Comments on New Thinking about Propositions, by Jeff King, Scott Soames, and Jeff Speaks Peter Hanks University of Minnesota [email protected] Central APA, St. Louis Feb. 21, 2015 Thanks to Daniel Bonevac for inviting me to participate in this session, and thanks to Jeff, Scott, and Jeff for writing such a great book. I believe it’s going to have a lasting impact. I foresee many graduate seminars on propositions with King, Soames, and Speaks on the syllabus. I’m going to focus my remarks on the three positive views in Part II. The authors criticize each other’s positions in Part III, and they respond to these criticisms in Part IV. Some of what I have to say picks up on parts of this discussion, but I hope I can add something new. I’ll start with Speaks, then Soames, and then King.

Speaks Before I get to Speaks’s view directly I want to say a few things about Russell’s views about propositions between 1903 and 1910. I think there are some lessons to be learned from Russell that bear directly on what Speaks is up to. Russell’s 1903 theory of propositions and its problems are well known, but let’s briefly review them. He held that propositions are worldly states of affairs or facts in which objects are bound together by relations-that-relate.

So, for example, the

proposition that Barack loves Michelle consists of Barack and Michelle bound together by the loving relation.

A major difficulty for this view is that there are no false

1

propositions. Or rather, false propositions don’t exist. They merely subsist, like the golden mountain or the present King of France. Barack doesn’t in fact bear the relation of loving to Vladimir. So there is no actual, existing fact in which Barack loves Vladimir. Rather, for Russell in 1903, there is a non-existent one. In Russell’s 1903 ontology the proposition that Barack loves Vladimir is a shadowy, merely subsistent fact, which Russell variously called a “false Objective”, “objective non-fact”, “objective falsehood”, and “fiction”. By 1910 Russell could no longer stomach this ontology of non-existent facts. Around this time he gave up on propositions in favor of his multiple relation theory of judgment. The point I want to focus on, though, is the view of truth that Russell was forced into by his 1903 theory. As he rightly saw, this theory required him to give up the correspondence theory of truth for propositions.

There’s no sense in saying that a

proposition is true if it corresponds to a fact if propositions are those facts.

A

correspondence theory of truth requires two things to correspond to one another and here we have only one thing (e.g. the proposition/fact that Barack loves Michelle). So Russell had to analyze truth and falsity as simple, unanalyzable, monadic properties of propositions. In a famous remark he says that: It may be said — and this is, I believe, the correct view — that there is no problem at all about truth and falsehood; that some propositions are true and some are false, just as some roses are red and some are white. (Russell 1904, 523-4) In fact, though, a few years later he had a bit more to say about what truth and falsity amount to:

2

If we accept the view that there are objective falsehoods, we shall oppose them to facts, and make truth the quality of facts, falsehood the quality of their opposites, which we may call fictions. … Truth and falsehood, in this view, are ultimate, and no account can be given of what makes a proposition true or false. (Russell 1906, 48) A proposition is true when it exists, i.e. when it is an existing fact. A proposition is false when it doesn’t exist, i.e. when it is an objective falsehood or a fiction. Russell takes the distinction between existent and non-existent facts and identifies it with the distinction between true and false propositions. I think dissatisfaction with this account of truth had as much to do with Russell’s abandonment of his 1903 theory of propositions as its Meinongian commitments. In 1910 he identifies the bearers of truth with belief states, i.e. particular states in the heads of subjects (see Russell 1910). This allowed him to reinstate the correspondence theory of truth. A belief state is true if there is a corresponding fact and false if not. Putting aside Russell’s weird ontology of non-existent facts, the problem for his 1903 theory of propositions was that it forced him to abandon the intuitive idea that true propositions correspond to how things are. There’s something else it forced him to give up, which either he didn’t notice or didn’t care about. Russell’s 1903 view also abandons the idea that true propositions accurately represent how things are. Consider two rocks, A and B, with A sitting on top of B. For Russell, the proposition that A is on top of B is this pile of rocks. But a pile of rocks isn’t true or false. A pile of rocks doesn’t say anything about the way the world is, in any intuitive sense of ‘say’. It is not about anything. It doesn’t represent the world to be a certain way. To my mind, this more than

3

anything disqualifies Russell’s facts from being bearers of truth and falsity and hence from being propositions. In order for something to be true or false it has to be a representation, and Russell’s facts are not representations. Speaks’s view faces exactly the same problem.

On Speaks’s account,

propositions are global properties that are had by everything or nothing. The proposition that Amelia talks is the property of being such that Amelia talks. If Amelia talks then everything — you, me, this table, the St. Louis arch — is such that Amelia talks, and hence everything has this property. If Amelia doesn’t talk then nothing has it. This allows Speaks to say that a proposition is true if it is instantiated by everything and false otherwise (p.76). In other words, Speaks identifies truth and falsity with the properties of being instantiated or not being instantiated. But, as Speaks is well aware, this doesn’t change the fact that these global properties are not representations. The property of being such that Amelia talks does not say that Amelia talks, and it is not about Amelia or talking or anything else. Here I am just elaborating a problem that Soames raises for Speaks. Let me quote Soames and then Speaks’s reply: For Speaks, propositions are a certain kind of property, and truth is instantiation. This makes it difficult to capture the fact that truth is a kind of accuracy in representation. A map or portrait is accurate or veridical, when it represents its subject matter as being how it really is; a proposition is true when it represents things as they really are. This parallel seems to be lost when propositions are identified with properties that, as Speaks admits, aren’t intrinsically representational. (Soames, p. 167)

4

One might worry, though, that saying that propositions can have truth conditions without being about anything will force us to deny some platitudes about truth. For example, it seems plausible that something is true just in case it represents the world as being some way, and the world is that way — but that can’t be right if, as I think, propositions are true or false but don’t represent the world as being any way at all. … The right thing to say, it seems to me, is that these seeming platitudes are really platitudes — but only when we’re talking about the notion of truth applicable to sentences or beliefs. … Like many, I think that propositions are the fundamental bearers of truth and falsity and that the truth and falsity of sentences, mental states, maps, etc. is to be analyzed in terms of the truth and falsity of propositions to which these entities stand in the relevant relations. On a view like this, it will be no surprise if the right account of truth and falsity for propositions is different from the right account of truth and falsity for the non-fundamental bearers of truth and falsity. (Speaks, p.221) Here’s a platitude about colors: anything red is colored.

This platitude captures a

conceptual or a priori connection between the properties of being red and being colored. That is why it is a platitude. Given this status, one can’t say that it only applies to some of the red things. If it is a priori that red things are colored, then all red things are colored, no exceptions. I think the platitude about truth has the same status. It is a conceptual or a priori truth that something is true if it represents the world to be some

5

way, and the world is that way. There’s nothing here about whether something is fundamentally or derivatively a bearer of truth. This is an a priori principle connecting the concepts of truth and representation. It’s therefore no good to say that it only applies to some of the truth-bearers. If it is a priori that truth requires representation, then anything that is true or false has to be a representation. Speaks might deny that the platitude expresses a conceptual truth about truth. But then what makes it a platitude? Anyway, let me try a different tack. Let’s return to the parallels between Speaks’s approach and Russell’s. Both reduce propositions to nonrepresentational entities: facts for Russell, global properties for Speaks. Both identify truth and falsity with properties had by these non-representational entities: existence and non-existence for Russell, instantiation and non-instantiation for Speaks. There’s a general strategy here for coming up with an account of propositions: find some non-representational entities to serve as propositions and then identify truth and falsity with properties of those entities. Once this general strategy is on the table it is not hard to think of other examples of it. Suppose we’re happy with an ontology of facts (i.e. existing facts, not Russell’s merely subsistent ones). Speaks seems okay with facts (p.72). If there are facts then surely there are properties of facts, like the property of being interesting, or the property of containing Amelia, or the property of being the fact that Amelia talks. Let the proposition that Amelia talks be the property of being the fact that Amelia talks. More generally, identify the proposition that p with the property of being the fact that p. Then we can identify truth and falsity with instantiation or noninstantiation of these properties. The proposition that Amelia talks is true because the property of being the fact that Amelia talks is instantiated. That seems to work just as

6

well as Speaks’s view. And there are other strategies in the vicinity, such as Mark Richard’s recent proposal to identify propositions with properties of situations, e.g. the property of being a situation in which Amelia instantiates talking (Richard 2013). Then we can say that a proposition is true when it is instantiated by the maximal situation, the whole world.

Speaks himself mentions a related strategy in a footnote, on which

propositions are identified with properties of worlds (fn. 11, p.77), although he doesn’t explain why he prefers his official view. Here’s another example of this general strategy. Let’s identify propositions with ordered pairs of objects and relations. So, for example, the proposition that Amelia talks is the ordered pair . More generally, identify the proposition that Rno1,…on with the ordered pair of << o1,…on>, Rn> (cf. Soames 1987). Here’s a property had by these ordered pairs — call it the property of being T. An ordered pair << o1,…on>, Rn> is T iff instantiates Rn. Now we can identify truth with the property of being T. That would vindicate the old view that propositions are ordered pairs. I take it, though, that King, Soames, and Speaks all agree that ordered pairs aren’t good candidates for being propositions. Why not? The standard line is that ordered pairs can’t be propositions because they’re not true or false, and they’re not true or false because they’re not representations. If we’ve severed the connection between truth and representation then this line of argument no longer carries any weight. So why not identify propositions with ordered pairs? Again, that seems to work just as well as Speaks’s view. There is, of course, a Benacerraf problem for the ordered pairs account. There are many ways of identifying propositions with ordered pairs and it is hard to see why one

7

reductive scheme is any better than the others. One of the lessons of the Benacerraf problem is that propositions shouldn’t be identified with ordered pairs. At best, we can use ordered pairs to represent or model propositions. I think the same should be said about Speaks’s view and its variants. Non-representational entities can be used as standins for propositions, but if we want to know what propositions are we have to find things that are genuinely representational.

Soames Let me start with a worry about what Soames says about perception. According to Soames, when I perceive a poster as red, I predicate redness of the poster (p.93). Furthermore, for Soames, to predicate redness of the poster is to entertain the proposition that the poster is red. Note that predication is basic here. Soames uses his concept of predication to analyze entertainment, not the other way around.

Still, for Soames,

predication and entertainment are the same thing, even if predication is explanatorily prior. Here’s the problem. Suppose I perceive the poster as red and then form the belief that it is red. In this case, my perceptual experience ought to justify my belief. This is a basic role for perceptual experiences — justifying perceptual beliefs. Clearly, though, an act of entertaining the proposition that the poster is red cannot justify my belief that it is red. Entertaining a proposition doesn’t justify belief in that proposition. I can’t justify my belief that the Twins will win the next World Series by saying that I spent all morning entertaining that proposition. If, as some think, believing that p requires entertaining the

8

proposition that p, and entertaining that p justifies the belief that p, then every belief would be justified. Something has gone wrong. I think the problem is with Soames’s identification of entertainment with predication.

Soames analyzes the traditional concept of

entertainment in terms of his notion of predication. There are, however, important differences between entertainment and predication. On Soames view, when I predicate redness of an object o, I represent o as red. This act of representation can be accurate or inaccurate. In representing o as red, I do so accurately or inaccurately. This means that I can get things wrong. If o is not red, then I inaccurately represented o as red. That means I made a mistake. I got things wrong when I predicated redness of o. Now, compare this with the traditional notion of entertaining a proposition. To entertain a proposition is to enter into neutral cognitive contact with that proposition. It is to single it out or identify it so that one can judge it or take some other attitude toward it. As Dummett puts it, entertaining a proposition is “fixing one’s attention on it,” (1981, p.298). When I entertain the proposition that o is red, in this traditional sense, I don’t represent o as red. The proposition does that, not me. I don’t represent o as red by fixing my attention on something that represents o as red. The proposition might get things wrong, but I can’t. You can’t entertain a proposition accurately or inaccurately. You don’t make a mistake by entertaining a false proposition. The mistake occurs if you go on to judge the proposition, not when you merely entertain it. On reflection, it looks like Soames’s notion of predication mixes together elements of entertainment and judgment. Like entertainment, predication is supposed to be a neutral kernel inside all the other attitudes.

As Soames puts it,

9

predication/entertainment “is the attitude on which the others are, in one way or another, based,” (p.97). Like entertainment, predication is non-committal and neutral about how things are. But like judgment, and unlike entertainment, predication is supposed to involve representing an object as being a certain way, which is something that can be done accurately or inaccurately. Like judgment, predication is a kind of action in which you can get things wrong. So Soames’s notion of predication combines the neutrality of entertainment with the representationality of judgment. My main worry is that this is incoherent. On the traditional view, the act of entertaining a proposition is neutral but not representational. The act of judging a propososition is representational but not neutral. It’s no accident that neutrality and representation are kept separate on the traditional view. Combining them leads to incoherence. Here’s a way to draw out the incoherence. Suppose I predicate redness of o, in Soames’s sense of ‘predicate’.

Suppose also that o is not red.

It follows that in

performing my act of predication I inaccurately represented o as red. That means that I made a mistake. If I inaccurately represented o as red, then I got o wrong — I made a mistake about o. For that to happen, though, I must have taken a position about whether o is red. It can’t be that I made a mistake about whether o is red and yet remained completely neutral on the question of whether o is red. So in performing my act of predication I didn’t remain neutral about whether o is red. But on Soames’s account, when you predicate a property of an object you’re supposed to remain neutral about whether the object has that property. This is incoherent. You can’t represent o as red and at the same time remain neutral about whether o is red.

10

That said, it would be quite convenient to have something that does what Soamesian predication is supposed to do.

Let’s return to perception.

Perceptual

experience is independent of judgment and belief. Suppose I perceive a poster as red. As Soames observes, the fact that my perceptual experience “represents it [the poster] in this way is independent of whether or not one forms the perceptual belief that it is red,” (p.93). Call this the belief-independence of perception. If, as Soames thinks, perception involves predication, then the neutrality of predication can explain the belief-independence of perception. Furthermore, perceiving the poster as red justifies my belief that it is red. That’s captured by the fact that in having the perceptual experience I represented the poster as red. I can justify my belief that the poster is red by pointing to the fact that I, or my visual system, represented it as red in a perceptual experience. The justificatory role of perceptual experience is thus captured by the fact that predication is representational. Perception is both belief-independent and justificatory of belief.

These aspects of

perception are nicely captured by the neutrality and representationality of Soamesian predication. To my mind, this is too convenient. The philosophical puzzle about perception is about making sense of how a mental state could be both non-committal and justificatory of belief. How can it be that in perceiving o is red I take no stand about whether o is red, yet being in this state justifies my belief that o is red? If the perception is neutral about whether o is red then how can it provide a reason for thinking that o is red? It would be great if we could solve this puzzle by positing an act that allows me to remain neutral about whether o is red while representing it as red. But that seems too easy. If anything,

11

it relocates the problem. How is it possible for there to be such an act? If what I argued above is correct, it’s not possible. I think the right way for Soames to go here is to give up the neutrality of predication. Once that change is made then I am pretty much on board for the rest of his view.

King There is a concept of ascription at the heart of King’s theory that is remarkably similar to Soames’s notion of predication.

In fact, ascription looks exactly like

Soamesian predication, and it has the same problem of incoherence. On King’s account, we interpret syntactic relations as ascribing properties to objects. Because of this, we also interpret the unifying relations in propositions as ascribing properties to objects. This is what endows propositions with truth conditions. On King’s view, the proposition that Michael swims is a certain structure or fact in which Michael and the property of swimming are bound together by an amalgam of syntactic and semantic relations. The reason this structure is true iff Michael swims is that we interpret it to have these truth conditions. How do we do that? By interpreting the amalgam of syntactic and semantic relations to ascribe the property of swimming to Michael. And the reason we do that is that we interpret the syntactic relation in ‘Michael swims’ as ascribing the property of swimming to Michael. King’s explanation for the truth conditions of propositions starts with the idea that we take syntactic relations in sentences to ascribe properties to objects.

12

On the face of it, though, that’s a very odd thing to claim. Ascription is a kind of action. It sounds very weird to say that a syntactic relation performs an act of ascribing a property to an object. But that’s not really what King means. The sense in which we take or interpret syntactic relations to ascribe properties to objects is that when we, competent speakers of English, encounter a sentence like ‘Michael swims’, we ascribe the property of swimming to Michael. We take the syntactic relation to perform an act of ascription in the sense that we perform an act of ascription when we encounter a sentence with that syntactic relation. As King puts it: That speakers interpret syntactic concatenation in the ways they do consists in the fact that they spontaneously and unreflectively compose the semantic values of the concatenated expressions in the ways described. Hence, this is how my talk of R above [the syntactic relation in ‘Michael swims’] being interpreted by English speakers as ascribing the property of swimming to Michael should be understood. (p.49) When we encounter the sentence ‘Michael swims’ we compose the semantic values of ‘Michael’ and ‘swims’ in a certain way, namely by ascribing the semantic value of ‘swims’ to the semantic value of ‘Michael’. This is what it means for us to interpret the syntactic relation in ‘Michael swims’ as ascribing the property of swimming to Michael. For King, it’s this fact that ultimately explains how propositions have truth conditions. But what is this act of ascription that we are supposed to perform when we encounter ‘Michael swims’? It can’t be that we’re committing ourselves to the truth of this sentence just by looking at it and understanding it. Ascription must be neutral and

13

non-committal like Soamesian predication.1 At the same time, acts of ascription must be truth-evaluable, which means they must be representational. Otherwise it’s hard to see how interpreting a unifying relation to perform an act of ascription could endow the corresponding unified fact with truth conditions.

So in ascribing the property of

swimming to Michael I represent Michael as swimming, and therefore do something that can be evaluated for truth and falsity. Just like Soamesian predication, King’s concept of ascription is both neutral and representational. As I argued above, this is incoherent. But let’s bracket all of that and look at a different problem for King. King’s view is an example of what we might call an interpretivist theory of propositions. Interpretivist views explain how propositions are representational and have truth conditional by appealing to the way we interpret them. Consider the fact that there is a coffee mug on my desk. This fact doesn’t represent anything and it’s not true or false. But we could take it to represent something else. Suppose we took the fact that a coffee mug is on my desk to mean that it’s snowing in Minneapolis. Now the fact about my desk represents a fact about the weather in Minneapolis. If one day there’s a coffee mug on my desk and it’s snowing in Minneapolis, then we could point to the coffee mug and desk and say ‘It’s true’.

On an interpretivist theory of propositions, this is how

propositions come to be representational. Here’s how it goes: first we find some structures or entities or facts that aren’t inherently representational, but which are otherwise suited to be propositions (e.g. they exist, there are enough of them, we have cognitive access to them, etc.). Then we explain how we endow these things with

1

Elsewhere, King says that entertaining the proposition that Michael swims involves ascribing the property of swimming to Michael (King 2013, 80). This reinforces the idea that King regards ascription as neutral and non-committal. 14

representational features and truth conditions by taking them to have these features and truth conditions. In broad strokes this is how King’s theory works. There is a dilemma facing any version of this interpretivist strategy. Interpreting one thing to represent another thing is a sophisticated cognitive achievement. Doing so surely requires having beliefs and intentions and other propositional attitudes. But the propositional contents of these attitudes cannot be the ones generated by the acts of interpretation in which they are involved. The propositional contents of these attitudes must come from somewhere else.

Now we face a dilemma.

Either we apply the

interpretivist strategy all over again to generate propositional contents for these prior attitudes, or we don’t. Re-applying the interpretivist strategy leads to a regress, since it will require its own antecedent attitudes and propositional contents, which will lead to another iteration of the interpretivist account, and so on. On the other hand, if we don’t apply the interpretivist strategy then we need some non-interpretivist story about what these prior propositions are and how they are representational. But if such a story is possible, why did we need the interpretivist account in the first place?

Either the

interpretivist account leads to a regress or it undermines itself. There are some remarks in King’s portion of the book that suggest to me that he would grasp the second horn of the dilemma, i.e. the one on which we don’t reapply the interpretivist strategy. The remarks come in the course of King replying to an objection from Speaks about the propositional contents of perceptual experiences (pp. 193-5). King thinks that perceptual experiences have propositional contents with truth conditions. Like the contents of sentences, these propositional contents are facts bound together by a propositional relation. As King emphasizes, it is highly unlikely that the propositional

15

relation in the contents of sentences is the same as that in the contents of perceptual experiences. For this reason, the interpretivist story about how the contents of sentences acquire their truth conditions doesn’t carry over to the contents of perceptual experiences. Furthermore, for reasons brought to light by Speaks (pp. 157-8), King doubts that any interpretivist account will work for perceptual propositional contents. So it’s not acts of interpretation that endow the propositional contents of perceptual experiences with truth conditions. What does endow them with truth conditions? Somewhat tentatively, King suggests that “my account of the contents of perceptual experience can hold that whatever fixes the phenomenal character of a perceptual experience endows its propositional relation with semantic significance and brings the proposition that is its content into existence,” (p.195). So perceptual experiences have propositional contents, which can serve as the contents of the attitudes that go into generating the propositional contents of sentences, but these perceptual propositional contents do not owe their truth conditions to acts of interpretation. There are a number of things to be worried about here (do we really want a distinction between linguistic and perceptual propositions?), but let me focus what I think is the main concern. The intuitive pull of King’s account of the truth conditions of propositions comes from the idea that, in an extended sense, propositional relations ascribe properties to objects. If something ascribes a property to an object then it is true iff the object has that property. The explanatory efficacy of King’s account comes from this intuitive connection between ascription and truth conditions. I strongly suspect that any explanation of how something has truth conditions is going to have to appeal to acts of ascription or predication performed by people somewhere along the line. (In King’s

16

account this appeal gets buried in his overall interpretivist approach, whereas in Soames’s account it is right there on the surface. That’s one of the reasons I prefer Soames’s account to King’s.)

On King’s theory, these acts of ascription are in some sense

performed by propositional relations. As we saw earlier, that can’t be literally true. It would be utterly bizarre to say that a relation literally performs an act of ascribing a property to an object. The sense in which propositional relations ascribe properties to objects must come from how we interpret those relations, which ultimately is a matter of us performing acts of ascription.

This is how King’s interpretivist theory works.

Propositional relations perform acts of ascription in the sense that we interpret them to perform these acts. If we drop this interpretivist line then it is difficult to see how to make sense of the idea that a propositional relation ascribes a property to an object, and without that we no longer have an explanation of truth conditions. To sum up: in order to explain how propositions have truth conditions King needs to explain how, in some extended sense, propositional relations ascribe properties to objects, but the only way to do that is to appeal to the way we interpret propositional relations. For King, it looks like interpretivism is unavoidable, which puts him on the regress horn of the dilemma. There’s another brief remark by King that suggests that he would try to halt the regress by appealing to what he calls “proto intentional” states: I raised this worry in King [2007] and responded by noting that our prelinguistic ancestors could have had “proto intentional” states that weren’t relations to propositions and that these mental states could have been sufficient for them to bring language, and hence propositions, into existence. It seems to me quite plausible that various creatures have such

17

proto intentional mental states and so the idea that our prelinguistic ancestors did so as well is reasonable. (p. 60; see also King 2007, p.67) This would stop the regress right at the beginning. Our prelinguistic ancestors didn’t need beliefs and intentions in order to perform the acts of interpretation necessary for bringing propositions into existence. Rather, they had “proto intentional” states that could do the work. These states don’t have propositional contents, but otherwise they can explain how people could interpret one fact to represent another fact. This is way too convenient. If it were this easy to halt a regress then a lot of philosophical regress problems would become a lot less interesting. Here’s a way to halt the regress of justification: posit proto beliefs that can justify other beliefs but which themselves don’t require justification. Here’s a way to halt the regress of explanation: posit proto facts that can explain other facts but which themselves don’t require explanation. Similarly for the regress of causation: posit proto causes, which can cause other events but which themselves don’t require causes.

King’s appeal to proto

intentional mental states is another instance of this strategy: posit proto intentional states that can figure in acts of interpretation but which themselves don’t require acts of interpretation to give them propositional contents. If this is King’s only recourse here then I think we have good reason to look elsewhere. References Dummett, Michael 1981: Frege: Philosophy of Language. Cambridge, MA: Harvard University Press. King, Jeffrey 2007: The Nature and Structure of Content. Oxford: Oxford University Press. King, Jeffrey 2013. “Propositional Unity: What’s the Problem, Who Has it and Who

18

Solves it?”. Philosophical Studies, 165, pp.71-93. Richard, Mark 2013: “What are Propositions?” Canadian Journal of Philosophy, 43, Special Issue: Essays on the Nature of Propositions, David Hunter and Gurpreet Rattan (eds.), pp. 518-33. Russell, Bertrand 1903: Principles of Mathematics. New York: Norton. Russell, Bertrand 1904: “Meinong’s Theory of Complexes and Assumptions (III)” Mind, 13, pp. 509-24. Russell, Bertrand 1906: “On the Nature of Truth”. Proceedings of the Aristotelian Society, 7, pp. 28-49. Russell, Bertrand 1910: “On the Nature of Truth and Falsehood”. In Philosophical Essays. New York: Simon and Schuster, pp. 147-59. Soames, Scott 1987: “Direct Reference, Propositional Attitudes, and Semantic Content”. Philosophical Topics, 15, pp.47-87. Reprinted in Propositions and Attitudes, Nathan Salmon and Scott Soames (eds.), Oxford: Oxford University Press, 1988, pp. 197-239.

19