My student Alan T. Moore defends his dissertation today. (Good thing we proved he exists!) One striking idea from his dissertation is that much of our consciousness might be, in his terminology, paraphenomenal. A conscious experience is paraphenomenal to the extent it is uncorrelated with cognitive and behavioral processes. (That's my own tweaking of his formulation, not quite how Alan phrases it himself.)
Complete paraphenomenality is a possibility so bizarre and skeptical that I'm unaware of any philosopher who has seriously contemplated it. (It seems likely, though, that someone has, so I welcome references!) Complete paraphenomenality would mean having a stream of experience that was entirely uncorrelated with any functional, cognitive, or sensory input and entirely uncorrelated with any functional, cognitive, or behavioral output (including introspective self-report). Imagine laying the stream of William James's conscious experience atop the behavior and cognitive life of Moctezuma II, or atop a stone -- simply no relationship between the non-phenomenal aspects of one's cognitive life and one's outward behavior (or lack of it) and the stream of lived experience. Or imagine taking the philosophical zombie scenario and instead of denying the zombies any experience, randomly scramble which body has which set of experiences, while holding all the physical and behavioral stuff constant in each body.
Paraphenomenal is not the same as epiphenomenal. Epiphenomenalism about consciousness is the view that conscious experience has no causal influence, the view that consciousness is a causal dead-end. But most epiphenomenalists believe, indeed emphasize, that conscious experience still correlates with causally efficacious brain processes. On paraphenomenalism, in contrast, there aren't even correlations.
Complete paraphenomenalism is about as implausible a philosophical view as one is likely to find. However, partial paraphenomenalism has some plausibility as an interpretation of recent empirical evidence, from Moore and others. Partial paraphenomenalism is the view that the correlations between conscious experiences and cognitive processes are weaker and more limited than one might otherwise expect -- that, for example, presence or absence of the conscious experience of visual imagery is largely irrelevant to performance on the types of cognitive tasks that are ordinarily thought to be aided by imagery. If so, this would be one way to explain empirical results suggesting that self-reported visual imagery abilities are largely uncorrelated with performance on "imagery" tasks like mental rotation and mental folding. (See my discussion here and in Ch. 3 of my 2011 book.)
Especially strikingly to me are the vast differences in the experiences that people report in Russell T. Hurlburt's Descriptive Experience Sampling (e.g., here, here, here, here). Hurlburt "beeps" people at random moments throughout their day. When the beep sounds, their task is to recall their last moment of experience immediately before the beep. Hurlburt then later interviews them about details of the targeted experience. Some of Hurlburt's participants report conscious sensory experiences in almost all of their samples, while others almost never report sensory experiences. Some of Hurlburt's participants report inner speech in almost all of their samples, while others almost never report inner speech. Similarly for emotional feelings, imageless or "unsymbolized" thinking, and visual imagery -- some participants report these things in almost every sample, others never or almost never. Huge, huge differences in the general reported arc of experience! When functional cognitive capacities vary that much between people, it's immediately obvious (e.g., blind people vs. sighted people). But no such radical differences are evident among most of Hurlburt's participants. Participants often even surprise themselves. For example, it's not uncommon for people to initially say, before starting the sampling process, that they experience a stream of constant inner speech, but then report few or no instances of it when actually sampled.
In his dissertation, Moore finds very large differences in people's reported experiences while reading (some of the preliminary data were reported here), but those reported experiential differences don't seem to predict performance in plausibly related cognitive tasks like recall of visual details from the story (for people reporting high visual imagery), rhyme disambiguation (for people reporting hearing the text in inner speech), or recall of details of the visual layout of the text (for people reporting visually experiencing the words on the page in front of them).
When faced with radical differences in experiential report that are largely uncorrelated with the expected outward behavior or cognitive skills, we seem to have three interpretative choices:
1. We could decide that the assumed functional relationship shouldn't have been expected in the first place. For example, in the imagery literature, some researchers decided that it was a mistake to have expected mental rotation ability to correlate with conscious visual imagery. Conscious visual imagery plays an important causal functional role in cognition, just not that role.
2. We could challenge the accuracy of the subjective reports. This has tended to be my approach in the past. Maybe people who deny having visual sensory experience of the scene before them in Hurlburt's and Moore's data really do have sensory experience but either forget that experience or fail to understand exactly what is being asked.
3. We could adopt partial paraphenomenalism about the experience. Maybe people really are radically different in their streams of experience while reading, or while going about their daily life, but those differences have little systematic relationship to the remainder of their cognition or behavior (apart from their ability to generate reports). I wouldn't initially have been much attracted to this idea, but I now think it's an important option to keep explicitly on the table. Alan Moore's dissertation builds an interesting case!