The best theory of phenomenal consciousness is one that equates it with the contents of the so-called global workspace (a.k.a. working memory)—or rather, with a subset thereof. Why a subset? Because although conceptual information can be bound into the contents of perceptual and imagistic states, and made available in the global workspace, my view is that concepts don’t make a constitutive contribution to the phenomenal properties of the resulting states, but at best a causal one. Phenomenal consciousness is restricted to globally broadcast nonconceptual—fine-grained—information.
Put differently, phenomenal consciousness is access-conscious nonconceptual content. Phenomenally conscious states are those that are available for verbal report (and puzzlement), for planning, to serve as a basis for inferences, and to give rise to long-term memories. Note that there is a nice confluence here between the empirical evidence collected by cognitive neuroscientists and the first-personal nature of phenomenal consciousness as introduced by philosophers. For states that enter the global workspace will thereby be available to be reported on and puzzled over from the first-person perspective. This is because those states will be available to the systems (broadly speaking, executive-function systems) that are engaged when one reflects, introspects, and becomes puzzled.
The evidence for global-workspace theory comes from many sources. But some of the most elegant derives from studies using fMRI, EEG, or direct neural recording (or some combination thereof), in conditions where perceptual input is kept constant across conscious and unconscious conditions. Some experiments use very brief and faint stimuli, which are either backward-masked by a follow-up stimulus or have to compete for attention with a later stimulus (as in the so-called attentional blink). In both of these sorts of case, the stimulus parameters (intensity and duration) can be kept fixed, but people are only aware of the stimulus on about 50% of trials. One can then see the difference that conscious perception of the exact same stimulus makes in the brain. (Answer: global as opposed to local activity.) Other experiments use full-strength but bi-stable stimuli, as in binocular rivalry, where completely different images can be presented to the two eyes. One can take direct or indirect measures of the moment when the seen-image switches from one stimulus to the other, and see what happens in the brain concerning the very-same content in seen versus unseen conditions. (Same answer.)
Importantly, we now know that the activity in prefrontal / executive systems that is unique to consciously-experienced trials doesn’t result from response preparation, because the same findings emerge even when the responses are held constant as well (in conditions where people have to guess at a line-orientation whether they see it or not, in blindsight-like conditions), and also in completely passive viewing conditions where no response or any sort is required or made.
Of course, global workspace theory isn’t the only game in town. But frankly, most of the alternatives have little to be said for them. Perhaps the most plausible competitor is Ned Block’s fragile-short-term-memory account, which is supported by the alleged richness of conscious experience in comparison to what can enter working memory and be retained for verbal report. But in fact these data can readily be explained in terms of conscious perception of the gist of a scene, which is known to be processed swiftly and reliably independent of focal attention. So the gist is conscious, and some individual items are conscious, and the remainder can become conscious when specific currently-unconscious fragile-short-term-memory contents are targeted with attention following a retrospective cue. Strictly speaking, the evidence reviewed here is only evidence of the cognitive / neural correlates of phenomenal consciousness, not of what the latter itself is. But for anyone who has principled reasons for rejecting non-physical qualia, the stronger inference is warranted by default. In fact, however, we can do better. Global-workspace theorists can explain why the consciousness issue should get such a grip on us in the first place. I will say a little about that tomorrow.