I introduce the neologism of the title to refer (!) to a particular type of philosophical position—namely, one which contests whether a particular class of expressions is genuinely referential. It is supposed that there is apparent reference for the class in question but the position denies that such apparent reference is genuine reference. Thus we are subject to delusions of reference—or there is illusory reference. Reference occurs more narrowly than is generally assumed; it is harder to bring off than we tend to suppose. Reference is demanding and its demands can fail to be met, even when we naively think they are met. Here are some examples: reference to the non-existent, reference to abstract entities like numbers, reference to elementary particles and other unobservable entities, reference to selves (“I”), feature-placing sentences (“It’s raining”), definite descriptions (Russell), predicates and properties, logical connectives (Frege), any referring terms at all (Quine), ethical terms (expressivism), mental terms (eliminativism). In all these cases it has been supposed that the conditions for reference to occur are not satisfied—thus we have anti-referentialism (compare anti-realism). 
Let me distinguish three types of reason for claiming anti-referentialism. The first is simply that we cannot have reference without the existence of what is referred to: if a speaker refers to an object x, then x must exist. Thus whenever the existence of certain entities is denied we have anti-referentialism: fictional names, numerals and numbers, denials of the existence of the self or particles, denials of the existence of mental states. The expressions in question would be referential if the entities existed, but they don’t. By contrast, we have a second type of view: there is no question that the denoted entity exists, but the term for it is not really semantically a singular term. Thus, according to Russell’s theory of descriptions, there is a queen of England but “the queen of England” doesn’t refer to her—because it is simply not a genuine referring term. It quantifies, not refers. This is a bit like the “it” in “it’s raining”: it looks syntactically like a referring device, but that appearance is misleading—the word merely enables us to form a complete sentence without having any reference in itself. There is also the view that abstract singular terms, like “triangularity”, are not referring terms, being mere shorthand for the corresponding predicate (“triangular”). Third, we have the view that the necessary conditions for a word to refer to an entity are not met, even though the word is name-like and the entity exists. Thus you might have a name for a person and there are persons (including that one) but you cannot fulfill the conditions for achieving reference to that person, because you don’t have enough identifying information to single the person out from all others. Or it might be held that numbers exist and we have names that purport to stand for them, but the absence of causal relations between speakers and numbers precludes our referring to them, granting the correctness of a causal theory of reference. There is nothing wrong with the entities ontologically and our putative names are really names, but it is just that the conditions that enable reference are not met—we can’t refer to those entities with those names.
I am particularly interested in the third type of reason for anti-referentialism. We can envisage several types of rationale for adopting such a view, none of them uncontroversial: in addition to the two just mentioned (problems of identification and problems of causation), there are empiricist worries (we need sense experience of the entity) and worries about ostensive definition (we can’t point to the entity). Then too we have more radical and general worries, such as Quinean indeterminacy of reference and Kripke-Wittgenstein skepticism about reference—no reference is possible. What we need is some clear and uncontroversial theory of reference so that we can decide whether a particular case meets the conditions for genuine reference. I don’t have any such theory, but I do think it is important to distinguish two possible sorts of theory: restrictive theories and liberal theories. Restrictive theories will make reference difficult to achieve, so that only in certain special cases will we have genuine reference; while liberal theories will make reference easy to achieve, so that reference is virtually unlimited. The first type of view is typified by empiricism—as in Russell’s position that only entities with which we are directly acquainted can be genuinely referred to (“logically proper names”). The second type of view would allow that any entity of any type can be referred to, no matter how remote temporally or spatially, no matter how unobservable, no matter how causally inert, no matter how elusive to pointing, no matter how private. According to this liberal view, anything can be referred to and all reference is on a par. We might compare the two views of reference to two similar views of truth: a restrictive view and a liberal view. According to the restrictive view, truth only applies to limited and select subject matters, say the physical world of observable bodies (not to sentences about atoms or values or modalities, etc). According to the liberal view, truth is completely promiscuous and applies to every subject matter you care to mention—fictional truth, moral truth, mathematical truth, aesthetic truth, etc. Of course, in both cases—truth and reference—we can envisage intermediate positions, depending on theoretical predilection.
The case of reference to the mind is especially interesting because it is so unobvious what to say. First, let me make clear that I am not considering anti-realism about the mind—I am assuming realism but wondering about referentialism. Granted that we really have sensations, thoughts, feelings, and so on, can we refer to them? Are the conditions for reference satisfied in this case? It is natural to suppose that they are, but the question still needs to be asked. Wittgenstein poured cold water on the idea, preferring to view psychological utterances as expressive; but there is a question whether he put his finger on what is really problematic about such putative reference. So let us ask how (alleged) mental reference differs from other kinds of reference. Consider demonstrative reference to a particular animal, e.g. a cat—as in “that cat is stealthy” said while observing a cat about to pounce. What is notable is that such reference occurs in a context in which many other referential viewpoints are possible: referring from different angles, at different distances, with varying clarity of view, with the cat half-concealed—an indefinite number of referential perspectives. Yet a single animal is successfully picked out, despite all the variation. There is constancy of reference but variability of perspective. When we have the ability to pick things out like this we have the power to pierce through variation and home in, arrow-like, on singularity. We can also track the animal as it moves through space and time, preserving referential constancy. Reference is invariant under transformations of position and perspective. But notice that all this is lacking in cases where philosophers have doubted reference: obviously so in the case of things that don’t exist, but also for abstract entities, values, atoms, etc. We don’t have the multiple perspectives, the variations in context, the tracking through space and time, the singularity amidst diversity. And when such things are lacking we are apt to feel skeptical about whether genuine reference is occurring. That is not to say that we are right to reject reference in these cases; it is just to explain certain philosophical tendencies. These are the features that correlate with our comfort level about attributions of genuine reference; and they are not particularly doctrinaire, nor wedded to some sort of theory. They form our paradigm of reference.
What is striking is that in the case of supposed mental reference we also lack these features, quite conspicuously. Take “this thought” or “this pain”, where these purport to refer to mental states of the speaker. The assumption is that these terms refer to thoughts and sensations in just the sense in which “that cat” refers to a cat. But all the characteristic marks of a paradigm case of demonstrative reference are lacking: there is no variation of spatial perspective, no perceptual presentation, no possibility of concealment, no tracking through space and time. What we appear to have is a bare unmediated confrontation between a speaker and an inner state: the subject is aware of the inner state and simply utters the words in question, purporting to pick it out. But as Wittgenstein asks, what makes this reference and not merely mouthing a sound when you experience something inside you? Surely it is not sufficient to refer to an inner state that one utters a sound in its presence, or else a parrot could perform such reference simply by making a noise. What is needed to achieve reference is a whole background and context within which reference can intelligibly occur. It doesn’t occur by magic or sheer will. My suggestion, then, is that the features I gestured at (they need refinement and supplementation) are what are lacking in cases of apparent mental reference—and that is why attributing reference to mental words strikes us less than straightforward. We find ourselves uneasy with the concept of reference in this case, not quite sure what to say. It seems like a degenerate example of reference, or reference by courtesy. We can’t form a clear conception of what it involves, as we can for reference to ordinary material objects. It feels somehow “queer” (to use Wittgenstein’s word). I try to focus on a passing thought and intone the words “this thought” hoping thereby to ensnare that elusive particular: but do I stand in any intelligible relation of reference to my mind when I perform this ceremony? Could I go on to name the thought in question? Why don’t I have names for my mental states?
What should we say about this? We might decide to get tough with ourselves and declare that mental reference simply does not exist and is impossible, since the necessary conditions of reference do not obtain in this case. That seems clear enough and not without motivation: extreme, though understandable. Or else we speak of weak and strong reference, or some such irenic philosophical terminology. Or we airily suggest that nothing hangs on the question, that it is purely verbal, that it doesn’t matter: we can say that we refer to our inner states if we like, so long as we acknowledge the very different forms that reference can take (much the same can be said about truth). I won’t attempt to adjudicate between these responses; what I have wanted to do here is articulate the issue and explain the intuitions that people have had about reference. In our “craving for generality” (Wittgenstein again) we are apt to assume a greater uniformity in our concepts than is warranted by a careful consideration of the facts, and it can be salutary to point to differences in the conditions under which they are applied. No conceptual revision may be called for in the end, just greater sensitivity to variety; we can keep talking the same way but cultivate a keener awareness of differences. Sameness of word does not imply sameness of nature. We use the word “refer” in application to many areas of language, but it may not possess the unity we assume. We tend to picture reference as something like a beam of light or a cord connecting word to object, but it may have no more unity than the various moves in chess—lots of different ways to achieve a goal (winning the game or saying something true).
Here is one area in which these reflections might be helpful: the question of identity statements linking mental and physical terms, as in “pain is C-fibre firing” or “this pain is that C-fiber firing”. If we took the tough view, these would be declared meaningless, since one term is not a referring term at all—no true identity could be asserted by such a sentence. It would be like saying, “the true is identical to the beautiful” where “the true” is taken to denote the truth-value True (Frege). If we took the more deflationary view, then we would be coupling a word that refers in one way with a word that refers in another way: that would account for the oddity of this kind of identity statement. It would be misleading to compare it to a statement like “water is H2O”, where both terms are clearly referring terms. A better analogy would be, “I am this body”, in which there is real doubt about whether that is an identity statement, because of the peculiarities of “I”. The point is that we should be careful in trying to understand what such sentences express, given the uncertainties attaching to mental reference. Just because an expression looks referential doesn’t mean that it is referential. The philosopher’s term “reference” is a term of art and covers all manner of cases (or doesn’t, as the case may be). Our very idea of the mental may be shaped and distorted by the assumption that mental terms function like standard referring terms: they may be referential in only the most minimal and trivial sense (they allow us to “talk about” certain subject matters). There is really all the difference in the world between “that cat” and “this thought”: the relations between word and object are totally different in the two cases. Calling both “reference” without qualification is bound to invite misunderstanding. 
 I have tried to remain neutral on the question of mental reference, being content merely to set out the issue, but I must confess to an urge to reject the whole idea outright. I am acquainted with my mind, to be sure, but I don’t denote anything in it. However, I will resist this urge, purgative as it may be.