Necessity, Apriority and True Identity Statements Heimir Geirsson [email protected]

The thesis that the necessary and the a priori are extensionally equivalent consists of two independent claims: 1) All a priori truths are necessary and 2) all necessary truths are a priori. In Naming and Necessity1 Saul A. Kripke gives examples of necessary but a posteriori truths, so he disagrees with the second leg of the thesis.2 His examples are of two types; on the one hand statements involving essential properties and on the other hand true identity statements. My concern will be with examples of the second type and whether they refute (2). (2), however, is ambiguous and can mean one of three things: a) If p is a necessary truth, then one can know a priori that p is necessary. b) If p is a necessary truth, then one can know a priori that p. c) If p is a necessary truth, then one can know a priori that p and that p is necessary. Kripke maintains that we know a priori that if an identity statement is true, then it is necessarily true. Consequently, the issue at hand is how we come to know the truth of such identity statements, so it is clearly (b) that we should be concerned with.3 In order to refute (b), and thus (2), we apparently need to show that

2

d) there is a necessary truth whose truth I can only know a posteriori.4 I will argue that (d) is too strong and that we can refute (2) without an example satisfying (d). Section one of the paper introduces and clarifies Kripke's alleged counterexample to (2) that the paper will focus on. Section two will argue that if belief is a binary relation between a person and a proposition, then the example fails. It will also discuss G.W. Fitch's account of a ternary belief relation which seems to provide a counterexample to (2) and argue that Fitch's account is inadequate. Sections three and four introduce a new version of the ternary account of the belief relation and argue that it gives us a genuine counterexample to (2). Throughout the paper I will assume that names are direct designators and that the semantic content of a name is its referent. Consequently, my interest is confined to singular propositions, namely propositions consisting of an object, a property, and a relation.

I.

One of the examples Kripke gives of a necessary a posteriori truth is 3) Hesperus is Phosphorus.5 This statement, however, is not necessarily true. Consider a world in which Hesperus does not exist. In that world (3) does not express truth, so the proposition (3) expresses is contingently true, if true at all. Consider instead: 4) If Hesperus exists, Hesperus is Phosphorus. This statement appears to give us what we are after for the following reason. Let us view proper names as being direct and rigid designators.6 Then 'Hesperus' and 'Phosphorus' name the same objects in all possible worlds in which they exist. Since the semantic function of a name is exhausted by its reference and since the two names refer to the same object, (4) attributes self-identity to Hesperus, if it exists. Since self-identity is something Hesperus cannot possibly lack, (4) expresses a necessary truth.

3

The reason Kripke says we have an a posteriori truth is that we could be in a qualitatively identical epistemic situation where 'Hesperus' and 'Phosphorus' name two different objects. Consequently, we do not know a priori that (4) expresses a truth.7 Criticism of Kripke's example falls into two categories. On the one hand, it is argued that the proposition expressed by (4) can be known a priori. On the other hand, it is argued that the proposition we believe a posteriori when assenting to (4) is not the proposition expressed by (4),8 but rather a second proposition associated with the one expressed by (4).9 I will start by discussing the first line of attack using an argument of G.W. Fitch as a foil for the discussion.

II.

G. W. Fitch10 has argued that (4) is an example of a necessary a priori truth. His argument relies on the direct designation theory: 5) The terms 'Venus', 'Hesperus' and 'Phosphorus' are rigid designators.11 6) The objects of knowledge are propositions. 7) If S is a sentence containing a rigid designator a and S' is a sentence obtained from S by substituting a rigid designator b for a in S and a and b designate the same object, then S and S' express the same proposition. 8) We know a priori that if Venus exists, then Venus is Venus. So, 9) We know a priori that if Hesperus exists, then Hesperus is Phosphorus.12 Fitch's argument, let us call it F1, has been embraced by a number of philosophers.13 It seems impossible to refute F1 as long as we view knowledge as involving a binary relation between a person and a proposition.14 The traditional account of a priori knowledge is that a proposition is a priori if it can be known a priori and since "if Venus exists, Venus is Venus" and "if Hesperus exists, Hesperus is Phosphorus" express the same proposition and we can obviously know a priori that if Venus exists, Venus is Venus, we can know a priori that if Hesperus exists, Hesperus is Phosphorus. F1 can in fact be extended to all statements of the type

4

10) If N exists, then N is M where 'N' and 'M' are coreferential proper names. (10) expresses the same proposition as does 11) If N exists, then N is N which we know a priori. So, we know a priori the truth (10) expresses. This, however, could change once we view the belief relation as being ternary.15 If viewing the belief relation as being ternary will affect the argument it will most likely be because, given the ternary relation, the principle stated in (7) can not be applied in belief contexts. We should therefore revise F1 and eliminate talk about knowledge in favor of talk about beliefs. By revising F1 we focus more clearly on the source of the problem. The revision has the further advantage that it removes all talk of apriority which, for the moment, becomes an unnecessary distraction. Let us keep premises (5) and (7) as they are, but change (6) and (8) to 6*) The objects of belief are propositions and 8*) Jones believes that if Venus exist, then Venus is Venus. The conclusion becomes 9*) Jones believes that if Hesperus exist, then Hesperus is Phosphorus. Let us call the resulting argument F2. F2 depends on the substitutivity of coreferential names in belief contexts and it is not clear that the principle of substitutivity holds. Kripke, for example, appears to reject it with the suggestion that a Millian substitutivity doctrine for modal contexts can be maintained even if such a doctrine for epistemic contexts is rejected.16 But since he does not discuss why substitutivity in belief contexts fails, if it fails, the argument has not been refuted. In Naming and Believing Fitch discusses Kripke's example again and comes to a different conclusion. His present view depends on his theory of belief. Belief is analyzed as a ternary relation between a person, a proposition, and an attribution. The attribution allegedly does not affect the proposition believed although it can affect the believer's cognitive state. Consequently, although 'if

5

Hesperus exists, Hesperus is Phosphorus' and 'if Hesperus exists, Hesperus is Hesperus' express the same proposition, Jones might have two different cognitive attitudes towards it depending on what property he attributes to himself. The crucial part of the theory concerns the attribution. What is its nature? Fitch gives a clear answer. We attribute properties to ourselves (in a belief context) under a description that the speaker would supply if asked "Who is N?", where 'N' stands for the name in question.17 18 As an example, If I believe that Chisholm is a philosopher, then I believe the proposition by attributing the property being such that the one who has Q is a philosopher to myself, where Q is the property that I use to distinguish Chisholm from other objects.19 Since you might choose a different description for the attribution, both of us might believe that Chisholm is a philosopher, but we might be in different belief states. Now it becomes clear how Fitch treats identity statements. I can associate one "speaker sense" to 'Phosphorus', another speaker sense to 'Hesperus' and the third one to 'Venus'. If I do that, my belief that Venus is Venus is different from my belief that Hesperus is Phosphorus. Consequently, the inference in F2 does not go through. Once we know that F2 is flawed, one would expect that F1 is also flawed. But it is not clear that we can draw that conclusion. F1 ignores how a propositions is believed and only talks about that it is believed. In order to show F1 to be flawed we need to argue that the epistemic status of a belief (whether it is a priori or a posteriori) can be affected by how a proposition is believed. I will discuss this issue in the next section. Does the view that belief is a ternary relation have any bearing on the question of whether identity statements provide a counterexample to (2)? I may believe a proposition in such a way (under such descriptions) that it expresses an a priori truth. This happens when I associate the same speaker sense with both names. But, if I attach different speaker sense to the two names, I might also believe the same proposition in such a way that I know it a posteriori.20

6

Fitch's position is certainly interesting, but has he given us an account of beliefs of singular proposition? I don't think so. Fitch's view is peculiar for he adopts an account of belief developed by R. Chisholm who convincingly rejects the view that beliefs can be explicated in terms of a relation to a proposition.21 Fitch, on the other hand, is adopting Chisholm's theory to offer an account of belief which involves a relation to singular propositions. While this move is interesting, it is not supported. It becomes particularly difficult to see the alleged relationship between the proposition believed and the attribution to oneself. Fitch says that we associate a speaker sense to each name and that it is the sense of the description that the speaker would supply if asked "Who is N?", where 'N' stands for the name in question. Judging from the quotation above, the speaker sense would typically describe a property used (by the speaker) to individuate the object denoted from other objects. We have three options as of how to understand the attribution. If we understand the attribution in Chisholm's sense, then there is no need for singular propositions. Since Chisholm's theory eliminates the need for propositions a move that involves both proposition and attribution is ad hoc. If, on the other hand, we want to keep singular propositions, then there must be a relation between the object in the proposition and the description involved in the attribution. The relation seems to be as follows: When I assent to a sentence expressing a singular proposition I attach a description to the name in the sentence, thus forming a general proposition. It is the description in the general proposition I use when I attribute the property to myself. Consequently, it is the general proposition which characterizes my cognitive state with respect to the belief. Consequently, we have two propositions; the singular proposition and the associated general proposition. When I believe the singular proposition I attach my speaker sense to the name. The cognitive state I am in is partly characterized by this speaker sense. But since the speaker sense consists of a description (or descriptions), it follows that the cognitive state is partly characterized by the description(s). But we are now moving dangerously close to the description theories of names and away from the direct designation theory. If we call the singular proposition 'P' and the general, or the fully conceptualized proposition, 'P*', then P and P* are two different propositions. What we can say

7

is that, for me, the two propositions are associated such that when I assent to P I believe P* and when I utter P I use it as a shorthand for P*. But association is not identity. Note that the second option does not collapse into the description theory since the description in the associated proposition does not determine the reference of the name in the original (singular) proposition. What the second option does is cognitively isolate singular propositions, putting them beyond the reach of belief. They exist and we may assent to them, but the objects of belief are not singular propositions. Instead, what we believe when we assent to singular propositions are the associated, or general propositions. A third option might be to claim that one believes a singular proposition P by virtue of believing a general proposition P*. This option can only succeed if we break the connection between cognitive content and propositional content. In Fitch's theory this involves introducing a descriptive content which is not in the alleged object of belief (the singular proposition). The Putnam-FodorStich-Burge literature on "methodological solipsism" suggest very strongly that cognitive content and propositional content diverge in the sense that even if we can read a person's mental content we do not know what singular proposition is believed. A person who believes that "water is quenching" might be in the same narrow mental state as one who believes that "twater is quenching." But the third option does not support methodological solipsism literature. Instead, the third option introduces general propositions again, and there is no reason to suppose it not being the case that if your mental state is characterized by a general proposition, the object of belief is that same general proposition. So, as in the second option, singular propositions are left out of the picture. Another approach to the problem of how a Fregean might take the singular proposition as the object of belief and the associated general proposition, in which a suitable description replaces the object, as the content of belief, involves introducing a third party for the belief attribution. The subject itself might not be able to state the description, and thus not the general proposition. Instead, the description would be provided by the psychologist making belief ascriptions.22 But note that the psychologist only has access to the subject's narrow mental states, so the psychologist's belief ascriptions would fail to capture the wide content of mental states. Consequently, a view which

8

favors this option has to ignore the methodological solipsism literature which strongly suggests that even if we can read a person's mental content we fail to know what singular proposition is believed. The point is that we cannot accept any view to the effect that singular propositions are believed under a description. Once we do that we admit to either full fledged Fregean propositions as the real objects of belief, which is what the direct designation theory is trying to avoid, or we isolate singular propositions, putting them out of reach of belief.23

III. While I do not dispute that we apprehend singular propositions in different ways, we can not allow the mode of presentation of the object in the proposition to be descriptive. Once we do that we are saddled with either a version of the description theory or a view which forces us to introduce a distinction between the singular proposition assented to and the associated proposition which is believed. Since thoughts about objects seem invariably to be either descriptive or perspectival, the latter consisting, perhaps, of our ability to produce an image of the object thought about, and the modes of presentation can not be descriptive, the alternative is to introduce perspectival or nonlinguistic modes of presentation (NMP for short). Consider the following questions: Do frogs have lips? What is the shape of an elephants ear? How many windows are on the south side of your house? Studies conducted by Stephen Kosslyn and other cognitive psychologists have shown that people use images when required to answer questions about properties of objects that are "poorly associated with the objects," that is, parts people usually have not thought about before in connection with an object. In spite of its unpopularity in contemporary philosophy the picture theory of mental representation has been hotly debated in contemporary cognitive psychology. The picture theorists have argued that the best explanation of the results of a variety of experiments is that mental images are pictorial in nature. Here is not the place to discuss Kosslyn's theory. The nature of mental representation will most likely be revealed in a laboratory, but if Kosslyn and his team is heading in the right direction then we have some empirical evidence to the effect that mental representations are often stored in a format which does

9

not involve a description or a set of descriptions. Instead, we have in fact nonlinguistic representations.24 Introducing NMPs has the advantage that while we can be said to believe a singular proposition in different ways (depending on our NMP of the object in the proposition) we do not have to introduce associated propositions in order to account for the different ways in which we can believe singular propositions. Because we are dealing with only one proposition we believe it, not some surrogate proposition associated with the singular proposition. Consequently, we do not end up with a discrepancy between the proposition expressed by a sentence and the proposition associated with it where we assent to one and believe the other. An example might explain the direction in which I am headed. Suppose that Sam and Sue are looking attentively at a statue and that Sam is looking at its front while Sue is looking at its back. There is of course a sense in which Sam and Sue are seeing the same thing, for they are looking at the same object. In spite of that their experiences of the statue viewing are not at all the same. If we want to specify the manner in which Sam and Sue mentally represent the statue it is not sufficient to say that since they are viewing the same object their mental representations of it are identical. We have to take into account that they are appeared to differently and that their mental representation has to be specified relative to how they are appeared to. Suppose the statue Sam and Sue were looking at was David and that they form a belief expressed by 12)

David is huge.

Given their limited and different exposure to David they will in all likelihood think differently of it. When Sam entertains his belief he pictures David as he saw it; the front side. Sue, on the other hand, pictures David as seen from the back. Their different experiences of David will thus result in different mental representations of it, and though Sam and Sue believe the same proposition they do not believe it in the same way. Sam associates a different NMP with 'David' than does Sue. Assuming that the general idea presented above is correct, suppose you and I associate different NMPs with 'Hesperus'.25 If we sincerely assent to

10

13) Hesperus is a planet then there is a sense in which we have the same belief; namely, we believe the same proposition. But since our NMPs differ there is also a sense in which our beliefs are different. The object of our belief, the singular proposition, is the same, but the content of our belief, which depends on the NMP I associate with the name, is different.26 The proposition believed does not reveal how we mentally represent the object in it. Note that I am not giving up the view that propositional content expresses cognitive content. However, since my belief only captures an aspect of the proposition believed the propositional content will be richer than the respective cognitive content. One might object that the present account is open to the very same criticism raised against the accounts in section II. There is was alleged that the introduction of general propositions as the contents of beliefs placed singular propositions out of the reach of belief. The objector maintains that the introduction of NMPs + properties as the contents of beliefs also places singular propositions out of the reach of belief.27 This objection, however, fails, for all parties to the dispute are attempting to offer an account of belief of singular propositions. Hence, all parties agree that propositions are the objects of belief. The reason that singular propositions were placed out of the reach of belief in the accounts discussed in section II was that those accounts introduced general propositions as the content of beliefs. On the present account we are dealing with only one proposition, namely the singular proposition. Given that we are dealing with only one proposition, and given that beliefs are of propositions, the only candidate for the object of belief is the singular proposition. If I sincerely assent to both (13) and 14) Phosphorus is a planet and associate the same NMP with both names, then I see no reason why the names should not be interchangeable in my belief context without changing either the object or the content of belief. If, on the other hand, I associate different NMPs with the names, then interchanging the names in my belief context does not affect the object of my belief although it does affect the content of my belief. The truth of the principle stated in (7) then depends on whether it concerns the object or the content of belief. There are strong reasons why we should view it as concerning the object of belief, the

11

proposition believed. After all, the principle has to do with sentences expressing the same proposition, not the same belief. In addition, if you and I believe (13), then there is a sense in which we have the same belief, and that is captured by saying that we believe the same proposition. The same, of course, is true if I believe (13) and you (14). If we view things this way, then the principle holds for singular propositions. Since the restricted version of Fitch's argument, (F2), only concerns the objects of belief and the substitutivity of names in belief contexts, (9*) follows from its premises. But I think we have found good reasons to argue that F1 is flawed and that (2) is false.

IV. In Naming and Necessity Kripke argued that true identity statements involving proper names are necessarily true and that we can not know the truth of some of them a priori. This seems to establish his view that there are necessary truths whose truth we can only know a posteriori. But, as we have seen, this kind of reasoning can not be accepted if the belief relation is binary. The situation is different if the belief relation is ternary. Suppose I believe the necessary truth (4) expresses. Since Kripke maintains that the belief is informative the NMPs I attach to the names differ. But since the NMPs differ we cannot simply say that I believe the proposition and leave things at that. It matters how I believe the proposition and that can affect its epistemic status. Since, for any singular proposition, there are going to be differences between the propositional content and the belief content of a person who believes that proposition, we need to introduce a device to make it clear when we are talking about the proposition and when we are talking about a given belief of that proposition. If 'N' stands for a sentence expressing a singular proposition and 'S' stands for a person, let '(N)^' represent the content of S's belief that N. Then, the way in which I believe (4) can be represented by 4*) (If Hesperus exists, then Hesperus is Phosphorus.)^

12

What is the epistemological status of the proposition believed? I don't think we can answer that because the proposition does not have a definite epistemological status. In order to see that, consider the same proposition believed in a different way, by a different person. This person, unlike the one in Kripke's example, learned the name 'Venus' and then considered 15*) (If Venus exists, then Venus is Venus)^ attaching a single NMP to 'Venus'. As a result of how this person believes what (4) expresses it is an a priori proposition for him. He needed, of course, some experience to acquire the concept of Venus, but once the concept is in place he needs no further experience in order to know what (4) expresses. But I did not come to believe the proposition in this way and I attach different NMPs to the names. Given how I believe the proposition I need further experience to come to know its truth. Consequently, the way in which I believe the proposition lends itself only to a posteriori knowledge. F1 fails because it overlooks this possibility. Someone might object that I am making a similar mistake as many before me; that I am not discovering the truth of (4) but rather the truth of 16) 'Hesperus' and 'Phosphorus' represent NMPs of the same object which is, of course, a contingent truth. When I discover the truth of the proposition expressed by (4), I discover something about Venus. When, on the other hand, I discover the truth of the proposition expressed by (16), I discover something about myself, a discovery not very interesting from an astronomical point of view. It is very unreasonable to deny the claim that we mentally represent objects in different ways. It is also very unreasonable to deny that we make discoveries about objects, discoveries to the effect that what we thought were two objects is in fact one object. We seem, as a matter of fact, to be able to discover that without realizing that two different mental representations we have are of the same object. In the case of Hesperus and Phosphorus the story might be this. I am immersed in calculating the planetary motions and come to believe of the second planet from the sun that it has to be seen both first in the evening and latest in the morning. What I previously thought were two planets are in fact one and the same planet. I have now discovered that Hesperus and Phosphorus are identical.

13

Later in the day, when I reflect more calmly upon my discovery, I make the second discovery; namely that 'Hesperus' and 'Phosphorus' represent NMPs of one and the same object. Consequently, I can discover the truth of (4) without discovering the truth of (16). But this does not show that I can know the truth of what (4) expresses as I believe it without knowing the truth of what (16) expresses. The most obvious way to show that it can be done is to point out that I might come to know the truth of (4) while being ignorant as to what an NMP is and therefore not be in a position to know the truth of (16). Where does this leave Kripke's original example? Kripke argued that we could only discover the truth value of some necessary truths a posteriori. But, as we have seen, the proposition as such is neither a priori nor a posteriori. Its epistemic status depends on how it is believed. For this reason Kripke failed to establish the claim that there are necessary truths whose truth I can only know a posteriori. Consequently, his example does not refute (2). What we end up with on the account I have given is much weaker than (d), but nevertheless sufficiently strong to refute (2). What I have argued for is d*) there is a necessary truth and a way of believing it such that one who believes it in that way can know its truth only a posteriori. Note that (d*) does not talk about a proposition being a priori or a posteriori. Instead it relativizes the epistemic status to the way the proposition is believed. Furthermore, since we can only know the truth of a proposition when believed in a certain way a posteriori, I am replacing the traditional absolute sense of the a priori with a relative sense of the a priori.28 Those who have argued against Kripke that the second leg of the extensional equivalence thesis is true are therefore partly vindicated because there is more than one way to believe a singular proposition and, for the propositions of the type we have focused on, there is always a way to believe them such that the belief is justified a priori. But the important point is that there are also ways to believe them such that the belief can only be justified a posteriori. While the conclusion I have reached is weaker than Kripke's it should not be dismissed as being any less important. What was at stake was whether all knowledge of necessary truths is a

14

priori. (d*) establishes that it is clearly not the case. Whether our knowledge of a necessary truth is a priori or a posteriori depends on how the proposition in question is believed. The natural reply to this is to point out that traditionally we say that a proposition is a priori if it can be known independently of experience and that I have not shown that there are necessary truths whose truths we can only know a posteriori. But this reply does not take into account that belief is a ternary relation. Instead, it assumes that the only factors in a belief are the believer and the proposition believed. The third belief relatum requires us to introduce finer grained criteria for the justification of beliefs of singular propositions. Once we do that, (d*) is justified and the second leg of the extensional equivalence thesis is proven false.29

15

Bibliography Casullo, A.: 1977, 'Kripke on the A Priori and the Necessary', Analysis 37, 152-159. Chisholm, R.: 1981, The First Person, University of Minnesota Press, Minneapolis. Devitt, M. and Sterelny, K.: 1990, Language and Reality: An Introduction to the Philosophy of Language, The MIT Press, Cambridge. Fitch, G. W.: 1976, 'Are There Necessary A Posteriori Truths', Philosophical Studies 30, 243-247. Fitch, G. W.: 1987, Naming and Believing D. Reidel Publishing Company, Dordrecht. Geirsson, H.: 1991, 'The Contingent A Priori: Kripke's Two Types of Examples', Australasian Journal of Philosophy 69, 195-205. Kripke, S. A.: 1979, 'A Puzzle About Belief', ed. A. Margalit, Meaning and Use, D. Reidel, Dordrecht. Kripke, S.A.: 1981, Naming and Necessity, Harvard University Press, Cambridge. Loar, H.: 1988, 'Social Content and Psychological Content', Contents of Thought, ed. Robert H. Grimm and Daniel D. Merrill, The University of Arizona Press, Tucson. Loar, H.: 1988, 'Reply: A New Kind of Content', Contents of Thought, ed. Robert H. Grimm and Daniel D. Merrill, The University of Arizona Press, Tucson. Plantinga, A.: 1974, The Nature of Necessity, Clarendon Press, Oxford. Plantinga, A.: 1978, 'The Boethian Compromise', American Philosophical Quarterly 15, 129-138. Rollins, M.: 1989, Mental Imagery: On The Limits of Cognitive Science, Yale University Press, New Haven. Salmon, N.: 1986, Frege's Puzzle, The MIT Press, Cambridge. Salmon, N.: 1991, 'How Not to become a Millian Heir', Philosophical Studies 62, 165-177. Tichý, P.: 1983, 'Kripke on Necessity A Posteriori', Philosophical Studies 43, 225-241. Tye, M.: 1978, 'The Puzzle of Hesperus and Phosphorus', Australasian Journal of Philosophy 56, 219-224. Tye, M.: 1991, The Imagery Debate, The MIT Press, Cambridge.

16

1

Kripke 1981.

2

Kripke also argues that there are examples of contingent a priori truths, so he disagrees with the first leg of the thesis. For a discussion of those examples see Geirsson 1991.

3

See e.g. Kripke 1981 p. 101.

4

Albert Casullo has argued that even though we find an example which supports (b), it is not clear that it would go against the traditional account of the relationship between the necessary and the a priori. See Casullo 1977.

5

Kripke 1981 pp. 102-5.

6

Plantinga's discussion of rigid designators has made it clear that we can view proper names as being rigid designators without being direct designators. For his discussion on the issue see Plantinga 1974 and 1978.

7

Kripke 1981 p. 104.

8

Tye (1978) argues that we can not know that Hesperus is Phosphorus without first knowing that Hesperus actually exists. Since experience is required to acquire that knowledge, Kripke's example is of a necessary a posteriori truth. But this does not affect the conditional being discussed.

9 10

For the second line of attack see Tichý 1983. Fitch 1976.

11

I have already pointed out that it is not enough that the names be rigid designators. They also have to be direct designators. 12

In his argument Fitch uses 'Venus is Venus' and 'Hesperus is Phosphorus' instead of the corresponding conditionals. That has the disadvantage that we are then dealing with contingent propositions. 13

See e.g. Salmon 1986, p. 137 and Tichý, p. 233.

14 In his writings Kripke assumes that belief is a binary relation. Given that, it is difficult to speculate as to what he had in mind in the quotation above. 15 For arguments as to why we should accept the ternary relation, see and Salmon 1986. 16

Kripke 1979, pp. 239-83, note 9, p. 273.

17

See Fitch 1987 pp. 134-43.

Fitch 1987

17

18

Fitch 1987, p. 139. Fitch's motivation for introducing descriptions is his claim that we need descriptions to characterize the thought in the mind of a person. That, however, is not true. For a recent work on how images can function as representations see Rollins 1989. 19

Fitch 1987, p. 138.

20

Fitch 1987, pp. 175-6.

21

See Chisholm 1981.

22

This possibility was raised by an anonymous referee.

23 An anonymous referee raised the possibility that a Fregean might take the singular proposition as the object of belief and claim that the description which replaces the object in the general proposition which serves as the content of belief cannot be stated in a public language. Instead, it is a description in the subjects language of thought. This suggestion does not help the Fregean for two reasons. We are still dealing with two propositions, namely a singular proposition and a general proposition. The only difference is that, instead of dealing with one language, we are now dealing with two languages. And secondly, languages of thought are highly problematic entities. For a clear discussion of some of the problems they encounter, see Michael Devitt et al. 1990. 24

For a good discussion on mental imagery see Michael Tye 1991.

25

A lot more has to be done to develop a full fledged theory, but I believe I have said enough to provide a general direction as of where to go with such a theory. 26

Brian Loar has developed a view similar to this. See Brian Loar, "Social Content and Psychological Content," and "Reply: A New Kind of Content," in Contents of Thought, ed. Robert H. Grimm et al. 1988. 27

This objection was raised by an anonymous referee.

28

Salmon (1991) has claimed that the relative sense of a posteriori does not replace the traditional absolute sense where a proposition is a priori if it can be known a priori. But the traditional account of propositions did not allow for different ways of grasping one and the same proposition. As a result, there either was or there was not an a priori justification for believing the proposition. Once we acknowledge different ways of believing a proposition it becomes evident that there can at most be a justification to believe the proposition in this or that way. Absolute justification gives way to relative justification. 29 I thank an anonymous referee for Erkenntnis and especially Al Casullo for helpful comments. I also thank the Icelandic Council of Science for its support while writing the paper.