Self-Blindness and “I”
Hume’s point that the self is not introspectively detectable has always been met with intuitive acceptance. Its significance is more contentious. Does it show that the self does not exist, or that introspection is limited, or that the self is really just a congeries of mental states? A possible view is that introspection has blind spots and the self falls into one of them. Introspection doesn’t reveal the body or the brain either, but they clearly exist; perhaps the self is a real entity that happens to fall outside introspection’s possible field of acquaintance. God can see it perfectly well, but humans are blind to this aspect of their nature. They are introspectively blind to the self in the same way they are blind to other things—through lack of acuity, lack of coverage, and lack of receptivity. They are not completely blind to the mind, since introspection can reveal other mental phenomena, but the self eludes their introspective powers. Some people are totally blind in the ocular sense, some partially so, and everyone is blind to some things (elementary particles, remote galaxies, parts of the electromagnetic spectrum); well, humans are introspectively blind to the self. Hume could have made the same point about the senses: search through the data of smell, taste, and hearing and you will not find any presentation of an object given to these senses. They deliver information about qualities, states, events, and processes, but they don’t include the perception of a continuant object—the objective source of the phenomena in question. You hear the bark of a dog but hearing gives you no impression of the dog itself as an object existing through time. Only vision and touch offer impressions of continuant particulars (and some have contested even this); the other senses are blind to such entities. Their intentional objects include only fragmentary passing occurrences: smells, tastes, and sounds. Introspection seems to be like this: it never presents the continuing self to the inner eye but only its states and contents. It suffers from perceptual closure with respect to the self—as well as for the brain and body (not to mention the rest of the world). Every sense is limited in some way, and introspection is no different. Self-blindness is really just par for the course. It certainly doesn’t imply the non-existence of the self.
But Hume’s point does raise an interesting question about the word “I”: how does it refer? It seems not to be equivalent to a description, but it doesn’t correlate with a mode of acquaintance either, since we have no acquaintance with the self (accepting Hume’s point). It isn’t like reference to pain or color sensations: here we doknow what we are talking about—these things are immediately presented to us inwardly. In these cases we “see” what we refer to. But we are constitutionally blind to the self, so we are referring to something we can neither perceive nor describe. We are like a blind man referring to what he can neither see nor pick out descriptively. This is perfectly possible: he may say “that dog” while pointing forward and happen to pick out a particular dog in front of him. He has no acquaintance with the dog and cannot describe it uniquely, but the demonstrative enables him to achieve reference nevertheless. The dog is not referentially closed to him—just perceptually and descriptively closed. Reference can transcend acquaintance and description. Hallelujah! We can refer to what we cannot otherwise access (compare remote galaxies, elementary particles, future persons, and other universes). Similarly, according to this line of thought, we refer to the self in just this kind of way: “I” refers in roughly the way “that dog” refers for the blind man. And what is that way? By means of context, indexical mechanisms, and the semantics of content and character.  We are like someone cut off epistemically from an object yet able to deploy the apparatus of indexical reference to make reference to that thing. Otherwise we would be referentially impotent with respect to the object. What this means is that we have no perception (and no real conception) of the self that we so effortlessly denote all the time—thanks to the semantics of “I”. We have nothing of epistemic substance in mind when we use “I”, but that doesn’t stop us referring to the self. This is a species of ignorant reference—reference in the absence of knowledge (by acquaintance or description).  This may account for some of the peculiarities of the word “I”—in particular, its air of airiness. It seems totally devoid of content, a mere schema or skeleton, perhaps a pseudo singular term, not really denotative at all. The reason is that it is a case of blind reference—reference not backed by knowledge. It seems like a shot in the dark, a mere gesture at reference—like putting your hand over your eyes and enunciating the word “that” hoping to net something to refer to. Even if you succeed, you have nothing much to say about the thing you have referred to, with no mental act of identification to back up your stab in the dark. Introspection is blind to the self (and external observation does nothing to remedy the lack) but the indexical semantics of “I” enables you to hang onto reference by the skin of your teeth. It is the constitutional weakness of introspection combined with the elastic power of indexical reference that characterizes the use of “I”: a sort of blind strength.
Other indexical words conform to roughly the same pattern, particularly “here” and “now”. Imagine someone subjected to complete sensory deprivation: absolutely no input is received via the senses from the external world. Thus nothing is known by the deprived subject about what is going on around her: she has no perception of what is occurring at the present time, nor does she have any descriptive knowledge that could uniquely identify the place and time involved. Yet she pronounces the magic words “here” and “now”, outwardly or inwardly, and evidently makes determinate reference thereby: a certain place and time are picked out. The reference is not mediated by acquaintance and not by individuating description, but it proceeds nonetheless. This is an extreme case of what I am calling ignorant reference. There is clearly a lot of ignorant indexical reference going on in typical language use, and it can be used to anchor other reference such as with proper names. It might even be argued that these cases are like the case of the self in that they involve blindness to the entities actually denoted: we never really encounter places and times as such in our epistemic searches. We don’t perceive them directly with the senses and we don’t have adequate descriptive knowledge of them; but we can refer to them by exploiting the mechanisms of indexical reference. And they are always there to be referred to, like the self (and unlike the material objects of perception): there is never simply no such thing as time or space or self to reciprocate the referential act. The self, though contingent, is always there because the act of referring guarantees it—where there is a referrer there must be a self that is referred to by “I”. There cannot be cases of empty self-reference. There is no need to rely on perception to provide evidence of existence as a precondition of successful reference; existence comes with the referential territory. We exist in space and time and the self is always present whenever reference occurs: so “here”, “now”’ and “I” always find a target. Blind spots don’t undermine successful reference in any of these cases. There has been a tendency to link reference with knowledge (by acquaintance or by description) but the interesting fact about reference is that it is free of such epistemic constraints: it can proceed in blissful ignorance of the thing referred to. Hume’s point is correct (though not damaging to the self’s existence) but it is no bar to reliable and useful reference to the self. Self-blindness does not entail referential self-blankness: we can denote what we cannot see (or otherwise sense). And this is true even if we necessarily can never encounter the self. 
 This is consistent with a causal theory of reference: the self might be the cause of reference to itself. But if so, it is a cause of which we have no knowledge. The causal theory of names is typically formulated in terms of observable objects causing chains of reference leading up to a particular use, but a causal theory of reference by “I” would more naturally be formulated in terms of an elusive self that nevertheless causes acts of reference to itself. Even a totally transcendental or noumenal self could operate causally in the production of occurrences of “I”. Some causes of acts of reference might be completely invisible and unknowable, yet indispensable.
 The theory of reference has been dominated by consideration of reference to public middle-sized material objects—people, animals, cities, etc. But this is parochial and possibly misleading: we also refer to selves (construed as private mental entities) as well as to places and times. Thus ignorant reference might be more paradigmatic than knowledge-based reference. The emphasis on reference to ordinary material objects probably traces to an empiricist (or positivist) tendency to put sense perception at the center of cognition, including linguistic understanding. But once we see the limitations of perception-based reference as a model for self-reference we are free to recognize that reference can proceed in conditions of ignorance. We don’t (can’t) perceive (introspect) the self, but we can refer to it with the greatest of ease. In the case of the self we appear to have an extreme example of the divorce between the epistemic and the semantic: deep ignorance combined with infallible reference. No wonder “I” is deemed so philosophically problematic.