1. Introduction
Begin with E.M. Forster. Remember, in Howards End (Reference Forster1910), the brilliant imperative ‘Only connect!’ What sort of thing is this connection business? Perhaps our first paradigm of connection is found in physical proximity: when things touch they are connected. But even while noting that, we can already appreciate that the significance of the phenomenon, of connection, is not fully revealed in that form. When touch matters, it matters because it is constituting some more abstract form of connection. To make matters worse, there are questions about what touching could itself amount to.
So, refining our question: what could connecting be that it should matter? Does connecting matter at all after all? And if so, why? In Forster's case, the idea of connecting involved building a bridge between ‘passion’ and ‘prose’, between the beast and the monk that are in us all. Without such a bridge, we are ‘meaningless fragments’. Forster suggests that some special relation – a ‘rainbow bridge’ – between what might be conceived as initially isolated, elemental parts of our selves is required, and will suffice, for a range of deep values: beauty, exaltation, happiness, love, and salvation:
Beauty ‘the bridge would be built and span their lives with beauty’
Exaltation ‘and both [prose and passion] will be exalted’
Happiness ‘Happy the man who sees from either aspect [the prosaic and the passionate]’
Love ‘with [the connecting bridge] love is born’
Salvation ‘the salvation that was latent in his own soul […]’
Most explicitly, Forster applies his proposal to the inner workings of a single mind or person or soul (Mr Wilcox). But he uses that individual case as a kind of model for the connections between individuals and other things – including those between people and between families – that can yield the same sorts of benefits collectively as they do in the individual case.
Whether or not the individual case – the case of connections between the parts of a single person – should have any priority, we can attend directly to the sorts of connection between ourselves and other things that are ultimately the broader theme of Forster's novel and of interest to anyone. So, again, what could it be to connect?
I propose that one deep form of connection is when something is about something. Being about is an important variety of connecting. When we think about someone, we're connected to them. Near or far, spatially or temporally, there they are, almost miraculously, with us! And when we engage each other about something, for example in discussion, then we are connected to that thing, as well as (thereby) to each other. Indeed our being so engaged depends on it – otherwise we're ‘talking past’ each other, not in a discussion.
Consider now the significance of this form of connection. Without the being about anything of a thought, the thinker is rendered isolated, the thought exhibits a lack of friction. And if when we talk, we are not together talking about some (one) thing, then we are not having a conversation, we are not in that way in community, we are to that extent not in society. This encourages the conclusion that when one thing is about another, they have connected in an important way. Notice that the examples I gave are of thoughts and ideas and then of discussion and talk. Being about things is something a mind can be involved in and is also something a bit of language can do. So let's pursue this relation of aboutness further. How does it work?
In his Philosophical Investigations (Reference Wittgenstein1953), Wittgenstein asked, ‘What makes my image of him into an image of him?’ He then wondered, ‘Isn't my question like this: “What makes this sentence a sentence that has to do with him?” […] And what makes our conversation about him a conversation about him?’
Notice, first, the range of ways in which we characterize the phenomenon: Wittgenstein is (translated as) speaking of images being ‘of’ someone and of sentences ‘having to do with’ things. I'm using being ‘about’ as the basic case. But there's also betokening, denoting, going proxy for, indicating, meaning, referencing or referring, representing, signaling, standing for, and symbolizing. And then notice, second, that like many before him and since, Wittgenstein in effect unifies questions about what makes for aboutness in the case of language and what makes for aboutness in the case of thought.
Kit Fine is explicit in his Semantic Relationism (Reference Fine2007):
The simplest and most natural view is that there is no more to the content of my belief than there is to the content of my words […] [it is] odd to suppose that there should be any fundamental difference in the general representational character of language and thought.
And here's Stephen Schiffer, in his Remnants of Meaning (Reference Schiffer1987):
Both mental states and sentences have what is called intentionality or representational content: a particular sentence means that worms do not have noses, and a particular state is a belief that worms do not have noses. What the theorist ultimately wants, of course, is a general theory of content, a theory of linguistic and mental representation.
Adopting a slightly different angle, Jerry Fodor, in his Psychosemantics (Reference Fodor1987), worried about whether intentionality is a natural phenomenon:
I suppose that sooner or later the physicists will complete the catalogue they've been compiling of the ultimate and irreducible properties of things. When they do, the likes of spin, charm, and charge will perhaps appear on their list. But aboutness surely won't; intentionality doesn't go that deep […]. If aboutness is real, it must really be something else.
Whether Fodor is right about the fundamentality of intentionality, notice that he, too, seems to take for granted that it is simplex.
Out of tune with this symphony, I will propose instead not one but two things for aboutness really to be! The contrasting unificatory presupposition is reasonable: there are good reasons for supposing that any account of what makes a word mean what it means must serve equally as an account of what makes a thought be about what it represents. After all, we seem to be able to put our thoughts into words. If you say something, I can believe it. But I nevertheless recommend that we reconsider the presupposition.
What controversy there has been about the relation between the linguistic and the mental case has concerned priority, over whether
(i) having language is a precondition of thought, so that linguistic meaning to that extent precedes the meanings of thought,
or instead
(ii) we in effect invest our languages with the sort of semantic content that our thoughts already possess, so that thought has priority.
My proposal rejects both sides of that debate: they both presuppose that the relevant linguistic phenomenon and the case of mental meaning can be ordered, in terms of priority, along the relevant dimension. On the suggestion here, neither precedes the other: they are independent.
I will not, here, show how I accommodate the point about communicating our thoughts in language; but I appreciate the need to do so. To make a start on developing the view that will face that challenge, we will begin by reviewing two key philosophical texts, one by the German logician Gottlob Frege and another by the American philosopher Saul Kripke.
2. The Fregean Turn
In ‘On Sense and Reference’, a seminal paper first published in Reference Frege, Geach and Black1892, Frege confronted a philosophical issue: ‘The discovery that the rising sun is not new every morning, but always the same, was one of the most fertile astronomical discoveries.’
The general issue was about the possibility of believing one thing without believing another, about having a thought with one significance without having another, different thought, with a different significance. Frege's Puzzle, as it's now known, was specifically about how we might already believe that Hesperus is Hesperus without yet knowing that Hesperus is Phosphorus. We discover the latter, even while already having the former belief. This sort of phenomenon is widespread: Agnes Bojaxhiu, Lev Bronstein, Lewis Carroll, Samuel Clemens, Robert Galbraith, Margaretha MacLeod, etc.
Is what we learn, when we learn that Cicero is Tully just what we already knew, when we knew that Cicero is Cicero? No. But isn't Cicero's being Tully just a matter of his self-identity, his being himself? The way the world has to be, for Clemens to be Twain is just how it has to be for that individual to be himself.
Here is one way to pose the issue. If
(i) what the sentence ‘Cicero is Cicero’ means is just what the sentence ‘Cicero is Tully’ means, and
(ii) what we have in mind, when we believe that Cicero is Cicero, is just what the sentence means,
then
(iii) we did already believe that Cicero is Tully.
Frege's solution was to introduce a new kind of semantic value… for words. But how could it possibly help to solve a problem about believing one thing and not another to add another layer to the semantic value of words? What does the semantics of words have to do with the prospect of one or another rational arrangement of belief? Nothing very immediate.
Frege saw a difference in the cognitive potential of minds that do and don't already believe that Hesperus is Phosphorus, and inferred that the meaning of the sentence ‘Hesperus is Phosphorus’ – the way that sentence is about what it is about – has to be able to explain that difference in cognitive potential. That Frege should have thought such a maneuver could help is a reflection, I think, of a presupposition: no distinction in cognitive potential without some distinction in linguistic semantics.
View the path I'm proposing here as an alternative, an alternative to that taken in what has come to be called ‘The Linguistic Turn’ (Rorty, Reference Rorty1967). The nature and structure of linguistic meaning is not, contrary to a key idea of the Linguistic Turn, a good model for thought and how it represents. Frege appreciated something deep: we need theoretical materials adequate to explain distinctions as fine-grained as can be made by an arbitrarily powerful rational mind. We need this in order to understand the nature of rational thought: inference, for example, both practical and theoretical, also coherence and incoherence, informativeness, and practical and theoretical justification. We need it to understand how there's no irrationality in believing that Phosphorus is bright even while not believing – as we might or might not put it – that Hesperus is bright.
But he remained committed to a problematic unification of the linguistic case and the case of thought: confronted by subjects whose rational beliefs differ in a familiar way, he proposed a new theory of linguistic meaning. In that way, he steered us toward the linguistic turn. We should separate more sharply the way thoughts connect to reality from the way language does.
Even while sentences of the form ‘a=a’ and ‘a=b’ have the same semantics, the latter form can nevertheless be informative because the informativeness of a sentence does not supervene on its semantics. Even relative to a fixed background set of beliefs, different sentences – even with the same semantics – can be differentially informative.
But wait: how can this be? How can sentences with the same meaning be differentially informative? Neither semantically encodes information the other doesn't. Nevertheless, differential informativeness can be a result of differential causal effect. If one sentence can, perhaps in virtue of its form, systematically cause a true belief that another sentence cannot cause, then those sentences are apt differentially to extend our knowledge. There is nothing problematic in the idea that different sentences should produce different effects causally, even should they have the same semantics.
The extension of our knowledge by ‘a=b’ is not necessarily a matter of our engaging a different linguistic meaning than when we encounter ‘a=a.’ It does, however, entail that we're caused to get different contents in mind. The product of a testimonial exchange, for example, the resulting belief in a rational audience, is not a causal result just of their existing beliefs and of the semantics of the input utterance. Which sentence is used matters too.
As a general point, this is obvious: even should sentences in different languages have the very same content, they are liable to be differentially informative to monolingual speakers of those languages, respectively. But the point holds up in a more specific deployment: even should the two sentences be in the same language, and even should your interlocutor understand both sentences, they may be caused to know something in accepting a sentence of the form ‘a=b’ that they could not come to know in accepting ‘a=a’. Informativeness is a causal phenomenon; and causal effect is sensitive to the specific form in which a content is transmitted.
So, as I said, I think Frege set us off on the wrong path when he presumed that if a pair of sentences can (even for a competent speaker of the relevant language) correspond to – or ‘give rise to’, or ‘be the expression of’ – different contents of belief, then they cannot have the same semantics. There's an implicit aspiration, not initially unreasonable, in this ultimately unsustainable insistence. But the semantics of sentences is not up to the task of articulating all relevant cognitive variability.
Frege remained committed to a problematic and still widely shared presupposition – call it ‘monosemanticism’: that the notion of content introduced to explain the nature of rational thought has to serve also as the semantic value of the expressions of a natural language. According to monosemanticism, the being meaningful of thoughts and the being meaningful – the having a semantic value – of linguistic expressions are fundamentally akin and correlated phenomena. That sort of monosemanticism was a good place to begin in thinking about content. It is valuable in accounting for certain apparently acceptable claims. (E.g., Ally believes everything that Billy said.) But we are no longer at the beginning of thought about content. It is time to reconsider the presupposition: in its place, I recommend polysemanticism, according to which there are multiple varieties of intentionality – so that a cognitive state's being contentful and a sentence's having a semantic value are fundamentally different – indeed independent – phenomena.
One significant consequence of the polysemanticist approach is a sharper separation between the Millianism/Fregeanism issue in philosophy of language and the externalism/internalism issue in philosophy of mind. A sign of the depth of the difference between cognitive and linguistic varieties of representation is in the fact that cognitive representation is internalist while linguistic representation is generally externalist. While duplicates – though just what it is to be the duplicate of a thinking thing is moot – will have propositional attitudes that represent alike, the referents of any names tokened by any duplicate human beings – which names might well even have a Millian semantics – might not be shared. And though the best linguistic theory may ultimately make the starred sentence below false, we can all still understand the true thought one might have tried to voice with it.
(*) ‘Lois believes Superman flies’ is true
even while
‘Lois believes Clark Kent flies’ is false.
Indeed, the very idea that Millianism and externalism are kindred philosophical positions, as they are often seen, should always have been odd: whereas the coarse-grained individuation of linguistic contents produced by Millianism permits substitution under what might seem relatively weak conditions (weaker than those imposed by Fregeanism), externalism's finding internal state insufficient for fixing content makes it a relatively demanding theory of content identity (and more demanding than internalism).
3. Kripke's Picture
As we pursue giving up the monosemanticist presupposition, we take up next a text that appears to oppose Frege's view. I think that the arguments Kripke brings to bear against a Fregean view of the semantics of names, however adequate they might be against their actual intended target, are not compelling as arguments against a Fregean conception of the contents of thought. I don't, incidentally, think Kripke himself tries to use them that way.
Perhaps the single best example, for seeing this, is the notorious Gödel/Schmidt case that Kripke gives in Naming and Necessity (Reference Kripke1980):
Let's take a simple case. In the case of Gödel [...] practically the only thing many people have heard about him — [is] that he discovered the incompleteness of arithmetic. Does it follow that whoever discovered the incompleteness of arithmetic is the referent of ‘Gödel’? […] Suppose that Gödel was not in fact the author of [the] theorem. A man named ‘Schmidt’, whose body was found in Vienna under mysterious circumstances many years ago, actually did the work in question.[...] On the view in question, then, [...] since the man who discovered the incompleteness of arithmetic is in fact Schmidt, we, when we talk about ‘Gödel’, are in fact always referring to Schmidt. But it seems to me that we are not. We simply are not.
This is an important passage and I think it makes a decisive point against a certain conception of how we refer with names. Recall the character in Carroll's Through the Looking Glass (Reference Carroll1871):
‘When I use a word,’ Humpty Dumpty said, in rather a scornful tone, ‘it means just what I choose it to mean — neither more nor less.’
The passage from Carroll is funny because of Humpty Dumpty's pretension. In fact words do not mean just what we then choose for them to mean. And similarly the name ‘Gödel’ refers to Gödel, whatever we might have meant to do with the word. Now then, in light of the distinction enabled by our rejection of monosemanticism, what should we say about the thoughts of the ‘ordinary person’ in Kripke's case?
Kripke could perhaps have been more explicit about this. But I think he does not intend to be making a point about the nature of belief content. Also elsewhere, in ‘A Puzzle About Belief’ (Reference Kripke and Margalit1979), Kripke talks explicitly about our practices of belief ascription, something we do with words. But ignoring the particulars of Kripke's texts for the moment, just ask yourself the following question: if, by explicit stipulation, the property discoverer of the incompleteness of arithmetic is exactly what one has in mind on a given occasion, and whatever word one then uses, what is that thought about? Just suppose that on a given occasion a subject's thought is descriptive – denying that the names have a descriptivist semantics does not entail that such thoughts are not so much as possible. A thinker is working on the incompleteness of arithmetic, for example, and forms the judgment that its discoverer was brilliant. Who is the thinker judging to be brilliant? It is not plausible, in this case, that the referent of the subject's thought should be the thief, Gödel. It seems rather to be about, well, the discoverer: Schmidt. Whatever the subject might go on to say, and whatever might then be referenced by any words they use on that occasion, the thinker seems to refer in thought to Schmidt.
In Kripke's example the subject uses the word ‘Gödel’. And in using that name they make linguistic reference to Gödel. But that does not entail that they believe that Gödel is brilliant. After all, by stipulation, it's discoverer of the incompleteness of arithmetic that the subject has in mind, and so it's the discoverer of the incompleteness of arithmetic that they believe to be brilliant. That was Schmidt.
So there's a way to accept Kripke's insight about names and other linguistic devices and how they represent even while preserving Frege's insight into the way rational minds work – even should Frege himself have misdeployed that insight and left himself open to Kripke's rebuttal.
Kripke claims that we do not associate with proper names properties that we believe to pick out an individual uniquely. We often know that most all the properties we associate with one name are the very ones we would also associate with other names. True: we should all agree that a word's being meaningfully used to refer to something does not depend on a user's associating with that word a property that they believe to pick that thing out uniquely. On one version of polysemanticism, for example, a good ‘picture’ of linguistic reference is drawn as Kripke does, in terms of a network of causal relations, without appealing in any problematic way to the meaning of a user's thoughts.
Suppose that on a given occasion you do have in mind just the property leading theoretical physicist and you don't believe it picks anyone out uniquely. You would then not use any mental analogue of ‘the’ with that concept; you would use an indefinite operator, the mental analogue of ‘a’. You might on a given occasion think this: no doubt a leading theoretical physicist will be asked to serve on that blue-ribbon commission. But whatever definite referring you might then do with the name ‘Feynman’, in such a case it is not plausible that your thought is referring differentially to Feynman rather than to Gell-Mann. Of course, normally, if you did then use that name competently, it will be because you in fact have the property leading theoretical physicist named ‘Feynman’ (or something like that) in mind: one needn't defend a metalinguistic theory of names to think that occasionally subjects have is in this context uniquely referred to by [N] in mind when using a name. But that's not offered now as a general theory of the meaning of names; it is a plausible speculation about what might sometimes be in the mind of a subject using a name.
And contra Kripke's claim in the ‘Puzzle About Belief’, it does not seem possible to wonder whether Cicero is Tully, so to speak, if the content you have in mind, associated with the ‘Cicero’ term, is exactly the content you have associated with the ‘Tully’ term. An identity issue presented to you as exactly the same content, twice over, flanking an identity concept, cannot provoke uncertainty.
We do not have to be committed to a metalinguistic theory of the meaning of names in order to recognize the role of metalinguistic commitments in the transition from an attitude to its expression. A speaker will tend to communicate their belief that an F is G with a name ‘N’ only if they associate the F with ‘N’. All of this is consistent with the most radical Millianism about the semantics of names.
If the way minds manage the task of being about things is fundamentally different from the way words manage that task, then Frege's proposal, though he pitched it as about the way language works, can be accepted as a viable claim about the way minds work. And Kripke's insights about the way language works need not be seen as incompatible with the new conception of the contents of thought.
In Kripke's example, if we have discoverer of the incompleteness of arithmetic in mind, and use the word Gödel, then our thought is about Schmidt and our sentence is about Gödel.
Aboutness bifurcates.
4. Conclusion
I've revisited Frege's ‘On Sense and Reference’ and Kripke's Naming and Necessity from, if you will, a polysemantical point of view. The leading ideas so far are that Frege's seminal insight actually concerns the nature of thought and its contents and that expressing that insight in the way he did, as about the semantics of proper names, is problematic – ultimately, I believe, for the reasons Kripke gives. At the same time, Kripke's arguments are compelling specifically when applied to the form in which Frege did express his insight – they are not telling against a Fregean conception of thought content.
Once we separate linguistic semantics from the intentionality of thought, then the fact – as a Millian semantics would have it – that ‘If Lois believes that Superman can fly then she believes that Clark can fly’ is a true sentence of English is rendered a different sort of philosophical challenge. Our contrary intuition is about the possible condition of Lois's mind, about how she might rationally represent reality in thought. And the truth of such sentences does not determine the contents of her mind – if you'll allow me perhaps to misuse words just now in this way! – determined as they are by their meanings.
There is accordingly a sense, awkward though it may be to express, in which ‘belief’ reports do not report beliefs: my apologies, I have no better word and yet we all know what I mean.
Polysemanticism is a useful framework for Millianism: whether or not it is best to allow the truth of ‘Lois believes that Clark can fly’, the intuition that drives resistance can be preserved. Perhaps the most refined form in which to put the point is that the thought driving resistance to the truth of that sentence is itself just that, a thought. It's an intuition about Lois and whether she believes something. Without using ‘that Clark can fly’ to express that thought, without, indeed, using any other ordinary English sentence involving a name to specify it, we can anyway engage, in our own thought, the thought that Lois might lack. Whether to use one or another sentence in expression of her belief is a decision we will make collectively, answering to the sorts of pressure characteristic of linguistic convention. But whether Lois actually has one or another thought is not similarly a matter of convention.
Now if, in the grip of a view according to which the contents of thought are just what's expressed by ordinary sentences, we try heroically to discover the linguistic expression that reflects her mental circumstances, we might ourselves say that she ‘does not believe that Clark can fly’, or we might on the other hand say ‘that Clark can fly’ is indeed something that ‘she stands in the belief relation to’. Each of those moves can be pushed to confront well-known problems – the first will eventually have to resist Kripke's powerful arguments about the semantics of names, the second will eventually give up articulating mental differences between Lois and, say, Superman himself in terms of the contents to which they do and don't stand in the belief relation to, respectively. Better, I think, to admit that any expression of her thought in an ordinary sentence is not well suited to making explicit her mental circumstance and then go on to admit even that the first admission is itself not quite well expressed (because it used the phrase ‘mental circumstance’). If you understand this, you understand polysemanticism.
Sense was recruited for one job, and then given another. We think with the sort of thing a sense is introduced to be; but we don't need language to have sense in order to understand how we think that way. It would help my pitch for polysemanticism to recommend, finally, a particular form in which it might be developed.
Return to Fodor, who, recall, insists that aboutness has to really be something else. I said above that I would offer two things for aboutness really to be. Our philosophical mystification in the face of the phenomenon of intentionality is I think in part the product of confusion about what the relation of representation could, ultimately, be. Unless we think that relation is primitive, unless we think one thing's being of or about another is not a condition whose obtaining is in virtue of any other condition's obtaining, then we should wonder just what (other) relation might subserve it.
There are in fact two relations, with each of which we are more or less familiar, that are, though quite different from each other, both plausibly helpful in understanding in virtue of what one thing might be about another.
Consider for example the instantiation relation that holds between collections of properties and the particulars that might instantiate those properties. If an individual has a bunch of properties, then anything that is somehow constituted by those properties is in one important way distinctively related to that thing (those constitutive properties select it, through its instantiating them, from all the things there are). From this perspective, the relation of instantiation can be understood to ground a variety of intentionality. By getting sphericity in mind, for example, a subject can in a way represent inter alia soccer balls. By being spherical, those objects can be the intentional objects of a thought with sphericity as an appropriate constituent part.
But there is also, differently, the phenomenon that Grice (Reference Grice1957) distinguished as natural meaning: smoke represents fire in virtue of a systematic causal correlation between those phenomena. In this case the representation relation is subserved through causal mechanisms. Properties of fires can stand in for them representationally; but causal correlates of fire can do that too, if in a recognizably different way.
The aboutness of language is, I think, to be made out in terms of causation, compatibly with the causal ‘picture’ that Kripke himself proposes. The aboutness of thoughts, by contrast, is to be made out in terms of satisfaction, in line with the model Frege offered (though he mistakenly offered it as accounting for the reference of words).
Returning then, finally, to Forster: we connect in more than one way. We can think of loved ones – remembering long-lost ancestors as easily as we anxiously await offspring – and we can consider whether worms have noses or whether to seek out a saucer of mud. We do that kind of thing by getting properties in mind that the relevant individuals instantiate.
But we also connect in a very different sort of way, with language, by naming things and referring to them and talking about them. We do that by taking advantage of causal relations that have been established between the words we use and the things they (thereby) stand for.
The possibilities for beautiful and loving relationships, the prospect of a kind of exaltation, and of salvation, and the preconditions for happiness, are all accordingly in place, in more than one way. We need not be alone in the nightmare scenario postulated by the solipsist: our thoughts provide for a satisfying connection, and our language provides for yet another. If you want to get about, understanding its multiplex character is the beginning of wisdom.
