David Henderson

Draft, Comments Welcome

I. My Concerns  1

II. Casting About for a Promising Approach to the Subject and to What Is A Priori 2

III. The Promising Approach to A Priori Conceptual Truths. 3

IV. Objections and Refinements  5

 

I. My Concerns

 

          Actions are done for reasons.  The reasons are beliefs and desires, which are physical states that causally interact in a rather special way.  Their interaction exhibits a characteristic pattern: it is rational, at least in certain important respects.  My concern in this paper is to say something about how we know these things.  Is all this known or knowable a priori?  And, if so, what elaborations of these claims are also so knowable?  (For example, can we know a priori to what extent beliefs, desires, and actions must be rational?)

This issue arises most insistently when I read the work of Donald Davidson.  Famously, Davidson writes of “constitutive (or synthetic a priori) laws [or principles]… within a conceptual domain” (1980b, p. 221).  We are to find such principles at work “constituting ideas” and “sustaining concepts” (1980b, p. 221).  In the conceptual domain of intensional psychology, there are said to be such constitutive principles associated with the assignment of content: “the content of a propositional attitude derives from its place in the pattern [of states]”; and that pattern is said to involve “a large degree of consistency,” or rationality, and correctness (1980b, 221).

            Davidson explains that these constitutive criteria can be denied or repudiated.[1]  However, in doing so, one would have “changed the subject.”  Apparently, when we decide not to accept a certain constitutive principle or criterion, we cease to use the old concept that was thereby constituted.  Using different concepts, we then discourse on a different subject (Davidson 1980b, p. 216).[2]

            There is something compelling in the idea that, in various domains, there are certain central principles that cannot be denied without doing something that might be called “changing the subject.”  It is also plausible that these matters might be understood in terms of “concepts.”  At the same time, along with Davidson, I wish to make sense of all this in a way that does not presuppose the analytic-synthetic distinction.  Davidson’s discussion, with its somewhat cryptic or at least compressed, talk of “concepts” held in place, or “ideas” constituted, by “whole set[s] of axioms, laws or postulates” seems not to take us far in the direction of a clear positive alternative.[3]

Suppose we have in hand a workable general understanding of the subject, concepts, and the a priori.  What can be said of the rationality of beliefs, desires, and actions?   It does seem that certain elements of rationality are central to our understanding of propositional attitudes and the ways that they cause actions.  But, while it may be a priori that beliefs, desires, and actions are rational in certain respects, are those respects enough to make it a priori that agents are predominantly or largely rational?

            I will reveal my suspicions at this point.  I believe that we must indeed attribute significant rationality to agents.  However, I also believe that much of this requirement is a function of the nature of current theory, and most of what is required here is not a priori.  Our attributions of these contentful states are informed by what in the way of rationality (and irrationality) is explicable or understandable in terms of our psychological theories (Henderson 1993).  But, when we get a better general handle on what is a priori, I think that we also have good reason to think that little of the rationality indicated by our best theories is a priori.

II. Casting About for a Promising Approach to the Subject and to What Is A Priori

Formulating our issues in terms of changing the subject seems fecund.  It calls to mind those perspectives that have been fruitful in understanding how scientists working within significantly different theories and research traditions can nevertheless talk to each other about a shared subject, and in understanding how experts and nonexperts can manage to exchange questions and information about a shared subject.  Central to our managing in these contexts is what Kitcher (1993) calls “the referential use” of a concept.  In this use, one employs a concept with a deference to reference as fixed either by a community’s expertise or by a tradition of interaction with kinds in the world. One thereby brackets or treats as negotiable most of what one antecedently thinks one knows about the things referred to.  Our concept then refers to a (perhaps ill-understood) kind (quantity, or individual) with which relevant portions of the community interact(ed).  The subject is that kind (quantity, or individual).  The payoff of this referential use—and the attendant understanding of concepts and subjects—is substantial: we can thereby keep alive multiple research traditions in a field, learning from each other. We thereby avoid becoming lodged in local epistemic-theoretical maxima.  We can appropriate results from various historical and contemporary communities, and the cumulative character of scientific work is thereby sustained.  We remain open to learning just how badly we have been mistaken about a subject.   This usage also is important for understanding that epistemic and semantic situation in which we can be lead to abandon what was once held so central in our theories about a subject.  An account that accommodates these aspects of the semantics of concepts is crucial in understanding the subject about which we theorize.

            Even so, there may be limits to the understanding that comes from focusing exclusively on the referential use.[4]   At least in many of its forms, work of this sort seems particularly unpromising ground for understanding a priori truths.  The reason is that such work understands concepts in an externalist fashion, and commonly affords the epistemic agent little access to the content of those concepts that he or she may have acquired.  Still, it seems that such referential understandings of “the subject” must be central when the issue becomes whether one has changed the subject or instead come to say deeply different things about an old subject.  

            Recent work by Chalmers and Jackson promises both to build on the sort of externalist semantics introduced by Putnam (1989, 1975) and Kripke (1972) and to do so in a way that yet finds a basis for significant a priori truths.  It aspires to identifying a less externalist core to the semantics of concepts.  While I have my doubts about certain details, and about the ultimate success of the a priorist aspirations, pursuing this program will help us improve our appraisal of the prospects for a priori truths (including any having to do with the rationality of actions and their causes).

III. The Promising Approach to A Priori Conceptual Truths

Various thought experiments lead us to appreciate that content—that which determines the referent or extension of a term—does not supervene on physical states internal to an individual agent.  We all know the story and the judgements that support this conclusion.[5]   Our judgments reflect two functions associated with our concepts.  The first is a function from possible worlds (or centered possible worlds)[6] to what has come to be called the content of a given concept:

(D)  F: Wc®Content

(D) represents the partial determination of content by elements of centered worlds that are beyond the skin of any individual speaker.  Content here is basically that which settles the referents or extension of the concept in possible worlds.  Such content is the second function associated with a term; it is a function from possible worlds to extensions or referents:

(C)  F: W®R

For illustration, Consider the English concept “water.”  C for water picks out H2O as the extension of “water” in any world:

Cwater    F: W®the H2O in each world

This function is itself determined by “water” being a rigid kind-concept, and by it picking out H2O in the actual world.  It is common to use the phrase, “watery stuff” to stand for whatever it is that makes certain stuff the referent of “water” in the actual world. Unpacking and articulating what makes for “watery stuff” is a delicate matter.  In any case, for “water”, C is determined by the fact that, in the actual world, the watery stuff is H2O.  Of course, in another possible world, the watery stuff might be something else—the XYZ of philosophical lore. What the “watery stuff” is in the actual world determines the content of “water.”  The function D (above) from possible worlds to contents (the Cs above) thus yields the actual content for “water” as its value when the value of its argument is our world (the actual world, or any H2O world for that matter).  We might summarize Dwater:

If the actual world is an H2O world, then [W®the H2O],

It the actual world is an XYZ world, then [W® the XYZ]

 

D and C provide the inspiration for distinguishing two intensions associated with a concept.[7]   One determines the extension of the concept in any possible world (considered as counterfactual)[8]C does this.  However, as Kripke, Putnam, and many others have pointed out, one can acquire and employ a concept without knowing its extension and without knowing C.

According to Jackson and Chalmers, what one knows by virtue of having acquired a concept is how the extension of that concept is fixed in the actual world.  Knowing this much, we can consider each possible world under the supposition that it is the actual world.  (Call this considering possible worlds as actual.)  Knowing how the reference of a given concept is fixed in the actual world, we then can know what the extension of the concept is in each possible world considered as actual.  Supposedly then, for any acquired concept, one knows the extension of that concept in possible worlds considered as actual by virtue of knowing how the extension of that concept is fixed in the actual world.

If this is correct, then by virtue of having acquired a concept, we know a function from possible worlds considered as actual to extensions (in those worlds).  We know this independently of any knowledge about what world is the actual world.  We can know this without knowledge of what is the actual content, C, for that concept.  Functions from worlds to extensions are termed “intensions.”  So, we know an intension.  Chalmers calls this the primary intension of the concept—as it can be known prior to knowing C, which comes to be termed the secondary intension of the concept.[9]  He writes:

The primary intension of a concept is a function from worlds to extensions reflecting the way that actual world extension is fixed.  In a given world, it picks out what the referent of the concept would be if that world turned out to be actual.  … The primary intension of a concept … specifies how reference depends on the way the external world turns out, so it does not itself depend on the way the external world turns out (Chalmers 1996, p. 57).

Put simply, the central idea is that the primary intension is that, “which fixes reference in the actual world” (p. 59).  Following Chalmers, I will here speak of the primary intension, or simply P.   Chalmer’s and Jackson’s claim is that we can know a concept’s primary intension simply by virtue of having acquired a concept.

Block and Stalnaker (forthcoming) provide useful complimentary treatment of primary intension.  They define it in terms of a kind of composite of the functions D and C, with coordinated values being assigned to the argument places:

The primary intension, like the secondary intension, takes just one argument into extensions, though in this case the argument is centered worlds.  The value of the primary intension for a given centered world as argument is defined to be the same as the value of the two-dimensional intension [the functions C and D] for the pair of arguments consisting of the centered world and the same world, uncentered (Block and Stalnaker, forthcoming).

In effect, if one knows D, then without knowledge of what world is the actual world, one can know what the extension of an understood concept is in any given possible world, on the supposition that that world is the actual world.  “Slightly more intuitively, to get the value of the primary intension on a given world, ask what the extension would be in that world considered as actual.”

Jackson and Chalmers seek to identify a class of a priori truths that can be arrived at by a kind of conceptual analysis.  The project begins with a recognition of the externalist character of the content of concepts (and of the problem it poses for traditional conceptual analysis).  The general idea in distilling out the primary intension of a concept has been to skim off the externalist element in what was traditionally conceived of as “meaning” or “content”—relegating it to the secondary intention and leaving us with something more internal.  The result is to be something that, free of externalist complications, is known or understood in acquiring the concept.  Primary intension is supposed to be just this.  Further, primary intension, as that which determines the extension of the term in any world considered as actual, holds independently of what world is the actual world.  Thus, insofar as it can be articulated, primary intention promises to provide truths that (a) are knowable solely on the basis of having acquired a concept, and (b) are knowable without any further information about what the world is like.  This is pretty much what it takes to have a priori knowledge.  Thus, primary intension is thought to provide a reformed and rehabilitated basis for what can be known a priori.

 

IV. Objections and Refinements

There is something in what Chalmers and Jackson propose.  But, it is less than they hope.  For many concepts, nothing can do quite what primary intention is advertised as doing.  The reference-fixing role assigned to it commonly cannot be served by what serves the epistemic role assigned to it.

Why does primary intension (P) seem to be the fountainhead of significant a priori knowledge?  Part of the story is that, for any concept, one can know that a formulation that correctly reflects its primary intension would be true of all in its extension in any world considered as actual. So, one can know that “a correct formulation” of it holds true in the actual world without any knowledge of what world is the actual world (See Jackson’s aspirations 1998, p. 51).

But, we must be very careful regarding just what qualifies as a priori on this line of thought.  The following would seem to be a priori: any formulation that correctly reflects what fixes the extension of a concept in worlds considered as actual, provides truths about what satisfies that concept in the actual world.  We can know this much, independent of any knowledge about which world is the actual world.

However, it is tempting to then suppose that correct formulations provide us substantive a priori knowledge about what satisfies a given concept.  Unfortunately, such high a priorist aspirations overlook something central to the a priori as an epistemological matter, and it is after all an epistemological matter.  What is a priori is what can be known by an individual epistemic agent to be true by way of competent reflection upon what is minimally required for the acquisition of the relevant concepts.  I do not think primary intension (P) can provide this.  We may know that a correct formulation of a concept's primary intension would hold of all in that concept's extension, but this avails us little in the search for significant a priori knowledge unless a correct formulation of that primary intension is appropriately accessible by virtue of having acquired the concept.  I doubt that correct formulations of primary intension generally are so accessible.

There is an important sense in which P, as that which fixes reference in worlds considered as actual, is not as cleansed of externalist elements as Jackson and Chalmers seem to think.[10]  As a result, the primary intension, P, for a given concept need not be accessible to individuals who have acquired that concept, as it  would need to be, were it to issue in a priori knowledge.  Significant knowledge of the primary intension (adequate to yielding significant a priori knowledge) need not be learned when acquiring a concept.  Recall that primary intension is to reflect a stripping away of certain externalist elements of content--for example, those determining that the concept “water” refers to H2O rather than to XYZ.  Such contributions of the world to intension are said to make for secondary intension.  Primary intension is not conditioned by the world in such a way.  Still, primary intension retains an externalist character.  Consider  Chalmer's formulation--as it seems as guarded and precise as any: primary intension is “a function from worlds to extensions reflecting the way that actual world extension is fixed” (1996, pp. 58-9).  Full and correct formulations of primary intensions must then be enough to fix extensions of concepts in the actual world, and in worlds considered as actual.[11]  Whatever the details, commonly the primary intension of a concept would seem to "reflect" (or "turn on," or "be sensitive to") the sorts of matters discussed in Kripke’s discussions of reference fixing and Putnam’s discussions of “meaning.”  Even when we are considering just how reference in the actual world is fixed, this retains a healthy externalist element.

For many concepts, a full and substantive correct formulation of primary intension commonly would make reference to such things as the historical interaction of a linguistic community with stuff having a certain crudely appreciated but consistently keyed-upon character.  It must make reference to patterns of deference across historical stages of a community and across differences in learning.  And, if the formulation--along with P itself--is to fully reflect how reference is fixed in the actual world, it must include enough on these matters to fix the distinct referents of concepts in the actual world—with H2O as the extension of “water,” CH3CH2OH as the extension of “grain alcohol,” the stuff with atomic number 79 as the extension of “gold,” and the stuff with atomic number 42 as the extension of “molybdenum,” ….   Obviously, all this includes rather a lot “outside the head or skin” of individual language users.  P retains rather a lot of externalist dross.

But, then, primary intension is not something an individual language learner, qua competent language user, must know, have acquired, or have anything approaching full cognitive access to.  I will continue to use the term ‘primary intension’, or 'P', for the function that reflects how reference is fixed in the actual world.  We have found that P cannot play the epistemic role assigned to it, for it (or a full and correct formulation of it) is not something learned by individual language users in acquiring a concept—it is not something to which one has access by virtue of having acquired a concept--and so cannot be the basis for reflection drawing only on what is thereby learned.

What is epistemically accessible in a fashion that might make for claims with an a priori status is whatever any individual must have learned in order to “hook into” the linguistic community, acquiring the concept thereby.  More will be required in connection with some concepts than with others.  But, in many cases, this will be little more than the realization that there is some pattern of referring to a kind in the community.[12]  One must recognize that individuals can often acquire concepts on the basis of very little learned information.  This (commonly paltry) informational requirement can be seen as a minimal individualistic component of primary intention, or Pi.[13]

Summarizing my misgivings: it is not P as such that is epistemically accessible—and that might make for certain claims being a priori.  Rather, it is the minimal individualistic component of primary intensionPi.  Again, for any given concept, a correct formulation of P would be true of all that satisfies that concept (in the actual world). But, correct formulations are not epistemically accessible in a way that allows one to parlay this general point into a set of a priori truths reflecting the contents of correct formulations of P for particular concepts.  Instead, for any concept, a (sometimes) very thin component of P, the minimal information learned by an individual in acquiring the concept, Pi, is all that is epistemically accessible simply by virtue of having acquired the concept.  It is correct formulations of Pi that constitute the a priori payoff of concepts.

It behooves us to get a clearer view of just what the contents of Pis are like.  Concepts vary somewhat.  We can think of general concepts as exhibiting four patterns, derived from two dimensions.

One dimension is represented by the distinction between descriptive concepts and rigid kind concepts.  For rigid kind concepts, their P and their C diverge.  Descriptive concepts are not rigid.  As a result, the P and C for such terms are almost identical (C being in effect a copy of P over the space of uncentered worlds).  Such variation in concepts matters little for our purposes here.  Our issues turn on how the Pi for a given concept are related to its P, rather than on the relation between its P and C.

Commonly, it is not required of someone acquiring a concept that they know much about what goes into fixing the reference of the concept in the actual world.  Commonly, little is required beyond what is needed for the intention to defer to a community usage that is itself poorly glimpsed.  And, when more is required, it is often rather short of the kind of knowledge that would really fix reference in the actual world.  The second dimension of variation has to do with how much or how little is required.  In high-deferential concepts, little more is required than what is minimally needed to underlay a deference to how those around me employ the concept—especially the relevant (historical and contemporary) experts (whoever they might be).  “Elm” and ‘arthritis’ provide paradigm high-deferential concepts. “Molybdenum” is another high-deferential concept.  In contrast, for low-deferential concepts, rather more is demanded of one acquiring the concept.  Certain simple functional concepts, such as “pump” seem a case in point.  “Water” is less deferential than many other substance terms.  (Obviously, this second dimension is less dichotomous than the first.)  With some simplification,[14] we can view the matter as in Figure 1.

Figure 1: Kinds of Concepts and some examples

 

 

Descriptive Concepts—for which extension in a world is fixed by the satisfaction of descriptive characterizations by things in that world.

Rigid Kind Concepts—for which the extension in all worlds is fixed by what is the extension of the concept in the actual world

High- deferential

E.g.: Concepts referring to common functional components of systems that can be talked about while very crudely understood—the "PCV value" in an car’s engine.

E.g.: "molybdenum," "ozone," "E. Coli,"

Low-

Deferential

E.g. Common functional components which are taken to wear their functional role on their face—a pump.  Many artifactual kinds: cars, and plates.

E.g.: ‘water’—where the features that allow one to fix on the referent of the concept  in the actual world are widely known, and sensitivity to these are demanded of one who as acquired the word.[15]

 

 

For a high-deferential concept, Pi carries little of the information in or about P.  To have acquired a concept such as “molybdenum” it is probably sufficient to know that it is some stuff that the relevant chemical experts (whoever they might be) manage to pick out.  It seems likely that P for molybdenum turns on the interactions of certain historical experts with a stuff—a stuff that is homogeneous in its essential microstructure.[16]  However, before knowing anything approximating the historical story "reflected in" the P for ‘molybdenum’, and before knowing the theoretical understandings that arose from the ensuing investigation, I could have acquired the concept.  Despite such ignorance, I could competently assert of my first custom bicycle frame (Milan, circa 1981) that it was made from a steel hardened with molybdenum.

For a low deferential concept, Pi carries somewhat more of the information in or about P.  To have acquired a (descriptive-kind) concept such as ‘being a pump’ one at least knows the functional principle behind something qualifying as a pump.  To have acquired the (rigid-kind) concept of  “water,” it seems that one must have learned that water is a potable liquid, and one must have learned enough to pick out salient cases in one’s local environment.[17]  This is to require that we learn enough to take a (somewhat active role or) local place in the patterned interaction with a stuff in our environment, an interaction that constitutes the P for “water.”  But, it is not to require that we have learned enough to correctly articulate the P for “water.”

In general, when high-deferential concepts are in play, there is very little that can be known a priori. As we have seen, the a priori is tied to Pi.[18] For example, with “molybdenum,” one can know that molybdenum is the stuff that the relevant experts (do or did) interact with in association with the concept (where just who the relevant experts are is commonly left to be sorted out). [19], [20] 

What constitutes a change in subject when a high-deferential concept is in use would seem to allow wild variation in what is said or thought about the (unchanged) subject.  I could claim that I have arthritis in my thigh.  I could claim that molybdenum is an acid used in the curing of steel.  In both cases, I would be quite wrong and could rightly be corrected, because in so claiming or believing I have not changed the subject.[21]

One might think that low-deferential concepts give rise to a significant body of a priori knowledge. It is true that they give rise to more substantial a priori truths than do high-deferential concepts.  But, if one comes with expectations such as Jackson’s and Chalmers’, one will be disappointed.  The crucial question is just what one must have learned in acquiring such low-deferential concepts as “water,” and “temperature?”

The Pis for typical low-deferential concepts are meager in comparison with what is Jackson and Chalmers portray as the correct and full formulations of the relevant  Ps.   At least for concepts such as “temperature” and “water,” Pi seems to have a lot to do with the ability to pick out contextually prominent samples (salient cases in local environments).  In acquiring this ability, one learns a crude phenomenology (in the scientists sense).[22]  For example, all who have acquired the term ‘temperature’ (and/or the coordinate terms for ranges in a commonsensical temperature scale—‘hot’/‘hotter’ and ‘cold’/‘colder’) thereby have acquired certain understandings of the functional-role(s) associated with ranges or changes of temperature.  For example, one probably knows that enough change from hot to cold, or in the opposite direction, can induce phase-changes in certain familiar materials.  In particular, one probably knows something of water changing from liquid to a gas, or to a solid (or the reverse).  One might know that when certain  (locally prominent) things (such as paper, or leaves) get hot enough, they burn.  One might know that when things get hot enough, they cause discomfort when touched—even injury.  And, provided the individual lives in a cold climate, he or she learns that when it is cold outside, this can cause a distinctive discomfort—even injury.

We might sum up such information as follows: what is minimally known by anyone who has acquired the concept of “temperature” is probably something like the following.

1.      Temperature is a “quantity.”  (Which is only to say that material objects can have more or less of it.)  The ones that have relatively more are warm or even hot, and those with relatively little of it are cool or even cold,

2.      Some general knowledge associated with environmentally salient, yet quite rudimentary, phenomenological regularities, and

3.      Whatever scattered information is needed enable one to pick out certain locally prominent hot things, and cold things—that is, samples of high and low temperature.

I think that this pretty much exhausts the minimal individualist component of primary intension (the Pi) for "temperature."

If this is roughly correct, then we have grounds for taking a deflationist stance with regard to some pretensions to significant a priori knowledge.  Compare my reconstruction of the Pi for "temperature" with Jackson's (1998, p. 58-9) treatment of “temperature” as a functional-role concept whose P is a matter of what fulfills “the T-role in the ideal gas laws.”  Failing to distinguish P and Pi (perhaps abetted by an over-estimation of the contents of correct formulations of P), he must conclude that the ideal gas laws are (within the relevant ranges of temperatures and volumes) a priori!  However, as we have seen, it is Pi that is epistemically relevant, and the Pi for (even a low-deferential concept such as) “temperature” is much thinner than suggested by Jackson’s presentation.

One might wonder whether there is a significant distinction between P and Pi, at least in the case of low-deferential descriptive concepts.  To this point, I have said little to question Chalmer’s and Jackson’s strong characterizations of the Ps for various concepts. Yet there is reason for concern on this point.  To begin with, there is a certain vagueness in many presentations of primary intension.  On all accounts, primary intension is that which fixes the extension of the concept in the actual world, and thus in worlds considered as actual.  Or, rather, it is a function reflecting how reference in the actual world is fixed.  Now, as I have understood primary intensions to this point, full articulate formulations of the primary intension of a given concept would provide a nicely filled in account of those aspects of things in the world that make a certain thing, quantity, or what-have-you, the referent of the relevant concept.  In this, I have sought to honor Chalmer’s and Jackson’s faith that (correct formulations of) primary intensions would contain significant substantive and interesting information.  By drawing on primary intension, we supposedly are to be able to know that water is the potable stuff filling the lakes and rivers—or something like this—and that temperature is that quantity that fulfills the T-role in the ideal gas laws.  If this falls out of knowledge of primary intension, then a correct formulation of it must include the substantive characterization of water as the dominant potable stuff in the lakes and rivers, and of temperature as that quantity that (in a mole of a gas) is related to pressure and volume according to the formula pV=KT (where K is a constant).  I have also sought to accommodate the suggestion (from Krpike and Putnam, echoed clearly in Jackson and faintly in Chalmers) that what fixes referent of some concepts is not some description of the properties of a stuff or what-have-you, but particulars regarding the historically situated interaction of (elements of a) community with a stuff, quantity, or what-have-you.  When we take this into account, while still looking for primary intensions to deliver significant a priori knowledge, we get the result that (correct and full formulations of) primary intension must include rather a lot of story regarding the historically situated interactions of a community with the stuff (or whatever) in question.   This is the line I have taken above—supposing that, if one knows the primary intension of the term, one would (have internal access to) a substantive characterization that would determine, for each world, what comprised the referent of the relevant concept.

However, a more modest understanding of the relevant Ps is possible.  On this alternative understanding, the distinction between P and Pi becomes tenuous for such concepts: One might think to treat the P for low-deferential descriptive concepts in terms of a ramsification of the crude phenomenology that I have associated with their Pi—a ramsification that is then taken with a large “grain of salt.”  (This approach is quite in keeping with the influence of Lewis () on this literature.)

On this alternative understanding of the Ps for these concepts, the distinction between P and Pi may not be of great importance, but one must then carefully resist excessive estimates of the content of the P for low-deferential functional-role concepts.[23]  On the immodest understanding of the Ps for these concepts, the distinction between P and Pi is of great importance, for Pi, and not P, is epistemically accessible as a basis for a priori truths.  In either case, in appraising the extent of the a priori, we can focus on Pi and ignore P.  For, either P is epistemically inert (as in the immodest understanding), or P is something of a reflection of Pi for low-deferential concepts (as in the modest understanding).[24]

V. Applications to Belief, Desire, and Action

Like “temperature,” “belief” and “desire” are low-deferential functional-role concepts.  Upon reflection, the parallels are striking.  First, there are parallels in what we must learn in acquiring these concepts.  One who has acquired these concepts must be able to pick out locally salient cases—although one need not be able to formulate general truths that fully characterize the functional states at issue.  While their Pi is sufficiently substantive that one who has acquired these concepts can then apply them with reasonable success to easy cases in their local environment, their Pi remains a quite limited reflection of the states referred to.  Second, these concepts are functional-role concepts.  They deal with states that are the states they are (beliefs and desires with such-and-such contents) by virtue of their role in a pattern of causal interactions.  Thus, just as “temperature” is a functional-role concept embedded in an evolving theory (or family of related theories), so the concepts of intentional psychology are embedded in an evolving family of theories or understandings.[25]  Theories provide our best, but nondefinatory, understanding of these functional roles.

It is quite plausible that the concepts of “belief” and “desire” are minimally associated with a crude phenomenology that one must appreciate in order to have acquired the concept.  To get some handle on the outlines and extent of that phenomenology, one must think in terms of what the simplest folk, contemporary and historical, are supposed to know or have known about “beliefs and desires.”   Clearly, it falls quite short of being a “ramsification” of our best theories, or even of central elements of our best theories.

 My earlier reconstruction of the Pi for “temperature” was intended as an approximation.  In the same register, I now suggest a plausible approximate formulation of the Pi for “belief” and “desire.”

First, it is common to remark on the pervasive association of these concepts with the simple rationalizing form of action explanation.  We typically explain an action A of an agent by noting that the agent desired some end, E, and believed that by A-ing, he or she could (probably) bring about E.  Such "atomic rationalizing explanations" focus on the most local and rudimentary of belief/desire causal interactions.  They are ubiquitous. Suppose someone did not so much as know that beliefs and desires interact to cause actions in the local way reflected in atomic rationalizing explanations.  We could not count such a person as having understood or acquired the concepts of belief and desire.  So, this much is a part of the Pi for these concepts.

Second, we all know that, even though an agent desires some end, E, and believes that by A-ing he or she could (probably) bring about E, yet that agent might not A.  We commonly find this explicable in terms of an overriding causal chain featuring yet other beliefs and desires.  Such explanations may refer to the desire for some end, G, stronger than the agent’s desire for E.  We all know that desires come in varying strengths, and that a strong desire can dominate another, weaker, desire with the effect of precluding some effects that weaker desire might otherwise have had.  The knowledge that desires vary in strength, and that stronger desires dominate weaker desires, seems to be a component of the Pi for “belief” and “desire.”

Third, it is common to recognize other causes for an agent failing to take an action that is indicated by certain of that agent's beliefs and desires.  For example, the agent may believe that there are more effective, or less costly, means to a given end.  To recognize this much is to recognize that sometimes agents choose among prospective actions on the basis of beliefs about the relative effectiveness of certain means to a desired end.  It is plausible that one who did not know of such interactions would not be counted as having acquired the concept of “belief.”

Finally, it seems a commonplace that one can “know” or believe something, without that belief having its characteristic causal impact on thinking and decision because one temporarily failed to recall what one nevertheless continued to know or believe.  Some distinction crudely corresponding to that between standing and occurrent beliefs and desires seems to be a commonplace.  While I am not at all confident of its status, one might at least entertain including this much in a reconstruction of the Pi for “belief” and “desire.”

Summarizing these points, we might begin a rough catalog of the components of the Pi for "belief" and "desire" with the following four items:

1.      Beliefs and desires causally interact in a fashion to produce actions in conformity to atomic rationalizing explanation,

2.      Desires vary in strength, and that stronger desires dominate—interfering with (to the point of blocking) some of the common causal effects of other (weaker) desires,

3.      Sometimes agents choose among prospective actions on the basis of beliefs about the relative effectiveness of certain actions for a desired end,

4.      An agent may have standing beliefs that do not become occurrent beliefs in a given stretch of thinking, and, as a result, such beliefs do not come to effect the outcome of that thought as they might otherwise.

Certain crude understandings having to do with the interaction of beliefs with other beliefs should be added to this list

5.      Most people are “fairly good” at seeing the “obvious” implications of their beliefs—when something makes it worth their while to think about the relevant matters and when they then call many of the appropriate beliefs to mind.  (There is variation across people here.)  (One probably knows enough to generate some concrete examples of good ways of reasoning that can be expected.  One can rate some cases for “obviousness.”)

6.      People can commonly be led to reason badly.  Some ways of thinking that would be good are just too difficult for most of us.  (There is some variation on what reasoning people find (too) difficult.)  

7.      Some internal and external contexts increase peoples’ susceptibility to bad reason.  (Again, one can probably generate a few examples.)

There are presumably a few other claims that represent components of Pi.  There are claims about peoples’ desires, how some are commonly keyed to certain biological matters, and (perhaps) how they can be influenced in certain very general ways.

         I believe that the vague and schematic claims in my catalog provide a reasonable reflection of the limited sort of knowledge that constitutes the Pi for “beliefs” and “desires.”[26]  The sketchy, schematic character of the commonsense generalizations employed above is importantly to the point.  As one employs the concepts of “belief” and “desire,” learning more about the subject in the course of those applications, one in principle could come to very different views of the family of functional systems that constitute actual believers and desirers.  On those many matters about belief and desire left undetermined by these schematic claims, one could come to believe quite differently from what we have come to believe. 

Most of us know rather a lot more about actual believers and desirers than has been included in Pi.   What more we know is not a priori, and we could have come to believe rather different things without thereby having come to investigate a different subject.  Again, this is parallel to how we have gone well beyond the Pi for “temperature.” [27]

         All of us believe much stronger things about the rationality of believers than is entailed by the Pi for “belief.”  In order to illustrate the scope for variation in belief about belief and desire, to illustrate the scope for beliefs that differ from our own more developed understandings without constituting a change in subject, I now explore just how different one’s beliefs about the rationality of belief and desire might be while remaining compatible with Pi.  To do this, I set out what I take to be an empirically false belief/desire psychology.

         In keeping with catalog items 1-4, beliefs and desires must at least interact in little clusters or eddies to cause and rationalize action in the fashion reflected in atomic rationalizing explanations.  Further, the range of occurrent beliefs and desires that might causally condition what action is preformed will commonly be wider than mentioned in any one atomic rationalizing explanation.  Multiple occurrent desires of varying strengths may come into play.  Beliefs about the relative effectiveness of differing means for satisfying particular desires may also become occurrent and operative.  But, granting all this, we might envision these processes diverging sharply from what we think is normatively appropriate.

         Pi requires that occurrent beliefs and desires settle into a decision that is controlled by the “stonger” desires.[28]  However, it says nothing about how an agent’s standing beliefs and desires become occurrent.  By what dynamic do beliefs and desires get “called into cognitive play.”  We really do not know all that much about this today—witness the recalcitrance of the frame problem—although we suppose that agents commonly manage to “call to mind” many of their beliefs and desires that would be relevant.  Of course, we have some ideas about limits to this—or systematic difficulties that folk might have in automatically activating all the standing beliefs and desires that are relevant.  But, compatible with Pi, one might believe that agents were really very bad at this.  So, the set of beliefs and desires that enter into decision-making in any one episode might be a rather small and half-hazard subset of the set of standing beliefs and desires that are contentfully relevant and should be considered.

         In keeping with Pi as represented above, we must make sense of the idea of desires with differing strengths.   But, this does not require as much as might be thought.  The formulations above only require that some scaling of desires be possible internal to the individual episodes of reasoning issuing in a choice or decision.  Just as reasoning might be inappropriately selective in the beliefs and desires that happen to feature in any one stretch of decision-making, so strength might be a property primarily of occurrent desires internal to any one episode.  That is, there is no need to suppose that standing desires are comprised in a robust (if somewhat evolving) preference structure.  Perhaps agents have a set of standing partialities and aversions that do not generally have relative strengths.  The relative strengths of occurrent partiality or aversion might then be almost wholly contextually engendered—as desires become occurrent within a decision-context they come to have relative strengths.  For example, walking certain streets in Amsterdam might induce certain episodic desires in some men that are episodically stronger than the desire for an honest relationship with the person with whom they individually share their domiciles.  Pictures of children with distended bellies may induce strong concern for world hunger, while the smells issuing from certain restaurants may weaken it for purposes of episodic processing.  Perhaps all decisions of agents are so contextually conditioned.   Recall the “semi-autonomous structures of the mind” to which Davidson appeals when attempting to make sense of weakness of the will (1980a).  The present suggestion (again not intended as an open empirical possibility) is something of the limit case of this—each episode of decision-making might comprise its own semi-autonomous structure in which occurrent desires interact according to their contextually emergent relative strengths, within episodic eddies, in combination with what contentfully relevant beliefs happen to become occurrent.

         If we were to judge the agents just envisioned using our standards of rationality, we would find them quite irrational.  We might well take advantage of their shifting episodic preference structures to systematically drain them of resources.  Perhaps this would be facilitated were we do discover ways of manipulating the contexts of decision to frame decisions in ways that manipulated those contextually emergent preference orders.  (This would provide a truly horrific variant on the framing tendencies suggested by Tversky and Kahnaman 1981—one in which the contextual or episodic frame engenders the preferences in a fashion much deeper than the mere coloring of enduring preference orders that is envisioned in that literature.)  Further, not all of such an agent's desires—aversions and affinities—would be considered in any one episode.  And not all the agent's beliefs that are relevant to evaluating the costs and benefits associated with any one course of action need be to called to mind.  One might say that such an agent never really produces an “all things considered” judgment, as the agent would never even approximate considering all the relevant matters (as gauged in terms of their own standing belief and desire states.  I believe that one would say that so inept a practical reasoner would be predominantly irrational.

         I have argued that what is a priori, what cannot be denied without changing the subject, what must be believed in order to acquire and use our concepts is fairly minimal.  It is tied to what I have termed individualist primary intention, Pi, a small component of what has been called primary intention, P.  For many concepts—high-deferential concepts such as “molybdenum,” for example—Pi is exceedingly thin: it can comprise little more than the understanding that a set of experts (whose identity is undetermined) refer to a kind of stuff, disease, or what have you, and the intenson to refer to whatever they refer to.  In other cases—low-deferential concepts such as “temperature”—Pi is more substantive.  I have presented what I take to be a plausible reconstruction of the Pi for “temperature.”  While this Pi is found to be rather more substantive that what was seen in the case of “molybdenum,” it is nevertheless quite sketchy and schematic.  It must be this sketchy if it is to represent that minimal understanding possessed by anyone who has acquired the concept, even those who come historically early, and those who are not expert.  Indeed, this limited character of the Pi of such concepts is needed when making sense of communication, regarding a shared subject, across differing research traditions--a consideration that is particularly compelling in the present context.  My general remarks on the Pi for various concepts, and my treatment of "temperature" as a useful illustration, are intended as correctives to the overreaching aspirations for significant a priori knowledge found in writers such as Chalmers and Jackson.  Finally, using "temperature" as a model for our concepts of "belief" and "desire," I develop a plausible parallel reconstruction of the Pi for these concept.  I find that, while it is a priori that agent's are rational in certain very limited respects, what is a priori here is not enough to insure that people are predominantly rational.
References

Block, N. and Stalnaker, R. Forthcoming: “Conceptual Analysis, Dualism, and the Explanatory Gap,” Philosophical Review.

BonJour, L. 1998: In Defense of Pure Reason. Cambridge: Cambridge University Press.

Chalmers, D. 1996: The Conscious Mind.  Oxford: Oxford Univ. Press.

________. Forthcoming: “The Components of Content,” Mind.

Davidson, D. 1980a: “How is Weakness of the Will Possible?” in Davidson, Essays on Actions and Events. Oxford: Clarendon Press, pp. 21-42.

________. 1980b: “Mental Events,” in Davidson, Essays on Actions and Events. Oxford: Clarendon Press, pp. 207-25.

________. 1980c: “Philosophy as Psychology,” in Davidson, Essays on Actions and Events. Oxford: Clarendon Press, pp. 229-44.

_______. 1980d: “Towards a Unified Theory of Meaning and Action,” Grazer Philosophische Studien 11: 1-12.

________. 1984a: “Belief and the Basis of Meaning.” in Davidson, Inquiries into Truth and Interpretation.  Oxford: Clarendon Press, pp. 141-54.

________. 1984b: “On the Very Idea of a Conceptual Scheme,” in Davidson, Inquiries into Truth and Interpretation.  Oxford: Clarendon Press, pp. 185-98.

Graham, G. and Horgan, T. 19xx: "Southern Fundamentalism and the Ends of Philosophy,"

Henderson, D. 1993: Interpretation and Explanation in the Human Sciences.  State University of New York Press.

________. 1994: “Conceptual Schemes after Davidson,” in Preyer, Siebelt, and Ulfig (eds.), Language and Philosophy: On Donald Davidson’s Philosophy.  Kluwer.

________. 1991: "On the Testability of Psychological Generalizations," Philosophy of Science 58, 586-606.

Jackson, P. 1998: From Metaphysics to Ethics: A Defense of Conceptual Analysis.  Oxford: Clarendon Press.

Kitcher, P. 1993: The Advancement of Science.  Oxford: Oxford Univ. Press.

Kripke, S. 1972: Naming and Necessity.  Cambridge, MA: Harvard Univ. Press.

Lewis, D. ()

Peacocke, C. 1992: A Study of Concepts. Cambridge, MA: M.I.T. Press.

Putnam, H. 1975: “The Meaning of Meaning,” in Putnam, Mind, Language, and Reality: Philosophical Papers, Vol. 2.  Cambridge: Cambridge Univ. Press, 215-71.

________. 1989: Representation and Reality.  Cambridge, MA: M.I.T. Press.

Quine, W. 1961: "Two Dogmas of Empiricism," From a Logical Point of View, 2nd ed.  Cambridge, MA: Harvard University Press.

Tverski, A. and Kahneman, D. 1981: “The Framing of Decisions and the Psychology of Choices,” in Science 221: 453-8.

 

 

 



[1] Davidson both insists that it would be distorting to think of these principles as either analytic or synthetic and then goes on to write of them as "synthetic a priori" truths.  The tensions in his presentation here reflect the difficulty of the issues.

[2] Of course, Davidson has argued that there are no incommensurable sentences.  Sentences are integrated components of a language and languages must be holistically commensurable.  However, there are apparently sentences whose interpretation or translation necessitates the non-homophonic treatment of certain of their components.  He seems to believe that the sentence, ‘People are predominantly rational in their beliefs and actions’, as understood by Davidson and his linguistic community, is so secure that its denial is unassertable (when taken homophonically).  Were someone to utter the sentence ‘People are not preponderantly rational in their beliefs and actions’, this would call for treating some of the component terms in non-homophonic fashion.  Perhaps they have certain something different in mind by talk of “rationality.”  Are the same normative standards being employed?  Would the speaker insist on their satisfaction at a similar level, or are they setting the “rationality” bar set at the same height.  If the differences did not seem to come here, Davidson presumably would conclude that the speaker was not really talking about beliefs and desires, but something else.  Perhaps the other is Stephen Stich, and he is talking about tokens in B-boxes (whatever that would be) but not beliefs. In effect, Davidson would insist that the new concepts can be captured in terms of the old language and understandings, of which the vocabulary and theory of propositional attitudes was but a part.

So, given that we are confronted by someone who holds true sentences different from those we would, when should we conclude that they are discoursing on a different subject.  What sorts of sentences matter deeply here?   What sorts do not?