In this paper I discuss some objections raised by Donnellan (1977) and b y Soames (2005) against the examples of contingent a priori truths proposed by Kripke (1980) and by Kaplan (1989). According to Kaplan and Kripke, the mechanism of direct reference alone can guarantee that some contingent truths can be known without any relevant experience. Both Donnellan and Soames claim that the examples brought up by Kaplan and Kripke can only be considered as real pieces of knowledge if the knowledge involved is de re, i.e., if we know the object (i.e., the referent of the proper name or of the demonstrative expression) of which something is predicated. I shall argue that Soames and Donnellan's considerations do not undermine the possibility envisaged by Kripke and Kaplan.
1 Introduction
One of the most interesting results of the so-called direct reference theories is the distinction between epistemic and metaphysical modalities. As Kripke argues in Naming and Necessity (henceforth simply NN), the notions of aprioricity and necessity are not intensionally equivalent, since the former has to do with truth in possible worlds, while the latter has to do with the justification of beliefs. But although these notions are not intensionally equivalent, it could turn out that they are extensionally equivalent, i.e., that all a priori truths are necessary and vice-versa. However, Kripke argued in NN that, if one accepts the thesis that names in natural language are rigid designators (and there are independent arguments for this thesis), then the notions of aprioricity and of necessity are not even extensionally equivalent, i.e., there are cases of a priori truths that can be known only a posteriori, as well as cases of contingent truths that can be known a priori. In Kaplan’s work on demonstratives we find a similar result coming out of the conceptual apparatus that he introduces for dealing with the phenomenon of indexicality.
In this paper I first review the essential aspects of Kripke’s and Kaplan’s examples
of necessary a posteriori and contingent a priori truths. I concentrate on the contingent
a priori. Next, I discuss the arguments against these examples raised by
In Lecture I of NN Kripke famously introduces the example of Goldbach’s conjecture as a case in which the parallel between epistemic and metaphysical modalities might collapse. Presumably, since there is no proof or disproof of the conjecture so far, there is no a priori knowledge of it. But since it is a mathematical conjecture, if it is true or false, it is necessarily so. Now suppose that a very powerful computer can tell us that an incredibly large number is an exception to Goldbach’s conjecture, a number so large that it surpasses the human capacity of effectively doing or checking the computation. In this case, since our confidence that the computer yields the right result is based a number of physical and mechanical hypotheses, our knowledge that Goldbach’s conjecture is false would be a posteriori, although we have a necessary truth here (or falsity, depending on how you take it). This example is not quite convincing as it stands, and has not convinced many people. It can, at best, work as a sort of pressure for the clarification of the notion of a priori. Kripke himself recognizes that, perhaps, a more refined definition of a priori knowledge could avoid this conclusion.
Much better are the examples coming from Kripke’s interpretation of identitystatements and from his essentialism, which are discussed in Lectures II and III. According to Kripke, if proper names like ‘Hesperus’ and ‘Phosphorus’ are rigid designators, then an identity-statement like ‘Hesperus is Phosphorus’ expresses simply that the objects referred to by the first and the second names is one and the same. That is to say, if true, the statement says that Venus is self-identical, being therefore a necessary truth. And, if false, it says of two different objects that they are one and the same, which is necessarily false. However, this is hardly something that we could know a priori, since at least some research or astronomical observation has to be done to discover that the heavenly body named Hesperus is the same named Phosphorus. How can we then account for the aposterioricity of this statement, given that it is necessary? According to Kripke, there are other possible worlds that are phenomenologically identical to ours, in which, the body named ‘Hesperus’ has exactly the same appearance as our Hesperus, the body named Phosphorus has exactly the same appearance as our Phosphorus, and nevertheless these are different bodies. Hence, given the available astronomical evidence, some research has to be done in order to determine which of the phenomenologically equivalent worlds is the actual world. Other kinds of examples come from the consideration that some properties might be essential to individuals and, nevertheless, we cannot discover these properties without some sort of empirical research. To use another famous example by Kripke, I have to do some investigation in order to discover, e.g., that this table in front of me is made of wood and not of ice or some other substance. But given that the table is made of wood, there could not be a situation in which it were made of some other substance and continue to be this very same table. That is to say, to be made of wood is a necessary property of this table, but one cannot know a priori that it has this property. Hence, if it is true that the table is made of wood, it is necessarily true (and, if false, it is necessarily false), but this true can only be known a posteriori. Notice that, contrary to the Goldbach’s conjecture case, this is a case of a necessary truth that necessarily can only be known a posteriori.
We also have as a result of the separation between epistemic and metaphysical modality the dual case, i.e., examples of contingent truths that can be known a priori. Kripke’s classical example is that of the standard meter bar in Paris, i.e., what we call meter was once established conventionally to be the length of that particular bar at a certain time t0. Hence, Kripke argues, we can know a priori that the length of that bar at t0 is one meter, since this requires no measuring at all. Nevertheless, this truth is clearly contingent, since in other possible worlds the bar could have a different length.
Something essential to these examples is that the names are rigid designation. In the case of the identity, if the names involved are not rigid designators, i.e., if they abbreviate some (non-rigid) definite descriptions, then the identity is of the form D.D.1=D.D.2, but if D.D.1 and D.D.2 are different descriptions, the statement is not necessary but contingent, since D.D.1 and D.D.2 could capture different objects in different possible worlds, and if the descriptions are the same, we have a necessary truth that is also a priori.1 And the bar example only works if we have on one side of the identity-statement a rigid designator (‘one meter’) and on the other side a nonrigid definite description (i.e., ‘the length of the bar at t0’) that is used to fix the reference of the rigid designator. If ‘meter’ were an abbreviation of ‘length of the bar at t0’, then the identity would be simply ‘length of the bar at t0 = length of the bar at t0’, i.e., would express a necessary and a priori truth, and if the definite description were a rigid one, the identity expressed would be one between two rigid designators, being therefore necessarily true or necessarily false. 3
In Kaplan’s theory of demonstratives we find equivalent examples, both of necessary a posteriori and of contingent a priori truths. His most interesting examples come from the semantic properties of the dthat-operator, which was introduced by him as a rigidifier for definite descriptions, and works as a kind of paradigm of demonstratives. According to Kaplan’s theory, a complete demonstrative is something of the form
D[α] in which we have the combination of a demonstrative expression D (like ‘this’ or ‘he’), with a demonstration α (typically a pointing). The demonstration is a nonlinguistic element that accompanies the linguistic demonstrative expression. The combination of both, when placed in a context, rigidly designate a certain individual (a person, an object, a location, etc., depending on the demonstrative). Now, according to Kaplan, there is a strong analogy between demonstrating an individual and describing it from a certain perspective, so that a demonstration can be seen as a kind of definite description. (In the same way that we can consider a definite description in different counterfactual circumstances, we can also consider the same demonstration in other counterfactual circumstances. E.g., the same pointing to Venus early in the morning can be imagined in a counterfactual situation in which Saturn is the body occupying that same position early in the morning. The pointing would be something analogous to ‘the brightest heavenly body seen from here’.) Motivated by this analogy between demonstrations and definite descriptions, Kaplan introduces the term 1 It is less clear what would happen if the names were both abbreviations of rigid definite descriptions. ‘dthat’, which operates on definite descriptions, and rigidify them.2 Hence, ‘dthat[the brightest heavenly body seen from here]’ works as a rigid designator of Venus, in the same way that ‘that[α]’ (pointing to Venus) does.
Using this conceptual apparatus, Kaplan builds a formal system, together with the corresponding semantic notions of model and validity, in order to capture some intuitions related to indexical expressions. For example, the fact that the sentence is true whenever uttered in any context (i.e., at any place and by any person), but the truth thereby expressed is not a necessary one (since this person could be in other places in other counterfactual situations). Similarly, the sentence
expresses something true whenever uttered (or simply thought, as in Descartes’ Cogito) by any person, but it does not express a necessary truth since there is a possible world in which the person who utters it does not exist.
How can sentences be true whenever uttered and, nevertheless, express contingent truths? Kaplan’s explanation is that aprioricity and necessity have to do with different semantic dimensions of indexical expressions, namely, with what he calls character and content, respectively. The character of an expression is, in Kaplan’s conception, a general rule that associates an extension to each context of use.3 It provides the rule for using the indexical. The character of an indexical is always the same in all contexts, although the extension might change from context to context. For example, the character of ‘I’ is the same for any person, although the extension changes from person to person. Similarly with the character of ‘here’: this is the same anywhere, although the extension changes from place to place. The character of a sentence is a composition from the characters of its parts. Hence the character of ‘I am here now’ is the same in all contexts, and the content is a different proposition in each context. In each and every context of utterance, the proposition expressed by ‘I am here now’ is true (since the agent inevitably finds himself in the location and at the time in which the utterance takes place), although the truth expressed in each context is contingent. The following passage is a good summary of Kaplan’s view on aprioricity and necessity: 2 In the “Afterthoughts”, Kaplan himself detected an ambiguity in the dthat operator. It can be conceived as a linguistic device which syntactically is a singular term, and which requires a special (non-linguistic kind) kind of demonstration or it can be taken as an operator, that requires another linguistic expression (definite descriptions) for its completion. 3 Actually, Kaplan’s suggestion is little more complicated. The character actually associates a certain intension with each context. In the case of indexicals, since they rigidly refer to an individual in a given context, this intension is a constant function, i.e., a function that associates the same individual to all possible worlds. Kaplan is actually thinking primarily of sentences in which indexicals occur: the character of such sentences associate to each context a certain intension, which is, in Kaplan’s view, a russellian proposition. The russellian proposition as a whole is an intension, but it contains some extensional element (object), which is the indexical’s contribution.
It is the character (or the sentence, if you prefer) that is logically
true, producing a true content in every context. But it is the content
(the proposition, if you will) that is contingent or necessary.
dthat[α] = dthat[β] (where α and β are non-rigid definite descriptions to which we attach the dthatoperator). The following is a logical truth:
(dthat[α] = dthat[β]) ↔ (α=β) It is a logical truth because it is true in any context that the rigidified descriptions α and β will designate the same individual iff the non-rigidified descriptions already designate the same individual. On the left-hand side of the bi-implication we have an identity-sentence between two rigid designators, and if it is true, it is necessarily true, and on the right-hand side we have an identity between non-rigid definite descriptions, which is contingent. Now is a priori if and only if is also a priori, but the latter cannot be a priori, since α and β are non-rigid; hence, is a posteriori, and necessary. (The example can be changed to D[α] = D[β], where α and β are demonstrations, like ‘he[pointing at a person in an old picture] is he[pointing at a person in a recent picture]’). The example is in several aspects similar to Kripke’s ‘Hesperus=Phosphorus’, since here we also have an identity between rigid designators, and the truth of it cannot be known a priori (in Kaplan’s case, we cannot know a priori that the demonstrata of two different demonstration is the same). An example of the dual case is the following sentence: dthat[α]=dthat[β]
α=β dthat[α]=dthat[β]
dthat[α]=α (where α is a non-rigid definite description). It is, according to Kaplan, true in every context of use, being therefore a priori. However, since we have on one side a rigid designator (i.e., it refers to the same object in all possible worlds) and on the other side a non-rigid designator (i.e., the reference might change in different worlds), the truth expressed is a contingent one, since in another world the referents of dthat[α] and of α might not coincide. (This example is in several ways analogous to Kripke’s meter bar case.) Something essential to Kaplan’s argument is the assumption that logical truth and aprioricity coincide in his system, i.e., the bearers of logical truth and of aprioricity are the same entities, and the thesis that demonstratives, when combined with a demonstration in a context, become a rigid designator. (Kaplan provides some independent arguments for the latter in (1989).) As we shall see below, the first assumption will be challenged by Soames. 4
In a recent book (2005), Soames takes up a vigorous defense of referentialism against descriptivism, i.e., the view that proper names refer by means of an associated descriptive content. The paradigms of this perspective are Russell (for whom ordinary proper names are abbreviations of definite descriptions) and Frege (who, according to a current interpretation, advocates the view that a proper names expresses the sense of a definite description, and it is this descriptive sense that “captures” the reference of the name. However, along his defense of referentialism, Soames elaborates some critical points about Kripke’s and Kaplan’s examples of contingent a priori that were discussed above. Soames intends these points to be of marginal effect, but I think it worth discussing them, since they bring into light some crucial aspects of the contingent a priori in Kripke’s and Kaplan’s style. 4.1
In section XIX of (1989) Kaplan defines logical truth in his language LD (“Logic of Demonstratives”) as truth in all LD-structures, and an LD-structure is something of the following kind:
U = <C, W, U, P, T, I> where C is a non-empty set of contexts, W a non-empty set of possible worlds, U a non-empty set of individuals, P a non-empty set of spatial positions, T is the set of positive integers (thought as temporal instants), and I (the interpretation-function) associates to each predicate and each functional symbol of LD an appropriate intension. Each context c∈C consists in a possible world w∈W, an agent a∈U, a location p∈P, and a time t∈T.
As we saw, Kaplan claims that the bearers of logical truth and aprioricity are different entities (namely, content and character, respectively), and this is what explains the possibility of contingent a priori and necessary a posteriori truths. However, in other parts of his text Kaplan identifies the character of an expression with its meaning, i.e., that part that remains fixed from context to context, and which any speaker must understand. In other words, the meaning (i.e., character) is the bearer of logical truth, and, hence, of aprioricity.
Soames’ first critical point is directed to the identification of character as the bearer of logical truth. For, as Soames points out, character corresponds to the intuitive notion of meaning, and meaning is something that changes from model to model:
The inclusion of the interpretation function-which is indispensable in model
theoretic semantics-ensures that the interpretations, and hence, meanings of
the nonlogical vocabulary will vary from model to model. This is
significant, because the nonlogical vocabulary includes all primitives of the
language-all names, predicates, and function-signs-other than the standard
logical symbols, the modal and the tense operators, and the special indexical
terms and operators introduced by Kaplan in his formal system.
Soames’ point would be correct if in Kaplan’s system we had names in the logical
vocabulary, whose meaning can change from model-structure to model-structure.
However, an important feature of Kaplan’s system is that it does not include names in
the non-logical vocabulary, but only predicates and functional symbols. Is this a
failure of Kaplan’s system? I don’t think so. The technical apparatus of his formal
system was introduced for a particular theoretical purpose, i.e., for studying a
particular semantical phenomenon. This is how he justifies the introduction of LD:
Just to be sure we have not overlooked anything, here is a machine against
which we can test our intuitions.
What would happen if Kaplan had included names in the non-logical vocabulary? Then some adjustments would have to be made in his notion of validity (or in the interpretation-function I) since Kaplan accepts Kripke’s idea that names are rigid designators.
Soames’ second critical point is against Kaplan’s thesis (crucial for his examples) that bearers of logical truth and bearers of aprioricity are the same entities. Since aprioricity is an epistemic notion, items of aprioricity must be items of knowledge (and, hence, items of our epistemic contemplation). Soames points out here that Kaplan’s claim that the bearers of aprioricity and of logical truth are one and the same is not consistent with the distinction that the latter introduces between “objects of thought” and “cognitive significance” in the following passage of section XVII of (1989): Is character, then, the object of thought? If you and I both say to ourselves, […] “I am getting bored” have we thought the same thing? We could not have, because what you thought was true, while what I thought was false.
What we must do is disentangle two epistemological notions: the objects of
thought (what Frege calls “Thoughts”) and the cognitive significance of an
object of thought. As has been noted above, a character may be likened to a
manner of presentation of a content. This suggests that we identify objects of
thought with contents and the cognitive significance of such objects with
characters […] According to this view, the thought associated with
‘dthat[]=dthat[]’ and ‘dthat[]=dthat[]’ are the same, but the thought (not the
denotation, mind you, but the thought) is presented differently.
Of all criticisms made by Soames to Kaplan, this is the weakest one. For Kaplan (following Perry) clearly states that cognitive states are individuated by characters, and not by contents, being therefore the real objects of belief. In the passage quoted by Soames Kaplan indeed uses the term “objects of thought”, but since Kaplan never says anything else about the epistemic import of them, I think that “objects of thought” should rather be taken as a purely technical term, without the epistemic import that Soames reads in it. Soames is right that, perhaps, Kaplan’s terminology is unfortunate here, since he wants to distinguish the bearers of two functions that Frege attributes to one single entity, and Frege uses the name “thought” for this entity. But although Kaplan’s terminology is misleading (and, perhaps, a little inconsistent), he certainly does not mean that contents are the subject of epistemic attitudes.
So far the critical points discussed are concerned with specific details of Kaplan’s treatment. But Soames’ third critical point affects not only Kaplan’s example of contingent a priori truths involving the dthat operator, but also Kripke’s meter bar example, as we shall see. Soames considers the following example (adapted from Quine’s classical example):
(1) Dthat[the youngest Chinese spy] is the youngest Chinese spy According to Kaplan’s characterization, the truth of this statement can be known simply on the basis of the truth of
(2) The youngest Chinese spy is the youngest Chinese spy (which is certainly a priori) and the semantic properties of the dthat-operator. (A relevant element in this example is that, supposedly, nobody knows who the youngest Chinese spy is, although we know that there should be one, if the set of Chinese spies is non-empty.) In other words, (1) can be known a priori. According to Soames, however, the example seems to require something impossible, namely, that solely on the basis of (2) (which is a tautology), we could come to know of a certain individual that he is the youngest Chinese spy. In other words, Soames’ point is that simply by a priori manipulating a conceptual apparatus, we could not achieve de re knowledge about the individual referred by ‘dthat[the youngest Chinese spy]’, but this is required, according to him, to know the truth of (1).
Now there is something strange in Soames’ claim, since Kaplan himself does not see de re knowledge as something required for knowing the truth of (1). Indeed, he seems to say quite the opposite:
There is nothing inaccessible to the mind about the semantics of direct
reference, even when the reference is to that which we know only by description.
What allows us to take various propositional attitudes towards singular
propositions is not the form of our acquaintance with the objects but is rather
our ability to manipulate the conceptual apparatus of direct reference.
(4) I should give a lecture today According to Kaplan and Perry, Dirk’s basic cognitive relation is with the character associated with ‘I’ and ‘today’. However, if we want to report the next day in indirect speech what Dirk believed the day before, it would not do simply to place (4) within the scope of the epistemic verb, as in
(5) Dirk believed that I should give a lecture today ((5) clearly gives the wrong report) but we should have to say something like (6) Dirk believed that he should give a lecture yesterday In (6), what is of relevance in the sentence embedded is not the character (since it changed from (5) to (6)), but the content.4
This is probably the reason why Soames points out that the truth of (3) requires that Dirk knows of the person that is the youngest Chinese spy that he is the youngest Chinese spy. But since knowing of a certain person that he has a certain property always requires some sort of experience, it follows that there cannot be a priori knowledge here. Kaplan’s examples of contingent a priori involving the dthat operator do not work simply there cannot be a priori truths here.
Basically the same objection is repeated for the example of the contingent a priori presented by Kripke. For here we have something like
(7) Lee is the youngest Chinese spy (if there is one) 4 Genoveva Marti pointed out that not any description of the same Russellian proposition would do within the scope of the epistemic verb in (6). There might be limits to the way of describing the same proposition. where ‘Lee’ is a rigid designator introduced using the definite description ‘the youngest Chinese spy’. Kripke’s idea is that the truth expressed in this sentence can be known without any investigation about Lee’s professional life or without checking the list of Chinese spies. Now to test the claim that (7) can be known a priori, Soames imagines two possible scenarios concerning the epistemic contact that a subject might have with the youngest Chinese spy. In the fist scenario, there is no epistemic contact at all, i.e., the subject has no idea at all about who could be the youngest Chinese spy. According to Soames, in this circumstance there can be no knowledge of the truth of (7), and a mere linguistic stipulation (or a “baptism ceremony”, as Kripke describes it) cannot change that. In the second scenario, the subject already has, on the basis of some perceptual experience and collected information, some kind of knowledge or belief that a certain person is the youngest Chinese spy. In this case, the subject can know the truth of (7), but this can hardly be called a priori knowledge. That is to say, either we have some sort of knowledge about a certain person that she corresponds to a description (and in this case we have a posteriori knowledge), or we have no previous knowledge at all, and in this case, we cannot, according to Soames, say that there is any knowledge of (7) at all.
I do not think that Soames’ critical point is well grounded. The change in perspective by means of which character loses preponderance to content (i.e., to a Russellian proposition) is pointed out by Kaplan as a semantic phenomenon that arises within indirect belief-reports, but this does not necessarily imply something about the nature of belief (and, hence, about the nature of a priori knowledge). Although we are, in reporting a third-person belief, forced by our semantic apparatus to privilege the proposition instead of the character, this does not necessarily mean that belief is a relation between a subject and a proposition. On the contrary, there are several examples illustrating the fact that one can have a priori knowledge of the character, without knowing the corresponding proposition. Perry presents the example of a crazy fellow who believes that he is David Hume. However he might be mistaken about the proposition expressed by his utterance of ‘I am here’, he nevertheless has a priori knowledge corresponding to the character of ‘I am here’. Dirk might suffer from amnesia and have completely lost any clue he might have about his identity and about the fact that he is in Rio de Janeiro, but nevertheless he has a priori knowledge corresponding to ‘I am here’. If we try to report what he believes, we will have to present the proposition that Dirk is in Rio de Janeiro, but that doesn’t mean that there is a cognitive relation between him and this proposition.5
Soames’ last objection repeats to a large extent some remarks made by Donnellan thirty years earlier in (1977), in an article where he addresses both Kripke’s and Kaplan’s examples of contingent a priori. Donnellan distinguishes between knowing that a sentence is true and knowing which truth the sentence expresses. In the case of Kripke’s and Kaplan’s examples, it could be, according to him, that a subject knows 5 Actually, there is a stronger and a weaker version of this claim about the epistemic preponderance of the character. The weaker is that the subject of our propositional attitudes is the character taken together with a content. The stronger is that character alone is the subject of propositional attitudes. Perry seems closer to the weaker claim. But Kaplan seems closer to the stronger claim, when he says, at the end of a section entitled “Epistemic Remarks”, that “This doesn’t prove that the cognitive content of, say, a single sentence or even a word is to be identified with its character, but strongly suggests it” (1989, p. 532). that the corresponding sentences are true, but does not know which truth they express. He illustrates this point with a comparison: someone that does not understand German at all might be informed by a native speaker that the sentence ‘Schnee ist weiss und der Himmel ist blau’ is true. Based on this report and on the authority and reliability of the German speaker, he might know that the sentence is true, although he has no idea of which truth it expresses. The examples of contingent a priori à la Kripke and Kaplan are, according to Donnellan, largely similar to the above, i.e., the speaker knows that the corresponding sentences are true, but doesn’t know which truth they express. But if this is knowledge at all, it is, according to Donnellan, only a trivial and uninteresting kind of meta-linguistic knowledge, and this is certainly not what Kripke and Kaplan had in mind.6 In order to say that the speaker effectively knows the truth expressed by a sentence of the form ‘N is D.D.’ (where ‘N’ is a rigid designator, and ‘D.D.’ is a non-rigid definite description that originally fixes the reference of ‘N’, but does not give the meaning of it) or a sentence of the form ‘dthat[D.D.]=D.D.’, something more is required according to Donnellan: the speaker must know which object the sentence is about, i.e., the referent of ‘N’ or of ‘dthat[D.D.]’. In other words, we can only say that a speaker knows which truth the sentence expresses if he has de re knowledge. Consequently, in order to correctly claim that Dirk knows (7), we must be able to quantify within the epistemic verb, as in
(8) ∃x(Dirk knows that x is the youngest Chinese spy) To dramatize the situation, Donnellan presents the following example. Suppose that, today, we rigidly baptize as ‘Newman’ the first person that will be born in the XXIInd century. We have no idea today of who this person will be, although we assume that there will be one. Now suppose that one hundred years from now we meet this person, who was in fact baptized ‘John’ by her parents. In this case, as Donnellan argues, it would seem quite strange to tell him ‘I knew one hundred years ago that you would be the first person born in the XXIInd century, and that your name would be “Newman”’. The reason why this is so strange is, according to Donnellan, that we never had, before his birth, real knowledge (besides trivial meta-linguistic knowledge), since we never had de re knowledge corresponding to ‘Newman is the first person born in the XXIInd century’.
However, there are some possible replies to Donnellan’s point against Kripke’s and
Kaplan’s examples. First, one could say that, in cases like that of Newman, what we
really have is a double baptism. We could know a priori that Newman is the first
person born in the XXIInd century, and a posteriori that John is the first person born
in the XXIInd century, and afterwards discover that Newman=John, in the same way
that the same mountain can receive different names when seen from different
prespectives. The fact that someone was baptized as ‘John’ does not eliminate the fact that he
was also baptized ‘Newman’ before that, and the knowledge that comes with this
baptism.
6 As
Second, something that seems to me essential to Kaplan’s examples of a priori knowledge is that it is not of the proposition expressed, but of the character, i.e., someone might know the truth of what is being said (or simply thought), even if he does not know the proposition expressed in that context. As we saw, someone suffering from amnesia might correctly think ‘I am here’ without having any idea of the proposition expressed. In the same way, someone might have a priori knowledge of ‘dthat[the youngest Chinese spy] is the youngest Chinese spy’ without knowing which proposition is being expressed-hence, without de re knowledge. In the same way, I might see someone reflected in a mirror glass, without knowing that it is myself, and have thoughts and cognitive attitudes towards this person, without knowing exactly which proposition is expressed by ‘he is seated there’. Someone that wants to argue along Donnellan’s lines would have to explain why in some cases we don’t need to know the proposition (as in ‘I am here’), while in other cases we do need to know it.
Finally, there is something to be said in favor of Kaplan’s and Kripke’s examples
using a notion developed in
This is not the proposition that was expressed by whoever uttered u that day. If I had known the details of the context of this utterance, e.g., that John wrote the note on January 1rst 2003 and left it in Mary’s front door, I would know the official content of it, namely, the content of ‘John was in front of Mary’s house on January 1rst 2003’. But I don’t know these details, and so (9) is the only understanding that I have about u. In this sense, (9) captures a minimal aspect of the meaning of u , namely, that which a speaker that has no idea at all about the context of utterance understands by u. (Perry calls this the ‘reflexive truth conditions of u’ because it is necessarily spelled out in terms of the utterance itself.) Now suppose we read in a wall a graffiti saying ‘I love Mary’ (let’s call this utterance p). The reflexive truthconditions would be something like
(10) p is true iff the utterer of p loves Mary Now in (10), although there is reflexivity, the reflexive truth-conditions are not automatically satisfied as in (9). That is to say, for some utterances their reflexive truthconditions are automatically satisfied, while for others the reflexive truth-conditions are not automatically satisfied.
Now we see here that, for some utterances, their reflexive truth-conditions are always satisfied no matter what, i.e., we know a priori that some utterances are true without knowing which proposition (official content, in Perry’s terminology) they express. However, knowing the reflexive truth conditions of an utterance by no means correspond to simply blindly knowing that a sentence is true without knowing which truth it expresses, as Donnellan would say. In order to know that an utterance is always true, I have to be a competent speaker and know their reflexive truth-conditions, and some utterances are always true and have complicate reflexive truth-conditions. By no means this is a kind of knowledge that is as trivial and uninteresting as simply knowing that a sentence that I don’t understand is true.
There is indeed, at first sight, something surprising in the very idea of contingent a priori truths since, in principle, a truth is contingent in virtue of a fact in the world. But in the examples advanced by Kripke and Kaplan here discussed, the corresponding sentences are made true not because of a fact in the world, but because of some linguistic stipulations. On the other hand, the surprise might disappear if we keep in mind that something might be known a priori as a product of our conceptual and semantical apparatus. In my view, rigid designation is a very special semantic phenomenon, and we might expect some very special epistemic consequences as well. There is a certain analogy between the phenomenon that Kripke and Kaplan are dealing with and the Kantian synthetic a priori: in the same way that the latter can be seen as a product of the conceptual apparatus imposed by our intellect in order to understand reality, the Kaplanian and Kripkean contingent a priori can be seen as a product of the semantical apparatus imposed by our language to talk about the world.