Against Kripke on Unrestricted Exportation

Abstract

This paper defends the doctrine of Unrestricted Exportation (UE) against Saul Kripke’s attack on it. According to UE, the exportation step from the de dicto belief report, S believes that α is F, together with the premise that α exists, to the de re report, S believes of α that it is F, is valid. By presenting an alleged counterexample, Kripke tries to show that UE has much more implausible consequences than its advocates would accept. By going through the details of Kripke’s scenario, I argue that UE does not commit us to the consequences Kripke associates with it.

Cet article défend la doctrine de l’exportation sans restriction (Unrestricted Exportation, UE) contre l’attaque menée par Saul Kripke. Selon l’UE, l’étape de l’exportation du rapport de dicto «S croit que α est F» (assorti de la proposition que α existe) au rapport de re «S croit de α que c’est F» est valide. En présentant un prétendu contre-exemple, Kripke essaie de montrer que l’UE a des conséquences beaucoup plus invraisemblables que ses avocats ne le reconnaissent. En examinant les détails du scénario de Kripke, j’affirme que l’UE ne nous engage pas aux conséquences que lui associe Kripke.

1. Introduction

In his article, “Unrestricted Exportation and Some Morals for the Philosophy of Language,” Saul Kripke argues that the doctrine of unrestricted exportation (henceforth, UE) “has far more sweeping consequences than its advocates have ever hinted.”¹ According to UE, the exportation step from the de dicto belief report that

1. (1) S believes that α is F

(where α is a singular term), together with the premise that α exists, to the de re belief report that

1. (2) S believes of α that it is F

is implicative.²

To show how counterintuitive UE can be, Kripke invites us to consider a subject, say Ralph, who believes that Philby is a spy, and has at least one false belief, say p. Given these assumptions, Kripke argues, the de dicto belief report (3) would be true.

1. (3) Ralph believes that the y that is Philby if p, and is the Eiffel Tower if not p, is a spy.

Assuming UE, from (3) we get the de re belief report that

1. (3*) Ralph believes of the y that is Philby if p, and is the Eiffel Tower if not p, that it is a spy.

Now, since p is false, the definite description ‘the y that is Philby if p, and is the Eiffel Tower if not p’ picks out the Eiffel Tower. So, the UE advocate will be committed to holding that (4) is true.

1. (4) Ralph believes of the Eiffel Tower that it is a spy.

This is an implausible consequence, Kripke argues, and the view from which it follows should be rejected. Here is Kripke’s characterization of the situation:

On [UE], we are unable to say what we would say intuitively. Intuitively, because of [Ralph’s] misconception that p is true, the belief ascribed in [(3)] is actually a belief about Philby, not a belief about the Eiffel Tower. For it is Philby whom [Ralph] thinks satisfies the description in [(3)]. This, however, is precisely what we cannot say on the views we have been considering.³

I agree with Kripke that (4) is implausible. I also share Kripke’s view that the standard Gricean way in which advocates of UE try to rebut the well-known ‘shortest spy’ argument against UE is unavailable here. The defender of UE argues that even when the de dicto belief report that

1. (5) Ralph believes that the shortest spy is a spy

is true on the basis of Ralph’s general beliefs that there are spies, and that no two of them are of the same height, and that they are finite in number, the de re belief report that

1. (6) Ralph believes of the shortest spy that he is a spy

is strictly speaking true, though it is inappropriate or misleading to assert (6) in normal conversational situations.⁴ According to the advocate of UE, therefore, the infelicity of (6) should be accounted for pragmatically—for example, by arguing that, although it is strictly speaking true, it conveys some false implicature. Such a Gricean maneuver, even if accepted for ‘shortest spy’-like cases, is unavailable in Kripke’s case. It is totally unmotivated to say that Ralph, and each of us who believes that Philby is a spy and has at least one false belief, believes of the Eiffel Tower that it is a spy. “It is,” as Kripke puts it, “strange that our language should contain locutions of such misleading surface form.”⁵

Despite this much agreement, I disagree with Kripke that this implausibility is a consequence of UE. To show that it is not, in the following section, I introduce another scenario that leads to a similarly implausible consequence, but one that no one would regard as a counterexample to UE. Although it is evident that this example does not threaten UE, I spend a little more time than may seem necessary explaining why this is the case. My motivation for going into such detail will be obvious in the third section. There, I will show that, despite its mesmerizing colouration, the relation between Kripke’s example and UE is just the same as that between my example and UE; thus his is not a counterexample to UE either. In the final section, I reply to a possible objection to my defence of UE against Kripke’s alleged counterexample.

2. A Non-Consequence of UE

Let us consider a scenario wherein an argument against UE does not work. Ralph is at a party, looking at Philby, who is holding a martini glass. Noticing Philby’s suspicious behaviour, and supposing that he is the only man drinking a martini at the party, Ralph mutters, ‘the man drinking a martini is a spy.’ We can say:

1. (7) Ralph believes that the man drinking a martini is a spy.

Assuming UE, from (7) we get

1. (7*) Ralph believes of the man drinking a martini that he is a spy.

Ralph is mistaken, and Philby is drinking water. Unbeknownst to Ralph, however, there is another man, Jerry, who is the only man drinking a martini at the party. Thus, the description ‘the man drinking a martini’ picks out Jerry. Therefore, the UE advocate will be committed to the truth of

1. (8) Ralph believes of Jerry that he is a spy.

This is highly implausible. Ralph does not have the faintest inkling of Jerry’s existence, and, a fortiori, has no opinion about what he is drinking. Intuitively, the belief ascribed in (7) is a belief about Philby, not a belief about Jerry, for it is Philby whom Ralph thinks satisfies the description in (7). This, however, is precisely what we cannot say on UE.

Obviously, no one would regard this as a counterexample to UE.⁶ Let me spend some time, however, explaining why this is the case. For this purpose, I must talk a little about the nature of the problem of exportation and the claim made by in UE.

In discussing the problem of exportation and trying to spell out the conditions under which the argument from (1) to (2) is valid, it is assumed that (1) is a de dicto belief report in the sense that it represents S’s belief in a way that corresponds to how S conceives of the world. The ‘that’-clause in (1), in other words, is assumed to represent the content of S’s belief. This idea is nicely captured in Ernest Sosa’s formulation of UE, in which he uses the technical belief ascription structure ‘S believes ^┌α is F ^┐’ rather than the ordinary ‘believes that’ structure used in (1):

S believes about x that it is F (or believes x to be F) if and only if there is a singular term α such that S believes ^┌α is F ^┐, where α denotes x.⁷

Here is Sosa’s definition of this technical notion:

“S believes ^┌α is F ^┐” will be short for “S has a belief (in a proposition) that, given normal circumstances, he could correctly express in our language by asserting the sentence composed of α followed by ‘is’ followed by ‘F’.”⁸

After discussing and defending UE, in the last footnote, Sosa warns us:

Note that I have never supposed that ‘S believes ^┌α is F ^┐’ is equivalent to ‘S believes that α is F.’ This is a separate question that cannot be treated here.⁹

As Kripke points out, Sosa does not explain why he avoids using the ordinary ‘believes that’ structure and why these two structures might be non-equivalent.¹⁰ However, given Sosa’s definition of ‘S believes ^┌α is F ^┐,’ it is plausible to suppose that he is aware, as has been pointed out by many philosophers, that not every occurrence of the ‘believes that’ structure in ordinary language is de dicto in the relevant sense.¹¹ Roderick M. Chisholm, for example, writes: “In suggesting that we restrict the locution ‘S believes that p’ to this de dicto sense, I am not suggesting that the locution is in fact so restricted in ordinary English. Indeed, I believe that it is not so restricted.”¹² I am also inclined to assume that, in discussing the following examples,¹³ Sosa has the distinction between ordinary ‘believes that’ and his technical de dicto structures in mind:

Speaking of the present pope we could correctly say that his classmates believed that the future pope would not amount to much. And, speaking of Columbus, we could correctly say that he believed that the land we call ‘Cuba’ was in the Indian Ocean. In these examples we have substantives occurring merely as referring terms though inside the scope of psychological operators. That is, Columbus did not believe that what he thought of as the land we (would) call ‘Cuba’ was in the Indian Ocean, since he did not think of it that way. (And similarly for the pope’s classmates.)¹⁴

It is plausible, therefore, to assume that, by introducing the technical notion of ‘S believes ^┌α is F ^┐,’ Sosa makes it explicit that, in the premise of the exportation inference, what matters is a de dicto belief report in the sense introduced above.

Now reconsider (7). The question is whether (7) is a true de dicto report of the belief expressed by Ralph in uttering the sentence

1. (9) The man drinking a martini is a spy.

The first thing to notice is that, in uttering (9) in my scenario above, Ralph is not making an attributive use of the description ‘the man drinking a martini’ to talk about whoever fits that description.¹⁵ The belief he is expressing when uttering (9) is not the general belief that whoever is the unique man drinking a martini (at the party) is a spy. On the contrary, he has a particular person in mind, i.e., Philby, and uses the description merely as a means to pick him out and to assert of him that he is a spy—that is to say, in uttering (9), Ralph is using the description referentially to express his singular belief about Philby that he is a spy.

Now, in reporting Ralph’s belief about Philby using (7), we, as reporters of the belief, might be using the description ‘the man drinking a martini’ either attributively or referentially. If we use the description attributively, UE will imply (8): by exporting the description, we get

1. (7*) Ralph believes of the man drinking a martini that he is a spy

which implies (8) by substituting the attributively used description with the coreferential name ‘Jerry.’ In this situation, however, (7) is not a true de dicto report of Ralph’s belief in the sense that it represents Ralph’s belief in a way that corresponds to how he conceives of the world. Ralph’s belief as expressed in (9) is not a belief in the general proposition that there is a unique man drinking a martini who, whoever he is, is a spy. We can make this point more explicit by using Sosa’s notion of ‘S believes ^┌α is F ^┐.’ Ralph’s belief is not such that, given normal circumstances, he could correctly express it in our language by asserting the sentence ‘the man drinking a martini is a spy,’ using the definite description attributively. So, the belief report ‘Ralph believes ^┌the man drinking a martini is a spy^┐’ is false; and, therefore, UE does not commit us to (8).

We may, of course, make the same mistake that Ralph did about what Philby is drinking, hence, in reporting his belief by uttering (7), use ‘the man drinking a martini’ referentially to refer to Philby. Given that the description is used referentially, it will be true to say that, in Sosa’s words quoted above, Ralph “has a belief … that, given normal circumstances, he could correctly express in our language by asserting the sentence” ‘the man drinking a martini is a spy.’ The belief report ‘Ralph believes ^┌the man drinking a martini is a spy^┐,’ therefore, is true. In this case, however, (7) does not imply (8). Exporting the referentially used description from (7), we get (7*), in which ‘the man drinking a martini’ is also used referentially to refer to Philby. It is thus inadmissible to substitute it with the non-coreferential proper name ‘Jerry’ to get (8). What we get is:

1. (10) Ralph believes of Philby that he is a spy,

which is intuitively correct.

To sum up, in reporting Ralph’s belief in (7), the reporter uses the description ‘the man drinking a martini’ either attributively or referentially. If the description is used attributively, the inference from (7) to (8) is valid but unsound because its premise (7), understood as a de dicto belief report, is false. However, if the reporter uses the description referentially, then (7) is a true de dicto belief report. In this case, the inference from (7) to (8) is invalid, but it is not an instance of the exportation argument.

In the next section, I draw upon the lessons learned from this non-counterexample to neutralize Kripke’s attack on UE. Before that, however, I want to make a clarification about my take on, and my use of, Keith S. Donnellan’s referential/attributive distinction to ward off a possible objection to my diagnosis of this non-counterexample (and the parallel diagnosis I provide in the next section of Kripke’s alleged counterexample to UE). In explicating why the case illustrated in this section is not a counterexample to UE, I assumed that Donnellan’s distinction has semantic significance. The claim that an utterance of (7), in which the description ‘the man drinking a martini’ is used referentially by the reporter, is a true de dicto report of Ralph’s singular belief about Philby that he is a spy, presupposes that the referentially used description semantically refers to Philby. Likewise, the move to (10) through substitution of the description in (7*), where it is used referentially to refer to Philby, with the proper name ‘Philby,’ rests on a semantic reading of Donnellan’s distinction.

It is important, however, to note that if such a semantic reading of Donnellan’s distinction is denied, my diagnosis of this non-counterexample will not be affected and the case I devised in this section will still not represent a counterexample to UE. This is because if the semantic significance thesis is rejected, then we should hold that, semantically speaking, the belief reports (7) and (7*) must be interpreted as attributive. On such a pragmatic interpretation of the distinction, even when the reporter uses the description ‘the man drinking a martini’ referentially, the belief semantically ascribed to Ralph in (7) will be the general belief that whoever is the unique man drinking a martini is a spy. As pointed out in my discussion of attributive use of the description, on this reading, UE does imply the counterintuitive result (8), but since (7) will then be false, we should not try to apply UE to it in the first place. The same point about the significance of Donnellan’s distinction must be kept in mind throughout the next section, in order to see why the fact that Kripke denies the semantic significance of the distinction¹⁶ does not undermine my objection.

3. Against Kripke

Let us now return to Kripke’s alleged counterexamples to UE. My claim is that the implausibility of the belief report (4) in Kripke’s case is not a consequence of UE, for exactly the same reason that the implausibility of (8) in my example in the previous section is not.

Before presenting my argument, let me point out a difference between Kripke’s example and my case discussed above. As Kripke says, to ascribe that

1. (3) Ralph believes that the y that is Philby if p, and is the Eiffel Tower if not p, is a spy,

we do not need to assume that Ralph, who believes of Philby that he is a spy and has the false belief that p, is “clever enough to think of the tricky sort of definite description in”¹⁷ (3), that is, ‘the y that is Philby if p, and is the Eiffel Tower if not p.’ As Kripke argues, “a belief need not be explicitly expressed or thought of to be properly ascribed to a person.”¹⁸ For (3) to be a true report of Ralph’s belief state, it suffices that he will respond properly to (3) if it is put to him. So, in Kripke’s case, unlike the ‘man drinking a martini’-case, we need not assume that the subject, Ralph, is actually using the description referentially by uttering the sentence:

1. (11) The y that is Philby if p, and is the Eiffel Tower if not p, is a spy.

Because of this difference between the two cases, I follow a two-step strategy in my response to Kripke. First, I assume that the subject is clever enough to use the description in (11). I show that, in such a case, Kripke’s example will be exactly parallel to the non-counterexample discussed in the previous section, and it is inconsequential for the same reason. Second, I show why Kripke’s case will not have a stronger force against UE if this assumption is relaxed.

So, let us start with a case in which Ralph is clever enough to form the definite description used in Kripke’s case. Ralph believes that p and believes that Philby is a spy. Based on these beliefs, he takes the description ‘the y that is Philby if p, and is the Eiffel Tower if not p’ to be satisfied by Philby, whom he has in mind, and, for whatever reason, he utters the sentence (11) to express his belief about Philby that he is a spy.

In using the description in (11), Ralph’s situation is precisely similar to what was going on in the previous example involving the description ‘the man drinking a martini.’ In both cases, Ralph has a particular man, Philby, in mind, and he uses a definite description he believes to be satisfied by that man as a means to speak about him. Ralph uses the description referentially to refer to Philby.

In such a case, Kripke’s example is as inconsequential as my case in the previous section. In reporting Ralph’s belief as in (3), if we, as reporters, use the description attributively, the absurd consequence (4) follows, but in this case (3) is not a true de dicto belief report: (3) is not a report that represents Ralph’s belief according to how he conceives of the world. This can be seen straightforwardly if we use Sosa’s notion of ‘S believes ^┌α is F ^┐.’ Ralph’s singular belief about Philby is not a belief he could, given normal circumstances, correctly express in our language by asserting the sentence (11), while the description is used attributively. Therefore,

1. (12) Ralph believes ^┌the y that is Philby if p, and is the Eiffel Tower if not p, is a spy^┐

is not true. So, UE would not, in this case, commit us to the implausible consequence (4).

In this example, too, we may make the same mistake as Ralph’s believing that p is true, and we may share his belief that Philby is a spy. Therefore, in uttering (3) or (12), we, too, may use the definite description ‘the y that is Philby if p, and is the Eiffel Tower if not p’ referentially to refer to Philby. In this case, while (3) is a true de dicto report, and (12) is a true ascription of Ralph’s belief, (4) is no longer implied by (3) or (12). By exporting the referentially used definite description, we may substitute it not with the non-coreferential term ‘the Eiffel Tower,’ but with a coreferential term such as ‘Philby’ to get

1. (13) Ralph believes of Philby that he is a spy,

which is intuitively correct.

The situation will be no different even if Ralph is not clever enough to form Kripke’s tricky description and to use it in his language or thought as a means for picking out Philby. This fact does not prevent us from reporting his belief using (3). Here we attribute (3) because Ralph will assent to the sentence (11) if it is put to him. And he does so because he believes that p and believes that Philby is a spy. Now, it is important to notice that, in assenting to (11), Ralph interprets the relevant definite description as being referentially used to refer to Philby. So, (3) is, as before, a true de dicto ascription of Ralph’s belief only if the definite description is used referentially to refer to Philby. Consequently, UE will not lead to the implausibility we see in (4).

So, once again, we are faced with a dilemma: if the description in (3) is used attributively, the inference from (3) to (4) is valid but unsound because the premise is false; if, on the other hand, it is used referentially, it is invalid but it is not an instance of the exportation argument.

4. A Reply to an Objection

My argument in the previous section, if successful, shows that, insofar as (3) is construed as a de dicto report of Ralph’s singular belief about Philby, Kripke’s attack on UE is unsuccessful. It may, however, be argued that this is not a serious objection to Kripke, since, besides this singular belief, Ralph also believes the general proposition that the y—whatever it is—that is Philby if p, and is the Eiffel Tower if not p, is a spy. Using the description attributively, (3) is a true de dicto belief report of this general belief. Therefore, on UE, we still get (4).

Although I find the attribution of such a general belief to Ralph in this situation in need of independent motivation, if not outright implausible, I will not challenge it here, because such a reading of Kripke’s alleged counterexample robs it of any new sting against UE. The main force of Kripke’s criticism against UE is that it is supposed to show that UE “has far more sweeping consequences than its advocates have ever hinted.”¹⁹ As is evident, these “more sweeping” consequences are supposed to concern the claim that UE forces us to say that a belief that is intuitively about Philby, ascribed to Ralph in (3), is of the Eiffel Tower.

On the present suggestion, however, the belief ascribed to Ralph in (3), that is, the general belief that the y—whatever it is—that is Philby if p, and is the Eiffel Tower if not p, is a spy, is intuitively a belief about Eiffel Tower. Assuming that Ralph has such a general belief and that this belief is accurately ascribed to him in (3), it may still be infelicitous to utter (4) and say of the Eiffel Tower that Ralph believes of it that it is a spy. This infelicity, however, is just another ‘shortest spy’-case, of which the advocate of UE is already aware and for which she has a solution. It is no doubt odder to say that Ralph believes of the Eiffel Tower that it is a spy. But this extra oddness is not something for which UE is responsible. This oddness is inherited from the oddness of Ralph’s general belief, according to which, depending on the truth-value of p, either Philby or the Eiffel Tower is a spy. Contrary to what Kripke seeks to establish, we are not, therefore, dealing with a case in which UE forces us to say of a belief that is actually about Philby that it is about the Eiffel Tower.

In sum, in order to argue, contra Sosa and other UE advocates, that the exportation step is not unrestrictedly implicative, Kripke puts forward an example in which UE allegedly leads us from a true de dicto report of a belief about a particular object to an obviously false de re report about another object. In this paper, I deployed Donnellan’s distinction between attributive and referential uses of definite descriptions to show that Kripke’s example either involves an inference that is invalid but not an instance of UE, or employs a false premise and so is unsound. Either way, it falls short of being a counterexample to UE.