It is often supposed that consciousness has something to do with self-reference. Thus it is held that to be conscious is to be conscious of oneself: consciousness is the same as self-consciousness. This view comes in several versions: referring to oneself with “I”, having higher-order thoughts, perceiving one’s own mental states, self-monitoring, introspective knowledge, inward attention. The trouble with these kinds of theories is that they presuppose abilities that not all conscious beings possess: a creature can be conscious of the world and yet not be conscious of itself—as when the creature consciously perceives its immediate environment. Consciously seeing an object seems quite independent of consciousness of oneself seeing it. There can be something it’s like to be a bat without the bat being able to self-ascribe mental states. Consciousness is more basic than self-consciousness, and logically independent of it.
However, there may be a more subtle way in which consciousness involves self-reference. Consider the perception of color. Our visual experience of the world is saturated with color, but color is subjectively constituted; so our visual experience of objective things is saturated with subjectively constituted properties. It is traditional to call color a “secondary quality”, in contrast to such “primary qualities” as shape and size, but there is a good sense in which that terminology fails to do justice to color: as a matter of phenomenology, color is just as “primary” as shape—just as salient, useful, and vivid. It is just that color is not as “objective” as shape—not the concern of physics, not part of the “absolute conception”, not ontologically removed from human experience. Color is determined by subjective responses; shape is determined by mind-independent reality (so, at least, it is commonly assumed). The standard view of color is that a color property is a disposition to produce color experiences in perceivers (or supervenes on such a disposition). Given that, we can say that an object is red, say, just when it is disposed to produce red experiences in perceivers: so being red consists in a relation to conscious subjects, viz. having the power to cause experiences of red in perceivers. To specify or explicate the property of being red we refer to experiences of red.
But then a conscious experience of a red object is an experience of a property that essentially involves conscious experience. We are experiencing a property that is constituted by experience (a disposition to bring about experience); we are consciously aware of a property whose nature itself incorporates conscious awareness. Let me say, somewhat loosely, that the property refers to conscious experience—reaches out to it, embraces it. In analyzing the property we certainly refer to experience (unlike the case of shape), so it is not too much of a stretch to say that the property itself makes such reference. Then the following can be stated: visual consciousness refers to a property that refers to visual consciousness. My experience refers to the property of being red (the cup in front of me is red) and redness itself refers to experience of red (including the experience I am now having). So visual consciousness is self-referential: it refers to something that refers to itself. This is because color is defined in terms of subjective experience: that is what color is. To put it differently, color is a projection of consciousness (not an objective trait of things); so when we perceive color we perceive a property that comes from within consciousness. Consciousness spreads color on the world, so perception of color is perception of what is so spread. To put it loosely, we perceive our own consciousness (more exactly, we perceive a property that is a product of our consciousness). Thus it is that consciousness is self-referential—even when it is the basic kind of sensory awareness of the world. The world of visual perception is a world suffused with our own subjectivity, so awareness of objects in that world involves awareness of that very subjectivity. We might express this by saying that perceptual consciousness is “covertly self-referential”—implicitly, consciousness is consciousness of consciousness.
An objection may be mooted: that may well be true for color, but what about the other senses? Well, the other senses also have their secondary qualities, so the same argument applies. Take taste: when something tastes bitter it is apprehended as having a property that is defined and conferred by a disposition to taste bitter—so when we taste a bitter object we are aware of a property that is constituted by experience of bitterness. Bitterness is response-dependent. Thus we can say that tasting an object is “self-tasting”: we taste a property that consists in experiences of tasting. To be aware of a taste is to be aware of a quality of consciousness—how things taste subjectively. Tastes refer to experiences of tasting, so to perceive a taste is to perceive something that refers to tasting. Thus in tasting something one refers back to oneself: one has spread tastes on the world, and then one reaps the benefit of one’s own dispensation. To change the metaphor, one sows seeds in the world and then one harvests the results of what one has sown. Consciousness harvests itself.
Another objection: what about consciousness of primary qualities? If shape is not definable as a disposition to produce experiences of shape, then experiences of shape are not experiences of properties that refer back to experiences. There is no self-reference in seeing something as square—yet this is still a conscious state. So we can’t say that consciousness in general is self-referential, even perceptual consciousness (this is before we get to conscious thoughts about shape). Here two replies may be made. First, it is not clear that the geometrical properties that we naively attribute to the world are really objective; they may be just as much creatures of our subjectivity as colors and tastes. This is because reality may not objectively conform to our innate perceptual geometry—maybe we impose that geometry on physical reality (hence all the discussion about whether the physical world is really Euclidian). Maybe we are spreading shapes too (our shapes). Second, experience of shapeembeds experience of color, so we are always self-referring even when seeing objective shapes. No conscious visual experience is ever free of the kind of self-reference I have described. Certainly, our experience of shape is inextricably bound up with our experience of color, and color is a reflection of our nature as conscious beings: in seeing both color and shape we see our own reflection. The color that stares back at us originates with us, and it is conjoined with shape.
A third objection: what about unconscious perception of color? If we perceive colors unconsciously, and yet colors refer to experiences (conscious or unconscious), then self-reference is not sufficient for consciousness. That may be—all I have claimed is that self-reference is a necessary part of consciousness. But the stronger thesis can also be defended: for it is not clear that there is such a thing as the unconscious perception of color as such—that is, perceptually representing redness unconsciously. Maybe all we represent unconsciously is a certain wavelength of light, not the color property itself. What would it be to taste a substance as bitter completely unconsciously? Could it taste horribly bitter unconsciously? Granted, there is unconscious perception, but it doesn’t follow that such perception can be directed to the same array of properties as conscious perception. Could there really be a creature that sees the full range of colors available to humans but never sees colors consciously? How could these properties be defined as dispositions to produce conscious experiences if these creatures had no such experiences? The more natural view is that properties constituted by conscious experience can only be perceived by beings capable of conscious experience. If there are no colors in a world without conscious beings, then color is essentially connected to consciousness; so unconscious perception of color must be either impossible or derivative. In either case, perceptual self-reference will suffice for consciousness, simply because perceiving properties that refer to conscious experience will always take us to conscious experience. Being red is precisely a disposition to produce conscious experiences of red.
What is interesting about the position defended here is that the outer-directedness of consciousness incorporates a kind of inward-directedness. When I am aware of my environment I am aware of it as instantiating properties that make reference to my own states of consciousness. That is, I am aware of it as instantiating properties that depend for their very existence on consciousness. It is not that I overtly refer to my own conscious experience whenever I see a red object—uttering the words, “I am seeing a property that is constituted by my subjective response to the world”—but I am seeing a property that in fact arises (necessarily so) from conscious beings qua conscious beings. I am seeing my own conscious constitution in front of me—projected, spread. My consciousness is thus a consciousness of my own consciousness—via my consciousness of an objective world. So consciousness does involve a kind of self-reference—it loops back on itself. 
 I have discussed perceptual consciousness, arguably the basic case, but the point carries over to cognitive consciousness: conscious thoughts about color also refer to properties that are subjectively constituted. And the same is true of conscious desire—such as a desire for a bunch of red roses. To put it in Kantian terms, we live in a phenomenal world that reflects our own mode of consciousness, so our consciousness of that world always refers back to itself.