actually in need of further explanation. In reflecting generally on scientific practice one might come to overestimate the extent to which armchair theorizing conducted independently of careful engagement with existing scientific literature (data, operational definitions, hypotheses, etc.) really determines the relevant concepts of interest.

Given the rapidly advancing state of the sciences, especially the neural and cognitive sciences, the phenomena of interest to those not immersed in this body of empirical literature—the folk and sometimes philosophers—may well be too crude, dated, or inaccurate to actually need analysis. Take the general, folk conception of ‘memory’ for example. This broad concept, upon analysis, actually fragments into a variety of more specific and more theoretically useful concepts, such as implicit and explicit memory, semantic, autobiographical, and procedural memory, long-term and short-term memory, and so on. But this fragmentation is in many respects a product of the development of empirical psychology and cognitive science, not the product of armchair reflection (see Baddeley and Hitch 1974; Miller 1956 for some examples of the scientific roots of contemporary ideas about memory). The naturalist can of course acknowledge that all of this may have required some philosophical work. But for the naturalist it is the kind of philosophical work that is part and parcel of scientific methodology involving careful engagement with scientific literature rather than mere armchair reflection.

According to CF, by contrast, an explanation for a cognitive or behavioral phenomenon will begin with an analysis of the concepts central to that phenomenon. Those concepts include beliefs, desires, and many other basic psychological concepts putatively employed in everyday, naïve folk explanations of behavior like hunger, thirst, pain, and so on. The naturalist rejects this proposal on the grounds that whether those folk concepts are relevant to the explanation—indeed whether they refer to anything at all—is a matter to be settled by the sciences, for the reasons just given.

It is perhaps because naturalism in philosophy is so frequently construed as a primarily metaphysical thesis that JPS’s dispute with naturalism seems to slide toward disagreements about ontology. As JPS explain their dispute with naturalism, the naturalist holds that the sciences, and not armchair reflection, ought to determine which entities we believe in while JPS must preserve a role for CA in generating the relevant ontology since that is a bedrock commitment of the Canberra Plan. As I see it, though, this interpretation of the dispute hinges on a confused picture of naturalism.

The naturalist’s concern ought to be methodological rather than ontological. Naturalists are committed to continuity between philosophy and the sciences because they believe the scientific method to be the most reliable method available for acquiring information about the world and hence for generating explanations of cognition and behavior. The naturalist’s objection then is not that we need science to tell us whether to posit beliefs and desires as entities but rather that we need to engage directly with the relevant sciences in order to determine whether those concepts are relevant to our best explanations of human cognition and behavior. The naturalist, to put it more simply, is not properly in the business of writing up lists of entities in which we should believe, but is concerned instead with developing continuity between philosophical and scientific explanations. Despite this confusion—which, I think, hampers as many naturalists as non-naturalists—we might reasonably suppose that JPS’s reply speak to both strands of naturalism in virtue of its defense of CA. So I will focus on showing why their reply to the naturalist’s challenge, the heart of which is their argument for the indispensability of CA, achieves much less than they claim.

As JPS understand the dispute, then, naturalists and eliminativists alike have objected to CF on the following grounds: in generating an explanation for any given cognitive or behavioral phenomenon it is the sciences, not CA, that reveals the ontology that we require. Whether beliefs and desires figure into the future scientific explanation for that phenomenon, then, is a matter to be settled in the laboratory rather than the armchair. JPS are, for this reason, “sure that [the objection from philosophical naturalism] lies behind much of the doubt naturalistically oriented philosophers have about beliefs and desires” (1990, p. 31).5 Of course naturalists need not actually be eliminativists about FP states, but many deny that CA is capable of revealing substantive truths about human psychological states, and that is the idea that the JPS must challenge to preserve the needed role for CA.

As JPS see it, philosophical naturalism poses no threat to either the existence of beliefs and desires or to CA. CA provides “excellent” evidence that we are sometimes in the states that satisfy what they call the “folk roles” for beliefs and desires (32). That is, JPS understand beliefs and desires in terms of the roles they play in our commonsense folk theory. The functional roles definitive of belief and desire are given by the roles we implicitly or explicitly presuppose in everyday predictive and explanatory practice, i.e., by their folk roles (18). Thus “to believe (or to desire) that so-and-so is to have a state in one playing the playing the role definitive of that belief between inputs, outputs, and other functionally specified states” (16). Beliefs and desires are thus functional roles that the folk use implicitly or explicitly in moving from observations of others’ behavior to hypotheses about their mental content (beliefs and desires) and to predictions of their future behavior.

Here is their argument for the existence of beliefs and desires:
  1. 1.

    It is sufficient for having beliefs and desires that one be in states which satisfy the folk roles.

  2. 2.

    We are sometimes in states which satisfy the folk roles.

  3. 3.

    Therefore, we have beliefs and desires.

Premises (1) and (2) are supposed to be relatively uncontroversial. (1), they argue, is a plausible account of the folk theory of beliefs and desires. And (2) will only be controversial if it is a genuine possibility that future neuroscience might ultimately give us reason to doubt that we are sometimes in states that satisfy the folk roles. But that, they claim, is virtually impossible. To accept the existence of beliefs and desires is merely to accept the idea that there are states that can be understood in terms of law-like relations among external circumstances, internal states and overt behavior and that facilitate our successful interactions with others by allowing us to move from our observations to explanations and predictions (16–17). We therefore have “excellent reason” to believe that the neurophysiological states needed to explain behavior actually occupy the functional roles definitive of beliefs and desires because beliefs and desires so construed are required for successful prediction and explanation: something must make possible our everyday successes in this way.
Moreover—and this is the crucial point—we are supposed to know this on the basis of CA. Here is the key section of their reply to the naturalist who is wary of the role of CA:

The concern of the naturalists is that in order to see the need for beliefs and desires, we had to indulge in some conceptual analysis. The science per se will tell us which roles are occupied by which neurophysiological states. The fact that this constitutes having beliefs and desires required common-sense functionalism, and so some conceptual analysis. We had to say something about how we conceive of beliefs and desires. It was, however, a very ontologically modest bit of conceptual analysis. We did not use our account of the folk conception of beliefs and desires as part of an argument that completed neuroscience is incomplete, but instead as part of an argument that beliefs and desires will very likely be found within what completed neuroscience tells us. The role of our account of the folk conception of beliefs and desires was simply to show which part of any likely complete neuroscience story is the part which says (though not in so many words) that there are beliefs and desires (32).

There are two problems with this response to the objection from philosophical naturalism. First, it misrepresents the naturalist’s concern about the role of CA in generating empirically adequate explanations of human agency. Second, it drastically overestimates the reach of CA in supposing that we can determine core components of explanations in the neural sciences of human cognition and behavior from armchair analysis of folk platitudes. I pose these in turn.

4 The limits of conceptual analysis

First, it is noteworthy that JPS introduce the objection from naturalism as the objection that “lies behind much of the doubt naturalistically oriented philosophers have about beliefs and desires” (31). It seems they have the motivation for naturalism the wrong way around, which suggests that they have not yet fully appreciated the force of the naturalist’s objection.

It is not that naturalists—by which, again, I mean those concerned with establishing continuity between philosophy and science—are motivated to reject beliefs and desires because they believe that CA is a waste of time. Rather they are motivated to think that all of the CA is a waste of time because, upon closer inspection, the relevant contemporary sciences seem to make little use of the mental state-roles—that is, the roles presupposed by the folk couched in terms of theoretical mental states—that philosophers like JPS would take to be legitimate occupants of the folk roles for belief and desire. This point is worth explicating.

A genuinely naturalistic criticism of CA is (or ought to be) primarily about methodology. Philosophical work on cognitive mechanisms ought to be continuous with scientific work that aims to explain the same phenomena. The idea is supported by the widely accepted observation that the naturalistic turn was inspired by progress in the neural and cognitive sciences (cf. Copp 2007; Darwall et al. 1992). If the FP states ultimately fail to appear in the relevant scientific research in the appropriate way then there is not much use in going to great philosophical lengths to save them. That is, unless we are strictly interested in describing how the folk think and talk about cognition in their own commonsense terms. But philosophical psychology is concerned with something richer than merely characterizing the way people talk. It aims to deliver empirically respectable, objective theories of the nature of cognition, including moral and social cognition, decision-making, and the like.6

Take for example the dispute surrounding the so-called Humean theory of moral motivation in metaethics and moral psychology, one of three core disputes constitutive of what Smith (1994) calls the central organizing problem in metaethics. The problem concerns whether an agent’s being motivated to act in accordance with her moral beliefs is brought about by the belief states themselves or only by the addition of a desire to act upon the belief. To engage in this dispute is to take on the question whether it is beliefs or desires that have causal efficacy in motivating agents. But this just assumes that one or the other of these states must figure directly into the best scientific explanation of the causes of moral motivation without looking first at the actual research to determine whether that assumption is plausible in light of existing data. It is an enormous assumption given the complexity of contemporary neural and cognitive science research on moral decision and motivation.

Even as a more general scientific matter, the question of whether the kinds of explanations emerging from contemporary cognitive science and neuroscience preserve the structure of our commonsense mental state concepts and FP explanations continues to be hotly contested in the philosophy of mind and science. There is a very difficult empirical project involved in determining whether and how one will actually find anything like beliefs, desires, wishes, fears, and so on in contemporary neuroscientific explanations for human and animal motivation, cognition, and behavior. Moreover, some scientists now openly acknowledge that even some of our basic scientific concepts are beginning to fragment into more accurate and more useful concepts for generating scientific explanations. Previously I mentioned memory as an example. Emotion provides a similarly compelling example. Here is the leading emotion researcher Elizabeth Phelps on the role of emotion in neuroeconomics, the emerging neuroscience of decision-making:

To date, most neuroeconomics studies have depicted emotion as a single, unified variable that may drive choice, often in contrast to a reasoned analysis of the options and their relative values (see, for example, Cohen 2005). This dual-system approach, although intuitively appealing, fails to consider the complexity of emotion or to capture the range of possible roles for emotion and affect variables in decision making. Adopting a more nuanced understanding of emotion will help clarify its impact on economic decision making and provide a basis for further understanding the complex interactions between emotion, value, and choice (2009, p. 234).

What is true of emotion, memory, and other basic mentalistic concepts could well be true of beliefs, desires, fears and other “intuitive” states. The naturalist need not deny that scientists discourse casually in the language of FP. Of course they do. But pointing this out is a far cry from successfully showing that scientists actually employ FP concepts as necessary theoretical elements in their work on motivation, memory and recall, long-term potentiation, decision-making, or whatever else neuroscientists might study. It is not just obvious from the start that FP states appear in the relevant capacities in the neuroscience of motivation, and so this is not an assumption that we should be forced to accept without evidence.

So what really casts doubt on the value of CA for understanding cognitive phenomena like moral motivation are not so much worries about ontology as worries about discontinuity between philosophical and scientific methods. Some recent scientific explanations of motivation, decision-making, and choice behavior are, as I see it, looking less and less like our high-level philosophical theories of motivation. They have little to say about beliefs and desires and much more to say about neural subjective value measurements, reward-prediction error signals, and movement selection thresholds (see, for example, Glimcher 2001; Glimcher et al. 2005; Gold and Shadlen 2000; Kable and Glimcher 2009; Shadlen and Newsome 2001).7 By reading naturalism primarily as a metaphysical rather than methodological thesis, JPS have missed the force of the naturalist’s objection to the indispensability of CA.

The second problem with JPS’s reply to the naturalist is that it drastically overestimates how much we can learn about the nature of cognitive mechanisms by a priori reflection on FP concepts and practice. For JPS, all that is required to defend the existence of FP states and hence the value of CA is evidence for the occupation of folk roles. And, as they see it, evidence abounds. If we were unable to predict the behavior of others on the road, for example, we would be unable to drive our vehicles without crashing (2004, p. 24). The folk clearly have tacit or implicit knowledge manifested in successful prediction and explanation of behavior on a regular basis, which is made possible by implicit (and possibly explicit) belief-desire hypotheses, or the putting of folk roles to use: “The empirical fact is that we are able to predict each other’s behaviour to a remarkable extent… We do this by, in many cases, working via belief-desire hypotheses” (18).

The key for JPS’s argument for the indispensability of CA is that we philosophers need good reason now, from the armchair, to doubt that neuroscience will ultimately do away with beliefs and desires construed as functional roles. This we are supposed to gather from reflecting on what must ultimately facilitate the success of this folk practice. JPS need to show that the philosopher is able to shed light on the mechanisms underlying folk practice—the practice of implicitly understanding the mental processes of others in a way that facilitates successful prediction and explanation—not through familiarity with the scientific study of human minds, brains, and behaviors but rather through a priori reflection on the kinds of mental state concepts that the folk must be employing in FP theory. That is what gets us the indispensability of CA.

But it seems to me that we could not possibly learn about the nature of implicit folk knowledge by reflecting on folk practice from the armchair for one simple reason. Knowing that FP makes successful folk practice possible by reflecting on folk platitudes is not the same as knowing how it makes that practice possible. This problem recurs at several points in their argument. To show that the folk must rely on belief-desire hypotheses, for example, they argue that the alternative is implausible:

Now the ability to move back and forth from behaviour in situations to belief-desire hypotheses in successfully explaining and predicting behaviour shows that we have implicitly mastered, whether or not we have always explicitly noted, the needed generalizations between the inputs in the situations, the behavioural outputs, and the beliefs and desires. The alternative is to suppose that we have arrived at our predictions by chance. But then the success of our predictions is also chance—and that is incredible (17).

But there is just no compelling reason to think that success by chance must be the one alternative to the use of belief-desire hypotheses for successful folk practice.

It might turn out, for example, that what we folk actually do when we predict and explain behavior is develop a kind of procedural knowledge: over time, repeated experience with a particular task or environment leads our nervous systems to automate responses to familiar patterns, much like how we learn to play sports and musical instruments, to walk and ride bicycles, and so on. We gradually come to have more fluid motor responses to how we feel resulting in more skillful behavioral output. Scientists might even be able to tell some story about how this is made possible by brain tissue: some kind of Hebbian process takes place so that when we encounter a previously-experienced situation the relevant synaptic connections are primed and the relevant neurons communicate more readily than before.

So the fact that we willingly stake our lives on our predictive success when we drive our cars may be evidence that we know we are generally successful in navigating from observation to explanation and prediction, but it really tells us nothing of why we are successful. We dread being in the car or on the road with new drivers precisely because those drivers have a limited experiential repertoire from which to draw. The fact that we could conceivably characterize all of this in terms of belief-desire hypotheses is irrelevant. The real issue is whether we must. But there are certainly available alternatives to belief-desire hypotheses, such as the one just given, which are not at all like success by chance.

Nevertheless, JPS insist that the functional roles of belief and desire are just too obvious to be denied: “It is beyond question that the functional roles are filled… to behave as if we have beliefs and desires is to provide overwhelming evidence that we in fact have beliefs and desires, precisely because to have beliefs and desires it to be understood purely functionally” (26).

But it is too quick to assume that because it is obvious that there is some function, Y, needed to bring about phenomenon Z that we must know how Y brings about Z. Prior to scientific investigation we may know almost nothing about the nature of Y, or about what function is required to bring about Z. JPS seem to think that in the case of intentional psychology we know the nature of Y on the basis of Z–Y is the (singular) function that needs to be filled. Yet there are many different ways in which the folk could have platitudinous knowledge: intuitions (Y1), propositional knowledge and deduction (Y2), some as yet unidentified sensory modality (Y3), and so on.

This important difference between that- and how-hypotheses can be illustrated using an analogous scientific case. Take for example the discovery of the mechanism underlying the action potential. Carl Craver provides a helpful account of that discovery:

[Hodgkin and Huxley] consider the possibility that ions are conveyed cross the membrane by active transport. They suggest that perhaps a number of “activation” particles could weaken the integrity of the membrane. They hint at a biological interpretation of their model according to which activation and inactivation particles move around in the membrane and somehow change the membrane’s resistance. They admit, however, that they have no evidence favoring their model of other possible models. This admission spurred research on the biophysics of the membrane and the search for ion channels. Nonetheless, well into the 1970s most neuroscientists regarded the talk of ion-specific channels as mere metaphor at best and boxology [vague re-description or “black boxing”] at worst (Craver 2007, p. 115).

What was crucial for explaining the mechanism of the action potential was not simply specifying what physical realizers filled the role of “activation particle” or “active transport” etc. Rather, it was a matter of pinning down how actually—by which function given the many possibilities—ions cross a cell membrane to initiate depolarization of a cell. Claiming that CA will elucidate the cognitive functions underlying successful folk practice (i.e. beliefs and desires as folk roles) while future neuroscience will merely reveal the realizers of those functions is very much like claiming that these scientists could have determined the existence of ion channels from the armchair, doing the biophysics only to reveal ion channel realizers. But that is false. The science revealed the function, not just the physical realizer. Alternatively, we might interpret JPS as claiming that the relevant functional role in this case is just the depolarizer. But the question isn’t what thing serves as the depolarizer but how a depolarizer actually depolarizes. Depolarization in this case is the product of the function, not the function itself.

The first premise in JPS’s argument states that, “it is sufficient for having beliefs and desires that one be in states which satisfy the folk roles” (18). But that is not quite enough. It is sufficient for having beliefs and desires that one be in states which satisfy the folk roles in such-and-such a way, i.e., in whatever manner is supposed to be characteristic of one’s undergoing beliefs and desires. How exactly do belief-desire hypotheses work? It may be obvious that we are sometimes in states that somehow manage to facilitate successful folk practice (i.e., folk roles), that is, that there is some internal mechanism that makes success possible and that makes our ascriptions of beliefs and desires intelligible when we describe, in the familiar language of FP, the behaviors of ourselves and others. But that is not the same as having evidence for the relevance of beliefs and desires—which serve to specify how and not merely that folk practice is successful—to empirically accurate explanations of cognitive processes. It is the latter that JPS need CA to deliver. For otherwise it really will be an empirical question whether people actually have knowledge of others’ mental states by way of belief-desire hypotheses, and we will have absolutely no special reason to doubt the possibility that neuroscience might eventually do away with beliefs and desires. As the naturalist has been arguing all along, we’ll have to do some empirical work to reveal the actual mechanisms responsible for our predictive and explanatory successes.

One of the more interesting features of this argument, I think, is that it can be tested against CF methodology directly. For it suggests that the purported first stage in the Canberra two-step, which is a priori in theory, will actually be filled with substantive empirical claims and scientific work in disguise when put into practice on a real philosophical problem. How else would the Planners come to assume that their particular hypothesis about which mechanism the folk require (namely belief-desire hypotheses qua functional roles) for successful practice is the right one given all of the plausible alternatives? In fact I think a closer look reveals that the Canberra two-step in actual practice often turns out to be armchair empirical psychology disguised as platitudes analysis.

5 Platitudes analysis or armchair psychology?

Smith (1994, p. 21–22) says that platitudes are a priori truths. That two possible worlds alike in all of their natural features must be alike in their moral features is a platitude; everyone agrees that the moral features of things do not “float free” of their natural features. A bit later he gives some examples of color platitudes. That there is no such thing as transparent matte white is a platitude, known a priori by competent users (29). The idea is that we come to master color vocabulary by coming to treat such remarks as platitudinous.

The issue I raised in the previous section concerns the nature of such platitudes. What Smith and other Canberra Planners need is that platitudes concerning mental states like beliefs and desires be of the same kind as these exemplary platitudes (such as those about whiteness). I have argued that this is not the case. What is required for successful folk practice may be described philosophically by these platitudes, but it isn’t just obvious that the folk must have mastery of these platitudes in order be largely successful at predicting and explaining behavior. It might be true that philosophers consider the claim, “there is no such thing as transparent matte white” a platitude in the sense that speakers who violate it might lack mastery of color terms from the standpoint of an external philosophical characterization of the folk, yet a person could conceivably live out her life without ever having her mastery of that platitude seriously tested, at least not in any way relevant to her having successful folk discourse or practice. Or, worse, she might actually employ the faulty concept to the appearance of some kind of muddled whiteness with some small degree of success (internally, from within the folk perspective), as when asked, for example, how the blinds in her office look in the morning sunlight. Philosophical description of folk success in terms of platitudes and folk mastery of such platitudes are two different things.

Moreover, do philosophers really know a priori that folk knowledge (of at least the implicit sort, though perhaps also explicit) of functional roles is platitudinous? It sometimes looks as though their methods are a tacit concession that philosophers have no such knowledge a priori. Consider, for example, how Smith arbitrates between competing accounts of desire, such as the phenomenological account and the dispositional account:

Do we really believe that desires are states that have phenomenological qualities essentially? That is, do we believe that if there is nothing that it is like to have a desire, at a time, then it is not being had at that time? I should say that, at least as far as commonsense opinion goes – and what else do we have to go on in formulating a philosophical conception of FP states – we evidently have no such belief. Consider for instance what we should ordinarily think of as a long-term desire: say, a father’s desire that his children do well. A father may actually feel the prick of this desire from time to time, in moments of reflection on their vulnerability, say. But such occasions are not the norm. Yet we certainly wouldn’t ordinarily think that he loses this desire during those periods when he lacks such feelings (1994, p. 108–109).

Note that the argument depends upon a controversial empirical claim—one in need of testing—precisely because it makes an assertion about what is normally the case.

There are two things going on in this passage at the same time that must be kept distinct: there is Smith’s reflection on what we evidently believe about the nature of desire and then there is his reflection on why we believe what we believe about desire. At first, it is just the latter that seems to require the support of empirical evidence. To know what we normally believe we have to reflect on what most of us think about the nature of desire by thinking through a hypothetical case of characterizing it in others (in this case in fathers). In fact though the former requires empirical consideration too, which is why Smith says that “evidently” we have no such belief about the nature of desire. What is evident are the norms governing the use of the term in the general population, or the way that people actually use the term when they use it in the same way that most everyone else does.

The first step in the Canberra Plan is to lay out the platitudes concerning the concept of interest, in this case desire. For Smith (1994, p. 31) this amounts to saying what is crucial to something’s being a desire (Nolan 2009, 274n). So the relevant part of the above passage from Smith is the question of whether we really believe that desires have phenomenological qualities essentially. And Smith’s answer is that we evidently do not really believe this because of the way that most of us employ the term. That is, because of the norms governing the usage of the term “desire” in the linguistic community. But the only way that we can have access to those norms is to go out into the world and observe the way that people use the term. One cannot have access to linguistic norms, by which is meant here standards determined by linguistic practice, from the armchair. At most one can determine how people might possibly use the term incoherently, as by using “bachelor” to refer to married men, for example. But that is not the same as determining how most people in fact use the term. People ought not speak incoherently; irregardless, they often do.

This is not the only instance in which Smith’s argument for the dispositional account of desire and against the phenomenological account invokes fairly substantial empirical considerations. He argues that there are two serious problems for the phenomenological conception. Consider his objection to the phenomenological conception, which is that it gives an implausible epistemology of desire (1994, p. 108). The idea behind the objection is that any adequate account of desire must recommend an epistemology that is both prima facie plausible and that allows subjects to be fallible about their desires. Smith thinks that the phenomenological conception fails to meet both requirements. The epistemology of the phenomenological conception is based on the epistemology of sensation, which holds that a subject is in a state with phenomenological content if and only if she believes she is in a state with that content. For example, she is in pain if and only if she believes that she is in pain. Consequently, “if we think of desires on the model of sensations, it is plausible to hold that a subject desires to ϕ if and only if she believes that she desires to ϕ” (105). We can represent the biconditional as PD (which stands for “phenomenological conception of desire”):

(PD). A subject desires to ϕ ← → she believes that she desires to ϕ.

He argues that PD fails in both directions. From left to right it is plainly false, while from right to left it violates the fallibility constraint, i.e., any adequate conception of desire must be able to account for the fact that people are sometimes wrong about the desires that they have. But the interesting point for my purposes has nothing at all to do with whether PD is a plausible account of desire but only with the methods that Smith uses to reject it. Both of his arguments against PD proceed by counterexample. Let’s look at the first.

Smith argues first that PD is false left to right because there are perfectly realistic cases in which it fails to hold. We are asked to consider John, a narcissist who each day buys a newspaper from one particular newsstand while on his way to work. Though John might just as easily get the paper at any newsstand, he continues to patronize just the one stand in particular. As it turns out, that particular stand has mirrors placed in such a way that shoppers cannot help but look at themselves as they make their purchases. If we were to tell John that perhaps he buys his paper there because he enjoys looking at his reflection, he would vehemently deny it, “and it wouldn’t seem to John as if he were concealing anything in doing so” (106). Suppose however that without the mirrors, John would no longer visit the shop. From this Smith concludes that we have reason to reject PD left to right:

If all this were the case, wouldn’t it be plausible to suppose that John in fact desires to buy his newspaper at a stand where he can look at his own reflection; that, perhaps, he has a narcissistic tendency and buying his newspaper at that stand enables him to indulge it on the way to work? And wouldn’t it also be plausible to suppose that he does not believe hat this is so, given his, from his point of view, sincere denials? If this is agreed, then we have reason t reject the principle left to right… (106).

Let us just concede to Smith for the moment that this case is rather plausible. The real issue anyway is not whether the case is plausible but whether the truths about the nature of desires that it reveals are platitudes about the concept that we find in folk discourse or practice. This is what Smith means when he says that we need an alternative account of desire in order to “make sense of desires as states with propositional contents and that thus allows us to make sense of our commonsense desire attributions” (111). Is Smith’s attribution of desire in John’s case really commonsensical or platitudinous?

I think it is not. We have seen that even for Smith platitudes are instances of knowledge, implicit or explicit, which are such that in coming to treat that knowledge as platitudinous we come to have mastery of the concept (31). Recall that Nolan characterizes Smith’s view of platitudes as coming close to the nonnegotiable analytic truths end of the spectrum. It is surely not anything like an analytic truth that desires are such that they make us do things that we might not consciously realize we desire to do. Nor could it be known a priori, with no experience with desiring or watching others desire. If this is a truth about the nature of desire at all it is surely known a posteriori in virtue of some experience with desiring and watching other people desire.

But is it platitudinous anyway? I doubt it particularly because the example is so controversial. The mere fact that most would say to John that he frequents that particular newsstand because he likes to view his own reflection is evidence that that is what most people would think. Without conducting a survey, it seems quite within the realm of possibility that the average person doubts that John had absolutely no idea why he preferred one newsstand to the other, at least for the first few trips. Similarly, it is not just entirely obvious that John lacks any awareness whatsoever of his enjoyment in looking at his reflection. Even within the field of psychology this is a significant empirical claim about the nature of desire, and one calls out for some testing. That is probably why Smith argues by counterexample rather than simply asserting the obvious fact that desires are such that they make us do things even when we have no idea whatsoever that we desire to do them. That is not an analytic truth, a platitude, or even a piece of common sense. It is rather a deliverance of armchair psychology, which ought to be regarded as empirically suspect until more rigorously tested.

Let me be clear about what I am not arguing here. I am not arguing against the possibility that one facet of the nature of our desires (whatever those turn out to be) could be that they sometimes make us do things of which we are utterly unaware. This could turn out to be an interesting fact about desire and it will still be the case that we arrived at the truth via careful observation, experience with desires and desirers, and assessment of cases. The one thing that seems clear is that some member of the folk could be adept at applying the term without any knowledge whatsoever of the fact that sometimes desires work entirely subconsciously. Such a person would be exactly the type to criticize John for his narcissistic tendency to gaze into mirrors. There is just nothing incoherent about a conception of desire lacking the subconscious element, and so it cannot be a platitude the accommodation of which is required for an adequate conception.

Second, and more importantly, I am not arguing that Smith will fail to abide by the methodological standard of CF that he endorses any time that he appeals to a behavioral example or a hypothetical case. We can agree that a posteriori considerations have some role in CA. It is just that any such considerations will have to reveal a rather obvious truth about the nature of the concept in question in order for it to legitimately reveal a platitude. And on JPS’s account this will generally happen in the course of considering what must be the case about the nature of desire in order for folk practice to be so successful.

But there is just nothing obvious about the purported features of desire that Smith takes these hypothetical cases to show, neither as an a priori matter nor as a fact about what the folk must do tacitly. He seems to concede this much when he asks whether the explanation for John’s behavior is plausible rather than declaring that it is obvious or necessary. And that is why I think the method used betrays the need for experience and empirical investigation. The whole point of engaging in empirical science at all is to test empirical claims that aren’t perfectly obvious. Successfully supporting a scientific hypothesis amounts to showing that it holds generally and so can be applied to help explain other untested cases, precisely like John’s.


FP will also stand for “folk psychology.”


Or, alternatively, have their references fixed by the theories in which we find them (see Jackson and Pettit 1990, p. 37).


The generic label “naturalism” in this dispute is a bit misleading since the Canberra Planners take their method to be naturalistic, too. Indeed the Canberra Plan is sometimes called “naturalistic analysis” (see Braddon-Mitchell and Nola 2009). For the sake of consistency, however, I will continue to follow JPS’s lead in using the term “philosophical naturalism” to pick out their opposition, a non-CF form of naturalism that is the source of the particular objection to CA that I consider in this paper.


According to Jackson & Pettit (1990, p. 53–4, n24), the naturalist position under consideration informs the approach of Devitt (1984), and Sterelney and Devitt convinced them of the need to address it.


The source of the quotation is an article authored by Jackson and Pettit (“In Defence of Folk Psychology”, Philosophical Studies5,1990) and not Smith. But it has since been reprinted in the collective volume, Mind, Morality, and Explanation (2004), where they explicitly declare their commitment to a collective projective. For that reason I will continue to treat the argument presented in Jackson and Pettit (1990) as a part of JPS’s collective defense of the Canberra Plan.


This is evident from the growing body of philosophical literature concerned with the application of data from psychology, psychiatry, and neuroscience to disputes in philosophical psychology. Recent scientifically informed philosophical work on moral motivation is an excellent example (e.g., Kennett 2002; Kennett and Fine 2009; Schroeder et al. 2010).


Of course one could try to show that each of these ideas preserves the structure of higher-level FP concepts but that is quite clearly a difficult empirical project the success of which we cannot just take for granted.


Copyright information

© Springer Science+Business Media B.V. 2012