The Objects of Intentionality



1. The word “object” has a number of different meanings in philosophy, marking several kinds of distinction. When used in the phrase “intentional object” it expresses the idea of what a mental state is “directed toward”: the intentional object of my belief that London is large is London, since that is what my belief is about (plus being large). In this sense, everything and anything can be an object—properties, events, processes, Fregean functions, universals, numbers, gods and ghosts—since all these things can be what a belief is “directed towards”. Objects, in this sense, are simply correlative with the idea of content. They are not the same as “intensional entities”—such items as Fregean senses or intensions in the technical model-theoretic sense. They are not “abstract”, since London is an intentional object—what someone is thinking about. Nor do they necessarily exist, since I can be thinking about Zeus or Pegasus. And it would be equally wrong to suppose that they necessarily don’t exist, since London exists. The merely intentional object doesn’t exist, as that phrase is typically used—it is thought about but has no reality. Pegasus is only an object of thought. It is futile to ask after the “ontological status” of intentional objects: some of them are things like London and Bill Clinton—things that exist—while some are things like Zeus and Pegasus—things that don’t exist. The phrase “intentional object” does not single out some special type of object (in another sense of that word), but simply gives expression to the idea of what a mental state may be about. If you ask me what the object of my search is, I may say “the fountain of youth”—this is simply what I am searching for—and you can agree that this is what I am searching for even if you doubt the existence of such a thing. This sense of “object” could just as well be expressed by the words “target” or “focus” or some neologism like “intentionalandum”; there is certainly no valid move from describing something as an intentional object to describing it as an object in some of the other philosophical senses—as existent, or complete, or concrete, or whatever. This notion is customary in the philosophical tradition, being roughly equivalent to Husserl’s noema, and it has an established use in analytical philosophy. I am laboring it here because I am going to make heavy use of the notion in what follows, and I don’t want to be misunderstood; specifically, I am not going to be arguing for the utility of “non-extensional entities” like Fregean senses. If anything, intentional objects belong at the level of reference (though, of course, senses can also be intentional objects if they are thought about, as opposed to being the means whereby other things are thought about).

            The notion of an intentional object can be introduced in a slightly more formal way, as follows. Suppose a sentence of the form “aRb” is true, where “a” is the name of a person, “R” is a psychological verb, and “b” is a singular term: for instance “the Greeks worshipped Zeus” or “Bill is thinking about Susan”. Then we can say that the “b” part gives the intentional object of the person’s attitude. To speak of an intentional object is just to speak of the truth of such a sentence, so that if such a sentence is true then it follows that the person’s attitude has a certain intentional object; in other words, the notion captures certain grammatical facts. If I say that “b” specifies an intentional object, I am simply saying that there is a true sentence of the form just outlined. I don’t myself think it is particularly useful or illuminating to formulate the notion in this formal way, but it serves to pin it down for those who suspect the notion of harboring some ominous ontological agenda (say, to resurrect Meinong, as they understand him--or misunderstand him).

            I need to make another distinction, between de re and de dicto intentional objects. Again, this is not a distinction between classes of object, since the same objects (in the other sense) can be now de re and now de dicto intentional objects; the distinction has to do with the nature of the intentional relation—or the truth conditions of intentional sentences, if you prefer. Suppose I am aiming my bow and arrow in the direction of a field, and in fact my father is standing in my line of fire—but I don’t know it is my father: then my father can be said to be the de re intentional object of my aiming—he is the object that my aim is directed (literally) toward. Why would I be aiming at my father in this way? Because my father looks to me like a tree stump and I am doing some harmless target practice. If you ask me what I am aiming at I will say “that tree stump”, pointing my finger in the direction of my father—this is the de dicto intentional object of my aiming. The de dicto object obviously corresponds to the content of my intention—how I am representing the world. And of course this object does not exist, since there is no tree stump in my line of fire; I have simply mistaken my father for a tree stump. My father exists and is the de re object of my mental state; the tree stump does not exist and is the de dicto object of my mental state. If we ask what I am aiming at in this situation, we get two answers—a man and a stump--only one of which is a real object. In one sense I am aiming at something that does not exist; in another at something that does. This point will be important later, when I argue that this kind of case is a lot more common than people have thought.


2. I now want to become a little more controversial. I am going to sketch a portrait of non-existent objects—which many philosophers find repugnant. I am not going to defend this notion here, though that has been done elsewhere. I don’t actually think the sketch ought to be very controversial, once it is clearly understood; but experience has shown me that many people find the idea of non-existent objects intensely distressing. My purpose here will be to put this idea to use in unexpected ways, not to provide a full justification for it; so the paper will have only a conditional interest for those subject to the distress just mentioned.

By a non-existent object I mean such things as Zeus or Pegasus or Vulcan or Allen Smithee—things that are thought about, maybe even believed in, but which don’t exist. These are objects of thought, but not existent objects of thought; and perhaps it will ease some of that distress if I observe that “non-existent object” is simply short for “non-existent intentional object”. Some philosophers think to make some argumentative headway here by asking, in a sneering tone of voice, “How can there be non-existent objects?”--clearly intending some other notion of object. But this rhetorical question loses all force if we ask instead “How can there be non-existent intentional objects?”--since that cancels the other sense of “object”, in which existence is already presupposed. There is obviously no strain whatsoever in saying “The object of Jones’s search is entirely nonexistent”—that is, the intentional object of his search does not exist. In any case, I intend to be speaking of intentional objects that don’t exist—corresponding to sentences of the “aRb” form in which “b” is an “empty” singular terms. These are simply cases in which “aRb” is true but “b exists” is false (obviously, then, I reject the idea that a sentence of the former form can be true only if a sentence of the latter form is). Of course, there are a great many cases like this, whenever so-called “empty terms” occur in the language. And the source of these cases are such familiar phenomena as misperceptions, hallucinations, dreams, mental images, deceptive verbal testimony, novels, theoretical error, and so on. Many of our intentional objects fail to exist, and life goes on regardless.

            Let me then simply state what I take to be the chief characteristics of non-existent objects. First, they are not mental—except, of course, when they are (for example, that non-existent pain I deceived you into believing I had yesterday). Zeus is not a mental object, but a divine object—which fails to exist. No doubt Zeus’s properties are conferred by human acts of mind, like the characters in a novel; they certainly don’t come from his mind-independent reality. The properties of fictional characters are the upshot of creative mental acts, but it would be wrong to assimilate such characters to mental entities like beliefs and intentions—Zeus is not himself a belief (otherwise he would exist). Second, non-existent objects are not senses, even if these are taken to be non-mental: I am not speaking of the sense of “Zeus” when I speak of him—and if I were, again he would exist. Such objects belong in the realm of reference. In fact, I think there is no problem in saying that “Zeus” refers to Zeus: this is perfectly correct and natural English. And it surely follows from the fact that the Greeks worshipped Zeus that they referred to him—for how could they worship him (have that mental attitude) unless they could make him the object of their thoughts, and hence refer to him?

            It is possible to make true statements about non-existent objects: that they don’t exist, that they are sometimes believed to exist, that they have such properties as being divine, irascible, and powerful. Indeed, I think anything can be truly predicated of them except existence: Sherlock Holmes is a detective, a man, a concrete entity (not an abstract one), has causal powers (he is a noted pugilist), lives in London, and so on. He has the properties he is represented as having, though only by virtue of being represented as having them. He is a perfectly ordinary individual—except for the small matter of not existing. In other words, there are true subject-predicate statements with “Sherlock Holmes” as their subject term.

            Non-existent objects are no more “bundles of qualities” than existent objects. It is not that they somehow reduce to their properties, while existent objects do not: they instantiate properties, they don’t collapse into them. It is often argued, correctly, that objects cannot be sets of properties because that would make them abstract entities, and they are not; but this point applies equally to non-existent objects. Also, of course, if they were such sets or bundles, they would have to exist, since the sets or bundles do—but they (the objects) don’t. The non-existence of these objects amply distinguishes them from their existent fellows; it is a symptom of confusion to try to find some other way to distinguish them—and these ways always end up negating the very non-existence they are trying to capture (that they are mental or are senses or reduce to sets of properties).

            Non-existent objects exhibit a marked degree of indeterminacy: not every predicate either applies to them or its negation does. This is a very familiar point: it cannot be said either that Hamlet has a mole on his left shoulder or that he does not; the number of lines under his eyes is utterly indeterminate; the color of his underwear is not just vague but radically unsettled. This is really a result of the conferred status of the properties of such objects: when an object is merely intentional there is nothing to confer properties on it but the mental acts of its entertainers. Non-existence and indeterminacy go together.

            There is also a clear sense in which there is no appearance/reality distinction for merely intentional objects: they have no properties beyond, or other than, those bestowed upon them. As a result, we cannot make sense of widespread error about their properties—it cannot be that we have got it all wrong about Zeus. This is because what Zeus is consists in what we take him to be, so he has no nature that might diverge from the one we attribute to him. Of course, there is room for local errors, based on false testimony, mistranslations, and so on; but there is no room for the idea that everyone had it wrong all the time—any more than Conan Doyle might be fundamentally mistaken about his own fictional creation. There is thus no “reality” apart from the appearance—truth and believed truth cannot be pulled apart. Non-existent objects are individuated by they way they are represented, while existent objects can generally be quite other than the way they appear and are described.

            I have said that non-existent objects can be referred to; so semantic predicates apply to them. There are truths of the form “x refers to y” even when y does not exist. Presumably, it is obvious that this precludes a perfectly general causal theory of reference--if this means that the object of reference has to play a causal role in producing instances of the name (or some such). A causal theorist of this (simple) kind will have to contest my claim about reference and non-existence. I do not intend here to contest his contestation; I am going to assume that reference to the non-existent is possible, and that as a consequence a (simple) causal theory of reference is wrong for such reference. My claim will be that such reference is extremely widespread, and is in fact the primary case.

            What we need to take from this brief discussion of non-existence is that some intentional objects don’t exist but that otherwise they are just like existent objects. It is not that I accept an “ontology” of non-existent objects—if that means that I believe that they enjoy some sort of existence. I accept no such thing: merely intentional objects do not exist in any sense—that is their whole point. What we have is terms in our language that refer to such non-existent objects of thought: that is all. I am separating reference from existence, not claiming that we have reference to some special exotic brand of quasi-existing entities. Nor am I introducing some such notion as subsistence: I am using the single notion of existence and saying outright that some intentional objects lack this property—tout court. It is not that there are two types of “being”, one stronger than the other; there is just being and non-being.


3. With these preliminary matters clarified, I shall argue now that reference to the non-existent is central, ubiquitous, harmless, and theoretically illuminating. Non-existent objects are not things to be afraid of—to quarantine and banish—but things to welcome and put to work; they are our friends not our enemies. They are not merely peripheral anomalies, to be disposed of and disrespected; they are the foundation of all intentionality—the basic case. I expect this to sound shocking now, but by the end of the paper I hope to have rendered the claim banal and innocuous (or at least tolerable). The claim to be established, then, is that whenever there is intentionality directed toward an existent object there is intentionality concurrently directed towards a non-existent object, so that reference to the non-existent is always embedded within reference to the existent; but the converse does not hold. Moreover, there is a sense in which reference to the existent is dependent on reference to the non-existent.

            There are two ways to argue for this claim: the argument from error, and the argument from individuation. The former way is less interesting, so I shall spend less time on it. In the case of my father and the tree stump the non-existence of my de dicto intentional object results from the perceptual error that characterizes the case: there is simply nothing there that is a tree stump, though it seems to me that there is—so “that tree stump” does not refer to an existent thing (though it does, according to me, refer to a non-existent thing). I have made a mistake of existence, supposing that something exists that does not; this arises simply because I take a predicate to be satisfied in my environment that is not so satisfied, namely “tree stump”. In effect, my perceptual state affirms the proposition that there exists a tree stump in my visual field, but there is nothing of that kind there, so my perceptual state is incorrect; hence if I were to give a name to what I seem to see, that name would fail to refer to an existent thing. This is to say that in a case of error my intentional object (or one of them) fails to exist—I have a non-existent intentional object. Now suppose that such perceptual error were widespread, even universal; then it would follow that the majority, or even all, perceptual states introduced non-existent intentional objects—in addition to the real ones that constitute the de re objects of perception. Some predicate would be affirmed to be instantiated in the sensory field, and this affirmation would be false. But is there such general error in perception?

            Well, a case can be made that there is; but I am less interested in this than in what follows from it—for it would entail my main claim. Three potential sources of error are worth considering, all familiar: solidity, color, and constancy. Suppose that objects are not solid, but that they look solid; they have gaps between the atoms, but they look to be continuous substances—a respectable, if highly debatable, view. Then all perception would affirm that there is a solid object out there, but that would be false, since solid objects do not exist. Thus existential error would infect every perception, and any singular terms employed to denote these ostensible solid objects would be existentially empty—“that solid object” would never have an existent referent. Therefore the de dicto intentional object of these perceptions would never exist—just like the case of my father and the tree stump. Or consider color: if you hold any view according to which the perception of color is in error, because the way color looks does not match the way it is, then the same result will ensue. Thus, if color looks non-dispositional and non-relative, yet is really dispositional and relative, then all color perception is of non-existent objects (as well as existent ones—the de re objects). Color perception affirms that there is a non-dispositional and non-relative color quality instantiated there, but there is not—there are no such objects in the world. You do see existent objects, but you see them as having qualities that they don’t have; hence you also “see” an intentional object that fails to exist. There is a true existential proposition here and a false one, and the false one generates non-existent intentional objects. Putting it in the formal mode: for any sentence of the form “aRb”, in which “b” refers to an existent object, there is a sentence of the form “aRb*”, in which “b*” refers to a non-existent object (or fails to refer to an existent object, if you prefer)--where “R” is some perceptual verb like “see” or “seems to see” or “visually represents”.

The same point can be made with respect to perceptual constancy—of distance, shape, color, or whatever. If things in the distance look smaller than they are, then perception contains an error as to size; and this will give rise to false existential propositions--such as that there exists something in the visual field that is both a person and smaller than the coffee cup on my desk. Now I know that many people will resist this description of the case, holding that people in the distance do not look smaller than nearby cups—after all, I have no tendency to judge that they are thus smaller. But I believe this is wrong: I think distant things do look smaller (in one good sense of “look”)—it is just that we automatically correct for this on the basis of background knowledge. Distant objects indisputably take up less of the visual field, and I think this translates into a difference in how they look; it is just that we don’t take how they look all that seriously, since we know that distant objects tend to look smaller than they really are. Anyway, my point here is that if you agree with this account of constancy, then you should agree, too, that perception introduces non-existent intentional objects all the time, since it is riddled with illusion—in this case, that things get smaller the further away they are. A nearby gun shot sounds much louder than a distant one, though we don’t judge the sounds to be objectively different in volume; but it is nevertheless true that there is an error built into these auditory perceptions—since the sounds are not in fact different in loudness, though they appear to be.

So these relatively humdrum examples of perceptual error actually entail my main claim; that claim ought not then to seem as dramatic as it might have appeared on first statement. But it would be far more interesting if we could establish the claim without assuming these controversial theses of perceptual error, and this is what I propose to do next. I shall do this by considering temporal and counterfactual cases in which plausible individuation conditions for intentional objects force us to introduce non-existent objects in addition to existent ones. First, then, consider the case of Fido and Rex, a real dog and an unreal one. At time t our subject S is seeing Fido in a perfectly normal way: Fido exists and is causing S’s experience as of a dog meeting Fido’s description. Now suppose that we remove Fido from the scene, but that we preserve S’s (qualitatively defined) experience—we use some deftly placed electrodes in his cortex or whatever. Let us suppose that S has been referring to Fido and carries on using the name introduced when Fido is removed. Now we, knowing what is going on, will not use S’s doggy name as he does, because we know he is no longer in the presence of Fido; we know that S is now speaking of a non-existent dog when he says “that dog”. Suppose we call this non-existent dog “Rex”, being well aware that Rex is not identical to Fido (S keeps using “Fido”). Then we can say that “Fido” and “Rex” have different references, one existing and one not. S’s intentional object is Fido at time t and Rex at the later time. The question I am interested in is whether “Rex” has reference before the removal of Fido. Suppose that we re-introduce Fido after an interval: does “Rex” still refer after this moment? In other words, does S’s experience have two intentional objects during the Fido periods, both Fido and Rex? One description of the case says that when Fido is removed a new intentional object enters into S’s experience, namely Rex, and that when Fido comes back that new object ceases to figure in S’s experience, being replaced by Fido. Another description—the one I favor—says that no new object enters S’s class of intentional objects when Fido is removed, and there is no subtraction of any object when Fido comes back: Rex is S’s intentional object all along (as well as Fido when he is around). Restoring Fido as S’s de re intentional object does not make Rex cease to be an intentional object of S’s—Fido adds, he does not subtract. I think this sounds intuitively right, but why is that? Because there is a qualitative continuity in S’s experience: it seems to him just like there is a single dog throughout the period. If the whole period had consisted of hallucination, then we would have said that Rex was there throughout; but making both ends of the period veridical doesn’t abrogate the continuity that encourages this description of the case. And isn’t it extremely odd to suppose that adding Fido could eliminate an intentional object? How could Fido supplant Rex? But if Rex is S’s intentional object after the period of hallucination, is he not also S’s intentional object before it—since there is an experiential continuity throughout? Didn’t S have an experience as of a single dog the whole time—though, of course, no such single dog exists? But if so, his intentional state took a non-existent dog as its intentional object.

Here is another case: You see what you take to be a man crossing the street, but in fact this is a rapid succession of distinct men cunningly arranged by God to look like a single man. God splices all these distinct men seamlessly together so that there is an appearance of manly unity. You call the man you seem to see “Bill”. The de re object of your seeing is a succession of (say) ten men, but there seems to be a unity to the object of sight, which is why you introduce the one name. I suggest that in this case there is a single (de dicto) intentional object, as well as ten (de re) intentional objects; and this single object does not exist—there is no unique man who performed the sequence of movements you observed. Your perceptual experience affirms that there exists a unique man who crossed the road, but no one did—Bill does not exist. So the identity conditions for your de dicto intentional object are not the same as for the de re object(s): it persists over time in question, while they do not. If this were happening to us all the time, then we would constantly be presented with non-existent intentional objects of perceptual experience.

The logic of these arguments resembles the well known case of the statue and the piece of clay that composes it. The clay is made into a statue and then reduced to a mere lump again. Does the lump persist throughout these changes? It would be extremely odd to say that the creation of the statue put the piece of clay out of existence, and there are good reasons to distinguish statue from lump. In the same way, if we start with Rex, then add Fido, and finally remove Fido again, leaving Rex to bark once more, it seems odd to suppose that Rex disappears during the middle phase; rather, we have Rex and Fido in the middle phase. Both of these objects are present in S’s total intentional state: if the non-existent object is an intentional object in the phase when the experience is non-veridical, then this object is not eliminated when the experience turns veridical and a new intentional object is added. And these two objects cannot be identified, since one exists and the other does not. Imagine that all your experiences up to now were hallucinatory—hitherto you have been a brain in a vat—but now you are suddenly hooked up to the world normally and actually see real things. Can we plausibly maintain that your erstwhile non-existent intentional objects are now no longer objects of thought for you, having been replaced by existent objects? How can adding existent intentional objects subtract from your stock of intentional objects? What happens, rather, is that hooking you up to real things creates a condition of double reference in you.

Now for the crucial counterfactual case: the argumentative structure is basically the same as the temporal case. Suppose you are a normal perceiver in the actual world—you perceive what you seem to perceive. Now consider the counterfactual world in which you are a brain in a vat but have the same experiences as in the actual world. In that possible world you have a range of intentional objects that do not exist but for which you may have names. So in the actual world you have names for existent objects, but in the counterfactual world you have names (that sound just the same) for non-existent objects. Question: do you also have those non-existent intentional objects in the actual world? In other words, does removing the existent objects create new intentional objects in the counterfactual world? My suggestion is that those non-existent intentional objects were already present in the actual world, alongside the existent intentional objects. Intuitively, if I think away the existent object of my mental state, I simply reveal an intentional object that was there all along; I don’t create a fresh intentional object just by subtracting the existent object. And this is basically because the merely intentional object is qualitatively individuated (perhaps relative to a person), and these qualitative conditions obtain even in the veridical situation. We can put the same point in terms of reference: I refer to non-existent objects in the counterfactual world, existent ones in the actual world—but do I also refer to the same non-existent ones in the actual world? I think I do--and if so there has to be a kind of double reference in the actual world: my names have to refer to both existent and non-existent objects concurrently.

The argument I am giving can be compared to a similar view of senses and experiences themselves. If we subtract reference from sense we don’t eliminate sense: it would be very odd to say that senses are created in the counterfactual situation in which references are removed; better to suppose that there is both sense and reference in the actual situation. It is not that we invoke sense only when reference fails: it is common to both referential success and the lack of it. Similarly, we don’t suppose that experiences exist only when they are non-veridical: removing the object doesn’t create an experience; it simply leaves as a residue what was there already when there was a perceptual object. I am saying the same about non-existent intentional objects: they are not created by the removal of existent ones; rather, they were there all along, capably of being referred to. So, right now, I have a range of non-existent intentional objects in my mental sights—those that I would have if all my experiences were non-veridical. Indeed, for all existent intentional objects I now have, there is the non-existent object I would have if the existent object were not there—and I have this non-existent intentional object as things actually are. Put brutally, I am now experiencing an existent coffee cup and a non-existent one—the one I would be experiencing if there were no existent cup to be seen. And whenever I use a name to refer to a real person I am also referring to an unreal person—the one I would be referring to if the real person did not exist. This follows from the simple principle that removing intentional objects from a person’s mental landscape does not create new ones; it simply exposes to view what was there already.

I hope it is clear that my ubiquitous non-existent intentional objects are nothing like sense-data. Sense-data are mental entities that exist, but objects like Pegasus or an hallucinated tree stump are not mental and do not exist. Nor are they senses or concepts or modes of presentation or ideas; they are what such items are about—they belong to the level of reference (albeit non-existent reference). Merely intentional objects are just like ordinary objects, except for the small matter of their existence (as I observed earlier). I am emphatically not making the platitudinous point that in any intentional act there is both the object of the act and the act itself. My claim is that there is always a duet of intentional acts, each taking different objects (one of them non-existent). And my general thesis is that reference to such non-existent entities is actually the normal case, not a symptom of malfunction or abnormal circumstances: we are intentionally directed to non-existent objects all the time—without making any mistakes of existence.


4. What are the consequences of this view for the nature of perception and reference? As I have said, we now have a kind of double intentionality as a regular occurrence. But it would be wrong to suppose that these distinct objects are independently represented by the mind; the two references are clearly connected. Compare this with the token-reflexive account of indexicality, according to which an indexical term refers to itself in the course of referring to an extra-linguistic object. For example, “now” means “the time of this utterance of ‘now’” and “I” means “the speaker of this token of ‘I’”. The indexical term refers to an object by referring to itself; it asserts a relation between the object and itself as an utterance. There is a kind of concealed double reference here, but the two referents are related. Or compare seeing an object: we can be said to see the surface of the object as well as the object as a whole; so there are two objects of sight. But they are not independent objects, since we see the whole object by seeing the surface of it. I think that in the case of reference to non-existent objects there is a similar dependency, or better interdependency. An existent object causes me to see it; as a result I have a specific type of experience, which takes a non-existent object as its intentional object, in addition to the real object that I see. So it is in virtue of seeing the real object that the unreal one becomes an object of intentionality for me. But, on the other hand, I could not see the real object without also seeing the non-existent object; after all, that object can be present to me even when I am hallucinating. It is this act of intentionality, combined with a suitable causal relationship, which delivers the perception of the existent object; the latter piggy backs on the prior intentional relation to a non-existent object. We could define the real object of perception simply as that object which (suitably) causes in the subject an intentional state directed toward a corresponding non-existent object. In the veridical case these two objects will closely resemble each other—they will share many properties. The non-existent object that corresponds to the real cup I am seeing will also be white, shiny, cylindrical, etc. So the two are not unrelated objects, picked out by separate acts of intentionality; they are intelligibly connected. We might, indeed, speak of them as counterparts of each other—closely similar objects that differ in that one exists and the other does not. They certainly have an identical appearance, since non-existent objects can easily be mistaken for existent ones—that, after all, is what hallucination is all about.

            This way of looking at perception provides an answer to the question of what is in common between qualitatively identical experiences one of which is veridical and the other is not. The experiences are the same, to be sure, as is their representational content; but that isn’t all—they are also directed to the same non-existent intentional object. I myself have no problem with saying that this object is seen in both cases, so that non-existent objects can be seen (I don’t think sense-data can be seen, however, because they are mental items); and therefore I am happy to say that in both the veridical and the illusory cases the same object is seen. It is just that in the veridical case another object is also seen, and in just as full-blooded a sense—namely, the existent causally operative object. So my view is that two objects are seen in the veridical case and only one in the illusory case; but there is a highly significant “common factor” here because an identical object is seen in both cases. Sense-datum theorists made the mistake of saying that a common mental item is seen in both cases: that is wrong, but their instincts weren’t so far off, since there is something that is a common object of seeing in the two cases—a non-mental non-existent (material) object. This view gives no support to the idea that the objects of perception are mental entities, nor does it claim that existent material objects are somehow not really seen. There is no forced choice here: we can be said to see both—just as we can be said to see both the surface of an object and the whole object. It is of course true that on this view we can be said to be perpetually hallucinating, even in the most veridical of cases, since there is always a non-existent object of sight in the offing; but once it is appreciated that this is not incompatible with the idea that we are also seeing an existent object the sense of outrage ought to dissipate.

            Theories of reference will look rather different in the light of this conception. What theory of reference works best for reference to non-existent objects? Not a causal theory, as I noted earlier. In fact, since such objects are individuated in terms of our beliefs, some sort of description theory seems natural. Of course, a social dimension will need to be included, which leaves room for error about the object, but at bottom the community will refer by means of suitable descriptions. Zeus has to fit the descriptions that initially introduce his name, even if later speakers are misinformed about his properties (they come to think he is vindictive when he is benign, say). So Russell and Frege were basically right about names, if we construe them as talking about non-existent objects of reference (not that they were). The causal theory will figure in roughly as follows: the existent reference of a name is defined as that object that causes reference to a non-existent object, which is itself a matter of description-fitting. Thus, if I use the name “Saul Kripke” I refer to the real Kripke in virtue of the fact that he is the cause of my using the name to refer in addition to Kripke’s non-existent counterpart--which latter reference is descriptively mediated. If there is no real Kripke (he is the result of an elaborate hoax), then I refer only to this non-existent object; but if there is, then he is the cause of that other act of reference. Looked at this way, the causal theory is hardly fundamental: I can refer to things without the aid of a causal relation to them (if they are non-existent), and when causality does come in it piggybacks on description-based reference. Not that the causal theory of names is wrong; it is just not as central as might have appeared. It seemed central because we didn’t think much about reference to the non-existent, and we didn’t recognize the ubiquity of such reference.


5. Is reference environmentally determined? Straightforward reference to the non-existent can hardly be, since the environment cannot contain non-existent objects that interact with thinkers and speakers. We can perhaps construct a twin earth case for a term like “Pegasus”, by stipulating the same social history of the term and so on; but if the uses of the name in the two communities refer to different non-existent objects that will not be because of any impact of the non-linguistic environment on speakers (it may simply be a function of the difference of speech communities). The reference of the term can exert no causal control over tokens of the term, so this kind of causal relation cannot tie the name down to two distinct environmental referents. We normally allow that different tokens of “Zeus”, uttered by different people at different times, can have the same reference, but it is less clear what to say about different (causally isolated) communities on earth and twin earth—perhaps this is enough to make us say that distinct non-existent objects are referred to, perhaps not. In any case, this will not be a classic twin earth case.

            What about terms that do refer? Does “water”, which has distinct existent references on earth and twin earth, also have distinct non-existent references? According to me, “water” on earth refers to water (an existent substance) and to a suitable counterpart non-existent object (not identical to water, obviously): but is this the same non-existent object that my twin on twin earth refers to? Questions about the identity conditions of non-existent objects can be tricky, but I think a case can be made that they are the same. Let us suppose, plausibly, that “water” is associated with the description “the (unique) watery stuff”, on both earth and twin earth (where “watery” here just expresses the common appearance of the two liquids on earth and twin earth). Then the qualitative facts about the terms are identical—the speakers are in the same qualitative state when they use their term. But since non-existent objects are qualitatively individuated, the two sets of speakers will be referring to the same non-existent object. Think of it this way: Consider what the two sets of speakers would be referring to if H2O and XYZ did not in fact exist on their planets (it is all a huge illusion). These non-existent intentional objects would be associated with exactly the same qualitative states, and so would naturally be identified. Even if we let the identity of the communities figure in the individuation of these objects, it is still clear that the actual existent substances don’t play a constraining role in tying down the reference of “water” to these non-existent objects. In short, the non-existent reference of “water” is not environmentally sensitive—unlike the existent reference. So there is no twin earth case for such reference, and hence externalism is not true of it. (The identity conditions of the non-existent referents mirror those of what some people call “narrow content”.)

            This limits the scope of externalism. If we think of reference in terms of an inner and an outer shell, then the inner shell is not subject to externalism, while the outer shell still is. Externalism is never true for the inner shell, no matter what kind of term we choose, since this kind of reference is environmentally insensitive—it coincides pretty much with the speakers’ qualitative states. There is broad reference and narrow reference--the kind that varies from earth to twin earth, and the kind that does not—and the narrow kind is to the non-existent. This kind of reference is “in the head”—that is, supervenient on brain states and so on. The other kind of reference is not, being dictated by the existent environment. I take it that this corresponds to the intuitions that most people have had about twin earth cases; I have merely re-cast those intuitions in terms of a covert reference to the non-existent on earth and twin earth. That is what the semantic overlap consists in.


6. It might be wondered whether the apparatus of non-existent objects of reference could replace the notion of sense. Can we say that a pair of co-referring names also refer to two distinct non-existent objects, and that these objects account for the semantic difference between the names? That might seem tempting once we decide that non-existent objects are individuated qualitatively, since different descriptions will give different such references. Consider what we would refer to with “Hesperus” if Hesperus did not exist, and similarly for “Phosphorus”: it is doubtful that these would be the same thing, since the two appearances will generate distinct non-existent objects. So might not these references do the job customarily assigned to sense? This might seem like an attractive theoretical simplification, but I don’t think it will really work. The problem is that there can be informative identity statements between terms for non-existent objects, so that we will still need a layer of sense, over and above reference, to capture these cases. Clearly, we can assert things like “Clark Kent is identical to Superman”, and these can be both informative and true; so we need two senses to account for such cases, and there is only one reference to work with here. Distinct descriptions don’t always generate distinct references to the non-existent. So senses can be sliced more finely than non-existent objects, and hence cannot be replaced by such objects. Accordingly, the semantics I favor includes two levels of reference and a level of sense (which may itself have more than one level): a name will refer to an existent object and to a non-existent one, as well as expressing a sense.

            How do truth conditions work out under this kind of semantics? Less simply than we usually suppose, and contextual rules will need to be invoked to capture the way we normally evaluate sentences for truth. Take “Kripke exists”: is that true? Well, it is if we take the name to refer to existent Kripke; but not if we take it to refer to Kripke’s non-existent counterpart—the object we would be (solely) referring to if Kripke didn’t exist. Generally, we ignore this latter truth condition, and indeed there seems to be a fixed rule that excludes these non-existent counterparts from figuring in truth conditions. It might even be best to restrict the ordinary notion of reference to the former relation, holding only that there is always another intentional object accompanying the existent one—but this object is not strictly referred to (where reference and truth are taken as correlative). From my point of view this is a minor issue: what matters is that there is always a non-existent intentional object for any existent one, whether or not this is what our ordinary terms are taken to refer to. Theoretically, we could stipulate that all our terms primarily refer to the non-existent objects, so that existential statements all turn out false; and then recover the usual assignments of truth-value by means of a rule linking the non-existent to the existent object—i.e. take the existent object that is the counterpart to the non-existent one and compute truth conditions accordingly. We simply need to coordinate the two possible truth conditions in the right way. This is more complicated than we are used to, but semantics has a tendency to get more complicated than we thought--does it not? Certainly, the role of context in determining the intended truth conditions is well established; the sentence by itself is apt to underdetermine its own truth conditions in use.


6. So far I have focused on reference to ordinary physical objects, and my claim has been that such reference is always accompanied by a non-existent intentional object. But is this true for all types of reference? Interestingly enough, it is not. In particular, it is not true for reference to mental entities—and also, I think, not for reference to selves or abstract entities. Suppose I say “that pain”, referring to a sensation of pain in my leg: do I also refer to some non-existent counterpart to this existent pain? The problem with saying this is that we cannot make sense of the idea of what I would be referring if the pain did not exist, since I cannot make that kind of error about my pains. If we take all my actual references to my experiences and consider what to say about what I would be referring to were I a brain in a vat, then we find that these references are preserved in that counterfactual situation, since by hypothesis the brain in a vat has all the experiences that I do. Appearance and reality cannot come apart in this case, so we cannot generate a non-existent object by keeping the appearance constant and removing the reality. Accordingly, when a mental entity is the intentional object there is no argument to the effect that there is a counterpart non-existent object in the counterfactual case that must be admitted even in the actual veridical case. There really is only one intentional object here. And that is because mistakes about existence are not possible for such cases.

Much the same should be said about selves and numbers: whenever mistakes about existence are impossible we cannot produce a scenario in which we have the appearance without the reality, and so we cannot establish that another non-existent object is needed. There is nothing in the idea of what I would be referring to with “I” if I did not exist! So the type of argument I gave for postulating extra non-existent intentional objects has no purchase here. Putting it in the formal mode again: if our sentence form “aRb” contains a psychological relation and a reference to a mental entity (etc), then such a sentence cannot be true unless “b” refers to an existent object. Thus, I cannot be said to feel a pain unless the pain exists—though, of course, I can be said to worship a god who lacks all existence. So these are very different cases, and my claim applies only to the latter kind of case. We are always hallucinating non-existent material objects, I suggested (putting it provocatively), but we never conjure up non-existent mental states (in our own case). But I take it that this difference in the applicability of my claim should come as no surprise, in view of the antecedently recognized difference between intentionality directed to the mind and intentionality directed elsewhere—specifically, to the external world. Real existence is never entailed by the appearances in the case of the external world, but it is in the “mental world”. And this is why my argument has no application to the latter type of intentionality. It is the possibility of error that produces the non-existent counterpart intentional object.

            If all this is right, then there are two main types of intentionality: the kind that builds in reference to the non-existent and the kind that does not. When we have intentionality directed toward material objects we have the former type of case; when we have intentionality directed toward our own mental states (and selves and numbers) we have the latter type of case. So the latter type of case fits pretty well the conventional view of how reference works and what the objects of reference are (though, ironically, this has been the area in which some philosophers have wanted to deny that genuine reference takes place). But in the former type of case we seem to have a radical departure from received views, in that non-existence turns out to be central and universal. There are a lot more intentional objects than we thought, and they show a disturbing tendency not to exist. Disturbing, that is, until you have grown accustomed to their company; I now find them nothing to be alarmed about, even to be quite friendly. Brentano said that all consciousness is consciousness of; I agree, but would rephrase: all (or nearly all) consciousness is consciousness of nothing—that is, of the nonexistent (though also of something—the existent). I hope this sounds less outrageous now than it did when I started. It is really just a consequence of the fact that consciousness presents its (existent) objects in such a way that were they not to exist consciousness itself would not change its intrinsic character. It is as if the existent object fills a slot already prepared for it by consciousness, and this slot consists in reference to a non-existent object. When it comes to the essence of consciousness, then, non-existence is dominant over existence. In its intrinsic nature consciousness is primarily intentionally directed to what is not.


Colin McGinn