Meaning

Interview by Richard Marshall

'To start with, it’s good to be semantically aware so as not to be taken in by equivocation and other fallacies in philosophical arguments. And of course there are philosophical questions about language, such as the nature of reference or meaning, that people have made claims about by means of analyses of the semantics of particular words and constructions, such as names and propositional attitude ascriptions.There is also arguably a connection between semantics and ontology, a tantalizing possibility that doing semantics might be able to deliver insights into ontology.'

'The Frege/Strawson approach improves on Russell, in my opinion, by having all the troublesome existential entailments that we met above tucked away in presuppositions instead of out in the open in at-issue content (roughly, asserted content). To be more precise, I follow Frege in supposing that a definite description ‘the F’, or someone who uses it, presupposes that there is exactly one F. (Strawson, as you will recall, has the presupposition be of existence only.) And I don’t mean, by the way, to be pledging allegiance to Frege’s general theory of presupposition.'

'The two traditional theories of definite descriptions, the Russellian and the Fregean, which I’ve just been describing, differ in the status they give to the existence and uniqueness entailments they associate with definite descriptions: Russell says that this is asserted content and Frege says it is presupposed. If you think about it, you’ll see that there is another logical possibility that keeps the traditional association between definite descriptions, existence, and uniqueness: existence and uniqueness could be both asserted and presupposed! This, roughly speaking, is what Hawthorne and Manley say...'

Paul Elbourne is a philosopher of language. Here he discusses the importance of semantics to philosophy, deference, ‘donkey anaphora’, Russell's theory of descriptions, some problems with it, the Frege/Strawson approach, the relationship between presupposition, assertion and definite description in this, demonstratives as Fregean definite articles, Soames and the nature of propositions, and the importance of Kripke.

3:16: What made you become a philosopher?

Paul Elbourne: It's perhaps not completely clear that I am a philosopher! As opposed, say, to a theoretical linguist who happens to write about things that philosophers are interested in. But I suppose I've also been teaching philosophy for the last seven years, so maybe I qualify by now.It was pure chance that I ended up doing what I do. I read Greats (Classics and Philosophy) at Oxford as an undergrad, mainly because I loved Greek and Latin literature, and in that degree, although I enjoyed the philosophy (which I was lucky enough to study with Christopher Taylor and Jennifer Hornsby), my real passion was Indo-European comparative philology—reconstructing the lost Proto-Indo-European proto-language from Latin, Greek, Sanskrit, and its other descendants. Fascinating stuff. I did an Oxford master’s degree in Indo-European too and even published a few articles on it based on my master’s thesis. But during that time I also developed a keen interest in general linguistics, on which I was taking courses with Jim Higginbotham. So afterwards I went to MIT to do the PhD programme in linguistics. 

My original plan was to specialize in phonology (the study of how the sounds of words are represented and manipulated in the mind) and apply it to the history of Greek or Latin, but in my first year I had to take an introductory course in semantics. I probably would not have taken that course if it had been optional—a good argument, I think, for not allowing students too much freedom in their course choices! Anyway, I found to my surprise that I loved semantics and switched over to that. I ended up working with a mixture of linguists and philosophers: Irene Heim, Kai von Fintel, Michael Glanzberg, and Danny Fox—and Noam Chomsky, although he was not on my committee. I wrote a PhD thesis on the semantics of singular terms, roughly speaking. I concentrated on pronouns but also included some stuff on definite descriptions and names. That became my first book, Situations and Individuals, which eventually attracted some interest from philosophers. After my PhD I worked in linguistics departments for thirteen years, but at some point the Oxford philosophy department was looking for a philosopher of language who knew about the interface with linguistics. By this stage I was citing approximately equal numbers of philosophers and linguists in my papers and was going to workshops with philosophers, so it felt like a natural move. I applied and was lucky enough to get the job, even though one might suspect that I was really a linguist who knew about the interface with philosophy.

3:16: You’re interested in natural language semantics—the meaning of words—and this is almost a stereotype of what some people think philosophers are  up to—you ask them a question about x and they invariably answer,  ‘Well, it depends what you mean by x’, and then we’re off to the races.  Why is semantics so important for philosophy—why can’t it be left to just linguistics and psycholinguistics? 

PE:  A few reasons, I think. To start with, it’s good to be semantically aware so as not to be taken in by equivocation and other fallacies in philosophical arguments. And of course there are philosophical questions about language, such as the nature of reference or meaning, that people have made claims about by means of analyses of the semantics of particular words and constructions, such as names and propositional attitude ascriptions.There is also arguably a connection between semantics and ontology, a tantalizing possibility that doing semantics might be able to deliver insights into ontology. It’s easy to overstate this, however. I personally like the way that Chris Kennedy and Jason Stanley put it at the start of an article of theirs in Mind in 2007, which goes something like this. 

Suppose that Davidson is right and that ‘Will slapped Chris’ means ‘There was an event e such that e was a slapping and Will was the slapper in e and Chris was the slappee in e‘; then it’s tempting to argue that, since ‘Will slapped Chris’ is true, then there are such things as events. For how could the paraphrase of the meaning that I just gave possibly be true if there were no events? Now of course we can’t argue quite so straightforwardly. But we can arguably say that either we have just shown there are events or Davidson’s semantics for action sentences is false or ‘Will slapped Chris’ isn’t true after all. The idea is, then, that successful semantic theories can nudge us towards particular metaphysical conclusions: given that Will did indeed slap Chris, we are now under pressure to admit that there are such things as events, since Davidson’s semantics for action sentences is very elegant and explanatory.I myself am very dubious about this kind of claim, since I think that the naive deliverances of the relevant cognitive systems cannot be trusted. You can take sentences that have the most outrageous metaphysical implications and find that people will judge them to be true. Chomsky has some great examples here. The following is one of his:

The bank moved across the road after burning down.

Under the right circumstances, people will judge this to be true. But there is no actual object in the world that could both burn down and (subsequently!) move across a road. Or so Chomsky urges. If he’s right, we seem to be forced to admit that the relation between metaphysics and judgements of truth is more complicated than we might like. Perhaps this sentence can convey a true proposition by expressing a proposition that is somehow related but which contains metaphysical presuppositions which are false. So the possibility opens up that other cases where we thought we could move from meaning to metaphysics are similarly screwy. Not everyone shares my scepticism, however. In addition to Kennedy and Stanley, my Oxford colleagues Matthew Gotham (Linguistics) and Ofra Magidor (Philosophy) and others have written articles arguing that sentences like the one above can directly express propositions that are true and which do not involve any impossible metaphysics. 

3:16: There’s a theory of language that says we don’t need to know the meaning of a word to use it—we defer to experts—and another says we kind of negotiate meanings as we go along. What do you say?

PE: I don’t think that either of those theories will get us very far. We hardly ever have the sensation of negotiating meanings as we go along. Perhaps with newly coined words we do, if we are very self-conscious, but I rather doubt we do even there. And cases where we defer to experts are likewise pretty rare. If, like Putnam, I don’t know enough about trees to be able to tell apart elms and beeches (which is true, alas), then I might use ‘elm’ and ‘beech’ in the knowledge that there is a body of experts who will fix the denotation for me in a conventional manner if ask them. Sure. But I don’t think this extends to the vast majority of words: not to ordinary words like ‘table’ or ‘walk’ or ‘lovingly’ or ‘beside’ or ‘if’ or ‘that’ or ‘the’… These last three are words that philosophers are particularly interested in, of course. So if anyone is an expert on the meanings of these words, philosophers are. 

But philosophers notoriously disagree about all these cases! Suppose, very plausibly, that David Kaplan and Jeff King are experts on the word ‘that’. The problem is that they will tell you very different things about the meaning of the word ‘that’. (Or at least the Kaplan of ‘Demonstratives’ and the King of Complex Demonstratives would—I’m not sure what these guys believe now.)  So good luck to you if you want to use the word ‘that’ and are relying on the experts to provide a meaning for it! But of course no-one would ever think of consulting experts in cases like this, or even relying on the idea that experts could step in if necessary. What happens is that we just charge ahead and speak (or write) without thinking about the meaning of our words at all, almost all the time. Indeed it’s pretty clear that we don’t have any conscious knowledge of the meanings of our words, beyond vague notions along the lines of ‘dog’ having something to do with dogs: if we did have conscious knowledge of our word meanings, Kaplan and King would be able to settle their dispute about the meaning of the word ‘that’ simply by introspecting. But they can’t. They have to engage in long, arduous study, framing hypotheses and looking for data to confirm or disconfirm them. 

I am not convinced, then, by the idea that we can defer to experts regarding the meanings of our words.To cut a long story short, I think that the only theory in this area that has a chance of standing up over the long term is some variety of internalist, nativist theory like Chomsky’s: we come equipped with a biological endowment that predisposes us to learn languages of certain kinds, and this crucially includes innate guidelines that help us acquire word meanings, which we do in early childhood (not for all the words we know, but for thousands of them, including ‘that’ and ‘the’ and so on), with no detectable instruction or negotiation involved, whether with experts or anyone else. (Parents sometimes think they are helping their children learn to talk when they point at, say, a dog and say, ‘Ooh, look, it’s a dog. A DOG!’ But this kind of thing isn’t necessary. Did any parent ever say to their toddler, ‘The dog… now I can say that because there is exactly one member of that species that is salient to us at the present moment…’? I doubt even Russell did that. But young children learn to use definite descriptions just the same.) 

It’s still deeply mysterious exactly how all this happens, of course, and I would not want to go as far as Jerry Fodor, who notoriously claimed that all word meanings are innate, including the meaning of words like ‘carburettor’. I hope that a nativist position can be carved out that is not as extreme as that one, presumably involving principles or components of meaning rather than whole, ready-to-use, off-the-shelf meanings (for words like ‘carburettor’, at least).By the way, you can tell from my last two answers that the MIT Linguistics and Philosophy department is uncompromising in its teachings! Just as everyone who goes through the philosophy programme there comes out construing propositions as sets of possible worlds, in the manner of Stalnaker, so everyone who goes through the linguistics programme comes out a nativist internalist, in the manner of Chomsky… (I jest, but there is truth in my jest.)

3:16: One phenomenon that kind of leads us into issues you discuss in depth is ‘donkey anaphora’! So to start with can you tell us what this is and what it has to do with understanding the syntax and semantics of proper names and definite descriptions?

PE: Donkey anaphora is a kind of binding relation in which the quantifier phrase apparently doing the binding does not have scope over the item that is bound. To be precise, it is a phenomenon in which a singular term seems to covary with an indefinite noun phrase that does not have scope over it. Here’s a classic example, the one that the phenomenon was named after, from Peter Geach’s Reference and Generality, based on examples from the medieval logicians:

Any man who owns a donkey beats it.

(You can tell it’s a classic example because it involves beating things…) Intuitively, the pronoun ‘it’ seems to depend for its interpretation on the indefinite ‘a donkey’. But ‘a donkey’, as it appears there, does not have scope over ‘it’: it is deeply embedded in the relative clause and only has scope over ‘owns’. So the kind of binding relation that we’re used to from logic, whereby a quantifier has to have scope over a variable in order to bind it, does not seem to do the job here and there is naturally a question about what exactly is doing the job. It doesn’t work, by the way, if you try to interpret ‘a donkey’ as having wide scope in that sentence by force, as it were: then you get a reading, ‘There is a donkey such that any man who owns it beats it’, which is obviously not the reading we’re after. Broadly speaking, there are two main schools when it comes to analysing this. The first school, which was originated by the British linguist Robin Cooper in the late 1970s, says that we can get a handle on things by interpreting the pronoun in the above sentence as a definite description. For example, if ‘it’ in that sentence means ‘the donkey he owns’, then we seem to predict the following reading:

Any man who owns a donkey beats the donkey he owns.

And that seems to be not half bad, although there are questions about what happens with men who own more than one donkey. Stephen Neale and I have endorsed this idea, each spelling it out with our own favourite theory of descriptions. This school messes with pronouns, you might say, since we can no longer think of them just as variables, but leaves binding alone. The second school, which goes back to work by Irene Heim and Hans Kamp in the early 1980s, does the converse: it leaves pronouns as simple variables (more or less) but says that quantifiers don’t have to have scope over things in order to bind them. There has been a lot of work following up on this idea since then, largely under the rubric of ‘dynamic semantics’, by logicians like Jeroen Groenendijk and Martin Stokhof and many others. The details from this point on get very technical very quickly and I don’t think this is the best forum to go into them. I certainly would not be able to handle them myself at 3:16am! I think I’ve said enough already about how definite descriptions fit in, but you also ask about donkey anaphora and proper names, which is very interesting, I think. In Situations and Individuals I give an example of a donkey-anaphoric name. (That’s my story, anyway, and I’m sticking to it!)

Every woman who has a husband called ‘John’ and a lover called ‘Gerontius’ takes only Gerontius to the Rare Names Convention.

The second occurrence of ‘Gerontius’ stands in the same structural relation to ‘a lover called ‘‘Gerontius’’’ that ‘it’ in the previous example stands in to ‘a donkey’. And it seems to have a covarying interpretation, like a bound variable. If this is right (and it is!) then the theory that names are just devices of direct reference has encountered a serious problem and we seem to be pushed towards the view whereby they are definite descriptions, since definite descriptions can have donkey-anaphoric readings, as we just saw. This kind of example has not been much discussed, although I should note that Hawthorne and Manley say some sceptical things about similar examples in The Reference Book

3:16: You defend what you’ve labeled the Fregean/Strawsonian theory against its rivals. Its main rival is Russell’s, isn’t it, so first could you explain what Russell’s theory claims so we can see what you’re going up against?

PE: Sure. Russell’s theory of definite descriptions claims that a sentence of the form ‘The F is G’ means ‘There is exactly one F and any F is G’. There are various equivalent paraphrases, including a well-known one in logic. 

3:16: Why do sentences containing definite descriptions embedded under propositional attitude verbs and conditionals cause problems for this  approach?

PE: I should start by saying that I’m mainly a humble popularizer when it comes to this question, since the basic arguments here go back to my PhD supervisor Irene Heim, although I did make one or two changes along the way. Here’s an example with a propositional attitude verb:

Hans wishes that the ghost in his attic will be quiet tonight.

There is no reason that we know of why the definite description here should not be able to take scope under the attitude verb. This means that the Russellian theory seems to predict that this sentence will have a reading (perhaps among others) equivalent to the following:

Hans wishes that there will be exactly one ghost in his attic and that any such ghost will be quiet tonight.

But the original sentence absolutely cannot mean this! Among other things, this paraphrase seems to predict that Hans wants there to be  a ghost in his attic. But the original sentence does not imply this at all. Hans seems to be assuming the existence of a ghost in his attic and wishing only that it be quiet. Note that you can’t rescue Russell’s theory by just stipulating that the definite description has to have wide scope. Then you get the following reading:

There is exactly one ghost in Hans’s attic and any such ghost is such that Hans wishes it would be quiet tonight.

And that commits the speaker to the existence of ghosts, whereas it would be possible to say the original sentence without believing in ghosts.Perhaps the best way to bring out the problem with conditionals is with a pair of sequences of sentences, as I do in an article in Linguistics and Philosophy several years ago. Here’s the first sequence:

I do not know whether there are any ghosts in Hans’s attic. But if the ghost in his attic is quiet tonight, he will hold a party.

And here’s the second:

I do not know whether there are any ghosts in Hans’s attic. But if there is exactly one ghost in his attic and any such ghost is quiet tonight, he will hold a party.

It’s clear intuitively that the speaker of the first sequence is being inconsistent somehow whereas the speaker of the second sequence is not. But the two sequences are identical except that the definite description in the second sentence of the first sequence is spelled out with a Russellian paraphrase in the second sequence. So Russell’s theory is not giving us the correct paraphrase.

The basic idea, as you can see, is that Russell predicts that assertoric existential quantificational force will show up in places where we do not want it. In fairness, I should mention that Wojciech Rostworowski has a very good paper in Phil. Studies, in 2018, that throws doubt on the argument from propositional attitude ascriptions by means of some reasoning that I will not try to reproduce now. But even Rostworowski admits that the conditional cases are serious problems for Russell.

3:16: And do the problems remain even when the operator in question allows the context to be extensional?

PE: Yes, indeed! I was looking for a case that avoids the intensional and hyperintensional complications of conditionals and propositional attitude ascriptions, in case anyone attempted to rescue the Russellian theory by claiming that the facts I just mentioned had more to do with the weird semantics of conditionals and propositional attitude verbs than with the semantics of definite descriptions themselves. Here’s what I came up with (in an article in Phil. Studies a few years ago):

No boy sold the dog he had bought.

If we use the paraphrase that I gave before, it would appear that the Russellian theory predicts this to mean the following (with ‘he’ bound by ‘no boy’ the whole time):

No boy is such that there is exactly one dog that he bought and any dog that he bought is such that he sold it.

And if you think about that paraphrase by itself, forgetting the original sentence for a moment, you will quickly see that it is equivalent to the following:

No boy bought exactly one dog and sold it.

But now when we compare this final paraphrase with the original sentence, we see that they are not intuitively equivalent. Take the case in which no boy bought any dog. The paraphrase seems to be straightforwardly true in that case but the original sentence most definitely does not. Anyone interested in this argument should also read Frank Pupa’s paper in Synthese last year, which spends some time responding to it.

3:16: So how does the Frege/Strawson approach improve on Russell? Can it cope with all the Russellian problems?

PE: The Frege/Strawson approach improves on Russell, in my opinion, by having all the troublesome existential entailments that we met above tucked away in presuppositions instead of out in the open in at-issue content (roughly, asserted content). To be more precise, I follow Frege in supposing that a definite description ‘the F’, or someone who uses it, presupposes that there is exactly one F. (Strawson, as you will recall, has the presupposition be of existence only.) And I don’t mean, by the way, to be pledging allegiance to Frege’s general theory of presupposition. My view is that, whatever kind of thing presuppositions are, definite descriptions or people who use them introduce one of those things, to the effect that there is exactly one entity that satisfies the descriptive content.Let’s take the conditional examples. There the datum was that the speaker of this sequence sounds inconsistent…

I do not know whether there are any ghosts in Hans’s attic. But if the ghost in his attic is quiet tonight, he will hold a party.

while the speaker of this sequence does not:

I do not know whether there are any ghosts in Hans’s attic. But if there is exactly one ghost in his attic and any such ghost is quiet tonight, he will hold a party.

The second sequence models what Russell predicts about the first. To work out what the Fregean theory predicts, we observe, first, that the Fregean prediction is that the definite description in the antecedent of the conditional will introduce a presupposition to the effect that there is exactly one ghost in Hans’s attic. This sounds promising, but we have to check something, namely what happens to presuppositions introduced in the antecedents of conditionals. There is an issue here, because some presuppositions in some positions are not felt at the level of the whole sentence (they don’t project to that level, to use the jargon), because of factors that are still the subject of debate amongst linguists and philosophers. But fortunately this example doesn’t seem to be like that. Let’s compare another example:

If John stopped smoking, Mary will be happy.

The construction ‘stop doing so-and-so’ introduces a presupposition to the effect that the subject of the sentence used to do so-and-so. So ‘John stopped smoking’, in the antecedent, introduces a presupposition to the effect that John used to smoke. And intuitively this presupposition is felt at the level of the whole sentence: if the utterance of this sentence is to be felicitous, it must be accepted by speaker and hearer that John used to smoke. The Fregean theory makes the analogous prediction about the first sequence of sentences above, then: it predicts that anyone who says the second sentence in that sequence presupposes that there is exactly one ghost in Hans’s attic. We can now explain why such a person sounds inconsistent in saying the whole sequence: in the first sentence, they say they do not know whether p; and then in the second sentence they presuppose something that obviously entails p. So the Fregean theory seems to get the facts right here.

It would take too long now to go through the predictions about presuppositions in the other cases (the propositional attitudes case and the ‘no boy’ case), but I’ve argued that the Fregean theory makes the correct predictions about those too, for broadly similar reasons.As for whether the Fregean theory deals with all the problem cases raised by Russell, well, yes, I think it does, but it would take a long time to show that. In fact I have a whole book, Definite Descriptions, in which I try to show that! But perhaps I should say a few words about non-existent objects, which is where there is perhaps the most obvious payoff for other areas of philosophy. One of the motivations mentioned by Russell for his theory of descriptions is that by analysing definite descriptions as being non-referential we are relieved of the embarrassing responsibility to find objects for phrases like ‘the fountain of youth’ to refer to. This is in contrast to Alexius Meinong, who actually did posit objects for phrases like that to refer to. So Russell would analyse

The fountain of youth does not exist,

for example, as involving wide scope for negation and his distinctive paraphrase for the definite description. So it is predicted to mean something like this:

It is not the case that there is exactly one fountain of youth and any fountain of youth exists.

This paraphrase has the conspicuous virtues of being true, like the original sentence, and of not involving reference to any putative fountain of youth. Things are slightly trickier, perhaps, for the Fregean theory here, since it does actually make definite descriptions be the kind of things that can be used to refer. The trick, I claim, is that in sentences like this the description picks things out in non-actual possible worlds. The idea is that in ‘The fountain of youth does not exist’ we slip into talking about a fiction in the terms of the fiction, just as we do in sentences like this:

Sherlock Holmes is a detective.

If you present this last sentence to most people, they will say that it is true. One idea here, a simple version of an analysis presented by David Lewis, is sentences that adopt the fictional point of view involve quantification over relevant possible worlds, so that the logical form of ‘Sherlock Holmes is a detective’ would be the following:

All possible worlds w that are compatible with the Sherlock Holmes stories (as they are in the actual world) are such that Sherlock Holmes is a detective in w.

Similarly then, with only a slight change in the overall logical form, I would analyse ‘The fountain of youth does not exist’ as follows:

All possible worlds w that are compatible with the fountain-of-youth myth (as it is in the actual world) are such that the fountain of youth in w does not exist in the actual world.

This has ‘the fountain of youth’ pick out (or range over) individuals, but only in non-actual worlds that contain fountains of youth, so things don’t look so bad.Now some people don’t like possible worlds, of course, although I would argue that they or something like them are essential to natural language semantics as we find it and so we just have to live with them. Things might seem to get worse when we come to deal with examples like this:

The highest prime number does not exist.

I would deal with this example in exactly the same way as I do the last one, except that here, of course, I will have to use impossible worlds. This will be unwelcome to larger numbers of people, but I would maintain that there are no fatal objections here. All you have to do to avoid a fatal contradiction is refrain from being a modal realist. If you favour ersatz worlds or, like me, see talk of worlds in this context as the employment of a formal device to model certain cognitive capacities and mental objects (in good internalist fashion), I don’t think you should sweat very much over impossible worlds. 

3:16: What’s the relationship between presupposition, assertion and definite description in this and why do you think Hawthorne and Manley get this wrong?

PE: The two traditional theories of definite descriptions, the Russellian and the Fregean, which I’ve just been describing, differ in the status they give to the existence and uniqueness entailments they associate with definite descriptions: Russell says that this is asserted content and Frege says it is presupposed. If you think about it, you’ll see that there is another logical possibility that keeps the traditional association between definite descriptions, existence, and uniqueness: existence and uniqueness could be both asserted and presupposed! This, roughly speaking, is what Hawthorne and Manley say in The Reference Book. A little more precisely, they say that a sentence of the form ‘The F is G’ has assertoric content ‘There is an individual that is F and G’ (just like ‘An F is G’, then) and a presupposition ‘There is exactly one F’. (I am simplifying slightly here by ignoring some stuff they put in about how incomplete definite descriptions are handled, that is cases of ‘The F is G’ where there is more than one F.) More or less exactly the same theory, by the way, was proposed by the linguist Barbara Abbott in 2008 and something very close to it was advocated back in 1979 by Lauri Karttunen and Stanley Peters.

Part of the attraction of these theories is that they have in common with Russell’s theory the ability to deal with the fountain of youth and the highest prime number without making essential use of possible or impossible worlds, because they keep definite descriptions quantificational, not referential. And at the same time they can arguably deal with the ghost in the attic and the boys buying dogs because they include relevant presuppositional content. You can see that this is a very attractive position. It sounds like the best of both worlds.I have doubts about these theories, however. I argued in a paper in Linguistics and Philosophy last year that we should reject them on the basis of some pretty simple data. Suppose we are out in the countryside and looking into a field that contains exactly one cow, which is plainly visible. We haven’t spoken; we’re just contemplating the scene and have been doing so for several seconds. Then it’s fine for me to say the following to you, especially if you don’t know much about the breeds of cows:

The cow in that field is a Friesian.

But it would be very weird for me to say the following:

There is an individual that is a cow in that field and a Friesian.

And it would not just be stylistically strange. You can rephrase it any way you like. The point is that if I were to say anything with these truth and falsity conditions I would be perceived to be stating the obvious, since, by stipulation, we have been contemplating the cow in its field for several seconds. But no such feeling of stating the obvious attaches to the first sentence. Now of course the theories we are now thinking about don’t claim that assertoric truth conditions like the ones of the last sentence are all that is on the table when we use the first sentence. They also claim that there is a presupposition to the effect that there is exactly one cow in the field in question. But how could that help? As far as I can see, the last sentence and paraphrases of it are infelicitous because they involve asserting that there is a cow in the field (or asserting something that obviously entails that there is a cow in the field) when this is already perfectly obvious to speaker and hearer. If there is also supposed to be a presupposition to the effect that there is exactly one cow in the field, this might if anything seem to make matters even worse. It’s obvious that there’s a cow in the field, I’m presupposing that there is, but still I take the trouble to assert that there is! I definitely can’t think of any way that the presupposition might cancel out the infelicitous assertion. So for myself I still favour the Fregean theory, although I think that Hawthorne and Manley’s theory is an improvement over Russell’s.By the way, I don’t know of anyone who has published on definite descriptions in the last decade or so who supports the Russellian theory. The nearest to come to it is perhaps Frank Pupa, but even he adds a familiarity condition to the original Russellian semantics.

3:16: You argue that demonstratives like ‘this’ and ‘that’ are Fregean definite articles? Who are you disagreeing with and why do they not agree with you?

PE: Yes, I’ve argued that they are Fregean definite articles with some extra bells and whistles. The most prominent work in this area is probably still that of David Kaplan in his classic paper ‘Demonstratives’ and the follow-up ‘Afterthoughts’. He said that demonstratives are directly referential, which is to say that they contribute a referent to the truth conditions and nothing else. He had data like the following, which show that there are differences in how demonstratives and ordinary definite descriptions behave:

[Charles is from Charleston. Paul is from St. Paul. The speaker points at Paul, who is sitting in front of her.] If Charles and Paul had changed chairs, then the/this man being pointed at would be from Charleston.

If the speaker says ‘the man being pointed at’, her utterance is true. If she says ‘this man being pointed at’, her utterance is false. This kind of minimal pair, for a long time, put people off the idea that demonstratives could be definite descriptions.Lots of other people have written on demonstratives, of course. For clarity and breadth of coverage and for being a pioneering break with Kaplan, I would particularly point to Jeff King’s book Complex Demonstratives, which says that demonstratives are something like Russellian definite articles with some extra bells and whistles: they involve existential quantification, for example. You see that some familiar battle lines have opened up once again…

3:16: How do you respond to them?

PE: My view of demonstratives is based closely on that of the late Geoff Nunberg, who pointed out a number of examples of what are now called descriptive indexicals. (He applied his theory to indexicals as well as demonstratives.) An example with a demonstrative is as follows:

[With a gesture at Pope Francis.] That Church official is usually an Italian.

We do not understand this, of course, as meaning that Francis is usually an Italian. That would be absurd. We understand it as meaning that the Pope is usually an Italian. (In other words, it boils down to claiming that most Popes are Italian.) This is already taking us away from direct reference and towards definite descriptions, as you can see. The theory says that use of a demonstrative involves an index, which is some object demonstrated, and a relational component, which is some relation that applies to the index. In the above example the index is Pope Francis and the relational component might be something like ‘has the same job as’.  The two of them combine to give us a property ‘has the same job as Francis’. Then, in my version, ‘that’ comes along and takes both this property and the property of being a Church official as arguments and gives us back the individual that has both those properties. (And when I say ‘the individual’ I assume that we have a definite description meaning here.) The whole sentence claims that the individual who has both those properties is usually an Italian—in other words, that the Pope is usually an Italian. This kind of example does not fit comfortably into a framework in which demonstratives are directly referential.The same theory can be used to explain other data that take us far from the realms of direct reference. Take bound demonstratives, for example. It’s perfectly fine to say this:

Mary talked to no senator without thinking afterwards that that senator would cosponsor her bill.

We could analyse this by saying that the index, in Nunberg’s terms, is quite generally contributed to the semantics by an individual variable associated with the demonstrative and that the relational component is contributed by a free relation variable. We’ve seen what values they would get in the ‘Church official’ example. In the current example, we might say that the individual variable is bound by ‘no senator’ and the relation variable has the value ‘being identical to’ (which I think is built in as a possibility, in fact probably as a default). So ‘that senator’ means ‘the unique x such that x is a senator and x is identical to y’, where the ‘y’ is bound. (In my actual paper on this I am forced to say something more complicated at this point to account for some other data, but what I just said gives the basic idea.)

Kaplan, by the way, acknowledged the existence of examples in which things that looked like demonstratives appeared to be bound, but said that these might be homonyms of the words that he was analysing. But I think this is very unlikely. For a start, demonstratives seem to be able to be bound as well as referential in language after language worldwide. To say that this is all homonymy is to posit an awfully big coincidence.  Secondly, if we admit that there are these homonyms in English of the words he is interested in, he has not really finished the job when he tells us that the words he is interested in are directly referential; he also really owes us an account of the semantics of the homonyms and an explanation of why they cannot appear in his examples and give rise to non-directly-referential readings. They would look and sound just the same as the words he claims are there, after all!The mechanisms that I mentioned above in connection with bound demonstratives are also used to explain seemingly straightforward referential examples. Take the following:

That cat is grinning.

Imagine this being said with a gesture towards the cat in question—call it ‘Felix’. Then according to my theory the individual variable contributes Felix as the index and the relation variable contributes its default value of ‘being identical to’. So the complex demonstrative as a whole contributes the unique individual that is a cat and which is identical to Felix—Felix, in other words. There is much more to be said here, including some kind of story about when we get directly referential readings and when we get descriptive indexical readings, but let me now turn to King’s theory. King in his book claims that a sentence of the form ‘That F is G’ will have a logical form like the following:

F and O* are uniquely jointly instantiated (in w at t) in an object x and x is G.

The part in parentheses is optional and the property O* is determined by the intentions of the speaker and can sometimes just be a repetition of F. For ‘That animal is a donkey’, for example, as spoken by someone in w at t with a particular animal a in mind, the logical form would be the following:

animal and =a are uniquely jointly instantiated in w at t in an object x and x is a donkey.

This looks encouraging, but now note that the following sentence…

If that ghost in my attic starts to make scary moaning noises, my boring guests will leave.

… is now predicted to have the following reading, perhaps among others:

If ghost in my attic and ghost in my attic are uniquely jointed instantiated in an object x and x starts to make scary moaning noises, my boring guests will leave.

This is equivalent to the following:

If there is exactly one ghost in my attic and any such ghost starts to make scary moaning noises, my boring guests will leave.

We can now see, I think, that King’s theory makes demonstratives into Russellian definite descriptions, as I alleged above. And with that move comes the kind of inaccurate prediction that plagues the Russellian theory of descriptions. In particular, it is clear, I think, that the current conditional example does not actually have the reading just given for it.

3:16: Why is Soames wrong to think that propositions can’t be sets of truth-supporting circumstances – and what’s at stake in this dispute?

PE: What is at stake is, of course, the nature of propositions, which is a topic that should be of interest to a wide range of philosophers, since these are often taken to be the things that we believe and towards which we have other propositional attitudes. Soames and I share the assumption that propositions are the semantic values of declarative sentences, whether these occur in isolation or embedded under other operators; this means that there is a distinctively linguistic way into this question. My work on this is basically a reply to an article that Soames published in the Journal of Philosophical Logic in 2008; my reply came out in the same journal a couple of years later. Soames had in his sights the position that propositions are sets of truth-supporting circumstances, whether sets of possible worlds or sets of situations, where situations can be thought of as parts of possible worlds. 

His reasoning is slightly complex, and I don’t think I’ll try to go through it in detail now, but he basically assembled some assumptions that he claimed were plausible and put them together with the assumption that propositions are sets of truth-supporting circumstances and proved an absurdity from them, the idea being that the assumption about sets of circumstances therefore had to go. Now one of the assumptions that he put in his allegedly uncontroversial group was the assumption that names, demonstratives, and pronouns were directly referential or at least had uses in which they could reliably be assumed to be directly referential. After hearing about the work of mine that I have already described, you can perhaps imagine how I felt about that assumption! My article was basically a quick canter through lots of the tricky cases that I’ve described in this interview, where singular terms behave as if they were definite descriptions. The upshot, as far as I’m concerned, is that we are still free to believe that propositions are (or can be formally represented as) sets of possible situations. 

3:16: I’m writing these questions in the week that sadly Saul Kripke passed. How important was he for your work and for the field of philosophy you work in? 

PE: Saul Kripke was a titan in the field who revolutionized the study of proper names, reference, and meaning. We would not have anything like the insights into those areas that we have today if it had not been not for him. And I say this as someone who has spent a considerable amount of time arguing that we should turn the clock back, as it were, and return to some kind of descriptivism, as advocated long ago by Frege and Russell, and which Kripke famously argued against.3

:16: And finally, for the readers here at 3:16am, are there five books you could recommend that will take us further into your philosophical world?

PE: 

For complete newcomers to the field, allow me to be immodest and recommend my own Meaning: A Slim Guide to Semantics. In this book I try to give an entertaining introduction to the analysis of meaning that draws on philosophy, linguistics, and psychology, including neuroscience. I see the study of meaning as nestling at the intersection of those three fields; to pursue it, one should be open to insights from all three. 

                              

As a more advanced introduction, I would recommend Semantics in Generative Grammar, by Irene Heim and Angelika Kratzer, which introduced a number of very influential innovations in the study of formal semantics, while still being a lucid and stimulating textbook. 

I would recommend Stephen Neale’s Descriptions, still the best introduction to definite descriptions and their philosophical significance, and with some great material about pronouns and donkey anaphora too. 

New Horizons in the Study of Language and Mind is a very useful collection of essays by Chomsky setting out his internalist and nativist take on language. 

And finally for some fresh and bold theorizing on the kinds of things I’ve been talking about, readers should consult The Reference Book by John Hawthorne and David Manley. (I say ‘fresh’ but I now see it’s ten years old already! Doesn’t time fly?)

ABOUT THE INTERVIEWER

Richard Marshall is biding his time.

Buy his second book here or his first book here to keep him biding!

End Time series: the themes

Huw Price's Flickering Shadows series.

Steven DeLay's Finding meaning series

Josef Mitterer's The Beyond of Philosophy serialised

NEW: Art from 3:16am Exhibition  - details here