:: Article


Interview by Richard Marshall.

[Photo: Laura-May Abron]

Eliot Michaelson works primarily on core topics in philosophy of language, like meaning and reference. He also works on the border of philosophy of language and ethics, on issues pertaining to lying and insincerity. He maintains independent research interests in the nature of skills, the metaphysics of colour, and the ethics of eating. Here he discusses how words refer, Kripke, descriptivism vs causal theory, Frege, Barcan Marcus, problems descriptivists face, why Kripke et al fare better, hybrids and Evans, how indexicals refer, do we have a unified theory of reference if we’re not Fregeian, why not every meaningful expression refers, reference in the Matrix, psychedelic goblins, reference and metaphysics, Quine and Davidson on reference, the answering machine paradox, bullshit in politics, and whether we should eat fish. This is a good long trail somewhere. Winters coming. Take your time…

3:AM: What made you become a philosopher?

Eliot Michaelson: Seemed like a good idea at the time? In all seriousness though, I think most of it really was just chance. I started off in college with a desire to make my life a useful one and a persistent sense that I really had no idea how to do that. So basically, I had no idea what I wanted to study. I took courses all over the map: math, physics, history, anthropology, photography, creative writing. I learned a lot from those, but philosophy was where I felt like we were addressing the sorts of concerns I was only just starting to be able to articulate—concerns about how to live in a big, complex, messy world where one has only very limited control over the environment one finds oneself in. Only later did it really dawn on me that my first couple of courses had all been in political philosophy! Oh well.

After that first flicker of interest, I think it was mostly just a matter of having had some really awesome philosophy teachers. I went to two colleges as an undergraduate. The first, Deep Springs College, is a bit unconventional. For one thing, it’s on a cattle ranch in the middle of nowhere; for another, the students provide much of the labor force for the farming and ranching operations. So my time there was largely spent falling off horses and practicing the dying art of staring meaningfully into the distance. More importantly, I failed an awful lot and had to get used to dealing with that—something which has proven oddly helpful in the pursuit of philosophy. Anyway, at Deep Springs, I was lucky enough to have been taught by Elizabeth Kiss, Jeff Holzgrefe, Mark Greenberg, Kinch Hoekstra, and David Arndt. Most of those folks were just passing through for a term or two, but all of them went above and beyond as teachers when they were out there. I’m deeply thankful to all of them for that. After Deep Springs and a year spent volunteering for AmeriCorps, I transferred to the University of Chicago, which prides itself on being a little different but to me seemed pretty conventional by comparison. It was also hugely exciting, intellectually. At Chicago, it was Arnold Davidson, Michael Kremer, Michael Forster, Jim Conant, John Haugeland, and Josef Stern who really pushed me deeper into philosophy. I should flag my particular debts to Josef and Arnold: Josef for introducing me to a great number of the topics I now work on, and Arnold for teaching me that occasionally, though only occasionally, doing good philosophy requires trying to convince yourself that you really don’t give a fuck about what anyone else thinks of what you’re doing.

Even by the end of college though, I had a lot of misgivings about doing this as a career. I still do, to be honest. I wondered, and still wonder, whether this was a particularly useful thing to be doing with my time. I’d like to think that it is, but when I look at some of my friends who are, say, reporters, or who work for various interesting non-profits, and I see what they’ve managed to accomplish already in their careers, it makes me wonder. On the other hand, the further we get into our respective careers, the more I realize that those same friends all have some very particular personal qualities that make them good at their jobs—a degree of easy sociality, a basic capacity to work well as part of a team, the ability to occasionally keep an opinion to themselves—that I very often lack. So then I think that maybe I found myself in the right career after all, even if I didn’t get there for quite the right reasons. I should probably also mention that I was lucky enough to have some graduate supervisors who were very sympathetic to all these worries, even if they didn’t have any real answers for them. That, and the eventual realization that worries like these are actually pretty widespread in philosophy, definitely helped to keep me in the field once I was headed down that path. I’m not sure why that helps, but it does somehow.

[All art: Bob Dylan]

3:AM: You’ve written about the philosophical importance of linguistic reference. I guess a key question is what is the mechanism by which words can refer. Two of the big competing theories are the descriptive theory and the causal theory. The first is associated with Russell and Frege and the other by JS Mill and Saul Kripke. Just so we get the geography, can you sketch what these two approaches claim?

EM: Okay, so just one brief clarification at the outset: the descriptivist and causal theory that you mention are both theories of how names work in particular. The implicit assumption that theories about how names refer largely exhausts the landscape of how words refer can seem pretty natural nowadays—particularly if, like me, one has been raised to think that Kripke’s Naming and Necessity is sort of like the Bible, only better—but I don’t think it should seem natural at all. No theory of names is going to come close to being the full story of how words refer, and there are going to be interesting questions about whether particular theories of reference for any particular class of terms can be subsumed under a more general theory of linguistic reference, or whether the field is bound to be highly disunified. I worry that, the way that these debates have played out, we’ve often lost sight of those broader questions. But hopefully we can come back to that later.

So okay, let me say something about descriptivism and the causal theory. The basic claim of descriptivism is, unsurprisingly, that names refer by describing things in the world. Let’s stick to Frege’s version of descriptivism, in particular, which I take to be the richest and most interesting version. For Frege, names have two different sorts of semantic values, or what we might roughly call ‘meanings’. On the one hand, we have the reference of a name, the object that it picks out. Well and good. On the other hand, we have what Frege called a ‘sense’. This was something like a way of thinking about an object. That could be a way of describing the relevant object, or a way of perceiving it. Two important things to note here: first, these ways were supposed to be at least potentially share-able. So if some way of picking out an object in thought were to turn out to be ineliminably idiosyncratic, it wouldn’t have counted as a sense for Frege. Second, we often talk about senses as if they were definite descriptions. That might be a fine way of modeling them for certain purposes, but it’s important to bear in mind that these notions weren’t equivalent for Frege. We can also pick things out visually, for instance, and this was even one of the central examples Frege used to illustrate the notion.

For Frege, these types of meaning were related in the following way: sense determines reference. So the sense of a name is a bit like a recipe, or a set of instructions, for figuring out what a name refers to. Supposing, that is, that one has access to sufficient information about the world.

The really crucial question, to my mind at least, is: what determines sense? That is, in virtue of what does a particular use of a name ‘express’ (in Frege’s terms) a particular sense? Frege never really answers this question explicitly, but he does tell us that different speakers, with different information about the world or simply in different mental states, can express different senses with uses of the same name. This would seem to imply that senses are speaker-sensitive, in the sense that it is something about the speaker—probably her mental state—that determines which sense is expressed. But again, I don’t think that Frege was ever fully explicit about this.

On the other side of this debate, we find the so-called ‘causal theory’. Roughly, the theory is made up of two main theses. The first thesis amounts to the denial that names describe in any sense whatsoever. The second thesis holds that names refer to what they do in virtue of standing in a certain sort of historical relation to an object. Let me expand on each of these claims in turn.

The first thesis is often called ‘Millianism’, since J.S. Mill gives a particularly clear exposition of it way back when. The idea is that names have just one type of meaning, or semantic value: their referents. I find the most intuitive way into this to be the thought—one which Kripke stole from Ruth Barcan Marcus—that names function something like ‘tags’. You put a tag on something not to describe that thing, but to help you find it again. Don’t think about price tags here or you’ll get the wrong idea. Think about the labels you might have on the fuse box in your home. One of them might say ‘kitchen’ even though it actually controls the dining room. This would be misleading to an outsider, but once you’re in the know it doesn’t really matter: if the power goes out in the dining room, you know which switch to flip. ‘Kitchen’, the thought goes, is still serving as a tag for the dining room here even if it doesn’t accurately describe the dining room. Alternatively, you could just label your fuses with letters or numbers, so long as you know which room each of these corresponds to.

Barcan Marcus and Kripke argued for the Millian part of the theory largely on the basis of the fact that descriptions seem generally to shift what they apply to across possible worlds, whereas names do not. If one thinks of a name as a tag, as a pointer to a particular object, then one might think: ah, just what we expected! After all, the name is just pointing to an object; it’s not describing that object. So its reference shouldn’t vary even as the description of the object in question varies, sometimes radically so.

The second thesis aims to fill a gap left by the first one. Simply saying that names are devices of ‘pure’ reference, as the first thesis would have it, doesn’t tell us anything at all about how they come to refer to what they do. So we need to know more. Kripke and Keith Donnellan and Peter Geach were all toying with very similar ideas on how to fill that gap around roughly the same time. Basically, their common thought was that a name’s referent gets fixed by facts about its history. We assume that the name is introduced at a certain point in time. Typically, this has been called a ‘dubbing’ or ‘baptism’. Then we also assume that the name gets passed down from person to person. The pronunciation might change over time (presumably the ancient Greeks didn’t pronounce “Plato” much like we now do), but so long as we can, from a god’s eye perspective, trace a chain of uses back to some object, we’ve determined the reference of the name. It just refers to whatever we find at the start of the chain, the object dubbed at the very beginning.

Anyway, there are lots of issues with both these theories, but both are also interesting and worth taking seriously. I hope that will do for an initial sketch.

3:AM: What’s are the key problems faced by descriptivists – and are they fatal?

EM: Let me run you through my favorite of the standard objections to descriptivism, what’s usually called Kripke’s ‘semantic objection’. Take the name of a biblical prophet like ‘Jonas’. What do we know about Jonas? Well, he preached somewhere in the Levant. But so did a whole lot of other people. And supposedly he lived for a while in the belly of a whale. But we have good reason to believe that to be false, because no one can do that. So what’s the sense that should be associated with ‘Jonas’? If the name refers, it can’t just be a Levantine preacher. That won’t pick out anyone in particular; there are just too many possibilities. So it seems like the descriptivist either has to give up on the claim that the name refers, or she has to give up on her theory.

Are there ways the descriptivist can respond here? For sure. First, she can try appealing to more than just what the speaker or listener happen to know in order to fix the sense. That, however, is going to threaten to take away certain of the goodies Fregean descriptivism is supposed to buy you—a solution to what’s typically called Frege’s Puzzle, for instance. But maybe that’s an acceptable price to pay. Or she can adjust the relevant description to something like the Levantine preacher who supposedly lived for some time in the belly of a whale. Problem solved! Ah, but wait. Maybe there’s more than one of those. Or just get rid of the whale bit. So, suppose that all we ever learn of Jonas is that he preached somewhere in the Levant. Or maybe we don’t even learn that. Perhaps the name “Jonas” just appears, alone, on some shred of parchment in the vicinity of some early biblical manuscripts. It seems that we can still sensibly say “I wonder who this Jonas was.” Who are we wondering about? Well, Jonas.

Again, the descriptivist might come back and say: fine, two can play this game. Here’s my description for you: the person whose name we have found on this piece of parchment, supposing there is one. Kripke thought that, at this point, the theory had devolved into something like circularity. I’m not convinced. But here’s another worry one might have: the theory may have initially seemed appealing because, when we ask someone “Who are you talking about?” they tend to respond with descriptive answers like “Oh, Sarah the political philosopher.” Many, maybe even most, of those answers are even going to correctly pick out the object we’re trying to talk about. So it can seem appealing to think something like “Oh, okay, I guess that’s how we really get on to objects. The apparent simplicity of names was misleading.” The ultimate point of Kripke’s examples, I think, is to suggest that maybe we have a capacity to think about objects more or less directly—such that we, individually or collectively, may have forgotten any accurate information about them, let alone information that would be sufficient to actually identify them. All we need is the name, Kripke suggests, and we can formulate beliefs and questions about an object. And we can express those in words. On the other hand, the really dedicated descriptivist can just come back and say “Yes, and I can account for that by allowing that the description is just of the form: the thing called ‘NN’.”

It’s not exactly clear how this ends. If you’re like me and you think that our basic way of thinking about objects is via perceptual acquaintance with them, and you further think that we can lock onto objects in spite of a misperceiving a great many of their properties, then you might, like me, find the Kripkean picture appealing. On that picture, the semantic structure of names is going to mirror, more or less directly, the semantic structure of the object-directed pieces of our thought. Descriptions, on the other hand, aren’t going to be of the right form. So the match between thought and talk is going to have to be looser in many instances. This isn’t exactly the semantic argument we started from—and it’s hardly a decisive one given the rather heavy background assumptions that went into it—but that’s how I tend to think about these things.

There is another major line of argument against descriptivism that I haven’t touched on here: modal arguments. Basically, the thought is that the reference of names doesn’t shift when we consider another possible world, another way things might have been, whereas the satisfier of a description often will. So names can’t be descriptions. But there is some fancy stuff that the descriptivist can try in response, like sticking an ‘actually’ in the description. Things get pretty complicated at this point. Scott Soames has argued at length in Beyond Rigidity that, ultimately, this ‘actualizing’ move isn’t going to work. I won’t get into all of that here. For one thing, I just don’t know these debates well enough to try to summarize them accurately; for another, I’m by no means up to date on the state of play here. For whatever it’s worth though, I found Soames’ arguments pretty convincing when I read them back in grad school. Convincing enough that I didn’t feel like I had anything more to add really.

3:AM: Does Kripke’s causal account handle these problems better than descriptivism and is it overall a better approach?

EM: Sure, the Kripke-Geach-Donnellan-Barcan Marcus approach is going to handle these problems better. The modal stuff comes for free, since one half of the theory is just: these terms denote specific objects, and that’s all they denote. Their meaning is exhausted by their reference. So there’s no chance of the relevant object changing as we shift from world to world. And it’s not going to matter if we, individually or collectively, have only false beliefs or no beliefs at all about the referent. Those beliefs are going to have nothing to do with what fixes reference. On that score, the causal theorist clearly comes out ahead.

There are serious problems facing the causal theory, however. One classic problem, and one of my favorites to teach, is Gareth Evans’ classic ‘Madagascar’ case. Evans points out that the reference of a name can seemingly shift even if, historically, all we have is a bunch of people picking up the name from each other and then trying to use the name in the same way as the person they picked it up from. He takes this to have really happened with the name ‘Madagascar’ (I make no claims as to the veracity of the example): this name having once referred to part of the African mainland and now, thanks somehow to Marco Polo’s misunderstanding some Malay sailors he was chatting with, clearly referring to a very large island off the east coast of Africa. Assuming that Polo was trying to use this name in the same way that his informants were, this would seem to cut squarely against the causal theory.

Another problem, this time with the Millian half of the theory, has to do with the fact that it seems basically platitudinous that more than one person can bear the same name. So when I say “Robyn is coming to the party,” I could be referring either to the linguist Robyn Carston or to my erstwhile colleague Robyn Waller. What makes it the case that I’m referring to one of them or the other? As it stands, the theory tells us nothing about this kind of case.

Also note that we can say things like “There were two Robyns at the talk yesterday.” The most recent wave of descriptivists—i.e. ‘predicativists’, led in particular by the late Delia Graff Fara—have really hammered on this as an objection to Millianism about names. If all that names do is refer to people, then it’s highly unclear what one can say about this sort of case. Of course, there are things to be said in response—Robin Jeshion and Anders Schoubye have both done some very nice work on this recently, and Aidan Gray has just thrown a bunch more really interesting data into the mix. But, basically, I don’t think anything is settled here yet.

If anyone for some reason just wants to know where I stand: I’m an unreformed causal theorist. I’ve even started to think that names need to be individuated in an extremely fine-grained manner, such that I don’t bear the same name as any other ‘Eliot’, let alone an ‘Elliot’ or an ‘Elliott’. God forbid! I used to think that this sort of view was completely ridiculous, but now I’ve changed my mind. Which, I suppose, is a lot of what’s fun about doing philosophy.

3:AM: Is a hybid approach even better? What would this look like?

EM: I’ve grown up, philosophically speaking, in an era dominated by debates between Millians, predicativists, and, to a lesser extent variabilists (a whole other complicated view which I’ll summarize, inadequately, by saying that it’s basically committed to names functioning more or less like pronouns). So, to be honest, my knowledge of the advantages and disadvantages of hybrid theories is less solid than it ought to be. I find Evans’ original ‘Ibn Kahn’ case pretty compelling though. Evans asks us to consider a case in which we find a number of ancient mathematical proofs all signed by ‘Ibn Kahn’. Only later do we discover that this signature belonged not to the mathematician who produced the proofs, but rather to the scribe who recorded them. Evans wants to know to whom we would be referring if we were to say “Ibn Kahn proved such-and-such,” and he thinks that the answer is, pretty clearly: to the mathematician, not the scribe.

Evans uses this to motivate a theory according to which the referent of a name isn’t whoever was originally baptized with that name, but rather whoever is the dominant causal source of information associated with that name. What exactly it means to be the ‘dominant causal source’ isn’t an easy thing to figure out, and Evans himself changes his mind about how this notion ought to be understood. But, leaving this notion at an intuitive level: it certainly seems as though this sort of theory is going to do better with both the Madagascar and Ibn Kahn cases than a less reformed causal theory is going to.

Like I said though, I’m an unreformed causal theorist. So I think that, in spite of the prima facie appeal of Evans’ approach, there are other fixes worth considering. For my own part, I’m rather fond of some ideas that Louis deRosset has been developing recently. DeRosset begins by noting that, while we might well expect for historians of mathematics to go on as though nothing has changed when they discover that the signature ‘Ibn Kahn’ belonged to the scribe, it certainly isn’t inconceivable that they might change their practice and start to say things like “Whoever dictated to Ibn Kahn proved such-and-such.” Perhaps they might even give this other person a name. If that’s right, if that’s a real possibility, then it might turn out that the reference of a name can depend, in part, on future practice. So in the one case, a new use of the name ‘Ibn Kahn’ is inadvertently introduced by the mathematicians who discover these manuscripts. In the other case, they try and fail to introduce a new use of the name—and thus end up saying a number of false things. I prefer this sort of theory, partly because it keeps us from having to try to figure out what it means to be the ‘dominant causal source’, which I’m pretty convinced is a vexed task. But, that said, I can totally see why some people might want to try to make an Evans-like theory work here.

3:AM: How do indexicals refer? Can the same mechanism used for proper names work just as well for them or do we need to have something else?

EM: Will the same mechanism work? Basically, no. Most people nowadays tend to think that indexicals like ‘I’, ‘here’, and ‘now’ refer in virtue of some pretty simple and straightforward rules having to do with the context of utterance. So, for example, in some context C: ‘I’ refers to the speaker at C; ‘now’ refers to the time of C; ‘here’ refers to the location of C. Rules like these don’t appeal to anything like a chain of usage, so they certainly look pretty different from the kind of mechanism that fixes reference according to the causal theory.

Now, there are different ways of trying to work out this basic idea. One could, for example, be a descriptivist of a certain sort and think that what a sentence like “I’m hungry now” literally expresses is something like “The speaker of this utterance is hungry at the time of this utterance.” One can see how this represents a pretty natural extension of Fregean-style descriptivism about names. Frege really was aspiring for a unified theory of reference, and he provided some remarkably versatile tools for carrying out that project.

Most people nowadays reject this sort of view though, on the basis of some cases that maybe I shouldn’t get into. Let me put it this way: they tend to assume a purely qualitative version of descriptivism, rather than one that might allow for things like percepts to enter in. I think that stacks the deck against the descriptivist view here—but, then again, I’m not a descriptivist and I’d rather leave it to one of them to defend their own view.

Instead, the more popular move has been to follow David Kaplan and to think of indexicals as devices of pure reference, much like the Millian thinks of names. The Millian owes us a story about how names get their reference, and Kripke and others suggested that we should fill in that story by appeal to historical chains. Here instead, Kaplan suggested that we should appeal to rules that are sensitive to the context of utterance, and never to any other context. He called these rules ‘characters’ and suggested that we could think of them as a level of meaning. Still, for both Kaplan and the Millian about names, part of what’s distinctive about a referential term is that the only sort of meaning it has that contributes to the truth-conditions of a use or utterance of a sentence is its reference (Kaplan likes to talk about ‘sentences-in-context’, but that’s a bit overly technical for our purposes here). This stands in contrast to classical descriptivists, like Frege, who think that sometimes the sense is going to be relevant to truth and falsity as well.

So there is a way in which the Kaplanian about indexicals and the Millian about names are offering a unified story. They’re saying that, for referential terms, only their reference is ever relevant to the truth-conditions of the use of a sentence. These sorts of theories are usually called ‘direct reference’ theories. While there is a sort of semantic unity to direct reference theories, they tend to posit very different mechanisms for fixing that reference in the case of, say, names and indexicals. Suppose that we accept both Kaplan on indexicals and Kripke on names—then we accept that the reference of a name is fixed by its history, but the reference of an indexical is fixed by a timeless rule. That doesn’t look very unified; rather, we’re positing that very different factors determine reference for different sorts of terms. So what we end up with, basically, is semantic unity but ‘meta-semantic’ disunity.

3:AM: Out of all this, do we have a unified theory of reference?

EM: Well, clearly we do if we’re Fregeans! Otherwise, it’s trickier, and it’s going to really depend on what we mean by a ‘unified theory’. As I mentioned, there’s a sense in which, if we think of referential terms as directly referential, we get a certain sort of unity: unity of function. What a directly referential term does, on this sort of theory, is contribute some thing, an object or a property, to the truth-conditions of the utterance, or use of a sentence if one prefers. Or we can say that it makes the truth or falsity of the relevant utterance object- or property-dependent. There will be some issues on the margins here—having to do, for instance, with names or property terms that fail to refer to anything—but still, this looks like a pretty unified picture in a sense. Let’s call this ‘minimal unity’.

What such minimal unity leaves out, I think, is unity at the level of reference fixing, at the level of explaining why particular uses of referential terms refer to whatever they do in context. As I noted above, this is often thought of as a meta-semantic issue. Whatever we call it, it’s a question to which Frege gave a unified answer and to which direct reference theorists generally have not. In fact, Frege gave a maximally unified answer: for any referential term, its sense determines reference. The starting point for much of my work has been to ask: is there some intermediate level of unity that the direct reference theorist might hope to achieve? I think the answer is “yes”.

How might we get there? My suggestion is to shamelessly pillage certain resources developed by ‘externalists’ about natural kind terms like Hilary Putnam and Tyler Burge, and then to put those together with certain minimal elements drawn from Frege’s theory, but which are in fact compatible with direct reference theories.

Let’s take the externalism part first. Start by considering a term like ‘water’. Putnam and Burge argued that both this term and its correlate at the level of thought were going to refer to some natural kind in the world regardless of whatever false or incomplete descriptive material the speaker/thinker happens to associate with this term/concept. Reference isn’t going to be determined by how the speaker or thinker conceives of water, but rather by the fact that she stands in a certain sort of persistent causal web that involves water, as opposed to something else, playing a certain sort of role: namely, being the sort of stuff that we drink, etc. For the moment, let’s suppose that this is how our concepts get their referential features, their aboutness—not by being associated with any particular descriptive material, but by bearing a certain causal relationship to some stuff in the world.

Now let us strip down the Fregean theory to its bare bones: on Frege’s theory, senses are parts of thoughts. So just as individual expressions express their senses, complete utterances express our thoughts. I like this model, I just don’t think that we should think of anything like senses as determining the referential features of our thoughts. So suppose that our thoughts, and their sub-parts, have referential features purely in virtue of some causal connection to the world. Then we can adopt a direct-reference friendly analogue of the Fregean theory: directly referential expressions inherit their referential features from the referential features of the sub-parts of our thoughts they are used to express. For example, a use of the name ‘Rachel’ might be used to express a sub-part of a thought about my friend Rachel. My use of the indexical ‘I’ might be used to express a first-personal sub-part of a thought, a thought about myself. In neither of these cases was it necessary for anything like a description to enter into the picture. This, in a nutshell, is my preferred picture of reference.

One worry that’s likely to arise at this point is the following: what’s to stop me from expressing a thought about myself with the word ‘you’, or a thought about Rachel with the name ‘Eileen’? Note that an analogue of this question is likely to face the Fregean too, so long as, like Frege, she posits that the sense expressed by a given use of a term depends on the speaker’s mental state. Anyway, I think that this sort of objection can be met by distinguishing between trying to express a thought, or trying to refer, and succeeding in either of these acts. Restricting our attention to reference for the moment, I think that the key is to insist that particular terms impose limits on what one can succeed in referring to by means of them. So ‘you’ can be used to refer to whoever inhabits the second-person role in the context, and ‘Eileen’ can be used to refer to Eileens. The former cannot be used to refer to whoever inhabits the first-person role, and the latter cannot be used to refer to Rachels. One could try to work out this constraining role in different ways, but perhaps the easiest is to say that different referring terms trigger different sorts of presuppositions. Unless the presupposition is met, reference fails. If the presupposition is met, however, then the term just inherits the referential features of the part of the thought it was being used to express. So when I utter ‘Rachel’ while trying to express a thought about a particular Rachel, that use of the name refers to the particular Rachel I have in mind.

Actually, I think things are even a bit more complicated than this. The problem is that I think that we are often trying, unknowingly, to express multiple thoughts with our utterance of one and the same sentence—thoughts that we cannot tell apart from the inside. If that’s right, then we might want to adopt a theory based on what I call ‘filters’ rather than presuppositions. But what I’ve said already should be enough to get the basic idea. Now we should ask: how unified is a theory of the kind I just sketched? Well, I think it’s more unified than the kind of direct reference theory we started with. We can say the following, for instance, about any use of a referential term: if it has a referent, it has that referent in virtue of its inheriting that value from a sub-part of a thought that it was being used to express. We also know that it must have been used successfully, that there was a certain sort of match between the selected term and the referent itself. Where disunity remains is with regard to the presuppositions: what it is to be a Rachel or an Eileen looks like it might best be explained in terms of historical facts about name-use. What it is to inhabit the first- or second-person role in a context, on the other hand, isn’t going to depend on historical facts, let alone historical facts about name-use specifically. So an element of the disunity we see in more standard direct reference theories remains in mine. I think that’s all right though; if different sorts of referential terms are different sorts of tools for bringing about a particular sort of end, I think we should be open to their differing pretty substantially in their form. The resulting theory is still more than only minimally unified. Granted, it’s not as unified as the original Fregean theory. The hope, however, is that we’ve traded this lesser degree of unity for substantial gains in terms of plausibility.

3:AM: Do all meaningful expressions refer? If we take a sentence like ‘nobody runs faster than me’ it seems plausible that ‘nobody’ doesn’t! How do you deal with this kind of issue? I guess this raises the issue of the relationship between reference and meaning.

EM: I suppose that one could try to claim that all meaningful expressions refer, but I don’t know of anyone who’s really tried to defend that in its full generality. I’m also not entirely sure why one would want to try to defend that view. But suppose there were some reason to. Best I can tell, that sort of theory is going to have to deal with two sorts of challenges: what to say about terms like empty names, and what to say about terms like ‘nobody’. Let me treat these separately.

Even within the realm of standard referential expressions, like names, some of these don’t look like they actually refer. Take a name like ‘Santa Claus’ or ‘Anna Karenina’. Not so clear that either refers, though both certainly seem meaningful. Now, one option here is to say that the former refers to a mythical object and the latter to a fictional object, whatever exactly the difference there is supposed to be. This is an idea that Kripke defended in some later work, Reference and Existence, and which Amie Thomasson has also defended in a bunch of really interesting work. Another option, of course, would be to say that these terms just don’t refer. They’re meaningful, perhaps because they purport to refer, but they don’t refer. That’s an option that Mark Sainsbury has defended in Reference Without Referents. I’m with Mark on this one—though, unlike him, I think there might be good reasons to posit fictional objects even while thinking that we don’t ordinarily refer to them. I’m going to leave the details aside here though. Suffice it to say that, while I personally find these debates fascinating, they get very complicated very quickly.

Let’s move on to ‘nobody’. Who might ‘nobody’ refer to? Well, not to any particular individual clearly. When Bert Williams sang “I never did nothing to nobody,” what he was saying was: there’s no one that I’ve done something to. It’s a denial of having done something to someone. In effect, nobody looks to be functioning as a quantifier. It’s expressing something like: for all x such that x is a person, I’ve never done anything to x. There’s obviously a lot more quantification packed in here, but I’ll leave that aside.

Anyway, quantifiers are actually not so hard to treat as referential, contrary to initial appearances. One standard analysis of these terms, dating back again to Frege, is that they denote properties of properties. So ‘every person’ is a property which applies to another property so long as that latter property is a property that applies to every person. “Every person is a person” is going to be trivially true since the property of being a person is, of course, a property that every person possesses. ‘No person’, on the other hand, is going to be a property that properly applies to another property so long as there is no person who bears that property. ‘Nobody’ should, presumably, work in much the same manner.

So long as we’re comfortable with talking about terms referring to properties of properties, terms like ‘nobody’ won’t be so bad. It’s true, these won’t refer to individuals, in spite of being terms that seem capable of occupying the subject-place of a sentence, just like a name. This causes some complications in formal semantics, but those are surmountable. And the basic idea here isn’t actually that odd: these terms are basically covert quantifiers, and quantifiers pick out properties of properties. The logical connectives might prove worse, terms like ‘and’ or ‘if…then’. Though, if one is so motivated, one can try a similar strategy: taking these to be properties of pairs of propositions, i.e. the property of their both being true, or the property of depending on one another in a particular sort of way.

All that said, I really don’t think we should think of all language as referential. It seems to me that a lot of what we say may be irreducibly expressive, in the sense that what the language is doing is expressing something about our mental states, but where those states aren’t representational, or at least not in any straightforward sense. Maybe the easiest way to see this is with exclamations like ‘Ouch!’ or ‘Oops!’ I really just can’t see a way of cashing these out in representational terms. Others have thought similar things about ethical and aesthetic expressions and, more recently, the epistemic versions of ‘must’ and ‘might’. I think there’s a lot to be said in favor of these theories, though for my own money I’m most sympathetic to claim about epistemic modals. The point though is that, if any of these theses is right, then there are going to be substantial parts of our language that are non-referential. So ‘nobody’ and ‘nothing’ look like they’ll be far from the hardest problems facing the die-hard referential universalist. Thankfully, I’m not one of those.

3:AM: I am intirigued by the issues raised by virtual reality and the Nozick machine and living in the Matrix and psychedelic mind experiences and Putnam’s brain the vat and so forth. What does your preferred theory of reference and its relationship with meaning, truth and knowledge handle these kinds of experiences?

EM: This is a great question, but I’m afraid it’s one that I don’t tend to think about all that much. Putnam, at a certain point at least, seems to have thought that there was a pretty clear route from externalism about meaning to some sort of anti-skeptical argument. I can’t say that I’ve ever actually felt like I’ve fully grasped his argument, and I’m pretty deeply skeptical that such an argument can be made to work in a non-circular manner. Of course, it’s a helpful working assumption to think that whatever is on the other end of our thought and talk is objects and properties, since it feels like we have some grasp of what those are. But if they ultimately turn out to be patterns of bits rather than combinations of sub-atomic particles, I’m not sure that matters a lot from the point of view of the theory of reference. And I definitely don’t think that knowing that ‘water’ refers to water is going to tell us anything interesting about the nature of that property of being water. Chemistry is going to tell us about that, or maybe ultimately some other science. And if that science eventually tells us that being water really comes down to having a certain sort of data structure in a register somewhere running our whole universe as a simulation, then so be it.

3:AM: Can substantial metaphysical issues arise out of any particular theory of reference that one might adopt? For instance, might one theory justify belief in psychedelic goblins or unicorns? Would the kind of ontology being produced by the theory be a way of assessing whether the theory of reference being used was ok?

EM: First of all, let me go on the record as saying that any theory of reference that can offer a decent justification for our belief in psychedelic goblins will immediately become my new favorite theory of reference. I’d love to see that theory, and I’d definitely prefer to teach it to some of the other options. There are, of course, theories that will allow you to believe in fictional psychedelic goblins or fictional unicorns. But that is just so much less interesting.

Returning to the gist of the question though: sure, I think this is very often one of the things we do in trying to assess the viability of particular theories of reference. But it’s messy and doesn’t really tend to settle much of anything. The problem is that, usually at least, no one is saying anything quite like “Oh, if you accept my theory of the reference of names, then you need to start believing in real psychedelic goblins.” Even most philosophers would probably be inclined to reject such a theory on the basis of such an entailment—though, as noted, some of us might be sorely tempted to teach it.

More often, I think that theories of reference are used to argue that some folk metaphysical beliefs we had are, in fact, wrong. For example, Adam Pautz has recently argued, in part on the basis of some commitments that he finds plausible about the reference of property terms (the details of which I’ll skip over here), that there are no colors in the world. There are certainly color experiences, experiences as of a colored world. But when we look out into the world and we try to find the colors, it turns out there’s nothing that meets the conditions that he thinks something would have to have to count as a color property. In other words, when we look carefully at the world, we don’t find any colors. So we should think, argues Pautz, that the color terms don’t refer to anything. They purport to refer to properties in the world, but they don’t. So they’re sort of like ‘Santa Claus’ or ‘Anna Karenina’ on Mark Sainsbury’s way of looking at things, only this time about properties.

What should we make of Pautz’s argument? Well, on the one hand you might think it’s totally absurd. Of course there are colors! What could we be more certain of? After all, most of us see them right in front of us basically any time our eyes are open. On the other hand, if you grant Pautz his starting assumptions about property reference—and they’re not by any means implausible—then that plus the relevant bits of color science will quickly drive you to the conclusion that there are no colors in the world. The problem is that there is nothing in the world that turns out to be like what a color would have to be like in order for it to be a color. So, basically, no colors—just pervasive color illusion.

Now, I don’t particularly care for the sort of theory of property terms that Pautz likes, and for reasons that have nothing to do with defending the claim that there are colors in the world. But supposing one was attracted to that sort of theory and also to the thought that there are colors in the world, then one would face a stark choice here: give up on your metaphysics, or give up on your preferred theory of reference. Work in this part of the field isn’t usually going to tell us which way to go when facing this sort of dilemma. But at least it can help to clarify the sorts of interdependence that can arise between issues in the philosophy of language and metaphysics.

3:AM: Philosophers like Quine and Davidson take a pretty negative view about reference. Can you first sketch for us why they and others have taken such positions and then say how do you assess these approaches?

EM: To be clear, I’m no Quine or Davidson scholar. So I’m sure that some in the field could give a much more exhaustive answer to this question than I can. Still, since you ask, here’s my best shot: it seems pretty clear to me that the shift from thinking of psychology in behavioristic terms to non-behavioristic terms was one important element of whatever was going on. Quine and Davidson both clearly had some sympathies towards behavioristic approaches to psychology, and those approaches aren’t going to allow one to distinguish any number of different possible objects one could be referring to: an ordinary rabbit v. undetached rabbit parts, to use Quine’s well-known example. Nowadays, I think philosophers tend to be more confident that, perceptually, what we are tracking is objects and properties rather than pseudo-objects like undetached rabbit parts.

That said, I’m not so sure that this confidence is actually justified. Why, after all, should we take it that tracking rabbits is fitness-enhancing vis-à-vis tracking undetached rabbit parts? The answer, it seems to me, isn’t at all obvious. And, in fact, there’s a nearby variant of this worry that doesn’t look open to a response along these lines. This variant comes to us from David Lewis and is usually called ‘The Problem of the Many’. Lewis asks us just to focus on what it is to be something like a rabbit: what exactly are we perceiving when we perceive such a thing? Suppose that rabbits are identical with a certain bits of matter, at least at a given time. Well and good. So which matter exactly? One can take away any random atom of rabbit skin and still have a rabbit. Or consider a hair that’s in the process of falling off our hypothetical rabbit: is that still part of the rabbit? When exactly does it stop being a part of the rabbit? Even if we think that evolution has primed us to perceive and refer to ordinary objects, that doesn’t really cut off the basic worry here: there are just a ton of different possible things we could be perceiving or talking about. So which one is it? And why that one? Without a principled answer to these questions—and I have yet to find any that strike me as all that compelling—I think we should perhaps be at least a bit skeptical that our perceptual system is going to somehow ‘solve’ this problem. If linguistic reference depends, in turn, on perception, then so much the worse for linguistic reference.

My somewhat charitable interpretation of what’s gone on in the wake of Quine and Davidson is this: Quine and Davidson may well have pointed to an irresolvable problem at the foundations of the theory of reference. If that’s right, then perhaps the best we can do is to idealize away from that problem and think about how language would work if we were somehow able to fully identify the particular objects and properties to which our language refers. That can look like just ignoring the problem, but I’m not sure that’s right. Perhaps it’s more like putting the problem on hold. Maybe one day we’ll figure out how it is that our perceptual system is able to solve these underdetermination problems and lock onto some very specific things in the world. I’m not putting my money on that though. On the other hand, one could think that there is still to be much to learn about how our language functions by treating it as if these underdetermination problems could be resolved. That’s my hope anyway, though I grant that it might turn out to be in vain.

3:AM: What is the answering machine paradox helping us to examine? What’s the answer to the question is asks?

EM: Quick and dirty introduction to the answering machine paradox: if you’re Kaplan, you claim that the sentence “I am here now” is a logical truth. That means that you predict that no use of that sentence will be false, and no use of its negation, “I am not here now,” will be true. But it would appear that the latter sentence can be ‘used’ as a truth on an answering machine. That is, it can be recorded and played back, and in the right circumstances, that playback token at least seems to be true. So it looks like either Kaplan’s theory is falsified, or else we have to explain how these aren’t really ‘uses’ of the sentence, at least not in the relevant sense. To my mind, this hardly makes a paradox; rather, it’s just a challenge for someone who wants to hold onto Kaplan’s theory of indexicals.

Now, what’s to be learned from the paradox? For my own part, I think that part of what we’re doing in philosophy of language is trying to tease out the border between literal and non-literal meaning. Now, it’s worth noting that we’re probably not using that distinction here in quite the same way that it gets used in the world at large, at least best I can tell. The ordinary contrast is perhaps more squarely between literal and figurative language, whereas philosophers tend to think of things that are implied or hinted at as ‘non-literal’ in the relevant sense. Those are hardly figurative uses of language though, at least on the face of it. But that’s okay. This distinction between what’s literally meant and what’s merely hinted at or implied or gotten across by some other means seems like a rather interesting one, and it’s clearly derived in some way from the ordinary notion.

The question thus arises with the answering machine: is it literally telling you anything at all? And, if so, what? That’s separate from the question of: what do answering machine messages typically communicate? One possible distraction here is that the example is becoming really dated. Nowadays, most people either use voicemail or, increasingly, nothing at all—just texts or WhatsApp or whatever. I sometimes feel like we should just start calling this the ‘answerphone problem’ since it sounds more overtly old-timey. But voicemail messages on a cell phone are probably even more complicated, so let’s stick to old-timey answerphones.

We know that an answerphone message of “I am not here now” will typically convey that the person who regularly uses the line isn’t at the place where the answerphone is located at the time when the message is being played back. The question is whether the message, as played back, literally says this. Some answer “yes”, in which case explaining how this gets communicated is pretty straightforward. You just have to come up with the right set of rules governing the use of terms like ‘I’ and ‘here’ as they appear on answerphones. Some answer “no”, in which case we need some further story about how this message ultimately gets conveyed. And then there will turn out to be various costs to different ways of answering “yes” or “no”. For whatever it’s worth, I’m a yes-man.

In early work on the subject, like that of Quentin Smith and Alan Sidelle, I think that the hope was that reckoning these costs and benefits was going to help us think more clearly about two things: first, whether we had the right theory of indexicals, or whether we had been confused by a restricted data set; second, how we should think about the notion of literal meaning extending to things like recordings. Sidelle noted that if we make just a minor tweak to the standard Kaplanian semantics for indexicals—allowing that the relevant context for evaluating terms like ‘here’ and ‘now’ is the one in which it is heard, rather than the one in which it is produced—then things seem to work out much better in the case of the answerphone. Since in ordinary face-to-face speech these coincide, it’s easy to see why earlier theorists might have opted for the former option. Basically, Sidelle was thinking: look, we make a minor tweak and we get a theory with more explanatory power. Good deal, right? Well, actually, he didn’t quite think this. Rather, he thought this idea looked enticing but ultimately fell prey to too many other problems to be worth adopting. More recently, Jonathan Cohen has dusted off this suggestion and given it a run for its money. According to Cohen, most of Sidelle’s worries can be dealt with in one way or another. It’s cool work, even though I don’t ultimately agree with it.

Once that ball gets rolling, it’s relatively easy to come up with a bunch more examples that previous theorists hadn’t seriously attended to. One starts to think: okay, maybe these can point the way to a better overall theory that straightforwardly accounts for both. Or maybe: okay, these examples can teach us something about what happens when we use language in ways other than just talking to each other face-to-face. And maybe that, in turn, can help us to better understand what it means to offer up a theory of literal meaning. That was the sort of hope I had in wading into these debates. That was the sort of hope that I think Sidelle and Smith started with as well. Unfortunately, as the literature has developed, the focus seems more and more to have shifted to coming up with some really recherché cases and less and less on grappling with the core issues that made certain cases interesting in the first place. I’m hopeful that maybe that will change someday—like I said, I think it’s fundamentally an interesting topic—but I guess we’ll just have to wait and see.

3:AM: You’re interested in insincere speech in politics, and lying and bullshit presumably too. There have been some interesting developments in what lying is from people like Sorensen and Carson and Lackey – I think one of the main things is that the idea of deceit is no longer seen as crucial to lying. Is that right? Can you get us up to speed on what is new here and why lying offers good data for the sorts of philosophical issues you’ve interested yourself with?

EM: I’m glad we saved the good stuff for anyone who’s managed to read this far into the interview. Finally, let’s talk about bad linguistic behavior! Honestly though, I think this is really one of the more interesting parts of the field at the moment—not least because it’s clearly so relevant to our current political situation. I think it matters, and should matter, to a lot of Americans whether Trump is lying to us or whether he’s just full of shit. Likewise for the Brexiteers. Maybe it’ll only matter for the history books—well, at least until Trump finally gets around to banning those. Still, it matters. If we’re going to do better, we’ve got to start holding people to account, and hopefully in the right sorts of ways. My vague hope is that learning to attend to the various different ways in which our politicians lie to us, mislead us, and otherwise use language to manipulate us might one day help us to start making better, more informed decisions. I just don’t know how we can possibly hope to move forward in the complex, rather fucked-up world we live in if we’re basing our collective decisions not on good information, but on some bullshit that Trump decides to put in a tweet or that Boris Johnson decides sounds good on the side of a bus.

So look, basically for thousands of years philosophers have taken it that lying requires an intention to deceive. And not for no reason—this is a really intuitive thought! But, somewhat scandalously to my mind, it basically never occurred to us until the mid-00’s that there’s a pretty obvious set of counterexamples to this. Then, in quick succession, Thomas Carson and Roy Sorensen, finally pointed this out: bald-faced (or ‘bare-faced’, as they say in the UK) lies probably don’t seem to involve any sort of intention to deceive. After all, the whole point of such lies is that they are transparent to all parties. Still, it’s very natural to think of such utterances as lies. If that’s right, then an intention to deceive simply cannot be required in order for an utterance to count as a lie.

To get clearer on this, let’s work through an example. The most commonly cited one is due to Carson and basically runs like this. A professor has caught a student, let’s call him “Jared”, cheating. Let her have arbitrarily good evidence of the cheating: a video, excellent forensics, or whatever you like. Just no direct confession from Jared. Now suppose that the Dean has an all-too-believable fear of being sued by irate students (or, worse yet, their aggrieved parents), and thus refuses to punish any student for cheating in the absence of an admission of guilt by that very student. Finally, suppose that the Dean calls Jared into her office. Together, they look at the impeccable evidence of Jared cheating. The Dean now asks Jared whether he cheated and Jared replies “No, I did not cheat.” Carson’s thought is that Jared can’t intend to deceive the Dean regarding whether he cheated, since Jared knows that the Dean won’t believe him. After all, they just reviewed the evidence together. Jared’s cheating is already out in the open!

So basically, Carson relies on the widely-accepted (though not totally indubitable) principle that one cannot intend to do something that one takes to be impossible for one to do. So, for example, I cannot intend to jump over the moon. Why? Well, one plausible explanation would seem to run: because I am utterly certain that I will fail.

Now, Jennifer Lackey has recently offered a rather clever argument to the effect that Carson was overly hasty here. While Jared cannot intend to deceive the Dean, he can nonetheless intend to be deceptive to the Dean. What’s the difference? Well, Lackey thinks that one can be deceptive by concealing evidence of something. What’s the evidence here? A confession, Lackey says.

I think that the sense in which evidence is being concealed here is stretched beyond the breaking point. Since there is no confession, it cannot be concealed —at least in the sense that evidence can be concealed (and, remember, that’s what’s at issue). That’s a point that Don Fallis and Andreas Stokke have both made recently. Together, Andreas and I have argued that there might be a better way to try to defend a Lackey-esque position: substitute ‘tries’ for ‘intends’, the idea being that one can sometimes try to do something even if one believes that one will inevitably fail. So you can try to swim to that island in the distance even if you know you are a weak swimmer and you fully believe that you will fail. If the situation is desperate enough, what’s left to do except try? Maybe Jared is in a case a bit like this one. Neither Andreas nor I ultimately think so, but we think that the reasons are at least a bit more nuanced than they were with respect to Lackey’s original view. So if you want to try to hold on to this connection between lying and deception, we think that our suggestion is a better way of trying to do that than Lackey’s turns out to be.

Turning back to the real world, it’s important to note the following: even if you think that lying doesn’t absolutely require an intention to deceive, one should probably still think that an intention to deceive regarding P, in conjunction with having asserted P and believing P to be false, is going to be sufficient to make one’s utterance a lie. So taking Carson’s point doesn’t leave us deeply adrift with respect to the sorts of real world questions we face about the lying scumbags in charge of a shocking number of western democracies nowadays, not to mention their kindred spirits in industry, media, etc. If we have good evidence that someone has told us something that they don’t believe while intending to deceive us about that something, then we still have good evidence that they have lied to us. There are further questions to be asked about what it takes to ‘not believe’ something in the relevant sense—questions that Stokke has taken up in some recent work on insincerity, and which Jessica Pepp has taken up in some very nice work on lying—but, again, these are really attempts to deal with cases at the margins, not the sorts of cases that we generally face in our everyday lives when thinking about the likes of Trump or Johnson.

3:AM: Finally, I have a few friends who are vegetarians but they seem to eat fish (a little). You’ve looked at this rather strange phenomenon. Are there reasons for treating fish like this – are there grounds for being morally indifferent towards fish but not birds or terrestrial animals?

EM: First of all, we should distinguish between teleost fish—basically, almost any sort of fish with a backbone—and various other sorts of sea creatures. Sometimes we lump all sorts of sea creatures together as ‘fish’, but there are some significant cuts here. Cephalopods like octopuses appear to be extremely intelligent. Some teleost fish, like giant manta rays, may be as well. In contrast, the bivalves like mussels and oysters don’t have much in the way of a dense, central nerve mass. That—paired with their pretty limited repertoire of behavior—seems like pretty good evidence for their having only limited intelligence or experience of the world. It’s tricky to draw firm conclusions on such bases though: the environments in which these creatures operate are very different from our own, so we may simply be failing to recognize various complex behaviors. And we know from the cephalopods that certain sorts of complex distributed processing are possible. Octopi have relatively little of their dense nerve mass in a centralized location, yet those suckers are capable of some stunning displays of intelligence. As YouTube can well attest to!

So, are there reasons to treat bivalves differently from certain terrestrial animals we like to eat, like chickens? Maybe. Chickens pretty clearly feel pain. Bivalves, well maybe not. So, in terms of the moral risk we’re taking, collectively, I’d rather see us eating bivalves than chickens—if, that is, we’re going to insist on eating animals of one sort or another.

But let’s stick to teleost fish: so tuna, flounder, mackerel, salmon, etc. Do pescatarians have good reason to think that, although they shouldn’t eat cows and chickens, it’s perfectly permissible to eat teleost fish like these? I suspect not. By the way, before getting into this, let me just flag one thing: I’m going to set aside any worries about whether your individual choices regarding what to do make any difference, and how, if they don’t, we should think about the question of what to do. Those reasons are going to hold as much for land animals as for fish—so if you’re already pescatarian, then I’m assuming that you’ve already thought about this. If you haven’t, you should really read Mark Budolfson’s wonderful and completely vexing work on these questions. To call them hard questions would be an understatement. They’re downright maddening.

Okay, so suppose you’re a moral pescatarian. So you already think it’s impermissible to eat at least some non-human animals, but it’s okay to eat fish. You probably already reject the idea that there’s something intrinsically special about being human such that we are non-eatable but everything else is fair game. Rather, you more likely think that whether it’s morally permissible to eat some species depends on certain characteristics of that species—probably mental characteristics, like being conscious, having a certain degree of intelligence, feeling certain emotions, forming loving bonds with others, or maybe planning for the future. The problem with many of these characteristics is that we can only gather indirect evidence for them. So, most people are happy to think that baboons feel pain because when they suffer bodily harm we can observe them reacting in ways that we recognize as being similar to the way that we react when we are in pain. Sometimes we can also rely on evidence regarding neural structure—or, more specifically, exhibiting a similar neural structure to our own. But that’s a tricky one, since it’s pretty plausible that many mental characteristics could be realized, physically, in a number of different sorts of structures. So this is less definitive than it has sometimes been made out to be. Anyway, I take it that most pescaterians must be thinking something like the following: in contrast to typical farm animals like chickens and cows, teleost fish lack the mental characteristics that would make it the case that we shouldn’t eat them. The question, then, is whether this is right.

What do we know about the mental lives of teleost fish? Well, that’s a remarkably large category of creatures, and we know a lot more about some than we do about others. But, basically, within that genus, there’s evidence of some very complex social behaviors, recognition of individuals even when non-conspecifics, tool-use, counting, and forward planning. There’s even some recent evidence to the effect that giant manta rays pass the mirror self-recognition test, which is some of the best evidence we’ve yet figured out how to gather regarding whether an individual is self-conscious. So, basically, there’s evidence that teleost fish are a hell of a lot smarter than we tend to think they are. Are all species of teleost fish so smart? We have no idea yet. But the more we look—the more that marine biologists have figured out how to test for and observe fish in their natural environments, which is a rather difficult task—the more evidence of intelligence we find.

For some people, I suspect that the real question is whether fish feel pain. That’s a tricky one, since some people also think that in order to feel pain, you have to be conscious or even self-conscious. But those are hard things to test for, and many creatures that we typically think of as feeling pain—dogs, for instance—won’t pass the mirror self-recognition test. So if that’s the gold-standard for self-consciousness, and if self-consciousness is required to be able to feel pain, then perhaps we’d have to conclude that we’ve been wrong about dog-pain. That doesn’t seem like something we should accept, however.

What we do know about fish is that they have receptors which are sensitive to bodily damage, nociceptors, though the sorts of nociceptors they have differ some from the sorts that one finds in mammals like ourselves. And the distribution differs as well. So we have these animals that generally can’t stop swimming (or they’ll suffocate) and that don’t have limbs with which to probe or protect an injured area—what sorts of behavioral modifications should we look for to see whether they’re in pain? Scientists mostly try to discern whether a creature is in pain by looking for stress reactions, but we’re not entirely certain what those would be like for fish. That makes it really hard to get good evidence here. Basically, my own thinking runs: okay, we have good evidence that many of these animals are fairly intelligent—more intelligent than some land animals to which we readily attribute pain—and they have the basic machinery necessary to detect bodily damage. Why would we expect for them not to feel pain if we are fairly certain that similarly-constituted land animals do feel it? It’s really sort of like making a calculated bet at this point. I don’t like the odds here, and I suspect that most pescetarians just aren’t reckoning them very well. The fact that fish are so different from us can deceive us into thinking that they must not be very morally relevant. But the evidence we have strongly suggests that at least some of them are at least as intelligent, social, and self-aware as many of the land animals that pescatarians think it important to avoid eating—possibly more so. Moral caution would seem to point us towards not eating them.

It’s probably also worth mentioning that the environmental and labor issues surrounding fishing are probably even worse than those surrounding the raising and killing of land animals for food. Say what you will about chicken farming, it’s a remarkably efficient source of protein. Horrific also, but efficient. Wild fishing really isn’t even that. It’s basically just hunting in big, gas-guzzling boats. Trawling is hugely damaging to the sea floor, and the labor practices of many fishing fleets are absolutely atrocious. So there are ample reasons to think that there’s parity between fish and land animals in this regard—not all of them having to do with the animals themselves. If one is worried about the horrible labor practices of many slaughter houses, as one should be, then one should be equally concerned with the atrocious labor practices of many of the world’s fishing fleets and fish processors.

In short, then, I suspect that most pescatarians are either ignorant of contemporary fish science, or just aren’t thinking clearly about the evidence. It’s easy to think of fish as sort of weird and alien and thus not worthy of moral consideration, but I think that’s a serious mistake. And once one considers the environmental and labor factors here, things become even clearer. It’s easy to romanticize fishing, but the fishing industry is anything but romantic. It’s huge, wasteful, and brutal—both to the fish it catches and to the human beings it employs. If one is swayed by such considerations in the terrestrial case, then I really cannot understand how one could think differently when it comes to eating fish.

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

EM: My part of the field is pretty thoroughly dominated by articles these days rather than books, so this is a harder question than you might think. I’ve already mentioned some texts that are rather important to me above—Naming and Necessity, for instance—so I’ll leave those aside.

I think that Gareth Evans’ Varieties of Reference, which I don’t think I’ve mentioned explicitly, is a pretty amazing book. I find more and more in there every time I come back to it, more hard problems that have just sort of fallen out of circulation, more really clever ways of trying to tackle those problems. Evans’ prose isn’t always my favorite, but I don’t think one can do much better if one is just looking to read a really interesting book in philosophy of language.

Beyond the Evans, I guess I really like system-building—even systems that I myself am inclined to reject. That’s one of the main things that can really make a book worthwhile, to my mind: having the space to put together all the little pieces that one so often has to shove into the footnotes in a paper. With philosophy, the devil is almost always in the details. And, sadly, the details are very often buried in the footnotes. So when someone has a rich system that they’re building up, you want the details, you want to see a book-length treatment where those footnotes get their own chapters, or at least a few paragraphs. Three books on language that wrestle with the details in ways that I admire are:

Kent Bach’s Thought and Reference,

Robyn Carston’s Thoughts and Utterances,

and Emma Borg’s Minimal Semantics. I vehemently disagree with almost everything in each these works. But I have a lot of respect for each; each is an attempt to both motivate and work out the details of a large-scale system for thinking about a bunch of related problems having to do with language and meaning. For Bach, the system is a sort of Fregean-cum-neo-Gricean one. For Carston, it’s relevance theory. And for Borg, it’s semantic minimalism. By the end of each of these, I think one ends up with a good idea of the motivations for adopting each of these systems, what they can do, and what their limitations are. I think these authors are all rather honest about the last of these—which, for me, is probably the single biggest reason why I admire each of their work.

Beyond the philosophy of language specifically, the work that’s probably had the biggest influence on my thinking is Derek Parfit’s Reasons and Persons. That book is just a tour de force of philosophy. It’s pushed me to think a lot harder about how we should act, as individuals, when we’re facing collective action problems that we can exert only very minimal influence over. That’s a question which, it strikes me, is of supreme importance to how many of us live in the modern world, and which is also just amazingly philosophically rich and engaging. I also just really admire Parfit’s style of thinking through philosophical problems. I don’t know that it’s rubbed off on me at all, but I keep sleeping with it under my pillow in the vague hope that it might.

Finally, I’m going to mention a piece of literature that I’ve been thinking a lot about recently: China Mieville’s The City and The City. I find it helps my state of mind to always be reading something other than philosophy on the side, but sometimes I find myself feeling that some bit of literature has, in some way I can never quite put my finger on, pushed my philosophical thinking forward. At the center of this work is a notion of ‘unseeing’, a mode of perception that the residents of each of the two cities mentioned in the title make use of when faced with residents of the other city with whom they must physically interact. It’s clear in the novel that this sort of perception involves, or at least feeds into, substantial self-deceptions that these residents are forced to undertake in order to go about their daily lives and avoid arrest by the secret police. I’ve generally been interested in both deception and self-deception for some time, but I think that Mieville has hit on something rather significant about both human experience and the way that we construct our mental lives. I think, for instance, that his notion of ‘unseeing’ may well be a pretty apt description of how many of us who live in big cities perceptually interact with the homeless populations we see on a daily basis, on our way to work or to the pub or whatever. Or perhaps how those in the countryside perceptually interact with the extreme poverty of transient labor. I’d actually be happy to talk about this quite a bit more, but I’ll leave it there.

You asked for five books. I gave you six. But, as a philosopher of language I am thankfully well-aware that giving six is a way of giving five. So there you go.

Richard Marshall is still biding his time.

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

End Times Series: the first 302

First published in 3:AM Magazine: Saturday, December 9th, 2017.