Some Lessons from Kripke ’ s A Puzzle About Belief

The literature on Kripke’s A Puzzle About Belief has delivered convincing answers to the problem raised by Kripke. This is so both for referentialists and descriptivists. In this article I consider what I take to be the best responses of both parties and what we can learn from these responses. I argue, firstly, that the most basic cleavage when considering the semantics of belief-attribution is between theories that claim content to be transparent and theories that do not, secondly, that such substitutivitypuzzles cannot be of much use in deciding the issue between referentialist and descriptivist theories of belief-attribution and, thirdly, that the most basic challenge facing the descriptivist is to come up with a notion of content on which such content is epistemically transparent.


Introduction
In Kripke's A Puzzle About Belief (1988) he tries to disarm challenges to "direct reference" theories of proper names that are based on the apparent failure of substitutivity in propositional attitude contexts.I argue that, while both descriptivists and referentialists can successfully deal with Kripke's puzzle, there is much to learn from what I judge to be their best responses.I start by looking at Sosa (1996), who shows that the descriptivist has nothing to fear from the puzzle, as Kripke implicitly assumes Millianism in order to set up his puzzle.I will defend Sosa's arguments against Kripke's likely replies.
The referentialist emerges similarly unscathed, as Kripke makes an assumption that directly conflicts with referentialism.Denying this assumption provides the referentialist with a way to escape the puzzle.An examination of the role that this assumption plays in creating such puzzles allows for a number of matters to become plain.These are, firstly, that the most basic cleavage when considering the semantics of belief-attribution is between theories that claim content to be transparent and theories that do not, secondly, that such substitutivity-puzzles cannot be of much use in deciding the issue between referentialist and descriptivist theories of belief-attribution and, thirdly, that the most basic challenge facing the descriptivist is to come up with a notion of content on which such content is epistemically transparent.

Definitions, background and exposition
• Descriptivism is the view that "a singular referring term 'a' refers to an object o iff o uniquely satisfies some condition φ that a speaker S who understands 'a' has associated with 'a' as its propositional content" (Kallestrup 2003:1).
• Referentialism is the "view that 'a' has no descriptive content as part of its propositional content.What 'a' contributes to determine the proposition expressed by 'a is F' is simply its referent such that if S understands 'a', then S knows of that referent that 'a' refers to it" (Kallestrup 2003:2).This is also known as "Millianism".I will use these two terms interchangeably throughout.
Kripke's puzzle is aimed at a variant of Frege's puzzle, which can be stated as follows.
Consider the claims Lois Lane believes that Superman can fly and Lois Lane believes that Clark Kent can fly.The first belief-attribution is true, whereas the second is false.But a change in truth-value surely reflects a change in semantic value.This implies that the two claims and, by extension, Clark Kent and Superman, are not semantically equivalent.This would imply that Millianism is false.
Kripke tries to draw the sting from Frege's puzzle by creating a similar paradox but one which does not rest on Millianism in any way.The idea is that this would show that it is illegitimate to blame the paradox on Millianism and the force of the Fregean puzzle would be shown to be illusory.

Kripke's puzzle
Kripke's puzzle rests on two principles.The first is the "disquotation principle" (DP), which states that "if a normal English speaker, on reflection, sincerely assents to 'p', then he believes that p" (Kripke 1988:112).Kripke also states a biconditional form of the DP, namely that "A normal English speaker who is not reticent will be disposed to sincere reflective assent to 'p' iff he believes that p" (Kripke 1988:113).The biconditional DP implies that failure to assent indicates lack of belief, as assent indicates belief (Kripke 1988:113).The two versions of the DP are formulated for English, but it is assumed that the DP will hold in any language (Kripke 1988:113).
The second principle that Kripke states is the "principle of translation" (TP).It states that, "If a sentence of one language expresses a truth in that language, then any translation of it into any other language also expresses a truth (in that other language)" (Kripke 1988:113).
Kripke's puzzle concerns Pierre, a monolingual Frenchman who has heard of London (he calls it Londres).On the basis of what he has heard of London, he is inclined to assent to the sentence Londres est jolie.Using the DP and TP we can conclude: (1) Pierre believes that London is pretty.
Later Pierre moves to London and is assumed to learn English by "direct method" (Kripke 1988:119).Pierre's surroundings are assumed to be unattractive, so that he is inclined to assent to the sentence London is not pretty.Using the DP, we can conclude: (2) Pierre believes that London is not pretty.
The conclusion drawn above regarding Pierre does not imply that he withdraws his assent from Londres est jolie.Rather, he "takes it for granted that the ugly city in which he is now stuck is distinct from the enchanted city he heard about in France" (Kripke 1988:120).
Kripke proceeds by considering whether or not we can legitimately attribute any of the above beliefs to Pierre (Kripke 1988:120).Specifically, he considers the options of denying that Pierre holds (1) or (2), or denying that he holds either, but dismisses such options (Kripke 1988:120-121).Pierre definitely did, at one stage, hold the belief that London is pretty, and he never changed his mind or gave up his original belief (Kripke 1988:120).His new belief cannot be denied either, despite his French background.If Pierre suddenly gets an electric shock that wipes out all his memories of the French language he would be in the same positions as his English neighbours who assent to the utterance London is not pretty.Since their belief cannot be denied, his cannot either.But the electric shock could not, surely, have given him a new belief, and so his French past is simply irrelevant to whether he has the new belief or not (Kripke 1988:120).Denying him both beliefs combines the trouble associated with the above options, and so cannot be correct either (Kripke 1988:120).Rather, it would seem that Pierre holds both beliefs, i.e. that he has contradictory beliefs (Kripke 1988:122).But this option seems to lead to insuperable difficulties.We can assume that Pierre is a leading philosopher and logician, and "surely anyone, leading logician or no, is in principle in a position to notice and correct contradictory beliefs if he has them" (Kripke 1988:122).The above options, however, would seem to exhaust the logically available alternatives.
Kripke also constructs the so-called "Paderewski-puzzle".This is used to show that the above problem can also arise within a single language, using phonetically identical tokens of a single name (Kripke 1988:130).Peter is assumed to learn about Paderewski, the famous Polish musician and statesman.Peter mistakenly assumes that he has learned about two different people (Kripke 1988:130).Because of this, he is inclined to assent to Paderewski had musical talent, and to Paderewski had no musical talent.Using the DP, and, the homophonic TP1 , the following can be concluded: (3) Peter believes that Paderewski had musical talent and (4) Peter believes that Paderewski had no musical talent.
The above situation is parallel with the situation of Pierre in London.None of the logically possible options regarding attributing beliefs to Peter seem satisfactory.
Kripke presents his puzzles, but states that he has no firm belief on how to solve them (Kripke 1988:123).His main contention is that these are genuine puzzles and that none of the logically possible ways of attributing beliefs to Pierre or Peter seem sound (Kripke 1988:123).Yet, these puzzles seem intuitively similar to the puzzles regarding substitutivity that have traditionally been used to criticise referentialist views.In his conclusion to the paper, Kripke states the following: It is wrong to blame unpalatable conclusions about Jones 2 on substitutivity.The reason does not lie in any specific fallacy in the argument but rather in the nature of the realm being entered.Jones' case is just like 'Pierre's': both are in an area where our normal practices of attributing belief, based on principles of disquotation and translation or on similar principles, are questionable (Kripke 1988:133).
Kripke's main contention is that the "Pierre-puzzle" shows that the substitution-failures in propositional attitude contexts that are normally blamed on a substitutivity principle licensed by referentialism can be generated without using any such substitutivity principles.If correct, this contention would seem to disarm the argument from substitutivity failure as a criticism of referentialism.

4.1
Translation rejected -Sosa Sosa (1996) considers the question as to the exact principle responsible for Kripke's puzzle, and identifies what he terms the "Hermeneutic principle" as the culprit.The "Hermeneutic principle" can be stated as follows: "If a name in ordinary language has a single referent then it may correctly be represented logically by a single constant" (Sosa 1996:388).
The claim that this principle is assumed in the translation practices that Kripke refers to does not need much defence.It is quite clearly shown by Kripke's use of Londres/London, Paderewski/Paderewski, and others that this is being assumed.It is this principle, in conjunction with the TP, DP and compositionality, which leads to the puzzle.
Sosa then proceeds to state that Kripke's puzzle, and similar puzzles, can now be seen as a reductio of the above "Hermeneutic principle" (Sosa 1996:388).And yet Sosa does not stop there.He now proceeds to try and determine the assumption behind the "Hermeneutic puzzle" that is leading to all the trouble.He comes up with the following: …the presupposition I am challenging is as follows: ambiguity is a matter of multiple reference.If we were to require, for a term to be ambiguous, that it have more than one referent, then, it seems to me, we would presuppose Millianism; such a requirement excludes a Fregean position in which a name with a single referent is ambiguous in virtue of having more than one sense.And Kripke's project is precisely to recreate the difficulty without presupposing Millianism (Sosa 1996:391).
Sosa must surely be granted his claim that, if ambiguity is taken to be solely a matter of multiple reference (for proper names), then it is also assumed that the referents of proper 2 "Jones" in the above quotation comes from the example used by Kripke to demonstrate putative substitutionfailure.Kripke (1988:133) uses the typical example of Jones who believes that Cicero was bald, but does not believe that Tully was bald.This example demonstrates the same point as the one concerning Lois Lane's beliefs about Superman/Clark Kent's ability to fly.
names exhaust their meanings.In other words, referentialism is assumed, and is a hidden assumption behind the translation principle.In this case Kripke's puzzle, if we assume compositionality, is an unintended reductio of referentialism in the same way that Frege's puzzle is intended to be a reductio of referentialism.Hence the principle that the descriptivist rejects in resolving Kripke's puzzle turns out to be referentialism itself.
As stated earlier, Kripke does present some arguments against the rejection of the TP.It is a weakness of the literature on Kripke's paper that commentators often seem to focus solely on the puzzle itself, without considering Kripke's replies in the article to positions that he anticipated some might take.I will consider the most important of these objections below.

Statement of Kripke's Arguments
Three main objections will be considered.The first concerns what would happen if descriptivism is claimed to be correct.Kripke foresees the option of simply stating that London and Londres will be semantically distinct because of descriptivist semantics, i.e. that they will be semantically equivalent to two different definite descriptions.However, here he objects that, even if we assume that the same properties are used to define both London and Londres, the puzzle can still arise (Kripke 1988:126).If we assume that London is being defined with reference to, for instance, being the capital of England or with regard to being the city in which Buckingham Palace is located, the paradox is not necessarily averted.Kripke states that Pierre "need only conclude that 'England' and 'Angleterre' name two different countries, that 'Buckingham Palace' and 'le palais de Buckingham' (recall the pronunciation) name two different palaces, and so on" (Kripke 1988:126).The same might even be the case for "natural kind" terms.If a speaker can be bilingual, rational and linguistically competent without realising that terms like rabbit and lapin have the same meaning, then "natural kind" terms are also excluded from ultimately featuring in the meaning of proper names.
The second objection concerns the TP as such.Kripke considers the option of simply denying London as a translation of Londres (Kripke 1988:128).His first objection is that this would violate standard practice; London is routinely presented as the translation of Londres (Kripke 1988:128).This would inevitably lead to the "drastic" proposal "decree that no sentence containing a name can be translated except by a sentence containing the phonetically identical name" (Kripke 1988:128).Kripke then refers to his work in Naming and Necessity and the link he established between names and "natural kind" terms (Kripke 1988:129).If the link established there is accepted, then the ban on translating proper names would have to be extended to a ban on translating "natural kind" terms.If Pierre is neither a botanist nor a zoologist, then he can, presumably, be under the impression that lapin and rabbit, or beech and hêtre, refer to superficially similar but distinct species (Kripke 1988:129).If this is correct, then the "translation ban" would have to have even more ridiculous scope.
The third objection tries to show that even the widest conceivable ban on translation would not end the trouble.Here Kripke presents the "Paderewski-puzzle", explained earlier, which shows how analogous trouble can arise within a single language.

Reply to Kripke's Arguments
Kripke's first objection, can, in a sense, be dealt with quite easily.If one holds that a name acquires a meaning in terms of an associated definite description, then surely consistency demands that any names in this description also acquire meaning in the same way.Kripke foresees this possibility, and writes that such a position would have to hold that eventually an "ultimate" level would have to be reached where names are only defined in terms of "pure" properties; yet he knows of no convincing argument why this should be possible (Kripke 1988:127).
Kripke's point about such an "ultimate level" is a good one, but is hardly news.The descriptivist, just in virtue of the definition of "descriptivism", would have to hold that all names are ultimately definable in terms of definite descriptions that do not involve names.Whether this can be done or not is a thorny issue, but it is something only a very inattentive descriptivist would not have been aware of all along.This is not a previously unseen implication of descriptivism that can now be seen to back the descriptivist into even more of a corner than he is in already.Rather it is something the descriptivist would have believed all along, and will be readily conceded.This is not to say that showing this to be possible will be easy.It simply means that there is no consideration here that is specific to Kripke's puzzle that adds anything new to the debate.
Similar considerations apply with regards to "natural kind" terms.It is, of course, the case that the descriptivist would not be able to accept the Kripkean account of the semantics of "natural kind" terms, accept that "natural kind" terms can be used to define names, and still avoid Kripke's puzzle3 .Here the descriptivist would have to either reject Kripke's analysis of "natural kind" terms, or give some account of why they would not feature in the definite descriptions that define names.Since Kripke's arguments concerning natural kind terms in Naming and Necessity (Kripke 1980) are analogous to his arguments concerning names, and since the descriptivist believes that Kripke misconstrues the semantics of proper names, one would hardly expect the descriptivist to endorse the Kripkean analysis.In other words, the first option would seem to be the natural route for a descriptivist to take.But, however the descriptivist tries to block Kripke's argument, this is still hardly a new point, nor is Kripke's puzzle needed to reveal it.Once again, Kripke's point, while definitely constituting a challenge, does not force the descriptivist toward a commitment that he would not have had to hold anyway.
Kripke's second point is that any rejection of the TP would have to have a very wide scope, in banning all names, and maybe even "natural kind" terms, from being translated.This would seem to flagrantly violate standard practice.
There are a variety of different answers that the descriptivist could give to such an argument.One, of course, would be to ban all translations that involve proper names (or "natural kind" terms).But this would seem to be a very drastic step, one that is taken at a great theoretical cost.
This is not, however, the only option that is open to the descriptivist.A rejection of the TP does not, as Kripke seems to imply, involve a commitment to a TP that explicitly excludes all names (and "natural kind" terms).Rather the TP can be amended in other ways that still preserve the vast majority of our translation practices.Below I will present one such amended translation principle.If the reader is not convinced of it then the general point as to the possibility of other amended translation principles still stands.However, I do think that the following proposal has considerable merit.
Consider the case of a descriptivist who holds the view that homophonic translation will often, but not always, be successful.In other words he believes that the conditions that an object uniquely satisfies, and in virtue of which a name that refers to it acquires meaning, will often be preserved in belief reports.Now imagine that such a descriptivist holds the same belief regarding ordinary translation.The argument to motivate such beliefs would be a difficult one, but it is a possible position.Such a descriptivist would obviously only want to ban translation in those cases where the meaning will not be preserved; he would not wish to ban it in all cases.
A descriptivist taking such an approach would have the option of treating the TP not as an absolute rule, but as a methodological rule of thumb that is sure to have exceptions.Consider a version of the TP that runs as follows: • Amended Translation Principle: (TP'): If a sentence of one language expresses a truth in that language, then any translation of it into any other language is assumed to also express a truth (in that other language), unless there is clear evidence to the contrary.
The above translation principle would not treat the original TP as an immutable truth, but as a general methodological rule that does not always apply.In other words, and as a lawyer would say, it is a "rebuttable assumption".TP' rests on the assumption that the "standard" translation of names would often be valid on descriptivist grounds, i.e. that our "standard" translation of names would mostly succeed in preserving the meaning of the translated name.
In Fregean terms, this amounts to the assumption that "standard" translations are mostly sense-preserving.The "Pierre-puzzle", i.e. if a rational and competent speaker of two languages is willing to assent to beliefs that would be contradictory if the TP is used, is exactly the type of case where there would be a prima facie case that the TP does not apply.This is not to say that all such translations would be blocked.Rather it amounts to the claim that, if we have evidence of a speaker's rationality and linguistic competence, and still translation leads to seemingly contradictory belief-ascriptions, then this would be sufficient evidence to rebut the assumption in favour of the validity of this specific application of the TP.
It is necessary to insert a clarification at this point.I am not implying that only descriptivist premises can lead to the use of TP'.Although I will not do so, the argument in the second half of this essay can be used to give a motivation for TP', or a close variant of it, that is independent of such motivation.
The amended TP would allow us to gain all the advantages of using the TP, but without the puzzling consequences that arise in extreme cases.Here the reader might well object that the above step is somehow "too easy", and that introducing the equivalent of epicycles into the TP does not really address the problem.Below I will argue that a principled and motivated case for TP' over TP can be made.Kripke is too quick to assume that his TP does present an accurate picture of present translation practices and does capture our intuitions regarding translation.Rather the consideration of borderline cases makes it clear that, in practice, we do allow the existence of contradictory data to block the application of the TP.
The above claim can be illustrated with reference to a hypothetical case constructed in Over (1983)  It seems reasonable to suppose that the non-philosopher would use the "word salad" of French and English in reporting this belief.In fact, such a case seems exactly the type of case that the use of "word salads" is good for.If he tells his friends The poor idiot next to me thought he was in Londres, but that London was still some distance away, we would not take him to be ignorant of the principles of translation, or violating standard usage.Rather we would credit him with realising that the TP does not apply to belief-attribution in this specific case.
The above case, and others like it5 , seems to indicate that standard practice does recognise that the TP does not always apply.Standard practice does not support the TP, but rather TP', which in effect treats the TP as a provisional and rebuttable assumption.Hence Kripke's assertion that a rejection of the TP would entail a blanket condemnation of the translation of proper names is too quick.Rather it would call for some slight amendment of the TP like TP', which, as argued above, would be closer to reflecting standard practice, and our intuitions, than the TP.
Kripke's objection regarding "natural kind" terms can be dealt with in the same way.The standard translation of "natural kind" terms can be assumed to preserve the "meaning" of a "natural kind" term, unless evidence strongly suggests otherwise.Once again the idea of a "blanket ban" can be dismissed, and TP' can give us the advantages of the TP without running into paradox.
Similar considerations apply to Kripke's third argument concerning the "Paderewski-puzzle".As Kripke notes, reporting Peter's belief regarding Paderewski is to assume a homophonic translation of Peter's language into our own (Kripke 1988:131).Here a homophonic version of TP' can be used to justify such translation in general, but to keep the options of admitting exceptions open.
Kripke, of course, foresees the part of the above argument that rests on treating different speakers' use of a single language as constituting different idiolects of such a language (Kripke 1988:145).He objects that, before being confronted with the second use of Paderewski, the interpreter would have had no trouble with treating Paderewski as unambiguous (Kripke 1988:145).It is only after the second utterance that that we will suddenly object, and this seems a rather ad hoc way of avoiding contradiction.
The above reasoning might hold on Kripke's assumption of the TP, but on TP' this can be shown to be a principled move.If we operate on the assumption that the homophonic TP will generally hold, then this justifies treating the first utterance of Paderewski as unambiguous.But, on a homophonic TP', such an assumption is merely provisional, and can be falsified by subsequent evidence.Hence, if we have independent grounds for assuming Peter to be rational and linguistically competent, it would not be an ad hoc move to retroactively revise our original treatment of Paderewski as unambiguous.If TP is not treated as an absolute rule, but as a rebuttable presumption, then revision in light of subsequent evidence is exactly the type of behaviour that it would sanction.
In short, it can be concluded that the three objections anticipated by Kripke fail to provide any reason to dismiss an analysis of his puzzle that rests on a rejection of TP.His argument concerning names and "natural kind" terms used in definite descriptions that constitute the meaning of names only forces the descriptivist into positions he would have had to hold anyway.His claim that the TP represents standard practice, and that abandoning it would require a ban on the translation of all names, "natural kind" terms, or even homophonic translation, has been argued to be wrong on both counts.TP does not seem to constitute standard practice or accord with intuition in all cases.Nor does the rejection of TP imply a commitment to a blanket ban on the translation of all names, "natural kind" terms, homophonic translation, etc6 .
It has now been argued that Kripke's puzzle fails to disarm the objection to referentialism based on Frege's puzzle.It does this because the descriptivist would reject the referentialist commitments of the TP.None of Kripke's arguments against such a move point out any difficulty for the descriptivist that he would not have been committed to resolving already.

Kripke's unwitting endorsement of internalism
A number of philosophers have pointed out that Kripke makes an assumption which conflicts with his referentialism.If this assumption is denied, Pierre's situation is no longer puzzling to the referentialist.This assumption is one that Boghossian (1994:33) has called "epistemic transparency".He formulates it as follows: Epistemic content is transparent if, and only if, "[when]…two of a thinker's token thoughts possess the same content, then the thinker must be able to know a priori that they do; and (b) if two of a thinker's token thoughts possess distinct contents, then the thinker must be able to know a priori that they do" (Boghossian 1994:36) 7 .
The relevance of epistemic transparency becomes evident if we look at Kripke's construal of Pierre's situation.When discussing the possibility of Pierre having conflicting beliefs, Kripke describes Pierre as a "leading philosopher and logician" (Kripke 1988:123).This is done in order to discredit the possibility of Pierre having contradictory beliefs, since, surely, such a person would not have contradictory beliefs.Kripke also seems to assume that having contradictory beliefs can only be possible as a result of faulty reasoning when he writes that Pierre "lacks information, not logical acumen.He cannot be convicted of inconsistency, to do so is incorrect" (Kripke 1988:123).Kripke further states that we hold people with contradictory beliefs subject to greater censure than people with merely false beliefs.He attributes this standard to the assumption that anyone is, in principle, in a position to notice and correct any contradictory beliefs that they may have (Kripke 1988:123).
Kripke assumes that correcting contradictory beliefs is a matter of logical acumen, not a matter of acquiring information.But this seems to hinge on whether epistemic transparency is true or not.If a person knows a priori whether or not the propositions expressed by two token thoughts are the same, then the logical properties of such propositions, i.e. whether they are consistent or contradictory, are similarly a priori.In other words, epistemic transparency would imply that determining and correcting contradictory beliefs is a matter of logical acumen, rather than acquiring information.
Epistemic opacity (the denial of epistemic transparency) would imply that logical acumen is not sufficient to detect all contradictory beliefs.The person would not be in a position to determine and correct all potentially contradictory beliefs, unless he has acquired information regarding the content of the terms he used and thereby gained knowledge of the logical properties of the propositions expressed by two given sentences.
Epistemic opacity would seem to be a direct implication of holding an externalist conception of mental content.In such a conception of mental content, a "…subject's intentional states are individuated in part by certain sorts of facts about the physical and/or social environment in which he happens to be situated" (Boghossian 1994:34).If my intentional states are individuated in terms of (physical or social) external facts, then I cannot determine the logical properties of propositions expressed by token thoughts that are individuated in such a way without reference to, and knowledge of, these external facts.In other words, externalism would imply that determining the consistency of two token thoughts is sometimes a matter of acquiring information after all, and not, as Kripke assumes, only a matter of logical acumen.
The discussion above illustrates how Kripke's puzzle can be turned into a non-puzzle for the referentialist.Pierre can be ascribed both the beliefs that London is pretty and that London is not pretty.Kripke's objection that we can assume Pierre to be a leading philosopher and logician, and as such will never have contradictory beliefs, does not matter once epistemic transparency has been rejected.Rejecting epistemic transparency would entail that a leading philosopher and logician can have contradictory beliefs, without this being puzzling.
A number of philosophers have pointed out that Pierre's beliefs can coherently be described as contradictory if epistemic transparency is rejected.Lewis (1999:416) states that "[i]f the assignment of propositional objects characterises more than the believer's inner state, then there is no reason to suppose that a leading philosopher and logician would never let contradictory beliefs pass, or that anyone is in a position to notice and correct contradictory beliefs if he has them".After similar reasoning, Kallestrup (2003:112-113) concludes that "[s]ingular propositions are individuated by their objectual constituents independently of how they are conceptualised.It is therefore not to be expected that speakers who entertain such propositions can come to know their logical properties by deploying their conceptual apparatus from the armchair… If this is right, then it would seem that we have a neat explanation of why Kripke failed to solve his own paradox: Referentialism and Epistemic Transparency are jointly inconsistent".
Kripke's tacit endorsement of epistemic transparency is especially puzzling if one considers his endorsement, here and in Naming and Necessity, of referentialism as a more adequate "picture" of reference than descriptivism.A referentialist view of the semantic content of proper names seems to represent a paradigm case of externalist content leading to epistemic opacity.Lois Lane, for instance, believes of Superman both that he can and that he cannot fly.But, as stated by Kallestrup (2003:112), "[t]he fact that her beliefs have logically contradictory properties is not reflectively accessible to her; it can only be discovered by appropriate empirical means"8 .
Hence, it would seem that there is an easy way to block Kripke's puzzle: simply reject epistemic transparency, and the puzzle disappears9 .The referentialist can attribute contradictory beliefs to Pierre, safe in the knowledge that such attributions are made possible by the very nature of externalism.This is not, however, all that can be said on the matter.A strengthened version of Kripke's puzzle has been constructed that purports not to rest on epistemic transparency.It will be addressed below.

5.2
Kallestrup's "Strengthened Puzzle" Kallestrup (2003) has argued that a strengthened version of Kripke's puzzle which does not rest on epistemic transparency can be constructed.This attempt fails, but helps to clarify some issues regarding descriptivism and Frege's puzzle.
Having considered Kripke's puzzle, Kallestrup (2003) concludes that the referentialist can avert the puzzle if epistemic transparency is rejected.This does not mean, however, that the entire problem regarding referentialism and substitutivity has been resolved for cases like that of Pierre.Kallestrup attempts to show this by constructing a strengthened version of Kripke's puzzle that does not rest on epistemic transparency: First, he formulates an amended version of the Disquotation principle (DP').The DP' links dissent with absence of belief.
DP': "If a competent speaker S sincerely and reflectively dissents from 'a is F', then it is not the case that S believes that a is F" (Kallestrup 2003:113).
This can be used to construct a strengthened version of Kripke's puzzle.In Kripke's original puzzle, he constructs a situation in which the propositions expressed by the embedded sentences in belief ascriptions are contradictory.In other words, the embedded sentences London is pretty and London is not pretty, as embedded in Pierre believes that London is pretty and London is not pretty, are contradictory.The possibility of this being an accurate belief ascription is then challenged by Kripke on the basis that Pierre would be sufficiently rational not to believe contradictory propositions.This reasoning, however, has been argued to be fallacious based on the assumption of epistemic transparency.Kallestrup (2003:114) uses the DP' to formulate a belief-ascription where the propositions expressed by the entire belief-ascription, and not the propositions expressed by the embedded sentences, are contradictory.Kallestrup's strengthened puzzle specifically concerns Kripke's Paderewski-puzzle, but it does not differ from the Pierre-puzzle in any way that will affect the argument here.If a subject S assents to a is F and dissents from a is F, then, by an application of the DP and DP', the conclusion S believes that a is F, and it is not the case that S believes that a is F can be reached.If the two tokens of the word-type "a" are co-referential, as they are in the Paderewski-case, this belief-attribution clearly expresses a contradiction.
This goes much further than what the original DP allows us to deduce.Use of the DP allows us to conclude no more than S believes that a is F and S believes that it is not the case that a is F.This sentence ascribes contradictory beliefs to S, but is not, itself, a contradiction.DP' is much stronger, and allows us to conclude S believes that a is F, and it is not the case that S believes that a is F. We have gone from ascribing contradictory beliefs to S, which is possible, to committing to the truth of a contradiction, which is not10 .This means that a contradiction that is sufficient for a paradox or puzzle has been reached, without pronouncing on the subject's rationality or making hidden assumptions about epistemic transparency.
Kallestrup's article is intended to be a vindication of descriptivism (Kallestrup 2003:107).He claims that the descriptivist has a way of blocking the puzzle constructed above, but the referentialist does not.For the referentialist, the sentence S believes that a is F, and it is not the case that S believes that a is F would express two contradictory propositions.However, the logical form of these ascriptions would be different for the descriptivist (Kallestrup 2003:114).The singular term "a" would have different semantic content in the two propositions, since "a" would have different descriptive conditions, both of which are uniquely satisfied by the individual that these descriptive conditions are true of, associated with it in these two instances.In other words, the propositions expressed by S believes that a is F, and it is not the case that S believes that a is F would not be contradictory on the descriptivist account.This allows Kallestrup to conclude that the strengthened version of Kripke's puzzle has the unintended consequence of counting as a puzzle for referentialism, but not for descriptivism (Kallestrup 2003:114).
Kallestrup's strategy is an attempt11 to change the implication of Kripke's puzzle so that, contrary to Kripke's intention, it turns out to count against referentialism.His argument, however, fails.His "strengthened puzzle" contains a hidden commitment to epistemic transparency and is not a strengthened puzzle after all.
Consider DP', the principle which states that, "if a competent speaker S sincerely and reflectively dissents from 'a is F', then it is not the case that S believes that a is F" (Kallestrup 2003:113).Such dissent will presumably take the form or something equivalent to it, of a subject uttering a sentence like: It is not the case that a is F. Based on this utterance, DP' allows us to conclude that It is not the case that S believes that a is F.
However, DP' is only valid if we assume epistemic transparency to be valid!Consider the case where I know that a subject believes some proposition to be untrue and the subject expresses this by dissenting from a is F or simply stating It is not the case that a is F. When can I safely use this knowledge in order to conclude It is not the case that S believes that a is F, i.e. that he does not believe the proposition in question?Clearly, I would need to know that he is rational and that he knows the relevant logical properties of all the relevant propositions in question.In other words, I would have to know that he does not, by chance, happen to believe the proposition in question without realising that it is the same proposition that he believes to be untrue.If epistemic transparency is true, this would not be a problem, but epistemic transparency is precisely what Kallestrup is trying to keep out of his analysis.And yet, I cannot conclude anything stronger than what DP affords me, unless I know that the subject knows the relevant "meaning facts".Furthermore, DP' cannot be held to be universally true unless all rational subjects will always know these meaning facts, which can only be the case if epistemic transparency is true.This implies that Kallestrup's attempt to construct a version of the puzzle that does not rest on epistemic transparency relies on a principle, DP', which can only be universally true if epistemic transparency is true.If epistemic transparency is false, then it is a trivial matter to construct cases where S's dissent from 'a is F' does not imply that it is not the case that S believes that a is F.
In short, Kallestrup's "strengthened puzzle" poses no problem for the referentialist.In Kripke's original puzzle epistemic transparency was assumed in thinking that detecting contradictory beliefs is only a matter of logical acumen.Kallestrup's use of DP' is intended to pose a similar problem without assuming the truth of epistemic transparency.But DP' assumes the truth of epistemic transparency once again, and in quite an obvious way.The referentialist should reject DP' and hence emerges unscathed from this attack.
This type of argument is not novel.Soames (2005:398) has claimed that principles like DP' presuppose the truth of epistemic transparency.But there are some implications of this that have not received much attention.I will discuss some of these implications below.

Externalism vs. internalism
Kallestrup thinks that Kripke's implicit assumption of epistemic transparency is something that is peculiar to Kripkean puzzles.He tries to construct a puzzle that does not depend on the issue of epistemic transparency and fails to do so.I would argue that his failure is due to the fact that he is trying to do the impossible.Epistemic transparency is not something that only affects Kripkean puzzles, but is, in fact, what all such puzzles turn on.Content that is not transparent will generate the linguistic phenomena that strike the descriptivist as constituting substitutivity failures; content that is transparent will not, and these seeming failures will arise, or not arise, in virtue of these facts.
The above claim can be vindicated by noticing the following: all the puzzles, whether Kripkean or Fregean, involve a situation where an externalist would contradict the truth-value that the subject of a belief-attribution would assign to the belief-attribution.In the case of Lois Lane, for instance, an externalist would affirm that Lois Lane believes that Clark Kent can fly is true, whereas Lois Lane would deny it.This, in conjunction with the fact that our ordinary "way of talking" when attributing beliefs always agrees with the truth-value assignment made by any rational subject of a belief-attribution, then suffices to generate problematic beliefattributions like Lois Lane does not believe that Clark Kent can fly.Hence we can characterise the problem cases as those where an externalist would contradict the assignment of truthvalues of a rational agent.But we can only differ from the truth-value assignments of a rational agent if content is externalist.This implies that all externalist theories will result in attributions that contradict our normal practice of belief-attribution, while internalist theories need not have these results.This distinction12 , and not the descriptivist/Millian distinction, is the fundamental distinction in the semantics of belief-attributions.

5.4
The relationship between Frege's puzzle and Kripke's puzzle Kripke's puzzle attempts to arrive at a contradiction by stipulating that a subject is rational and then showing how the DP and TP lead to the subject having contradictory beliefs.This is supposed to be irrational, and hence a paradox arises.It is also, of course, possible to construct a Kripkean variant of Frege's puzzle which implies that a perfectly rational subject has contradictory beliefs.Kallestrup does this in his article.In his strengthened Frege's puzzle, he considers the case of a subject, S, who assents to a is F and dissents from b is F (Kallestrup 2003:114).However, it is assumed that the "a" and "b" are coreferential (Kallestrup 2003:114).Using the DP, and DP', we can conclude "S believes that a is F and it is not the case that S believes that a is F" (Kallestrup 2003:114).This, clearly, is a contradiction, not between the embedded sentences as in his original Kripkean Frege's puzzle, but between the propositions expressed by the entire belief-ascribing sentence (Kallestrup 2003:114).
I have already argued that Kallestrup's use of DP' is invalid as it assumes the truth of epistemic transparency.Kallestrup's strengthened Kripkean Frege's paradox fails for the same reason.If a rejection of epistemic transparency, and, hence, of DP', implies that Pierre's contradictory beliefs in Kripke's puzzle do not amount to a paradox, then someone like Lois Lane's contradictory beliefs do not lead to a paradox for the same reason.However, I mentioned Kallestrup's strengthened Kripkean Frege's paradox, since it casts some light on Frege's original paradox.
Kallestrup's attempt to construct a version of Frege's puzzle that does not rest on epistemic transparency is easily defeated by the referentialist.Yet we intuitively feel that Kallestrup's Fregean puzzle and Frege's original puzzle trade on the same difficulties.If this is correct, then the referentialist should have a similarly simple answer to offer to Frege.This is not quite the case, but it is instructive to see where the difference lies.Consider Kallestrup's conclusion, derived from S's dissent from b is F, coupled with the DP', the fact that a = b, and a substitutivity-rule.These principles are used to reach the conclusion that "It is not the case that S believes that a is F" (Kallestrup 2003:114).This claim is equivalent to the claim "S believes that a is F is false".This is the general form of statements like "Lois Lane believes that Clark Kent can fly is false", i.e. the datum of Frege's original puzzle.
The difference between Frege's puzzle for belief-contexts and Kallestrup's strengthened Kripkean Frege's puzzle is obvious: Frege's puzzle simply assumes the truth of the datum, whereas Kallestrup tries to derive the datum from dissent from a sentence like Clark Kent can fly, which has been shown to be unsuccessful.Here we might ask whether any disquotation principle whatsoever can be used to establish the truth of the datum, as used in Frege's puzzle, without presupposing epistemic transparency.I do not see how this is possible.We have no access to people's beliefs other than via their behaviour, and most importantly their utterances.There are no utterances that, coupled only with the normal DP, can guarantee a conclusion like the datum.The only way to move from utterances to conclusions like the datum is to assume that the person does not happen to hold the belief that that they deny having, without realising that this belief happens to have the same content as the one they deny holding.But to make this assumption is to assume the truth of epistemic transparency, and so cannot be allowed by the referentialist.We simply cannot, based on utterances, arrive at claims like "Lois Lane believes that Clark Kent can fly is false" without utilising principles like DP' 13 .The normal DP cannot allow us to reach such a conclusion.
This leaves the descriptivist in an odd position.If the datum is correct, then referentialism is false.But there is now a strict limit to the evidence that the descriptivist can offer to support the datum.If disquotational principles are not allowed, then all that is left for the descriptivist is to insist that the datum is correct since it fits the way we talk.We do, ordinarily, consider claims like the datum to be correct.Given the situation of Lois Lane, we simply do characterise her as lacking the belief that Clark Kent can fly.Data regarding how we talk, however, do not always amount to a conclusive case.Various referentialists, for instance Salmon (1986), have exploited this fact to argue that the datum is, in fact, false, and that our ordinary way of speaking is to be explained with reference to Gricean implicatures and the like 14 .It would be comforting to the descriptivist if some further grounds for his trust in the 13 Kripke's "biconditional DP'", which links absence of dissent to failure to believe, would be able to justify this conclusion, but it also rests on epistemic transparency in an obvious way.This is also pointed out by Salmon (1986:132). 14Salmon claims that we often take a true sentence like Lois Lane believes that Clark Kent can fly, which actually expresses a true singular proposition, to include the Gricean implicature that Lois Lane would assent to Clark Kent can fly.Here the implicature of the sentence would "lead speakers to deny it, despite its literal truthconditions [being] fulfilled" (Salmon 1986:115).He also attributes the way we speak to practices where it is often our purpose to convey not only the believed proposition, but also the way, or guise, under which it is datum could be provided.But this now appears to be impossible.No utterance can lead to the datum without the use of a disquotation principle that simply stipulatively decides the issue 15 .
The only way to fully demonstrate the truth of the datum would be for the descriptivist to provide a full descriptivist semantics of belief-contexts which answers all the traditional objections to this position, and which has the consequence that the datum is correct.But if this was possible we would not need to worry about propositional attitude contexts and the datum in the first place.The very point of such puzzles is to give us data that can be used to construct theories and decide the issue between descriptivists and referentialists.If we can only be sure about the truth-values of claims made in the puzzles themselves once we have answered the fundamental questions of semantics, it follows that the puzzles themselves are of no real independent use.The puzzles are only useful if they are independent; if this data becomes theory-dependent they lose their utility.Hence, the role of epistemic transparency should come as good news to referentialists.It guarantees that the descriptivist will be deprived of a positive argument for one of his most valued claims, namely the truth of the datum.

Epistemic transparency and the descriptivist
I have argued that it is the opacity of referentialist content that gets the referentialist in trouble.When it comes to descriptivism the opposite is true; it is the commitment to epistemic transparency that allows him to escape the puzzles, not his descriptivism as such.This is trivially obvious when we consider what would happen if a descriptivist chose to reject epistemic transparency.Such an unorthodox descriptivist would immediately incur substitution failures in propositional attitude contexts.
Descriptivism, as the term is standardly used, is simply a doctrine stating that the meaning of a term is to be identified with the descriptive conditions associated with a term by a speaker.The issue of epistemic transparency or opacity is conceptually distinct.This leads to a further interesting conclusion.Consider the orthodox descriptivist, i.e. a descriptivist who holds that the descriptivist content in question is epistemically transparent.Ordinarily such a descriptivist would hold that his success in dealing with puzzles concerning belief-contexts is due to only allowing substitutions that do not change the content of the propositions in question.This is incorrect, as the possibility of the unorthodox descriptivist demonstrates.The substitution-failures can arise despite the fact that meaning is preserved if the descriptivist content is not transparent.The success of the orthodox descriptivist is to be attributed to his commitment to epistemic transparency.
This places a further constraint on descriptivist theories that are supposed to gain respectability in virtue of surviving puzzles concerning belief-contexts.Such theories need a very strong commitment to epistemic transparency.If a speaker associates two synonymous descriptive conditions with a term, then he needs to always be able to tell that they are believed (115)(116)(117)(118).An important criticism of this strategy is in Schiffer (1987).Salmon's response is in Salmon (1989). 15A possible escape for the descriptivist would be to simply hold that it is nonsensical to argue that how we talk can be incorrect.One could argue that the job of semantics simply is to treat our common attribution of truthvalues as data and then to construct a semantics that gives a neat account of these data.This is not an option that can fruitfully be explored here.Most descriptivists, though, would not argue for the truth of the datum on such methodological considerations.
synonymous 16 , or he will fall victim to some theoretical substitutivity-puzzle and end up no better than the referentialist.If this is correct, then anyone who accepts that claims like the datum are generally true is committed to epistemic transparency.This means that the orthodox descriptivist owes us a theory of content that delivers such transparency.
One way such a theory can work is by accepting Russell's doctrine of "acquaintance", i.e. the view that we have a "direct, cognitive relation" to the fundamental constituents of meaning (Russell 1910:108).Russell introduces the idea of acquaintance in order to have a category of objects that one simply cannot be wrong about.It is impossible to be acquainted with an object without the object actually existing and it is similarly impossible to be acquainted with a single object, and still mistake it for two distinct objects.In our terms, the objects that I am acquainted with are epistemically transparent to me due to the nature of this direct, cognitive relation.If I am not acquainted with an object, i.e. if my relation to the object is mediated in some way, then these different mediating intermediaries are a likely source of trouble.They might actually relate me to nothing, or I might mistake two such intermediaries, leading to one object, for two intermediaries leading to distinct objects.This seems true of descriptive conditions, perceptual relations and just about anything else that can mediate between me and an object in some way.These intermediaries cause states of epistemic opacity.If the orthodox descriptivist is to avoid a commitment to Russellian acquaintance, then he has to claim that at least some of these mediating intermediaries never mislead me.I do not even see how an argument for such a contention would go.The only other way to keep the commitment to epistemic transparency would be to analyse these intermediaries into objects that I am acquainted with, as Russell does.
The orthodox descriptivist can avail himself of any number of more modern theories when constructing his semantics, but at some point he will need some theoretical notions that play the role of Russell's "acquaintance" in guaranteeing utter certainty.It is, in the final analysis, the doctrine of acquaintance that enables Russell and other descriptivists to avoid substitutivity-puzzles, and only it or a similarly dramatic equivalent will do.
This might not be bad news to an orthodox Russellian, but it will not please all descriptivists.The group of things that I am acquainted with, i.e. that I cannot be wrong about, is worryingly obscure and exceedingly small.Russell famously only included sense-data, the self (perhaps), and universals in this category, and, given the strict criterion for membership, it is hard to escape his conclusions.The orthodox descriptivist might well wish to partially adopt Russellian semantics, but might wish to jettison the epistemological concerns that lead Russell to write "Knowledge by Acquaintance and Knowledge by Description".If the above reasoning is correct, it follows that the orthodox descriptivist has much less room to manoeuvre here than is commonly supposed.Affirming the truth of the datum might score points against the referentialist, but comes at the cost of committing to something like Russellian acquaintance.
The news for the descriptivist, however, is not all bad.People commonly and without hesitation do accept the truth of the datum.This means (unless we explain this away as a pragmatic phenomenon, as Salmon does) that our common practices of belief-attribution treat in order to demonstrate his claim that Kripke "misapplies the principle of translation"(Over 1983:256)  4. Consider the case of Pierre, who goes back to Paris.He now returns to London, but does so on a train that goes via the new channel tunnel.He boards in Paris, and, in a very short time, is actually in England.The train stops at a station that is labelled Londres for the benefit of French visitors.Pierre now remarks: So this is Londres.I've always wondered where Londres was.A pity I don't have time to look around -I must get to London as soon as possible(Over  1983:253).
If we follow the TP, then clearly we would have to attribute to Pierre the belief that he is in London.But this would violate standard practice, and ordinary intuition.Any ordinary person, i.e. a non-philosopher who has never heard of Kripke, but knows about Pierre's confusion concerning London and Londres, would have no difficulty in ascribing to Pierre the belief that he is in Londres, but not in London(Over 1983:254).These beliefs would, in fact, explain his behaviour in not getting off the train(Over 1983:254).