Tractarian Propositions and Taking the 'X in Y' Construction Seriously

This is a sequel to the previous post, in which I continue to react to interesting recent "cognitive act theories" of propositions championed by Soames and Hanks, and work out my own views in relation to them.

In Soames's recent article on the Tractatus, he argues that Wittgenstein's conception of a proposition - that it is a propositional sign in its projective relation to the world - is incoherent. Soames's idea is that this doesn't actually specify a thing over and above the propositional sign. Just as Soames-in-relation-to-his-wife (one of his examples) isn't actually a separate thing or person from Soames, a propositional sign in its projective relation to the world isn't actually something other than the propositional sign. It is not some "larger" entity which includes the propositional sign as a part.

This then leads Soames to propose, on the basis also of a remark in the Tractatus that our thinking the sense of a proposition is our putting it in relation to the world, that Wittgenstein would have been better off identifying propositions as propositional signs together with cognitive acts. And this is similar to his own recent theory of propositions, on which they are just cognitive acts in abstraction from any particular signs. 

In this way, Soames presents Wittgenstein as groping toward a better view of the nature of propositions - better than those of Frege and Russell, for whom propositions were Platonic things independent of language use - and offering a more coherent way of doing this.

Recently Peter Hanks has argued that there is another way here: that we can instead look at the (to me confusing and confused) idea in the Tractatus that a propositional sign is a fact. (This is something Wittgenstein later came to think of as a category mistake - but the new act theorists of propositions are generally disposed to be less keen to convict philosophers of category mistakes. After all, the view that a proposition is an act - a thing done - itself sounds like a category mistake. So we're in a pocket of philosophy where there's a fair amount of tolerance of what can seem like category mistakes, being explained away in terms of unimportant intuitions that should be overcome as we better systematise out thinking.) Looking at it that way, Hanks argues that the propositional sign in its relation to the world may be seen as a "larger" fact which involves the fact that is the propositional sign, but where the elements of that sign/fact that are related to one another are also related to further elements, things out in the world. 

I share Hanks's sense that Soames's argument against the coherence of Wittgenstein's conception of propositions and how they relate to propositional signs is surmountable. But I confess that the idea of a sign as a fact has never appealed to me, and really does seem like a category mistake. 

I think there's a third way to understand this talk of a proposition being a propositional sign in its projective relation to the world. It may even be more faithful to the Tractatus, but maybe not. I think it probably is more faithful to Wittgenstein's ideas as they developed after the Tractatus though. More importantly, I think it may be the best way of thinking about these matters.

Soames complains that a thing in some relation isn't actually a separate thing. And this is meant to be an objection to the Tractarian notion of a proposition, the idea being that propositions are meant to be distinct things from propositional signs. 

But who said they had to be? The objection I have to both Soames's and Hanks's reconstructions of the Tractarian idea is that they both look for a way to avoid taking the 'X in Y' construction seriously, whereas part of what makes it a truly radical and fruitful idea may get lost that way. (The new act theorists want propositions to be inherently representational things. But perhaps part of Wittgenstein's thinking is that really we have signs, and we use them, and it is only in those uses that they bear representational properties and properties like truth. And so looking for a further object over and above the sign is unnecessary, and even a mistake.)

Now, there are different ways to think about how this conception can be expressed, and how expressions of it can be decomposed. You might think of it this way: there's a complex sign, which may get used in two different language systems. Then you might think that this sign is true in one of its projective relations to the world, and false in another. Here the 'in projective relation R' becomes part of the predicate, like 'true-in-L' in discussions of Tarski. But you can also put it into the subject, 'The sign in relation R' and then predicate truth of the sign in that relation. Is this a further entity over and above the sign? You can think of it that way, but perhaps there's another way here, where we just have the property of truth - not some more complicated projective-relation-involving predicate - and we just have the sign as our main entity. But we nevertheless predicate truth not of the sign simpliciter but we predicate it of the sign in a particular projective relation to the world. On this conception, it's not that we have an augmented predicate or a thing over and above the sign, rather we're just using a logical form which isn't just a simple subject-predicate proposition of the sort whose truth conditions can be given as: the proposition is true iff the property expressed by the predicate is possessed by the entity denoted by the subject. To squeeze the Tractarian conception of propositions into that form is to water it down. In particular, it is to water down its ability to get around the problem laying at the foundation of the new act-based theories of propositions - the problem about propositions needing to be inherently representational. Taking the Tractarian idea of a proposition more seriously, we don't have to find a thing that is inherently representational and then put that at the basis of our theories. We may insist that the primary truth bearers really are signs, but that these signs only bear truth in use, and that a given sign can bear truth in one use and falsity in another. 

References

Hanks, Peter (2019). Soames on the Tractatus. Philosophical Studies 176 (5):1367-1376.

Soames, Scott (2016). Propositions, The Tractatus, and "The Single Great Problem of Philosophy". Critica 48 (143):3-19.

Against Inherently Representational Anything

Soames, in his fascinating recent work Rethinking Language, Mind, and Meaning, begins by posing a problem for the study of meaning and language as developed by philosophers and logicians in the Twentieth Century.

For propositions to play the theoretical roles assigned to them - such as being the primary bearers of truth, being the objects of propositional attitudes like belief and desire, being the contents of mental and perceptual states, and being the meanings of some sentences - they, Soames says, must be inherently representational. That is, they must impose conditions on the world off their own bat, so to speak. 

But, argues Soames, the sorts of things that traditionally play the proposition role in modern theories do not seem to be inherently representational. Soames provides 'reasons to believe that no set-theoretic construction of objects, properties, world-states or other denizens of Plato’s heaven, could ever be inherently representational bearers of truth conditions in this sense' (Rethinking, Ch. 2).

This leads Soames to his new theory of propositions, on which they are cognitive acts of a certain kind. ('Suppose, however, we start at the other end, taking it as an uncontested certainty that agents represent things as being certain ways when they think of them as being those ways' (Rethinking, Ch. 2).) For example, the proposition that snow is white, on Soames's theory, is the act of predicating whiteness of snow. These cognitive acts, according to Soames, are inherently representational, which means that they could be able to play the role of propositions. 

(Soames says that, at bottom, it is language users that represent things as being certain ways, and they do this by performing cognitive acts, which acts may derivatively be said to represent. But although, in this way, they represent in a derivative sense, they do so inherently - and that is the crucial point for Soames, that makes cognitive acts fit to play the role of propositions.)

I am very heartened to see Soames realising that his earlier, broadly Russellian conception of propositions won't do and looking for an alternative. But his cognitive acts seem a bit mysterious to me. In this note I won't try to refute Soames's new theory of propositions, but let me say something briefly about what worries me. It's not so much that I think that there's no such thing as the cognitive act of predicating whiteness of snow, but I don't feel like this is the sort of thing that is fit to play the sort of basic explanatory role that Soames wants it to play. It feels too much like a black box containing important workings for that. Treated as something basic, it feels occult or magical. (Wittgenstein in the Investigations seems very concerned to avoid positing mental goings-on that are meant to play this kind of foundational role in explaining representation. This has influenced me and I think there's something right about Wittgenstein's conviction that this is not the way forward.)

There is another conception of propositions, on which they are sentences (of a certain kind) in use (or a certain kind of use). For instance, in the Tractatus Wittgenstein said that a proposition is a propositional sign in a projective relation to the world. (In a recent paper I sketch an account of propositions that treats them similarly.)

(Of course, it is nice to be able to say what two synonymous sentences in different languages have in common, and often it is said that they 'express the same proposition'. On the present approach, they may be said to have the same use, the same meaning. But ultimately it is not just the use or the meaning that represents - it is a sign in use, a sign with meaning, that does this. The "proposition" or "statement" or whatever you call it that two different sentences express is an abstraction from the particular propositions that are alike in meaning.)

It seems to me that this fundamental logical move of treating propositions not as things by themselves, as it were, but a certain kind of thing in a certain kind of context is very important, and constitutes the right way to avoid the twin pitfalls of having propositions be things that don't themselves represent, and of having them be explanatorily basic cognitive acts. 

Admittedly, sentences in use can't just be slotted in to all the roles Soames delineates without further ado. For instance, we probably don't want to say that if you believe something - perhaps without even representing it to yourself linguistically - the thing you believe is a sentence in a particular use. Soames's theory may have an advantage over my approach here in having this one kind of thing - propositions - playing all of these roles (although I doubt it, since Soames's theory has predictions which sound like category-mistakes, such as that one may 'perform a proposition', propositions being acts according to this theory). But I think that ultimately such a theory with one single sort of thing playing all these diverse roles will not be attractive overall, and that we may have to refine the picture somewhat of how these things like sentence meaning and the objects of belief relate to each other. I am focusing for now on linguistic meaning.

Sentences in use, you might say are 'inherently representational': but the whole idea of inherence doesn't really fit here. The whole point is that sentences are not inherently representational, but used in certain ways, they are. If you want to treat the sentence-in-use as a sort of thing by itself, then this is a kind of abstraction. Such a logical construction may be useful, but the resulting entity is not explanatorily basic: underneath, you have the sentence and you have all the stuff about how it is used. And so, at the base level, you don't have anything inherently representational.

This approach fits naturally with the ideas of representing and of a representation. Fundamentally, a representation itself is just a concrete thing - like a drawing or a sentence. And it represents not inherently, but by being used in a certain way. People represent things as being certain ways by putting representations to use.

This approach also furnishes a kind of explanation of the feeling of occultness or suspiciousness in theories which posit inherently representational entities in their explanatory base: they feel unsatisfactory because they hide the contextual rabbit in the hat of some posited object. If we are forced to treat this object as basic, we can never see under the hat.

References

Haze, Tristan Grøtvedt (2018). Propositions, Meaning, and Names. Philosophical Forum 49 (3):335-362.

Soames, Scott (2015). Rethinking Language, Mind, and Meaning. Princeton University Press.

Wittgenstein, Ludwig (1922). Tractatus Logico-Philosophicus. Routledge & Kegan Paul.