Is Content-Externalism Compatible with Privileged Access?

     Externalist theories of thought-content are sometimes arrived at by reflection upon Twin-Earth thought-experiments of the sort made famous by Hilary Putnam and Tyler Burge. The conclusion many philosophers draw from these thought-experiments is that certain types of thought-contents are individuated, in part, by environmental or socio-environmental factors.¹ This doctrine of “Twin-Earth content-externalism” implies that it is possible for thinkers that are alike in all intrinsic physical respects to differ in the contents of their thoughts by virtue of differences in their environments.² Hereafter, by ‘externalism’ we shall mean specifically Putnam-type Twin-Earth content-externalism and Burge-type Twin-Earth content-externalism.

      Externalism has seemed to some philosophers to threaten our privileged self-knowledge. We are self-conscious: we have the capacity to reflect on our mental life; we can, moreover, normally know what we are thinking in a direct and authoritative way. How, these philosophers ask, can this “privileged access” obtain, if what we are thinking is fixed, in part, by environmental or socio-environmental conditions to which we lack privileged access?

     The concern is that externalism is incompatible with our having privileged access to our occurrent thoughts, that is, to our conscious mental states of thinking that P.³ There is no received formulation of the privileged access thesis in question. But the following is as strong a version of the thesis as any we have encountered:

Privileged Access. When our faculty of introspection is functioning properly, we can know what we are thinking by introspection.⁴

The concern is that if externalism were true of a certain type of occurrent thought (e.g., the occurrent thought that water is a liquid), we would be unable to know by introspection that we were having that thought (that we were thinking that water is a liquid), and so, privileged access would be false.

     There are two leading lines of argument for this incompatibilist thesis in the literature. The first appeals to so-called “Twin-Earth switching cases,” cases in which someone unknowingly switches from one content-determining environment (Earth) to another (Twin-Earth).⁵ The second appeals to the fact that introspective knowledge is, in a broad sense, a kind of a priori knowledge in that the knower’s warrant or entitlement for belief is independent of evidence provided by sense experience or perceptual belief. The second line of argument attempts to show that the conjunction of externalism and privileged access implies that we can have a priori knowledge of truths that in fact we patently cannot have a priori knowledge of, even in the broad sense of ‘a priori’ in question.⁶ Our aim is to show that neither line of argument succeeds, and thus that no case has yet been made for incompatibilism. We shall examine the first line of argument in Part 1 of this paper and the second line of argument in Part 2.

     Here is a standard Twin-Earth switching case.⁷ Oscar, without his knowledge, has been traveling back and forth between Earth and Twin-Earth.⁸ With each move, he stays long enough to acquire the concepts of the locals. So, when he utters the sentence “Water is a liquid” on Twin-Earth, he comes to express the thought that twater is a liquid, just as the members of the indigenous population do. Oscar is, however, unaware that such shifts in his thought and speech occur.

     The difficulty this switching case is alleged to present for compatibilism is the following.⁹ Suppose that on a particular occasion while residing on Earth, Oscar occurrently thinks that water is a liquid. Given his travels, that he is thinking that twater is a liquid is a relevant alternative to his thinking that water is a liquid.¹⁰ The introspective evidence available to Oscar is compatible with its being the case that he is thinking that twater is a liquid; and, so, his introspective evidence does not exclude that relevant alternative. Thus, Oscar cannot know by introspection that he is thinking that water is a liquid.¹¹

     Notice that even if this argument is sound, it does not establish that if externalism is true, then privileged access is false. For the privileged access thesis allows that one may be unable to know what one is thinking by introspection if one’s faculty of introspection is not functioning properly. It would, however, be a truly desperate, ad hoc, move for compatibilists to respond to the above line of argument by claiming that Oscar’s faculty of introspection would be malfunctioning. Were that the only response available to compatibilists, the compatibilist position would be quite implausible. Compatibilists and noncompatibilists alike should agree that switching from residency in the English speaking environment to residency in the Twin-English speaking environment and then switching back again would not automatically result in the malfunctioning of introspection.

     Given the shared assumption that traveling Oscar’s faculty of introspection is functioning properly, the switching argument can be formulated as a reductio ad absurdum of the compatibilism thesis. For given the privileged access thesis, the assumption that Oscar’s faculty of introspection is functioning properly will imply that he is able to know by introspection that he is thinking that water is a liquid. Combining this reasoning with the reasoning stated above a contradiction follows: Oscar can and cannot know by introspection that he is thinking that water is a liquid. Proponents of switching arguments can thus maintain that, by reductio, we should reject either the privileged access thesis or the externalist thesis that implies that the content of Oscar’s thought shifts with his travels.

     As we shall see, there are, however, other assumptions at work in switching arguments that one might reject instead of privileged access or externalism. We shall argue that switching arguments fail to establish incompatibilism. But we shall also argue for a further claim: switching arguments fail even to show that externalism poses a prima facie threat to privileged access. More precisely, we shall argue that switching arguments fail even to show that there is a prima facie threat to privileged access that is both implied by externalism and such that it would be removed were externalism false. We shall develop and defend this approach by comparing and contrasting it with a recent compatibilist response to switching arguments.

II. Falvey and Owens’s Response

     As Falvey and Owens (1994) point out in defense of compatibilism, the following is a tacit assumption of switching arguments:

(RA) If (i) q is a relevant alternative to p, and (ii) S’s belief that p is based on evidence that is compatible with its being the case that q, then S does not know that p. (p. 116)

     Now principle (RA) has intuitive appeal. The principle does not require for knowledge that P that the evidence on which a belief that P is based be incompatible with all alternatives to P; it requires only that the evidence be incompatible with all relevant alternatives to P. Thus, the principle is, for instance, compatible with one’s knowing that there is a barn across the road, even though one’s belief is based on evidence that is compatible with its being the case that there is a papier-mache’ barn facsimile across the road. For that there is a papier-mache’ barn facsimile across the road may fail to be a relevant alternative. If, however, one is in phoney barn country and, so, it is a relevant alternative that there is a papier-mache’ barn facsimile across the road, then (RA) counts one as not knowing that there is a barn across the road, even if there is a barn across the road, if one’s evidence fails to rule out that relevant alternative. This result has intuitive appeal.

     Falvey and Owens readily acknowledge the intuitive plausibility of (RA). Indeed, they go so far as to say that “(RA) surely states a plausible necessary condition for a subject’s belief to constitute knowledge, where the belief is a belief concerning the external world, such as a perceptual belief. Indeed, it is this principle that appears to be operative in our inclination to say of Tom that he does not know that the object he is looking at is a barn if the area is full of papier-mache’ facsimiles [and his evidence does not exclude the alternative that he is looking at a papier-mache’ facsimile]” (1994, p.116). However, they maintain that introspective knowledge yields an exception to (RA).

     They defend this response by maintaining that “the plausibility of (RA) is grounded in a more basic principle,” (p.116) namely

(RA’) If (i) q is a relevant alternative to p, and (ii) S’s justification for his belief that p is such that if q were true, then S would still believe that p, then S does not know that p. (p. 116)

     Falvey and Owens may well be right that compatibilists can embrace (RA’) since its antecedent will not be automatically satisfied when the thinker is a traveler. Indeed, setting aside cases of causal preemption and the like, if Oscar’s faculty of introspective is functioning properly, then his introspective belief that he is thinking that water is a liquid won’t satisfy the antecedent of (RA’). And, as we noted earlier, there seems to be no reason to think that shifts in linguistic environments of the sort Oscar undergoes will automatically result in the malfunctioning of introspection. If, then, the antecedent of (RA’) is not satisfied in the Oscar case, proponents of switching arguments cannot appeal to (RA’) to argue that Oscar cannot know by introspection that he is thinking that water is a liquid.

     Switching arguments, however, do not appeal to (RA’); they appeal to (RA). Falvey and Owens claim that cases of introspective knowledge are exception cases to (RA) and defend this response by claiming that the intuitive plausibility of (RA) is grounded in (RA’) and that introspective knowledge does not violate (RA’). But a proponent of (RA) will reject the claim that its intuitive plausibility is grounded in (RA’). Principles (RA’) and (RA) are compatible. Setting aside cases of causal preemption and the like, proponents of (RA) can happily embrace (RA’). They will, however, maintain that there are conditions on the evidence required for knowledge about which (RA’) is silent. In cases in which S would not believe that P, were relevant alternative Q true, (RA’) is consistent with S knowing that P on the basis of any evidence whatsoever. Proponents of (RA) will maintain that there are cases about which (RA’) is silent that (RA) correctly rules out as cases of knowledge on the grounds that the evidence for belief is inadequate for knowledge.¹⁵ An alternative principle to (RA’) stated by an anonymous referee is likewise silent about this case. The principle is that an epistemic agent’s incapacity to rule out a relevant alternative to P on the basis of his evidence E blocks him from knowing that P on the basis of E if, under the circumstances, the incapacity makes it false that if P were not the case, the subject would not believe that P. In the scenario in question, the relevant counterfactual is true: were Oscar to have no brain, he would not believe that he has a brain. So, Oscar’s incapacity to rule out relevant alternatives to his having a brain on the basis of his evidence does not make the counterfactual false. Thus, the principle is silent about whether Oscar knows that he has a brain. Indeed, the principle will be silent about any case in which S’s belief that P is counterfactually dependent on P. But a proponent of (RA) will insist that not all such cases are cases of knowledge that P and that (RA) explains why.

     Of course, even if (RA) yields the right results in some cases about which (RA’) is silent, that would by no means vindicate (RA). For (RA) may yield the wrong result in other cases, and thus be false. Moreover, even if there were no compelling counterexamples to (RA), general theoretical considerations of overall coherence and simplicity may favor rejecting the principle. We shall not pursue the issue of whether (RA) is true, however. For, as will become apparent in due course, it is not at all our concern to defend or to reject (RA). We hold that whether compatibilism is true does not turn on whether (RA) is true.

     Falvey and Owens succeed in pointing out one option available to compatibilists in response to switching arguments. Compatibilists can reject (RA). If (RA) is indeed false, then switching arguments fail to establish incompatibilism. In defense of compatibilism, however, it is important to see that rejecting (RA) is not the only option available to compatibilists. Compatibilists can deny that switching examples can be used to show that if (RA) is true, then incompatibilism is true. Compatibilists need not follow Falvey and Owens in maintaining that introspective knowledge is an exception to (RA). For compatibilists need not follow them in conceding to proponents of switching arguments that the antecedent of (RA) is satisfied in switching cases. It is open to compatibilists to deny that the antecedent of (RA) is satisfied in such cases. If such a denial is defensible, then proponents of (RA) are not required to be incompatibilists.

     Whether the best option for compatibilists is to reject (RA) or, instead, to deny that its antecedent is satisfied in switching cases turns on issues that any proponent of privileged access must face, whatever her views about content. The issues are whether and, if so, in what sense, introspective beliefs are based on evidence and what that evidence is.¹⁶ Externalism is silent on these issues. As we shall see in due course, whatever resolution of these issues a proponent of privileged access can embrace that is consistent with (RA), a proponent of privileged access who accepts externalism can embrace as well. Whether there is a tenable resolution of these issues that renders privileged access compatible with (RA) by no means turns on whether externalism is true.

     Falvey and Owens in effect concede to proponents of switching arguments that switching cases can be used to show that externalism implies a prima facie threat to privileged access, namely that if (RA) is true, then privileged access is false; they then argue that the threat is a false one since (RA) is false. Our central aim is to make a case that switching examples fail to show that there is any prima facie threat to privileged access —even a false one— that is posed by externalism. Let us turn to some further assumptions at work in switching arguments.

     In addition to (RA), switching arguments also assume the following thesis:

The Introspective Evidence Thesis. Introspective knowledge of what we are occurrently thinking is based on evidence that we can introspect.

The Alternative Thoughts Thesis. For at least one type of occurrent thought that P, there is some possible circumstance in which one is occurrently thinking that P, one’s faculty of introspection is functioning properly, and the evidence that one can introspect is compatible with its being the case that one is thinking some relevant alternative thought that Q.

     Now, to be sure, if switching examples succeed in showing that externalism implies the alternative thoughts thesis, then externalism indeed implies a prima facie threat to privileged access, namely that if (RA) and the introspective evidence thesis are true, then privileged access is false. Compatibilists would, then, have to reject either (RA) or the introspective evidence thesis.

     Notice, however, that the reason that externalism would imply this prima facie threat if it implied the alternative thoughts thesis is that the alternative thoughts thesis, principle (RA), and the introspective evidence thesis jointly imply that the privileged access thesis is false. To see this, suppose that the thought that P is an instance of the alternative thoughts thesis. Then, there is some possible circumstance in which one is occurrently thinking that P, one’s faculty of introspection is intact, but the evidence that one can introspect is compatible with one’s occurrently thinking some relevant alternative thought that Q. Given the introspective evidence thesis, if one knows by introspection that one is thinking that P, one’s knowledge is based on the evidence one can introspect. According to (RA), since one’s introspective evidence is compatible with one’s thinking the relevant alternative thought that Q, one cannot know that one is thinking that P on the basis of that evidence. Consequently, there is a possible circumstance in which in one’s faculty of introspection is intact, but one cannot know what one is thinking by introspection. It follows that privileged access is false.

     We see, then, that principle (RA), the introspective evidence thesis, the alternative thoughts thesis, and privileged access form an incompatible quadrad. It follows that no one can consistently hold all four theses. A point we wish to underscore is that it also follows that proponents of privileged access, whatever their views about the nature of content —whether they are externalists, internalists, or neutral on the issue— must reject either (RA), the introspective evidence thesis, or the alternative thoughts thesis.

     Privileged access indeed faces the following prima facie threat: if (RA), the introspective evidence thesis, and the alternative thoughts thesis are all true, then privileged access is false. That threat is, however, not posed by externalism since it would not be removed were externalism false. The prima facie threat remains whether or not externalism is true. Proponents of privileged access, whatever their views about content, must maintain that this prima facie threat to privileged access is a false one. They must maintain that either (RA), the introspective evidence thesis, or the alternative thoughts thesis is false.

     We shall not attempt to determine which of these three theses a proponent of privileged access should reject. As we have repeatedly noted, our concern is to make a case that switching examples fail to show that there is some prima facie threat to privileged access that is both implied by externalism and such that it would be removed were externalism false. It is our contention that switching cases fail to show that in embracing externalism, a proponent of privileged access would narrow her options in relation to (RA), the introspective evidence thesis, and the alternative thoughts thesis. For all switching cases show, the same options would remain open. Any two of these three theses that a proponent of privileged access who rejects externalism can accept, a proponent of privileged access who embraces externalism can accept as well. Switching cases fail to show that the tenable options available to a proponent of privileged access who embraces externalism are anything other than exactly those available to a proponent of privileged access who rejects externalism.

     It is clear that the alternative thoughts thesis poses the prima facie threat that if (RA) and the introspective evidence thesis are true, then privileged access is false. Proponents of privileged access who embrace the alternative thoughts thesis must reject either (RA) or the introspective evidence thesis. We shall argue, however, that whether the alternative thoughts thesis is true turns on issues that are orthogonal to whether externalism is true.

     Switching cases succeed in showing that externalism implies the alternative thoughts thesis only if a certain assumption about introspective evidence holds. The assumption is this:

The Underdetermination Thesis. For at least one type of occurrent thought that P, whether one is occurrently thinking that P (at t) fails to supervene on the evidence introspectively available to one (at t).¹⁷

     Let us ask, then, what introspective evidence is available to someone having an occurrent thought. Falvey and Owens (1994, p.111) follow proponents of switching arguments in speaking of the introspective evidence as qualitative mental states. In Oscar’s case, the qualitative mental state is supposed to be common to his occurrently thinking, while on Earth, that water is a liquid, and to his occurrently thinking, while on Twin-Earth, that twater is a liquid.

     These philosophers seem to have in mind something, roughly, along the following lines: one’s introspective evidence that one is thinking that P is that one is having an auditory image with certain qualitative features. What are the qualitative features in question? Presumably, they include imagistic phonological and stress features. Whether they include syntactic features depends on whether the qualitative features are supposed to be intrinsic; for syntactic features are arguably not intrinsic features of the auditory image. But, in any case, let us count syntactic features among the introspectible qualities of the image. If the fact that he is having an image with certain imagistic phonological, stress, and syntactic features exhausts the introspective evidence available to traveling Oscar when he thinks that water is a liquid, then, to be sure, the introspective evidence available to him fails to rule out the relevant alternative that he is thinking that twater is a liquid. Thus, if the evidence in question exhausts the introspective evidence available to Oscar, then there is no question that externalism implies the alternative thoughts thesis.

     The first point we wish to stress, however, is that if Oscar’s introspective evidence consists of his having an image with these imagistic phonological, stress, and syntactic features, then, on any theory of thought-content worthy of consideration, the underdetermination thesis will be true. For on any theory of content worthy of consideration, Oscar’s thinking that water is a liquid will fail to supervene on the introspective evidence available to him. The reason is that any theory of content worthy of consideration will be “externalist” in at least this very weak sense: it will imply that the content of an auditory image state will not supervene on the qualitative features in question. Thus, even an internalist, narrow conceptual role theory of thought-content will imply the underdetermination thesis. Given this view of introspective evidence, the underdetermination thesis will be true on any sensible theory of content.

     The second point we wish to underscore is that given this view of what evidence is introspectively available to Oscar, on any theory of content worthy of consideration, we will be able to describe switching cases: cases in which Oscar switches from one content-determining condition to another for the type of auditory image in question (i.e., one with the imagistic phonological, stress, and syntactic features in question). For, according to any theory that implies that the content of the auditory image state fails to supervene on these qualitative features, whether Oscar is thinking that water is a liquid will fail to supervene on the introspective evidence, even when his faculty of introspection is functioning properly. Hence, there will be relevant alternative thoughts that switching Oscar’s introspective evidence will fail to rule out. Given this view of introspective evidence, then, any reasonable theory of content will imply the alternative thoughts thesis. For any such theory will imply, then, that the thought that water is a liquid is an instance of that thesis.

     We see that given this view of introspective evidence, any sensible theory of content will imply that if (RA) and the introspective evidence thesis are true, then privileged access is false. Thus, if the introspective evidence available to Oscar consists of his having an image with the qualitative features in question, then proponents of privileged access —whether they are internalists, externalists or neutral on the issue — should reject either (RA) or the introspective evidence thesis.

     Suppose that the introspective evidence available to a thinker is richer than the qualitative evidence discussed above. However rich the evidence is, if it is not incompatible with all alternative thoughts, then the underdetermination thesis will be true: it will be possible for two thinkers, T1 and T2, to have exactly the same introspective evidence, and yet for T1 to be thinking that P, and T2 to be thinking an alternative thought that Q. Moreover, switching cases will be possible: it will be possible for T1 to switch from the condition that determines that he is thinking that P to a condition that determines that he is thinking that Q, without any shift in the introspective evidence available to him. Such a switching case would fail to show that the alternative thought thesis is true only if the fact that T1 has switched would require that his faculty of introspection be impaired. But we see no reason why that would be so even on a non-externalist theory of content-determining conditions that implies that T1’s faculty of introspection would have to be impaired during the course of switching. For the alternative thoughts could remain relevant even after switching has ceased, and so even after T1’s faculty of introspective has ceased to impaired. Moreover, on standard views of what makes an alternative a relevant alternative, T1 need not actually be a switcher in order for the thought that Q to be a relevant alternative thought. It is enough that T1 is in the vicinity of switchers and, by luck, has escaped being one. Indeed, on some views of relevant alternatives, it is enough that there are credible rumors of switching, even if T1 is ignorant of them and no one is in a fact a switcher. Thus, on some fairly uncontroversial assumptions about relevant alternatives, if the underdetermination thesis is true, then the alternative thoughts thesis will be as well. If the introspective evidence available to a thinker falls short of ruling out all alternative thoughts, then proponents of privileged access should reject either (RA) or the introspective evidence thesis whether they are internalists, externalists or neutral on the issue.

     Need proponents of privileged access accept that the introspective evidence available to a thinker is compatible with some alternative thoughts? As we have seen, if Oscar’s introspective evidence consists in his having a mental state with the qualitative features mentioned earlier, then there is no question that his evidence will fail to rule out all relevant alternatives. Moreover, as we have also seen, whatever mental state is Oscar’s introspective evidence for his belief that he is thinking that water is a liquid, if his being in that mental state is compatible with his having some alternative thought, then, on any sensible theory of content, the thought that water is a liquid will be an instance of the alternative thoughts thesis. However, proponents of privileged access need not follow Falvey and Owens in conceding to advocates of switching arguments that the introspective evidence available to someone having an occurrent thought is a qualitative mental state.

     Consider again the introspective evidence thesis. According to that thesis, introspective knowledge of what we are occurrently thinking is based on evidence that we can introspect. One might take the evidence on which introspective beliefs are based to consist in introspective beliefs that provide evidential reasons. Intuitively, however, one’s evidence for believing that one is thinking that P does not consist in further beliefs at all. Intuitively, one has introspective access to certain mental states, and these states are one’s evidence for one’s introspective beliefs without providing a propositional justification for those beliefs. On this reliabilist view of introspective knowledge, introspective beliefs are warranted because of their causal ancestry, which will include the introspected mental states in question. The introspective evidence thesis enjoys intuitive support, we maintain, on the reliabilist reading of ‘evidence’ and not on the reading that construes evidence as beliefs that are evidential reasons.¹⁸

     It is open to advocates of a reliabilist account of introspection to maintain that the mental states that are a thinker’s introspective evidence in such cases are the occurrent thoughts the introspective beliefs are about. A proponent of privileged access can thus reject the assumption that Oscar has introspective access to his occurrent thought that water is a liquid only via his introspective access to a qualitative mental state of the sort described earlier. A proponent can maintain that Oscar has direct introspective access to the occurrent thought itself. If this view of his introspective evidence is correct, then Oscar’s introspective belief that he is thinking that water is a liquid is based on conclusive evidence, namely on his state of occurrently thinking that water is a liquid. His introspective evidence will thus exclude all relevant alternative thoughts since it will exclude all alternative thoughts. On this view of the introspective evidence, proponents of privileged access —whatever their views about content— can embrace both (RA) and the introspective evidence thesis. For the underdetermination thesis is false, and, so, the alternative thoughts thesis is as well.¹⁹ The fact that one’s belief that one is thinking that P strongly supervenes on the introspective evidence does not entail that one’s belief is self-evident or self-warranting. We happily concede that no belief is self-evident, that is, that no belief is an evidential reason for itself. On the reliabilist view we favor, the occurrent thought that P (and not the belief that one is thinking that P) is one’s evidence for the belief that one is thinking that P. The occurrent thought is the evidence for the belief without being an evidential reason for it.

     On the above reliabilist account of introspection, states of thinking that P are, in the first-person case, self-presenting: we are not introspectively aware of occurrently thinking that P by means of being introspectively aware of some distinct mode of presentation of our occurrently thinking that P; rather, we are directly introspectively aware of the thought itself as such.²⁰ In this way, our introspective access to our occurrent thoughts differs from our perceptual access to our environment. We have visual access to the scene before our eyes by having a visual experience that is (appropriately) caused by the scene. In contrast, we do not have introspective access to our occurrent thoughts by having experiences that are (appropriately) caused by the thoughts. Our introspective access to the thoughts is direct.²¹ We do not experience an occurrent thought by having an experience of the thought. We experience an occurrent thought by having the thought. Occurrent thoughts are themselves a kind of conscious state that we can directly introspect.²²

On this view of introspective knowledge, the concept of a thought that P is, in its first-person present-tense application, a recognitional concept.²³ Those who have mastered the concept can introspectively recognize occurrent thoughts that P as thoughts that P, rather than having to infer that they are thoughts that P from other of their features or from their effects.²⁴ In cases involving Cogito thoughts, there is a conscious act of recognition. But in the typical case, one’s recognition of what one is occurrently thinking does not involve a conscious act. One can recognize that one is thinking that water is a liquid, when the only occurrent thought one is having is that water is a liquid.

     We see, then, that on this view of introspection, the occurrent thought that P both reliably causes and selectively, conclusively confirms the introspective belief that one is thinking that P. If this view of introspection is tenable, then proponents of privileged access can hold on to both (RA) and the introspective evidence thesis and reject the alternative thoughts thesis on the grounds that the underdetermination thesis is false.

     Twin-Earth switching examples provide no reason to think that this view of introspection is unavailable to externalists. For to appeal to such switching examples to argue that externalist thoughts are instances of the alternative thoughts thesis, one must assume that such thoughts are not self-presenting. Compatibilists can reject that assumption. One option open to compatibilists, then, is to embrace this view of introspection and to embrace, as well, both (RA) and the introspective evidence thesis.

     It is beyond the scope of this paper to defend a theory of introspective knowledge. Our main concern in Part 1 has been just to defend compatibilism against switching arguments. Compatibilists cannot, of course, accept a view of introspective knowledge that entails the introspective evidence thesis, (RA), and the alternative thoughts thesis. But no proponent of privileged access can. For (RA), the introspective evidence thesis, the alternative thoughts thesis, and privileged access form an incompatible quadrad. Our main claim is that switching cases fail to show that in embracing externalism, a proponent of privileged access would narrow her options in relation to these three theses. For all switching cases show, the tenable options available to proponents of privileged access who embrace externalism are exactly those available to proponents of privileged access who reject externalism. For this reason we contend that switching cases fail to demonstrate that there is a prima facie threat to privileged access —even a false one— that is both implied by externalism and such that it would be removed if externalism were false.

     We turn now to the second leading line of argument that externalism is incompatible with privileged access.

     Michael McKinsey (1991) has described a recipe for making a case that externalist theses are incompatible with privileged access. Suppose an externalist thesis implies that certain thoughts are individuated, at least in part, by environmental factors. Let the thought that-P be a thought of the type in question. McKinsey’s recipe for trying to show that the externalist thesis is incompatible with privileged access is essentially the following: find some E such that (i) E cannot be known a priori, yet (ii) the externalist thesis implies that it is a priori true that if one is thinking that P, then E. According to McKinsey, if such an E can be found, it can be successfully argued that either one cannot know a priori that one is thinking that P or it is not a priori true that if one is thinking that P, then E, from which it may be inferred that the version of externalism is incompatible with privileged access.²⁵

     As will become apparent, we have no quarrel with McKinsey’s recipe. Any externalist thesis that implies that it is a priori true that if one is thinking that P, then E, where E cannot be known a priori, is indeed incompatible with privileged access. Our position is just that no externalist theses supported by Twin-Earth thought-experiments alone will serve as an ingredient for his recipe. For no such externalist thesis has as a consequence that it is a priori knowable that if one is thinking that P, then E, where E is not itself knowable a priori.

     In what remains of this paper, we shall first consider a version of externalism that Colin McGinn maintains is supported by Putnam’s Twin-Earth thought-experiment. McGinn’s version deserves special attention since it is, we believe, the strongest externalist thesis any externalist has claimed to be supported by that thought-experiment. We shall argue that a McKinsey-type argument cannot be used to show that McGinn’s externalist thesis is incompatible with privileged access. Then, we shall argue the same for the version of externalism that Burge claims is supported by his own Twin-Earth thought-experiment. After that, we shall respond to a charge made by Jessica Brown (1995) that Burge is committed to an externalist thesis that is incompatible with privileged access. We shall conclude by commenting on the prospects for employing Twin-Earth content-externalism to provide an a priori refutation of solipsism.

     McGinn (1989, pp. 30-36 and pp. 47-48) maintains the following thesis:

(M) If the concept of K is an atomic, natural kind concept, then one possesses it only if one has causally interacted with instances of K.

     Before considering whether this thesis is subject to a McKinsey-type argument, we should note that McGinn explicitly rejects the stronger thesis that if the concept of K is a natural kind concept, then one possesses it only if one has causally interacted with instances of K. McGinn maintains that while the concept of H20 is a natural kind concept, it is not necessary to have causally interacted with any instances of H20 to possess it. Possession of the concept does not even require that there be any H20 (McGinn 1989, p. 35). The concept of H20 is not atomic; for an atomic concept lacks conceptual constituents. The concept of H2O is a descriptive concept, and thus molecular. McGinn asks us to imagine a possible world that contains hydrogen and oxygen, yet no H20 since hydrogen and oxygen are scarce and widely separated. A scientist in such a world who has the concepts of hydrogen, of oxygen, and of bonding could develop a concept of H20 by theorizing (correctly, let us suppose) that hydrogen can bond with oxygen in the combination H2O.

     While McGinn rejects the stronger thesis in question for the reason just noted, he, nevertheless, embraces, as we said, the weaker thesis (M). Atomic natural kind concepts, he maintains (1989, p. 35), cannot lack an extension. McGinn explicitly claims that the concept of water is an atomic natural kind concept. Thus, he holds that one cannot think that water is a liquid unless one has causally interacted with instances of water.

     How might McKinsey’s recipe be used to argue that McGinn’s version of externalism is incompatible with privileged access? It might be thought that given (M), and the fact that the concept of water is an atomic, natural kind concept, it follows that one can know a priori that one is thinking that water is a liquid only if one can know a priori that one has causally interacted with water. It is plainly false that one can know a priori that one has causally interacted with water; indeed, it is plainly false that one can know a priori even that there is water. Hence, either privileged access fails or (M) is false.

     This line of reasoning is invalid. It is permissible to infer from (M) that one can know a priori that one has causally interacted with instances of water only if it is assumed further that it is a priori knowable that the concept of water is an atomic, natural kind concept. We think it is an open question whether it can always be known a priori whether a concept is atomic. But, in any case, on McGinn’s intended notion of a natural kind concept, we certainly cannot always or even typically know a priori that a concept is a natural kind concept. For on his notion, the concept of K is a natural kind concept only if K is a natural kind; and that some K is a natural kind is typically not something that we can know a priori. In particular, while water is a natural kind, we cannot know that a priori. It is epistemically possible that water might have turned out to be like air, or like jade. We could conceivably discover, for example, that there actually is no single sort of substance to which the concept of water applies. Indeed, it seems epistemically possible that the concept of water could turn out to be like the concept of phlogiston. That is enormously unlikely, of course, but it is epistemically possible. Thus, even if our concept of water is indeed an atomic, natural kind concept, we cannot know a priori that it is.

     Even if, as McGinn holds, (M) is a priori, it is not a priori knowable that this version of externalism is true of the concept of water. Twin-Earth thought experiments involving the concept of water rely on the empirical assumption that water is a natural kind. Such thought experiments are not intended to show that the concept of water is a natural kind concept. Rather, they presuppose that it is. The point of such thought experiments is establish a general thesis along the general lines of (M).²⁷ Consequently, even given the a prioricity of (M), the fact that we can know a priori that we are thinking that water is a liquid does not imply that we can know a priori that we have causally interacted with water or indeed even that water exists.²⁸

     The central point here parallels one that is often made about so-called object-dependent thoughts. Object-dependent thoughts are ones that involve the exercise of a concept that purports singularity of application and is such that there is in fact some actual (contingent) individual to which the concept singularly applies. Such thoughts are individuated in part by external factors, namely by the existence of the individual in question. Thus, suppose that the thought that Cicero is an orator is an object-dependent thought because the concept of Cicero is in fact a singular concept of an actual individual, namely Cicero. Then, in any possible world, w, someone can think that Cicero is an orator in w only if Cicero exists in w. Obviously, one cannot know a priori that Cicero exists; it is epistemically possible that Cicero is a fictional character. It does not follow, however, that one cannot know a priori that one is thinking that Cicero is an orator. What the theory of object-dependent thoughts implies is that if Cicero actually exists, then in any possible world, w, one can think that Cicero is an orator in w only if Cicero exists in w. Since one cannot know a priori that Cicero actually exists one cannot know a priori that there is some actual individual, Cicero, such that one’s concept is a concept of that individual. If the thought that Cicero is an orator is an object-dependent thought, then that is an a posteriori fact about it. One can know a priori that one is thinking that Cicero is an orator. What one cannot know a priori is that one’s thought that Cicero is an orator is an object-dependent thought; for one cannot know a priori that the concept of Cicero succeeds in picking out an actual individual. Similarly, while one can know a priori that one is thinking that water is a liquid, one cannot know a priori that one is having a thought involving a natural kind concept; for one cannot know a priori that the concept of water succeeds in picking out a natural kind. If (M) is true and the concept of water is indeed an atomic, natural kind concept, then the thought that water is a liquid is a kind-dependent thought. However, that the thought is a kind-dependent is an a posteriori fact about it, not an a priori one.

     The strongest version of externalism that is alleged to follow from Burge’s own (1979) Twin-Earth thought-experiment seems to us to be straightforwardly compatible with privileged access. Burge’s thought-experiment is intended to show that the following thesis is a conceptual truth:

(B) If one exercises a deferential concept in thinking that P, then if one is a member of a linguistic community, the content of one’s thought that P is individuated partly by socio-environmental factors.

     Jessica Brown (1995) charges that Burge is, nonetheless, committed to a version of externalism that satisfies McKinsey’s recipe.³¹ This charge is made as part of an ad hominem argument against Burge, who explicitly endorses a privileged access thesis. Thus, it is argued that he holds two incompatible theses. We will first present her case and then argue that she fails to show that Burge holds an inconsistent position.

     Burge (1982) maintains that it is possible to have water-thoughts without the existence of water. He asks us to imagine a world without water in which an entire community is under the illusion that there is a certain liquid, which they intend to call ‘water’, with such-and-such phenomenal properties (the phenomenal properties of water) that fills the oceans and lakes, that often falls from the sky, that can freeze solid or be heated into a gaseous state, and so on.³² Scientists in this waterless world have theorized that hydrogen can bond with oxygen in the form of H2O.³³ They have, moreover, (mistakenly) theorized that the (illusory) liquid that fills their oceans and lakes, etc. is H20. Burge claims that these scientists count as having the concept of water. He further maintains that even if Adam, a member of that community, is entirely ignorant of chemical theory, and lacks the concept of H20, he may nevertheless possess the concept of water in virtue of his linguistic, deferential relationship to these scientists. Burge claims, however, that if there are no other people in Adam’s world, and thus no such scientists, and if Adam is ignorant of chemical theory and agnostic about the application conditions for his word ‘water’, then it

seems incredible...to suppose that Adam, in his relative ignorance and indifference about the nature of water, holds beliefs whose contents involve the notion, even though neither water nor communal cohorts exist. (1982, p.116)

     Appealing to the above passage, Brown (1995) maintains that Burge is committed to the following thesis:

(1) Necessarily, if x has a thought involving the concept of a natural kind k and x is agnostic about the application conditions of the concept of k, then either x is in an environment which contains k, or x is part of a community with the concept k. (p.152)

     Now Brown does not attempt to argue that (1) itself commits Burge to denying privileged access. For she recognizes that one cannot know a priori whether a given concept is a natural kind concept. Instead, her strategy is to argue that Burge is also committed to a second thesis which, in conjunction with (1), commits him to denying privileged access. The second thesis is this:

(2) Necessarily, if x has a thought involving a non-natural kind concept, c, and x is agnostic about the application conditions of c, then x is part of a community which has the concept c.

     Brown points out that (1) and (2) jointly imply the following thesis:

(3) Necessarily, if x has a thought involving a concept c, and x is agnostic about the application conditions of c, then either x is in an environment which contains instances of c and c is a natural kind concept, or x is part of a community which has the concept c, whether or not c is a natural kind concept. (1995, p.155)

     Brown maintains that Burge is committed to (3) because he is committed to (1) and (2). We agree with Brown that he is committed to (1) but deny that he is committed to (3). The crucial question, then, is whether Burge is committed to (2). Here is what Brown says in justification of her attribution of (2) to Burge:

     Burge’s argument for the claim that [Adam, who, you will recall, is ignorant of the nature of water] could not have a water thought unless water or other speakers exist, seems applicable to thoughts involving non-natural kind concepts. Imagine that [Adam] is agnostic about the application of the word, ‘sofa’. For example, he may apply it firmly and correctly to what we call ‘sofas’, but be unsure about whether it also applies to broad single seat armchairs. According to Burge, if [Adam] is part of an English speaking community then, despite his agnosticism, he has thoughts involving the concept sofa. But if, counterfactually, [Adam] had been part of a community in which ‘sofa’ is applied both to what we call ‘sofas’ and to broad single seat armchairs, then [Adam] would have had chofa thoughts, where the concept of a chofa applies both to what we call ‘sofas’ and to broad single seat armchairs. Now imagine that there are no other speakers in [Adam’s] environment. How could [Adam] have propositional attitudes involving the concept of sofa? Since sofa is not a natural kind concept, [Adam’s] natural environment cannot help him to acquire the concept. There are no other speakers. Nothing seems to show that his attitudes involve the concept of sofa as opposed to chofa. (1995, pp. 153-154)

     We agree with Brown that, in the scenario she describes, one can argue on Burge-type grounds that “nothing seems to show that [Adam’s] attitudes involve the concept of sofa as opposed to chofa.” Indeed, there seems to be nothing that makes it the case that Adam’s attitudes involve the concept of sofa, and nothing that makes it the case that his attitudes involve the concept of chofa. It seems that Adam lacks both concepts. But in inferring (2), Brown seems to be tacitly assuming that Burge would accept that there is no concept that Adam expresses using his word ‘sofa’. Brown offers no justification for that assumption and no evidence that it is Burge’s view.³⁵

     Notice that in the scenario Brown describes, Adam applies the term ‘sofa’ “firmly and correctly to what we call ‘sofas’.” (1995, p.153) but he is “unsure whether [‘sofa’] applies to broad single seat armchairs.” The term ‘sofa’, then, as Adam uses it, definitely applies to some things (namely, sofas). Moreover, it definitely fails to apply to some things (e.g., things that are not furniture that he deems appropriate for sitting), and it is indefinite in application to some things (e.g., broad single seat armchairs). It seems, then, that, according to Brown’s own description of the case, the term ‘sofa’, in Adam’s mouth, expresses a concept, albeit one that is to some extent indefinite in its application. To be sure, this Adam case is different from standard cases of vague concepts. In a typical case of vagueness, the social norms governing the use of the term ‘sofa’ might leave it indeterminate whether it applies to broad single-seat armchairs, whereas in Adam’s case, there are no social norms. Adam only mistakenly thinks there are. The indeterminacy derives from Adam’s own usage, not from communal usage. However, despite this difference between the Adam case and standard cases of vagueness, the indeterminacy in the Adam case does not disqualify Adam’s term ‘sofa’ from expressing a concept. The case, as Brown describes it, seems to us coherent. If it is coherent, then, far from supporting (2), it is itself a counter-example to (2).

     In any case, principle (2) seems to us dubious. Why cannot an individual have a non-natural kind concept and be agnostic about the necessary and sufficient conditions for its application without being a member of a linguistic community? Brown does not say. Notice that were (2) true, then one could refute the view that one is alone in the world just by determining that one is agnostic about the necessary and sufficient conditions for a concept one recognizes is not a natural kind concept. Suppose, for example, having taken oneself to have just learned the concept of a priori knowledge, a concept that one recognizes not to be a natural kind concept, one wonders what, exactly, the necessary and sufficient conditions are for its application. Could one, from these considerations alone, without assuming one has actually learned the concept from others, conclude that other people exist? There seems no reason to think so. But if (2) were true, one could. So much the worse for (2).

     In his seminal 1975 paper, Putnam claims that mental states do not a priori entail the existence of anything (contingent and not itself a mental state) other than the thinker who occupies them. Perhaps he was wrong about mental states. But his claim is compatible with McGinn’s externalist thesis (M) and with Burge’s externalist thesis (B). For both theses are compatible with the following brand of skepticism:

Solipsism Skepticism. One cannot know a priori that solipsism is false.

     Consider (M). Thesis (M) is consistent with solipsism skepticism since it is consistent with the claim that one cannot know a priori whether any of one’s atomic concepts are natural kind concepts. Given this consistency, there is no a priori route from (M) alone to facts about the external world. Thesis (M) asserts only that if the concept of K is an atomic, natural kind concept then it is impossible to possess it unless one has causally interacted with instances of K. This a priori entails that one has so interacted (and hence that solipsism is false) only if it is also assumed that it is a priori knowable that the concept of K is an atomic, natural kind concept. But (M) is compatible with the negation of this assumption, for every atomic, natural concept one possesses.

     Consider Burge’s thesis (B). This thesis also seems compatible with solipsism skepticism. For (B) is compatible with the (plausible) claim that one cannot know a priori whether one is a member of a linguistic community. So, those philosophers who look to externalism to provide an a priori refutation of solipsism will not find it in (M) or (B) alone.

     Perhaps, however, some combination of Twin-Earth thought-experiments, privileged access, and other a priori considerations can yield an externalist thesis that enables us to refute solipsism on a priori grounds. If so, that is all to the good. We would welcome an a priori refutation of solipsism. It would leave our main point unaffected. For if solipsism can be refuted on such a priori grounds, then there will be an E such that (a) we can know a priori that if we are thinking that P, then E, (b) E is knowable a priori, and (c) E is a priori incompatible with solipsism. There being such an E is compatible with privileged access since, by hypothesis, E is knowable a priori. Thus, we happily leave open whether there is such an E.

     One promising externalist strategy for arguing against solipsism skepticism is to argue that at least one of a thinker’s concepts must be an atomic, natural kind concept.³⁶ If thesis (M) (or (MT) (see note 26)) is a priori knowable, it would then follow that we can know a priori that there is at least one mind-independent kind such that we have causally interacted with instances of it. That would refute solipsism on a priori grounds.

     Whether the strategy just described can succeed will be left open here. For the success of the strategy would pose no problem for privileged access. The reason is that even if we can know a priori both that (M)(or (MT)) is true and that at least one of our concepts is an atomic natural kind concept, we cannot know a priori of any of our concepts that it is an atomic natural kind concept.

     In the first half of the paper, we have argued that not only do switching arguments fail to establish incompatibilism, switching cases fail to reveal any prima facie threat to privileged access —even a false one— that is posed by externalism. In this second half of the paper, we have argued that no version of externalism supported by Twin-Earth thought experiments alone has as a consequence that it is a conceptual truth that if one is thinking that P, then E, where E is not knowable a priori. McKinsey’s recipe is perfectly fine. It is just that no externalist thesis supported by Twin-Earth thought-experiments will serve as an ingredient. We conclude that neither of the two leading lines of argument for incompatibilism establishes that externalist theses supported by Twin-Earth thought-experiments are incompatible with privileged access. If Twin-Earth content-externalism poses a threat to privileged access, it remains to be seen what that threat is. ³⁷

Brian McLaughlin, Rutgers University

Michael Tye, Temple University and King’s College London

Boghossian, Paul “Content and self-knowledge,” Philosophical Topics, 17 (1989), 5-26.

Boghossian, Paul “Externalism and inference,” Philosophical Issues, 2 (1992), 11-28.

Boghossian, Paul “What the externalist can know a priori,” Proceedings of the Aristotelian Society (1997), 161-175.

Brown, Jessica, “The incompatibility of anti-individualism and privileged access,” Analysis 53 (1995), 149-156.

Brueckner, Anthony, “What an anti-individualist knows a priori,” Analysis 52 (1992), 111-118.

Brueckner, Anthony, “Knowledge of content and knowledge of the external world’ Philosophical Review (1994) 103, 327-343.

Burge, Tyler, “Individualism and the mental,” MidWest Studies in Philosophy 4 (1979) 73-121.

Burge, Tyler, “Other bodies,” in Thought and Object: Essays on Intentionality, ed. A. Woodfield (Oxford: Oxford University Press 1982), 97-120.

Burge, Tyler, “Individualism and self-knowledge,” Journal of Philosophy 85 (1988), 649-663.

Burge, Tyler, “Our Entitlement to Self-Knowledge,” Proceedings of the Aristotelian Society XCVI (1996):91-116.

Chisholm, Roderick, Theory of Knowledge second ed. Englewood Cliffs: Prentice-Hall, 1977.

Dretske, Fred, “Epistemic operators,” Journal of Philosophy 67 (1970), 1007-23.

Evans, Gareth, The Varieties of Reference (Oxford: Claredon Press, 1982).

Falvey, Kevin, and Owens, Joseph, “Externalism, self-knowledge, and skepticism,” The Philosophical Review, Vol.103, 1 (1994), 107-137.

Goldman, Alvin, “Discrimination and perceptual knowledge,” Journal of Philosophy 73 (1976), 771-91.

Hill, Christopher, Sensations (Cambridge: Cambridge University Press, 1991).

Hill, Christopher, and McLaughlin, Brian P., “There are fewer things than are dreamt of in Chalmers’s philosophy,” forthcoming, Philosophy and Phenomenological Research.

Korcz, Keith Allen, “Recent work on the basing relation,” American Philosophical Quarterly 34 (1997):171-191.

Loar, Brian, “Personal References’ in E. Villanueva (ed.) Information, Semantics and Epistemology (Oxford: Basil Blackwell, 1990).

Loar, Brian, “Phenomenal properties,” Philosophical Perspectives 4 (1990).

Ludlow, Peter, “Externalism, self-knowledge, and the prevalence of slow switching,” Analysis, 55 (1995a), 45-49.

Ludlow, Peter, “Social externalism, self-knowledge, and memory,” Analysis, 55 (1995b), 157-159.

McGinn, Colin 1977 “Charity, interpretation, and belief,” Journal of Philosophy, 74.

McGinn, Colin, Mental Content (Oxford: Basil Blackwell, 1989). McKinsey, Michael, “Anti-individualism and privileged access,” Analysis 51 (1991), 9-16.

McKinsey, “Accepting the consequences of anti-individualism,” Analysis 54 (1994), 124-28.

McLaughlin, Brian P., Review of Stephen L. White’s The Unity of the Self, Journal of Philosophy (1997):638-44.

McLaughlin, Brian P., and Tye, Michael, (forthcoming - a) “Externalism, twin earth, and self-knowledge,” in eds. B. Smith and C. Macdonald Knowing Our Own Minds: Essays on Self-Knowledge (Oxford: Oxford University Press).

McLaughlin, Brian P., and Tye, Michael (forthcoming - b) “The Brown-McKinsey charge of inconsistency,” in ed. P. Ludlow Externalism and Self-Knowledge (Cambridge: Cambridge University Press).

Miller, Richard “Externalist self-knowledge and the scope of the a priori,” Analysis 57.1 (1997): 667-75.

Moser, Paul, Knowledge and Evidence (New York, NY: Cambridge University Press, 1989).

Peacocke, Christopher, “Entitlement, self-knowledge, and conceptual redeployment,” Proceedings of the Aristotelian Society XCVI (1996):117-158.

Putnam, Hilary, “The meaning of ‘meaning’,” in Language, Mind and Knowledge: Minnesota Studies in the Philosophy of Science, Vol. VII, ed. Keith Gunderson (Minneapolis: University of Minnesota Press, 1975).

Shoemaker, Sidney, “Introspection and the self,” Midwest Studies in Philosophy 10 (1985):101-20.

Shoemaker, Sydney, “On knowing one’s own mind,” Philosophical Perspectives 2 (1988):183-209.

Shoemaker, Sydney, “Introspection,” Cambridge Companion to the Philosophy of Mind (Cambridge: Cambridge University Press, 1994), pp.395-400.

Tye, Michael, “Externalism and memory,” Proceedings of the Aristotelian Society, Supplementary Volume (1998):000-000.

Warfield, Ted A., “Privileged self-knowledge and externalism are compatible,” Analysis 52 (1992):232-237.