The Epistemological Bases of the Slow Switching Argument

Authors


Abstract

One of the main arguments intended to show that content externalism undermines the privileged access thesis is the ‘slow switching argument’, originally proposed by Boghossian (1989). In this argument, it is supposed that a subject is unknowingly switched back and forth between Earth and Twin Earth: then it is claimed that, given externalism, when the subject is on Earth thinking that water is wet, he cannot know the content of his thought a priori, for he cannot, by mere reflection, rule out the relevant alternative hypothesis that he is on Twin Earth thinking that twater is wet. One of the controversies surrounding this argument stems from the fact that it is not clear which epistemological principle underlies it. Here, I examine two suggestions made in the literature as to what that underlying principle might be. I argue that neither of these suggested principles is plausible, and thus that the slow switching argument never gets off the ground.

1 Introduction

Content externalism is the general idea that the content of at least some of our attitudes is individuated by external factors. There are different types of content externalism. According to natural kind externalism, our thought contents are individuated by the natural kinds with which we have causally interacted. The typical example used to motivate this type of externalism is Putnam's ‘Twin Earth’, a planet that is exactly like Earth except for the fact that the liquid the Twin Earthians call ‘water’, which has all the superficial properties of water, has atomic structure XYZ rather than H2O. Intuitively, when a subject on Earth and her duplicate on Twin Earth utter the sentence ‘water is wet’, they express different thought contents, while their intrinsic properties are the same. On singular externalism, however, one's thought contents are individuated partly by the particular objects to which one refers (via an indexical or a proper name). For example, when Oscar sees a cat—let's call it a—and, referring to it, thinks that that cat is on the mat, his thought content is individuated by a. So, in the counterfactual situation in which he sees another cat, b, with the same superficial properties, he would have a different thought content, although he is intrinsically the same in the actual situation and the counterfactual one. Finally, according to social externalism, a subject's thought contents are individuated partly by the practice of her linguistic community.

In recent years, one of the fundamental issues about content externalism has been whether it is compatible with what is often known as the privileged access thesis, the intuitively plausible idea that in normal cases, if we think that p we can know that we think that p in an a priori manner. There are different lines of argument intended to show that content externalism undermines the privileged access thesis.1 Here, I shall concentrate on one such argument which was originally proposed by Boghossian (1989), namely ‘the slow switching argument’ or ‘the discrimination argument’.2 In this argument, it is supposed that a subject, Oscar, unknowingly and slowly is switched back and forth between Earth and Twin Earth. Then, it is claimed that, given externalism, when Oscar is on Earth thinking that water is wet, he cannot know the content of his thought in an a priori manner. For he cannot, by mere reflection, rule out the relevant alternative hypothesis that he is on Twin Earth thinking that twater is wet.

One of the controversies surrounding this argument stems from the fact that it is not entirely clear which epistemological principle is employed by Boghossian in framing his argument. In the subsequent literature, however, two different epistemological principles have been suggested as underpinning the argument. In this paper I examine whether these two principles state plausible constraints on knowledge at all. I try to show that both of them are either false or unmotivated, and thus that the slow switching argument cannot get off the ground.

2 The Slow Switching Argument Explained

The slow switching argument is applicable to all kinds of externalism, including natural kind, singular and social externalism. Following the dominant literature on this issue, I present the argument for the case of natural kind externalism.3 Suppose that a subject, Oscar, is unknowingly switched back and forth between Earth and Twin Earth. According to externalism, when Oscar on Twin Earth interacts with the watery liquid there, the word ‘water’ in his language will, after a while, comes to refer to XYZ rather than H2O. Similarly, the content of his attitudes on Twin Earth would also be different. For example, when Oscar on Twin Earth utters the sentence ‘water is wet’, he will express a thought about XYZ, not H2O. We do not have, of course, a word in English to report his thought content, but we can coin the word ‘twater’ and attribute to Oscar the thought that twater is wet.

Oscar undergoes series of switches between Earth and Twin Earth without being aware of it, and each time he is switched he stays on the planet long enough to acquire the relevant concepts. Now, in one of his Earth phases, Oscar thinks that water is wet, and, exercising his reflective faculties, he comes to believe that he is thinking that water is wet. Does this belief constitute a priori knowledge? Boghossian answers in the negative. Due to Oscar's several travels to Twin Earth, the hypothesis that he is on Twin Earth thinking that twater is wet should be treated as a relevant alternative to his second-order belief that he is thinking that water is wet. But he cannot rule out this hypothesis a priori, since things are phenomenologically the same for him in the actual situation in which he is thinking that water is wet as in the counterfactual situation in which he thinks about twater. Boghossian argues that in such a case Oscar does not know that he is thinking that water is wet. Considering a switching scenario involving the concept arthritis and its twin concept, tarthritis, Boghossian writes:

[The subject] has to be able to exclude the possibility that his thought involved the concept tarthritis rather than the concept arthritis, before he can be said to know what his thought is. But this means that he has to reason his way to a conclusion about his thought; and reason to it, moreover, from evidence about his external environment which, by assumption, he does not possess. How, then, can he know his thought at all—much less know it directly? (ibid.: 160)

Here, Boghossian is exploiting a kind of relevant alternative constraint on knowledge that can be roughly stated as follows:

(R) If (i) there is a relevant counterfactual situation in which q is true instead of p, and (ii) S cannot exclude the possibility that q is the case, then S does not know that p.

The crucial question here is how we should understand the phrase ‘S cannot (or can) exclude the possibility that q is the case’: different interpretations amount, of course, to different formulations of (R). Boghossian himself did not spell out this notion precisely, but two versions of (R) have been suggested in the subsequent literature, stating different necessary conditions for knowledge. The first is due to Falvey and Owens (1994), and the second to Vahid (2003) and Brown (2004). In the following, I briefly discuss these reformulations, examining whether they state plausible constraints on knowledge. I will argue that neither of these two principles is acceptable, and thus that the slow switching argument fails.4

3 Falvey and Owens on the Slow Switching Argument

Let us say that a logically possible proposition q is a relevant alternative to the true proposition p if q is false (in the actual world), and the possibility that q obtains is relevant in the given context. Then, according to Falvey and Owens (1994), the epistemological principle that is operative in Boghossian's argument can be stated as follows:

(R1) 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. (ibid.: 116)

(R1) can be treated as a version (or reformulation) of (R). Two questions immediately arise regarding this principle: Does (R1) state a plausible condition for knowledge in general? And, given that (R1) is true, could it be used to show that Oscar (the victim of slow switching) does not know the content of his water thoughts?

Falvey and Owens answer the second question affirmatively: Oscar's belief that he is thinking that water is wet is based on introspective evidence that is compatible with his thinking the relevant alternative thought that twater is wet. As McLaughlin and Tye point out, it seems that Falvey and Owens take this introspective evidence to be a qualitative mental state (more specifically, ‘an auditory image with certain qualitative features’ (McLaughlin and Tye 1998: 361)) which is supposed to be common to Oscar's ‘water'-thoughts on Earth and Twin Earth. So the antecedent of (R1), according to Falvey and Owens, is satisfied in Oscar's case and hence, given (R1), his second-order belief does not constitute an instance of knowledge.

But what about the first question? Does (R1) state a tenable constraint on knowledge? Falvey and Owens concede the plausibility of (R1) in the case of beliefs about the external world, such as perceptual beliefs. In their view, (R1) is the principle that is operative in fake barn scenarios. Suppose that Tom is riding through a village in which there are numerous fake barns that he cannot distinguish visually from real barns. As it happens, he sees a real barn and believes that there is a barn in front of him. We are inclined to say that his belief is not an instance of knowledge. His belief is based on the object's visual appearance, and this appearance is compatible with the possibility that there is a fake barn in front of him. According to (R1), then, Tom does not know that there is a barn in front of him. But Falvey and Owens push the matter further: Why does Tom not know that there is a barn in front of him, they ask, when his evidence is compatible with its being the case that the object is fake? The answer is that in such circumstances, Tom could easily be deceived by the evidence he possesses. More precisely, if there were a fake barn in front of him, he would still believe, on the basis of the evidence he has, that it was a real barn. So the plausibility of (R1), Falvey and Owens conclude, is rooted in a more basic principle, namely,

(R2) 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. (Falvey and Owens 1994: 116)

Notice that, as Falvey and Owens point out, (R2) does not threaten privileged self-knowledge in the switching cases. To see this, consider again our victim of switching, Oscar, who believes that he is thinking that water is wet. Suppose that p and q are, respectively, the propositions that Oscar is thinking that water is wet and that Oscar is thinking that twater is wet. Now, the antecedent of (R2) is not satisfied in the case of Oscar's belief. It is not the case that if Oscar were thinking that twater is wet he would still believe that he is thinking that water is wet. The crucial point here is that second-order beliefs inherit their contents from first-order thoughts. In other words, the content of second-order beliefs is determined by the environment just as the content of first-order thoughts is. So, if Oscar were on Twin Earth entertaining the concept twater, this concept (not the concept water) would appear in the content of his second-order belief, too. The result is that (R2) does not threaten Oscar's self-knowledge.

Now, what is the bearing of all this on the plausibility of (R1) in the case of reflective beliefs? Falvey and Owens claim that the antecedent of (R1) is satisfied in the case of Oscar's reflective belief (for his belief that he is thinking that water is wet is based on introspective evidence compatible with its being the case that he is thinking that twater is wet), and therefore (R1) does undermine Oscar's claim to self-knowledge. On the other hand, as explained above, (R2) is silent about the epistemic status of Oscar's reflective beliefs. But, according to Falvey and Owens, the plausibility of (R1) is rooted in (R2), so when these two principles issue in different evaluations vis-à-vis Oscar's self-knowledge, (R2) has more right to determine whether Oscar's self-knowledge is undermined or not. In short, although Oscar's belief that he is thinking that water is wet is based on evidence that is compatible with his thinking that twater is wet, this does not prevent his belief from being knowledge, since that evidence does not make him liable to error in the counterfactual situation. Thus, they conclude, (R1) is false in the case of reflective beliefs.5

So far, we have seen that there is a possible reconstruction of the slow switching argument which takes (R1) as the underlying principle. Naturally, there are two ways to block this argument, either by denying the plausibility of (R1) or by denying that its antecedent is satisfied in the case of Oscar's reflective beliefs. Falvey and Owens take the first option and I am in partial sympathy with their decision: (R1) does not state a plausible constraint on knowledge and this suffices to show that the envisaged version of the slow switching argument fails (I will argue why (R1) is not acceptable in the following section).

One might suggest, however, that the second option is also available to the proponent of content externalism. For example, it might be argued that our reflective second-order beliefs, of the form ‘I believe that I am thinking that p’, are not based on evidence, and the fact that they constitute knowledge is merely due to their self-verifying nature.6 Given this position, the antecedent of (R1) is not satisfied in the case of reflective beliefs as these beliefs are not evidence-based in the first place. Thus, (R1), even if plausible, poses no threat to the thesis of the compatibility of externalism and privileged self-knowledge.7 I don't deny that this line of response might also be available to the proponent of content externalism, though it involves controversial assumptions regarding the nature of self-knowledge. What I want to point out, however, is that once we deny the plausibility of (R1) we can resist the slow switching argument, as construed by Falvey and Owens, without invoking controversial theses about the nature of self-knowledge.

Now, let us see again why Falvey and Owens do not treat (R1) as a tenable principle in the case of second-order beliefs. As described above, they argue that since the plausibility of (R1) is rooted in (R2) and these two principles issue in different evaluations with regard to Oscar's reflective-belief, it is (R1) that should be given up. Let us grant, for the sake of argument, that the antecedent of (R1) is satisfied in the case of Oscar's second-order belief. It seems, however, that Falvey and Owens are rather too quick to deny the plausibility of (R1) in this case. As McLaughlin and Tye (1998) have pointed out, a proponent of the slow switching argument can claim that the plausibility of (R1) is not rooted in (R2). She may argue, instead, that these are two independently plausible principles that state different necessary conditions for knowledge, and although (R2) is silent about Oscar's second-order belief, (R1) does preclude it from being an instance of knowledge.

So it seems that we need to tell a richer story for why (R1) is not a tenable principle for knowledge. This is what I will undertake in the following. But before that, let us look more closely at McLaughlin and Tye's remark regarding (R1). As noted above, according to McLaughlin and Tye (1998), a proponent of the slow switching argument may reject the claim that the plausibility of (R1) is grounded in (R2). They support the independent plausibility of (R1) by designing an interesting example:

Suppose that Oscar (correctly) believes that he has a brain. An alternative is that his head is full of sawdust. This alternative is relevant, given the following scenario: There is a widespread false rumor that using magic, a Wizard has removed the brains of certain people in the community and replaced them with sawdust. Oscar like the other members of his community is ignorant of the causal connection between brains and behavior … Hearing the Wizard-rumor, however, he, like many others in his community, begins to doubt that he has a brain. To determine whether he has a brain, he removes the entrails of a chicken. As luck would have it, to his great relief, he gets the reading that he has a brain. As a result, his belief that he has a brain is restored. (ibid.: 356, fn. 15)

We are very strongly inclined to say that Oscar does not know that he has a brain. But (R2) cannot explain this intuition. If Oscar had sawdust in his head instead of a brain, he would not have any beliefs at all; and in particular he would not believe that he has a brain in his head. So the antecedent of (R2) is not satisfied here and (R2) is silent about whether Oscar knows that he has a brain. But proponents of (R1) may claim that (R1) can explain why Oscar fails to know: Oscar's belief is based on evidence that is compatible with his head being full of sawdust. So the antecedent of (R1) is satisfied in this case.

McLaughlin and Tye, in fact, appeal to the explanatory advantage of (R1) over (R2) to provide some support for the independent plausibility of (R1), but they don't make a final assessment of (R1). They concede that if (R1) is indeed false, then the slow switching argument would fail, but they maintain that rejecting (R1) is not the only option available for compatibilists. Their concern is to show that the slow switching argument would fail even if (R1) were true. I will not here pursue their response to the slow switching argument. In what follows, I concentrate instead on whether the above example can provide any support for (R1) and, more generally, whether (R1) is acceptable at all.

4 Is (R1) Plausible?

Two remarks about (R1) are in order. The first concerns the example designed by McLaughlin and Tye to show the explanatory advantage of (R1) over (R2). As we saw, they claim that (R1), but not (R2), can explain why Oscar does not know that he has a brain, and they take this as providing some support for (R1). Suppose, for the sake of the argument, that (R1) can explain Oscar's lack of knowledge.8 The problem is, however, that appealing to (R1) is not the only way available to explain our intuition regarding Oscar's case. There is another principle which can explain it as well. We can state that principle very roughly as follows:

The global reliability principle: For S's belief that p to be a piece of knowledge it is necessary that the evidence on which S's belief is based, or the method by which his belief is produced, is globally reliable; that is, that kind of evidence or method mostly yields true beliefs in similar cases.

In the above example, the evidence that Oscar possesses is not globally reliable, since the reading of chicken entrails is not a good indicator of truth; this kind of evidence results in false beliefs in most cases. So, the global reliability principle, as well as (R1), can explain our intuition in Oscar's case. Notice that there are independent and compelling reasons for the global reliability principle (see, for example, McGinn 1984; Goldman 1986; Brown 2004). Now, given that we are obliged, for independent reasons, to accept the global reliability principle, and given that this principle can explain Oscar's lack of knowledge, why should we appeal to (R1) to explain his failure to know that he has a brain? It seems that although (R1) can be an explanation here, it is not the best explanation available for us. Thus the above example cannot provide good support for (R1). Yes, if we could find a case in which (i) the subject's belief intuitively fails to constitute an instance of knowledge, and (ii) both (R2) and the principle of global reliability are unable to explain our intuition, and (iii) (R1) could explain the subject's lack of knowledge, then this case would provide support for (R1). But it seems difficult for the proponent of (R1) to find such a case.

Now, let us turn to the second remark. It seems that if singular externalism is correct then, given certain assumptions, (R1) would be vulnerable to counterexamples. In other words, the proponent of singular externalism would have independent motivations to deny (R1), motivations which are not inspired by the alleged fact that (R1) paves the way for the incompatibility of content externalism and the privileged access thesis. In what follows I will offer three samples of such counterexamples; but first a preliminary point must be made. As explained earlier, according to singular externalism the content of a belief expressed by a sentence including a singular term is, at least partly, individuated by its referent. There are two main approaches within the broad idea of singular externalism. According to the direct reference approach or the non-Fregean approach, the content of a belief expressed by a sentence including a singular term is a Russellian proposition in which the referent of that singular term enters as an ingredient. On the Fregean approach, however, an attempt is made to combine the notion of Fregean senses with content externalism in the manner proposed by Evans (1982) and McDowell (1977; 1986). For the following counterexamples to work it is sufficient to presuppose singular externalism in its broad sense, alongside other assumptions to be described below. For convenience, however, in designing the first two counterexamples I presuppose the direct reference approach, and exploit the device of Russellian propositions. These two counterexamples can nonetheless be restated in accordance with Fregean externalism as well. Here, then, are the promised counterexamples.

Counterexample (1)

Suppose that Judy and Trudy are indistinguishable twins who are in Oscar's neighbourhood. All of their superficial properties are the same such that one cannot discriminate them visually. One day Oscar sees Judy sitting on the chair, but as usual he cannot identify her. Oscar is careful not to make a false report, so he does not use the name ‘Judy’, nor ‘Trudy’. Instead, he uses the pronoun ‘she’ and referring to Judy utters this sentence: ‘she is sitting on the chair’. We have a strong intuition that the belief he expresses by this sentence is an instance of knowledge. This belief is true and is produced by perceptual experience in normal conditions. It is hard to see which perceptual beliefs could be counted as knowledge if this is not.

What is the content of this belief? According to direct reference externalism, the content of this belief is a Russellian proposition in which the referent of ‘she’, Judy, appears as an ingredient. We can show this proposition by an ordered pair <Judy, the property of sitting on the chair>. Let us call this proposition p. On the other hand, given that the twins live in the same place, it is a relevant counterfactual situation that Trudy, instead of Judy, is sitting on the chair. In this counterfactual situation, the content of Oscar's belief would be a different Russellian proposition, <Trudy, the property of sitting on the chair>. Call this proposition q.

According to one view about perceptual evidence, the perceptual evidence in each case is a collection of perceptual appearances.9 Assuming this view, the evidence on which Oscar's belief that p is based is compatible with its being the case that q. For, by hypothesis, Judy and Trudy have exactly similar superficial properties and so Oscar acquires the same visual appearances in both the actual and the counterfactual situation. Thus, the antecedent of (R1) is satisfied here and consequently, according to (R1), Oscar's belief must fail to be knowledge. But, intuitively, Oscar knows that she is sitting on the chair. Thus, (R1) appears to be false. The proponent of singular externalism can therefore conclude, independently of considerations connected with the privileged access thesis, that (R1) is not a plausible constraint on knowledge.10

This counterexample can be designed in another way: when Oscar sees Judy sitting on the chair, he does not use the pronoun ‘she’. Instead, at that moment he chooses a new name, say ‘Sara’, for the girl he sees. The referent of the name ‘Sara’ in this tiny baptism can be fixed either by ostention or by a definite description such as ‘the twin whom I now see’. Then, Oscar sincerely utters the sentence ‘Sara is sitting on the chair’. Considering Oscar's well-functioning sight and the process of introducing the name ‘Sara’, here again our intuition leads us to say that Oscar's belief that Sara is sitting on the chair is an instance of knowledge. In fact, Oscar knows that (i) the twin whom he sees is sitting on the chair, and (ii) the twin whom he sees is Sara. The first piece of knowledge is an obvious instance of perceptual knowledge, and the second results directly from the baptism performed by Oscar. It seems very odd to say that Oscar knows (i) and (ii), but does not know that Sara is sitting on the chair.

According to direct reference externalism, the content of Oscar's belief in this scenario, as in the previous one, is the proposition <Judy, the property of sitting on the chair> (proposition p). One relevant alternative situation to his belief that p is one in which Trudy, instead of Judy, is sitting on the chair and Oscar gives the name ‘Sara’ to her. Admittedly the content of Oscar's belief, in this situation, would be another Russellian proposition, <Trudy, the property of sitting on the chair> (proposition q). Now, if perceptual evidence consists of perceptual appearances, Oscar would possess the same evidence in both the actual and the counterfactual situation and therefore, according to (R1), his belief would not qualify as knowledge. But intuitively Oscar knows that Sara is sitting on the chair. So (R1) is false.

Notice that (R2) is vulnerable to neither of the two versions of this counterexample, for the antecedent of (R2) is not satisfied in this case. If Trudy were sitting on the chair instead of Judy, Oscar would no longer believe the proposition <Judy, the property of sitting on the chair>.

It may be objected that the above counterexample can work against (R1) only if we adopt the aforementioned view about perceptual evidence, according to which perceptual evidence consists of a set of perceptual appearances. But there is another view here. Some philosophers, inspired by content externalism, maintain that perceptual evidence is a contentful state whose content is determined by the environment. Consequently, the perceptual evidence that Oscar acquires when he sees Judy is different from that which he possesses when he sees Trudy. In fact, the content of his evidence changes in the counterfactual situation just as the content of his belief does. Given this view about perceptual evidence, the evidence on which Oscar's belief that p is based is incompatible with its being the case that q. So, the antecedent of (R1) is no longer satisfied in the above scenario, and therefore it cannot be treated as a counterexample to (R1). In what follows, I offer another counterexample to (R1) that does not depend upon a particular view of perceptual evidence.

Counterexample (2)

Suppose that Oscar is informed, by a reliable source, that there is now just one person in the room, and that the room is now extremely warmed by a heater. Considering this information and his previous experiences about how people feel in warm environments, he comes to believe that the person in the room is feeling hot (call this proposition p). Intuitively, this belief is an instance of knowledge. It is in fact a piece of inferential knowledge that is derived from Oscar's other beliefs which are, themselves, based on adequate evidence:

  1. His belief that there is one person in the room and that the room is warm (based on the information he has received).
  2. His belief that people feel hot in warm environments (based on his previous experiences).

So the evidence on which Oscar's belief that p is based consists of the two aforementioned justified beliefs. Suppose further that, unbeknownst to Oscar, the person who is in the room is in fact Ted. When Oscar utters the sentence ‘the person in the room is feeling hot’ he refers to Ted via a definite description, namely ‘the person in the room’. Now, it seems that it is possible for him to directly refer to Ted and, consequently, to acquire a belief with a Russellian content about him. Suppose that Oscar, at that moment, introduces a name, say ‘Roomy’, for whoever is in the room. In fact, he exploits that definite description (namely ‘the person in the room’) to fix the referent of the name ‘Roomy’. Then he sincerely utters the sentence ‘Roomy is feeling hot’. It seems plausible to say that, by this utterance, Oscar expresses a belief about Ted. The content of this belief, according to direct reference externalism, is a Russellian proposition, <Ted, the property of feeling hot>. Call this proposition p 1. Here again our intuition leads us to think of Oscar's belief that p 1 as knowledge. In fact, Oscar's belief that p 1 is directly and immediately derived from his belief that p. We can say, therefore, that the evidence he possesses for his belief that p 1 is his justified belief (or knowledge) that p, which is based, in turn, on its own evidence.

Now suppose that there is a relevant counterfactual situation in which all things are the same as in the actual one except for the fact that in this situation, Ned, instead of Ted, is in the room. In this situation, again, Oscar believes that the person in the room is feeling hot (proposition p). In addition, he gives the name ‘Roomy’ to whoever is in the room. But in this counterfactual situation the name ‘Roomy’ refers to Ned, so when Oscar sincerely utters the sentence ‘Roomy is feeling hot’, the content of his belief would be another Russellian proposition, <Ned, the property of feeling hot> (call this proposition q). Proposition q is, in fact, a relevant alternative to proposition p 1.

How does (R1) assess the epistemic status of Oscar in regard to his belief that p 1? It seems that the antecedent of (R1) is satisfied in this case. The evidence upon which Oscar's belief that p 1 is based is compatible with its being the case that q. As indicated above, this evidence is his justified belief that p, which is supported by two other justified beliefs. Clearly, all these three beliefs as well as their own grounds are present in the envisaged counterfactual situation in which q is true. Notice that in this scenario neither Ted nor Ned enters into the content of any of the evidence which Oscar acquires. Therefore, even if the content of perceptual evidence is individuated by the environment, the evidential status of Oscar would be exactly the same in both the actual and the counterfactual situation. So, in contrast to our intuition, (R1) precludes Oscar's belief that p 1 from being knowledge. Thus, we can conclude that, given singular externalism, (R1) does not state a proper constraint on knowledge.11

Clearly enough, the second counterexample, like the first one, does not work against (R2) since its antecedent is not satisfied in this case. In the counterfactual situation in which Ned is in the room, Oscar would no longer believe that ‘Roomy’ (= Ted) is in the room. Instead, he would correctly believe that ‘Roomy’ (= Ned) is in the room.

Although the second counterexample does not depend upon any particular theory about perceptual evidence, it does rely on a controversial view about the conditions under which one can use a definite description to introduce a proper name. It is usually held that in baptism the referent of a proper name can be fixed by a definite description, but there are two different approaches as to when this can be done. According to the restricted view, one cannot successfully introduce a name for an object, say O, by means of a referent-fixing definite description unless one has a kind of direct acquaintance or contact with O.12 According to the unrestricted view, however, having direct acquaintance with an object is not required for introducing a name for it by a referent-fixing definite description.13

It is clear that if the restricted view is correct, the second counterexample will not work against (R1). I cannot settle the dispute between these two views here, but do want to make one quick observation. It seems that the restricted view, in its strong formulation, prevents us from introducing names for theoretical entities—contrary to what is quite ordinary practice in science. And it is highly implausible to interpret all theoretical terms used in science as abbreviations for definite descriptions, not as genuine singular terms. Therefore, it may be argued that the restricted view should somehow be modified in such a way that direct acquaintance with an object would no longer be a prerequisite for introducing a name for it. I guess that assuming any version of this modified view, a counterexample to (R1) could be designed which would be analogous to the second counterexample.14

Finally, we can exploit the above scenario to design another counterexample which depends neither on a specific conception of perceptual evidence, nor on a controversial view about introducing a proper name by definite description.

Counterexample (3)

Consider again the above scenario in which Oscar knows that the person in the room is feeling hot (p). For simplicity, I use ‘the F’ as an abbreviation for the definite description ‘the person in the room’. Oscar, then, knows that the F is feeling hot. Now suppose that Oscar uses the operator actual to form another belief to the effect that the actual F is feeling hot (call this proposition p 2). It seems that this belief is an instance of knowledge, too. In fact, his belief that p 2 is directly and immediately derived from his belief that p (more generally, ‘the F is G’ logically entails ‘the actual F is G’). So Oscar's evidence for his belief that p 2 is his justified belief (or knowledge) that p, which is based, as we saw, on its own evidence.

Now consider the relevant counterfactual situation in which Ned, rather than Ted, is in the room. In this situation p 2 is no longer true, while the evidence upon which Oscar's actual belief that p 2 is based is still present: he, in this counterfactual situation, still believes that p on the same grounds as in the actual situation. So, according to (R1) Oscar's belief that p 2 (in the actual situation) fails to constitute knowledge, whereas my intuition leads me to the contrary. This counterexample, clearly enough, does not rely on any of the controversial assumptions presupposed by the two previous ones. It is noteworthy that (R2) is not vulnerable to this counterexample, either.

As described above, Falvey and Owens concede the plausibility of (R1) in the case of beliefs about the external world. These counterexamples, if successful, are interesting in that they can show the inadequacy of (R1) even in the case of these beliefs. In particular, Falvey and Owens take perceptual evidence to be a set of perceptual appearances,15 so they are obliged to accept the first counterexample to (R1) and thereby its failure in the case of perceptual beliefs as well.

In this section I examined one of the common readings of the slow switching argument according to which (R1) is its underlying epistemological principle. I argued first that the example offered by McLaughlin and Tye in defence of (R1) does not provide any support for it, and second that (R1) is subject to counterexamples. I now turn to another interpretation of the slow switching argument, which takes it to be based on a different epistemological principle.

5 Vahid and Brown on the Slow Switching Argument

According to Vahid (2003) and Brown (2004), the principle which underlies the slow switching argument is the following:

(R3) If (i) q is a relevant alternative to p, and (ii) S cannot discriminate between the actual situation in which p is true and the counterfactual situation in which q is true, then S does not know that p.16

They maintain that the discriminative ability is an independent requirement for knowledge which cannot be reduced to the reliability condition. So, on their view, the aim of the argument is to show that externalism undermines Oscar's ability to discriminate a priori between his water-thoughts and twater-thoughts and, a fortiori, his a priori self-knowledge.

As Brown clearly illustrates, in some of the key cases used in the literature to support the reliabilist account of knowledge (such as the fake barn example), possession of a reliable belief-forming process goes hand-in-hand with possession of an associated discriminative ability. Because of this, many philosophers have failed to recognize that reliability and discriminative ability can come apart. But Brown persuasively exploits the switching scenario itself to show that having a reliable ability to form true beliefs is not sufficient for the possession of an associated discriminative ability. Oscar's second-order belief that he is thinking that water is wet is produced by a highly reliable belief-forming process. This belief is an instance of cogito thoughts, second order beliefs of the form: I believe that I am thinking that p. As Burge (1988) argues, cogito thoughts are self-verifying and this guarantees their truth whenever they are produced. So, Oscar does not make a mistake about what he thinks even after the switch. But at the same time he cannot distinguish, by mere reflection, between the actual situations in which he thinks about water and the counterfactual situation in which he thinks about twater, as things remain phenomenologically the same for him in these two situations.17

But what exactly does having a discriminative ability consist of? Brown does not provide a full analysis of this notion, but she tries to throw some light on it. She points out that discriminative ability is connected to several other abilities: the ability to notice change, to make reliable judgements of sameness and difference, and to act differentially. The following example is illuminating:

Suppose that a subject has the ability to distinguish yellow and blue. If she does, then we would expect not only that she can reliably judge whether objects are yellow or blue, but also that she has certain other abilities. In particular, we would expect that she is able to notice if an object changes from yellow to blue. We would also expect that she can reliably say of any pair of objects from a collection of objects, all of which are either blue or yellow but not both, whether or not they are the same color. Last, we would expect that her ability to distinguish the two colors could be manifested in an ability to act differentially with respect to samples of those colors. (Brown 2004: 49)

She then argues at length to the effect that the victim of slow switching lacks these three abilities vis-à-vis her water-thoughts, and thus that she cannot know the content of her water thoughts in an a priori manner.

In what follows, I argue, by offering some counterexamples to (R3), that having a discriminative ability (in the sense that Brown has in mind) is not a necessary condition for knowledge. Fortunately, the counterexamples are already at hand. The same counterexamples that I proposed to (R1) also refute (R3). As expected, since (R3) states a stronger constraint on knowledge, it is more vulnerable to such counterexamples than is (R1). Let me briefly restate the first counterexample (here again, I presuppose direct reference externalism, though all that is required for the counterexample to work is singular externalism in its broad sense).

Suppose that Judy and Trudy are indistinguishable twins. One day Oscar sees Judy sitting on the chair, but as usual he cannot identify her. Intending to report what he sees correctly, he uses neither the name ‘Judy’ nor ‘Trudy’ in his report, but instead employs the pronoun ‘she’ and, referring to Judy, utters this sentence: ‘she is sitting on the chair’. We have a strong intuition that the belief he expresses by this sentence is an instance of knowledge. Given direct reference externalism, the content of this belief is the Russellian proposition: <Judy, the property of sitting on the chair>. Call this proposition p. On the other hand, in the relevant counterfactual situation in which Trudy, rather than Judy, is sitting on the chair, Oscar would refer to Trudy by using ‘she’, and the content of his belief would be another Russellian proposition, <Trudy, the property of sitting on the chair>. Call this proposition q.

Admittedly, Oscar cannot discriminate between the actual situation in which Judy is sitting on the chair and the counterfactual situation in which Trudy is sitting on the chair. To see this more clearly, recall the three abilities which Brown described as connected to the discriminative ability, namely the ability to notice change, to make reliable judgements of sameness and difference, and to act differentially. Oscar lacks all of these abilities with respect to Judy and Trudy. If Oscar sees Judy somewhere and, while his attention is distracted, she is replaced with Trudy, he would not notice the change. Moreover, if Oscar is asked whether the twin he is watching now is identical with whom he saw yesterday, he could not answer reliably. Finally, he cannot act differentially with respect to these twins. Therefore, according to (R3), Oscar's belief that p is not an instance of knowledge, although our intuition strongly leads us to the contrary.18

The proponent of (R3) might try to respond to this counterexample in the following way. Since Oscar cannot discriminate between Judy and Trudy, he does not know the proposition he expresses by ‘she is sitting on the chair’. But intuitively, Oscar knows something in this situation. We can explain this intuition in two different ways. We can claim that all Oscar knows is that the sentence ‘she is sitting on the chair’ expresses a truth. But he does not know which truth it expresses: Judy is sitting on the chair or Trudy is sitting on the chair. Alternatively, we can maintain that what he knows is merely some descriptive proposition, for example the proposition that the girl he is looking at is sitting on the chair. Exploiting either of these two strategies (or a mixture of them), therefore, we can acknowledge that Oscar knows something when uttering the sentence ‘she is sitting on the chair’ without agreeing that it is the singular proposition <Judy, the property of sitting on the chair> that he knows.19

I think that this response does not do justice to our intuitions regarding Oscar's case. Of course, when uttering the sentence ‘she is sitting on the chair’ Oscar knows various things, including some descriptive propositions as well as the meta-linguistic fact that that sentence is true. The question, however, is whether he knows the very proposition he expresses by uttering ‘she is sitting on the chair’. My intuition strongly forces me to answer this question positively. Oscar simply sees that she is sitting on the chair, and by hypothesis his visual faculty works very well. It seems not only that Oscar knows that the sentence ‘she is sitting on the chair’ is true, but that he also knows which truth it expresses: it expresses the truth that she is sitting on the chair.20

The point can be reinforced by employing the plausible idea that there is an intimate connection between knowledge, assertion and practical reasoning. Following Williamson (2000) it is widely accepted that knowledge is the norm of assertion, in the sense that one is permissible or warranted in asserting that p only if one knows that p.21 It has further been argued that knowledge is also the norm of practical reasoning: one can appropriately rely on p in practical reasoning only if one knows that p.22 (For my purposes here, I am presenting these theses as merely specifying a necessary condition on proper assertion and practical reasoning, though some are inclined to take knowledge both necessary and sufficient for proper assertion and practical reasoning).

Now, even if one might have doubts as to whether Oscar knows the proposition he expresses by ‘she is sitting on the chair’, it is pretty obvious that he is warranted in asserting that ‘she is sitting on the chair’. This sentence is plainly assertable under such circumstances. This intuition, combined with the thesis that knowledge is a necessary condition for permissible assertion, would, however, yield the conclusion that Oscar knows the singular proposition expressed by ‘she is sitting on the chair’.23 Likewise, it seems quite plausible for Oscar to rely on the same proposition when engaged in practical reasoning. For example, suppose that as a matter of a special obligation, Oscar must pay one thousand dollars to whoever is sitting on the chair on Friday. Now, Oscar can appropriately reason as follows:

She is sitting on the chair on Friday.

I must pay one thousand dollars to whoever is sitting on the chair on Friday.

So I must pay her one thousand dollars.

Thus, given that knowledge is the norm of practical reasoning, it follows that Oscar knows the proposition he expresses by ‘she is sitting on the chair’. I conclude, therefore, that pace (R3) Oscar knows the proposition <Judy, the property of sitting on the chair>.

As pointed out before, this counterexample can falsify (R1) only if perceptual evidence consists of perceptual appearances. But in the present case there is no such limitation. Even if perceptual evidence has externally determined content, this counterexample can refute (R3). In any event, Oscar cannot discriminate between the actual situation in which Judy is sitting on the chair and the relevant counterfactual situation in which Trudy is sitting on the chair, even though he possesses different evidence in these two situations.

Exploiting the above scenario, we can design a different and even more dramatic counterexample to (R3). Suppose that Oscar, when he sees Judy sitting on the chair, uses the definite description ‘the girl who is sitting on the chair now’ to introduce a new name, say ‘Sara’, for her. Here, the referent of the name ‘Sara’ is fixed by that definite description. Intuitively, in such a circumstance Oscar is in a position to know the conditional proposition that if the girl who is sitting on the chair exists then Sara is the girl who is sitting on the chair (call this proposition r). This intuition, again, can be strengthened by the thesis that knowledge is the norm of assertion and practical reasoning. Now, consider the relevant counterfactual situation in which Trudy, rather than Judy, is sitting on the chair. In this situation, clearly, proposition r is false and another conditional holds: if the girl who is sitting on the chair exists, then Trudy is the girl who is sitting on the chair (call this proposition s). Here, s is a relevant alternative to r, but Oscar cannot discriminate between the actual situation in which r is true and the counterfactual situation where s is the case, as he cannot discriminate between Judy and Trudy. So, according to (R3) Oscar does not know that r, but this is highly counterintuitive.24 We can conclude, therefore, that the slow switching argument, as construed by Vahid and Brown, does not succeed.25

6 Summary and Conclusion

The slow switching argument exploits an epistemological principle that can be roughly stated as follows:

(R) If (i) the counterfactual situation in which q is true instead of p is a relevant situation, and (ii) S cannot exclude the possibility that q is the case, then S does not know that p.

It is not clear, however, what the second part of the antecedent (namely, the inability to exclude the possibility that q is the case) consists of. To achieve a more clear formulation of the principle operative in the argument, two interpretations of (R) have been suggested:

(R1) 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.

And:

(R3) If (i) q is a relevant alternative to p, and (ii) S cannot discriminate between the actual situation in which p is true and the counterfactual situation in which q is true, then S does not know that p.

I argued that no good reason has been proposed for (R1), and moreover that the proponent of singular externalism has independent reasons to deny it. Regarding (R3) I argued that, given singular externalism, it can be refuted even more easily and clearly by the same counterexamples I offered against (R1).

In addition to these two principles, there is another principle in the vicinity:

(R2) 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.

As proponents of content externalism have argued, this principle, though acceptable, does not pose any threat for the privileged access thesis in switching cases. We can conclude, therefore, that the slow switching argument does not get off the ground from the outset. Finally, a more general lesson, which is of independent epistemological interest, can be drawn from this discussion: if singular externalism is correct, then neither (R1) nor (R3) states a plausible constraint on knowledge, even when knowledge of the external world is concerned.26

Notes

  1. 1

    There is a huge literature on this issue. For collections of papers, see, for example, Ludlow and Martin 1998; Wright et al. 2000; Nuccetelli 2003.

  2. 2

    This latter name is due to Brown 2004.

  3. 3

    Moreover, I only consider the standard version of the argument, in which a subject is switched between Earth and Twin Earth. Other switching scenarios, such as switching between Earth and Dry Earth (a planet in which there is no watery substance but the inhabitants suffer a permanent illusion of there being lakes and rivers full of a watery liquid that they call ‘water’), would deserve a separate discussion.

  4. 4

    Another type of challenge to the argument is that slow switching does not occur in the normal course of events: so the possibility that a subject might have been in a twin situation is not by and large a relevant alternative, and consequently the privileged access thesis remains intact in normal cases. For this kind of response to the argument see, for example, Warfield 1992; 1997; Brown 2004. For criticism of this response see, for example, Ludlow 1995; 1997; Bruckner 1997; Vahid 2003 and Goldberg 2006.

  5. 5

    Falvey and Owens 1994: 116–18.

  6. 6

    Burge 1988; 1996 appears to endorse such a view. For criticism of the idea that self-knowledge is not based on evidence, see, for example, Boghossian 1989: 164–7. See also Falvey and Owens 1994: 117 and McLaughlin and Tye 1998: 364.

  7. 7

    I thank an anonymous EJOP referee for this suggestion.

  8. 8

    Notice that this supposition itself depends on controversial issues about the nature of evidence. For example, if we take Oscar's perception of the entrails as part of his evidence for his belief that he has a brain, then (R1) would not be able to explain his lack of knowledge. For, on this view, Oscar's evidence would not be compatible with his head being full of sawdust. Obviously, if he had sawdust in his head, he could not perceive anything. So, given this view about Oscar's evidence, (R1) and (R2) are in the same boat regarding Oscar's case: none of them could explain his lack of knowledge. Yes, if the entrails and their specific arrangement, not Oscar's perception of them, are the evidence upon which his belief is based, then the antecedent of (R1) would be satisfied, since that specific arrangement is compatible with Oscar's head being full of sawdust (for a good survey of this controversy regarding the nature of evidence, see Kelly 2008). For the sake of argument, however, I take it as granted that (R1) can explain Oscar's failure to know that he has a brain.

  9. 9

    For example, Falvey and Owens appear to adopt such a view about perceptual evidence (see Falvey and Owens 1994: 116). There is, of course, an alternative and even more common conception of perceptual evidence, which I will consider later on.

  10. 10

    It might be objected that if Oscar knows the proposition <Judy, the property of sitting on the chair>, then it must be correct to report his epistemic status by saying that he knows that Judy is sitting on the chair. But it seems that he does not know that Judy is sitting on the chair, as he is not inclined to utter the sentence ‘Judy is sitting on the chair’ sincerely. This objection is, in fact, an instance of Frege's puzzle applied in Oscar's case. Thus, we can reply to the objection by employing solutions proposed by proponents of the direct reference theory to Frege's puzzle. For example, we can say that although the sentence ‘Oscar knows that Judy is sitting on the chair’ is semantically true, it is pragmatically defective, and the puzzle merely arises from confusing falsity with pragmatic deficiency (among others, Salmon 1986 and Soames 2002 have proposed this kind of solution for Frege's puzzle). Alternatively, we can adopt the ‘hidden indexical’ theory according to which the verb ‘know’ (and other propositional attitude verbs) expresses a three-place relation that holds between a subject, a proposition, and a mode of presentation. On this theory, although Oscar is in the knowledge relation to the proposition <Judy, the property of sitting on the chair>, the sentence ‘Oscar knows that Judy is sitting on the chair’ is false, since Oscar does not think of that proposition in the ‘Judy-y’ way (for such a theory about propositional attitude reports see, e.g., Crimmins and Perry 1989 and Schiffer 1992).

  11. 11

    We can design this counterexample in another way: Oscar does not introduce a name for the person in the room. Instead, he exploits the ‘dthat’ operator and sincerely utters this sentence: ‘Dthat [the person in the room] is feeling hot’. According to Kaplan (1978), by uttering this sentence, Oscar expresses a belief whose content is a Russellian proposition, namely, <Ted, the property of feeling hot>. The other parts of the scenario remain the same.

  12. 12

    See, for example, Soames 2003. Donnellan 1979 has maintained a somewhat similar view. According to him, when one lacks any kind of direct contact with an object, although one can introduce a name for it by means of a referent-fixing definite description, one cannot generate a singular thought about it.

  13. 13

    Harman 1977, Kaplan 1989 and Jeshion 2001; 2004; 2006 and 2009 have endorsed this view.

  14. 14

    Moreover, Jeshion in a series of works (2001, 2004 and 2006) has developed interesting arguments against the acquaintance condition. In a recent work (2009) she tries to replace the acquaintance condition with a more resilient one, ‘the significance condition’. Discussion of her arguments goes beyond the purpose of this paper. It is notable, however, that the second counterexample can be reconstructed in a way that works against (R1) even on Jeshion's moderate theory.

  15. 15

    Falvey and Owens 1994: 116.

  16. 16

    Brown clearly endorses (R3), although she does not use the exact wording (see Brown 2004: 37–41). Likewise, Vahid proposes a very similar idea (see Vahid 2003: 377–8).

  17. 17

    Brown 2004: 45–50.

  18. 18

    Interestingly, Brown herself employs a similar example in a different context for another purpose. She supposes that there are two identical twins, Anya and Tanya, in Sally's neighborhood. One day Sally sees Anya and thinks the perceptual demonstrative thought that that girl (Anya) is happy. She also correctly believes that she thinks that that girl (Anya) is happy. According to Brown, this second-order belief does not constitute a priori knowledge, since it is a relevant alternative that Sally might have seen Tanya and thought about her rather than Anya. And on the other hand, Sally cannot discriminate a priori between the actual situation in which she thinks that that girl (Anya) is happy and the counterfactual situation in which she thinks the different thought that that girl (Tanya) is happy (Brown 2004: 86–9). Here, Brown applies (R3) to Sally's second-order belief, and the result is that that belief is not an instance of knowledge. But what if (R3) were applied to Sally's first-order belief, namely, her belief that that girl (Anya) is happy? Obviously enough, an analogous result would follow. Sally cannot discriminate between the actual situation in which that girl (Anya) is happy and the relevant counterfactual situation in which that girl (Tanya) is happy (for the sake of the argument we can suppose that Tanya's happiness is incompatible with Anya's). So according to (R3), Sally's first-order belief is not knowledge, either. Being highly counterintuitive, this result shows that (R3) does not state a plausible condition for knowledge at all.

  19. 19

    The first strategy was suggested by Vahid and the second by Brown, both in conversation. As we have seen in Section 4, this counterexample can also be raised against (R1). Likewise, these two strategies against the counterexample as well as my reply to them (which I set out below) can be repeated in the case of (R1).

  20. 20

    To make our intuition more vivid let us slightly change Oscar's scenario. When seeing a girl sitting on the chair and recognizing that she is either Judy or Trudy, he utters ‘that girl, whether she is Judy or Trudy, is sitting on the chair’. Provided that ‘that girl’ refers to Judy, this sentence expresses a singular proposition which is false in the indiscriminable relevant counterfactual situation where Trudy, instead of Judy, is sitting on the chair. I think it is quite intuitive to say that he knows the content expressed by this sentence in the actual circumstances. It seems to me that resistance to such an intuition is merely motivated by a desire to save the idea that discriminative ability is a necessary condition for knowledge.

  21. 21

    This idea has been formulated and defended in various ways. See, for example, Williamson 2000; DeRose 2002; Hawthorne 2004; Stanley 2005 and Turri 2010; 2011.

  22. 22

    See, for example, Hawthorne 2004; Stanley 2005; Williamson 2005 and Hawthorne and Stanley 2008.

  23. 23

    An anonymous EJOP referee has correctly observed that the proponent of (R3) might take different morals from the fact that Oscar is warranted in asserting that ‘she is sitting on the chair’, namely, that what is asserted is not the singular proposition in question, but a descriptive proposition like ‘the girl I am looking at is sitting on the chair’, or that the knowledge norm of assertion is mistaken.I make two brief remarks about this objection. First, it is quite plausible to suppose that typically if a competent speaker sincerely and assertively utters a sentence, S, in a given context, C, she thereby asserts (among other things) the proposition semantically expressed by S in C (see, for example, Soames 2002: 3–4). According to the direct reference theory of singular terms, on the other hand, the semantic content of the sentence ‘she is sitting on the chair’, as uttered by Oscar in the envisaged circumstances, is the singular proposition <Judy, the property of sitting on the chair>. We can, thus, conclude that by uttering that sentence Oscar is asserting this singular proposition. Admittedly, sometimes one does not assert the semantic content of the sentence one utters, but this occurs only in special cases as when the utterance is intended metaphorically.Second, even if one rejects the knowledge norm of assertion and, instead, opts for a weaker thesis, say, that justification is the norm of assertion, it would still follow that given (R3), Oscar is not warranted is asserting that ‘she in sitting on the chair’. To see this, consider that given the way our example is constructed, Oscar is aware that he cannot discriminate between Judy and Trudy. Now, suppose, for the sake of reductio, that (R3) is correct and Oscar knows this presumed truth. Under such circumstances, Oscar knows that he does not know the proposition expressed by ‘she is sitting on the chair’. On the other hand, it can plausibly be claimed that if one knows that one does not know that ψ, one is not justified in believing that ψ (see, for example, Williamson 2000: 256). It would then follow that Oscar is not justified in believing the proposition expressed by ‘she is sitting on the chair’, and consequently, assuming that justification is the norm of assertion, he is not warranted in asserting that proposition. Thus, if (R3) is correct, even the more modest thesis that justification, rather than knowledge, is the norm of assertion cannot secure our intuition about Oscar's being warranted in asserting that ‘she is sitting on the chair’.I thank the referee for highlighting the issues raised in this footnote.

  24. 24

    This example may remind you of the famous debate about the possibility of contingent a priori propositions. It seems that both sides of the debate agree (correctly, I think) that Oscar is in a position to know proposition r. What is at dispute, however, is whether his knowledge is a priori or a posteriori. But as we have seen in the text, if (R3) is correct Oscar does not know r at all. Needless to say, for this counterexample to work against (R3), we don't need to take sides on that debate about the possibility of contingent a priori propositions.

  25. 25

    It might be suggested that what actually underpins the problem at hand is that externalism tends to undermine a fundamental part of our conception of a rational agent, according to which under normal conditions a rational agent can grasp the simple logical properties of his thoughts a priori and so would not make simple invalid inferences. For, assuming that he cannot discriminate between the concept of water and the concept of twater, Oscar is apt to equivocate and make simple invalid inferences from beliefs involving these two distinct concepts, where such mistakes cannot be corrected without using empirical information (see Boghossian 1992; 1994). Now, if this is the real source of the problem, rejecting (R3) is not of much help for the proponent of the content externalism.Notice, however, that this is a distinct problem (at least in the context of my discussion) from the one I am highlighting here, which concerns the implications of externalism for privileged self-knowledge. And, in fact, Boghossian himself has presented these problems separately: the problem concerning reasoning appears in his 1992 and 1994, while the problem about the compatibility of externalism and the privileged access thesis appears in his 1989. What I am concerned with in this paper is only the latter problem, which can be avoided, clearly, by rejecting the epistemological principle that underlies the slow switching argument. Discussion of how externalism bears on the issue of rational reasoning is beyond the scope of the paper. I thank an anonymous EJOP referee for pressing me on this point.

  26. 26

    I owe thanks to Jessica Brown, my colleagues at IPM, and an anonymous EJOP referee for their helpful comments on earlier drafts of this paper. My main debt is to Hamid Vahid whose valuable comments greatly improved the paper. I would like to thank ISCA for its generous help during this research.

Ancillary