Why have neuroscientists thought that consciousness disappears during dreamless sleep?
One reason comes from the reports that people give when they are woken up from NREM (non-Rapid Eye Movement) sleep, especially when the EEG shows slow waves in the delta frequency range (0.5–4 Hertz) during sleep stages 3 and 4 (so-called slow-wave sleep). When given the instruction, “report anything that was going through your mind just before waking up,” people tend to report short and fragmentary thoughts or not being able to remember anything at all (Nielsen 2000; Tononi & Koch 2008, p. 243). On the basis of such reports, scientists conclude that the sleepers were aware of little or nothing at all prior to being woken up, and hence that slow-wave sleep is a state of reduced or absent consciousness.
We need to be cautious here, however. The fact that one has no memory of some period of time does not necessarily imply that one lacked all consciousness during that time. One might have been conscious—in the sense of undergoing qualitative states or processes of sentience or awareness—but for one reason or another one was not able to form the kind of memories that later one can retrieve and verbally report.
This point is familiar to scientists who study the effects of anaesthetics (Alkire et al. 2008). At certain doses, some anaesthetics prevent memory formation while sparing awareness. Near the threshold of unconsciousness, some anaesthetics block working memory, but patients may still be aware and fail to respond because they immediately forget what to do. At lower doses, patients under general anaesthesia can sometimes carry on a conversation using hand signals, but after the operation they deny ever being awake.
Although dreamless sleep and anaesthesia are not the same condition, the general point that retrospective oblivion does not prove a prior lack of consciousness must be kept in mind whenever we are tempted to infer that consciousness is absent in deep sleep because people report not being able to remember anything when they are woken up.
We also need to think about the kinds of verbal reports that people are asked to make when they are woken up in the sleep lab. The instruction to report “anything going through your mind just before waking up” encourages you to direct your attention and memory to the objects of your awareness—to anything you might have been thinking about. But what about the felt qualities or phenomenal character of your state of awareness? A different instruction would be to report “anything you were feeling just before waking up.” This instruction encourages you to direct your attention and memory to the felt quality of your sleep. Did you have any feeling of being aware? Was your sleep peaceful and clear, or was it agitated, restless, or sluggish? Or do you have no impression of any feeling or quality of awareness? The point here is to guide people away from focusing exclusively on the intentional objects of consciousness, which may be absent in deep sleep, and to orient them towards the felt qualities or phenomenal character of awareness itself.
Another reason neuroscientists think that consciousness fades away in deep sleep comes from comparing brain activity during slow-wave sleep with brain activity during waking consciousness. For example, during wakefulness, when an electrical pulse is used to stimulate a small region of the brain, the pulse generates an EEG response that lasts for 300 milliseconds and that is made up of rapidly changing waves that propagate in specific directions over long distances in the cortex (Massimini et al. 2005; Tononi & Massimini 2008). During deep sleep, however, although the initial EEG response to the stimulation is stronger than during wakefulness, the response remains localized to the stimulated region instead of travelling to distant regions, and it lasts only 150 milliseconds. In short, whereas the waking brain responds to stimulation with a complex pattern of large-scale activity across many interconnected regions, the deeply sleeping brain responds with localized and short-lived activity. These findings are interpreted as showing that “effective connectivity”—the ability of neural systems to influence each other—breaks down in deep sleep. As a result, “large-scale integration” (Varela et al. 2001) in the brain cannot happen—that is, the brain cannot generate the kinds of dynamically-changing large-scale patterns of activity that are known to characterize consciousness in the waking state.
But what is it about the loss of effective connectivity and large-scale integration that makes neuroscientists think that consciousness disappears in deep sleep? To put the question another way, what is the connection between the presence of consciousness and the presence of effective connectivity and large-scale integration?
To answer this question, neuroscientists usually rely on the idea that a content of consciousness is a reportable content, and that reportable contents are ones that can be attentionally selected, held in working memory, and used to guide thought and action. Such cognitive processes—selective attention, working memory, sequential thought, and action guidance—require the large-scale integration of brain activity.
One of the more theoretically-principled versions of this idea is Giulio Tononi’s “integrated information theory” of consciousness (2008). According to this theory, any typical conscious experience has two crucial properties. First, it is highly “informative,” in the technical sense that it rules out a huge number of alternative experiences. Even an apparently simple conscious experience, such as lying on your back and seeing the clear blue sky throughout your whole visual field, is richly informative in the sense that it rules out a vast number of other experiences you could have had at that moment. Second, the experience is highly “integrated,” in the sense that it cannot be subdivided into parts that you experience on their own, such as the top and bottom portions of your visual field, or the color and the space of the sky.
Given this model of consciousness as “integrated information,” Tononi proposes that the level of consciousness of a system at a given time is a matter of how many possible states (information) are available to the system as a whole (integration). In the waking state, many possible states are available to the whole system (the system is rich in integrated information), whereas in deep sleep this repertoire drastically shrinks to just a few states (the system is poor in integrated information). Transposed onto the brain, the idea is that during slow-wave sleep there is a massive loss of integrated information in the brain. Effective connectivity breaks down, leaving isolated islands that cannot talk to each other (loss of integration), while the repertoire of possible states contracts to a few largely uniform states (loss of information). Hence, according to the integrated information model, deep sleep is a state where consciousness reduces to a very low level or disappears entirely.
Although the integrated information theory offers a useful way of thinking about the qualitative richness and coherence of consciousness in informational terms, the theory has serious limitations as a theory of phenomenal consciousness, so it would be a mistake to use the theory to rule out the possibility of consciousness during dreamless sleep.
Despite Tononi’s bold claim that “consciousness is one and the same thing as integrated information” (2008, p. 232), integrated information does not seem sufficient for consciousness. On the one hand, even simple systems have some degree of integrated information, so the equation of consciousness and integrated information implies that even simple systems, such as a photodiode, have some degree of consciousness. On the other hand, complex digital computers can possess a high amount of integrated information. Yet neither system is conscious (at least the attribution of consciousness to such systems seems highly implausible) (see Searle 2013). As Ned Block (2009) points out, the integrated information theory fails to distinguish between intelligence, in the sense of being able to solve complex problems by integrating multiple sources of information, and consciousness, in the sense of sentience or felt awareness (phenomenal consciousness). Since integrated information does not seem sufficient for consciousness—let alone identical to it—the presence or absence of integrated information cannot be the crucial mark of whether a state is conscious or not conscious.
We also need to keep in mind the distinction between “phenomenal consciousness” and “access consciousness.” To be phenomenally conscious means to be in a state that has some subjective or phenomenal character. To be access conscious means to be in a state where there is cognitive access to the contents of awareness. Whether a state’s being phenomenally conscious requires that it be cognitively accessible is currently a matter of debate (Block 2011; Cohen & Dennett 2011). Although large-scale integration in the cortex is crucial for cognitively accessed or reported conscious experience, it may not be crucial for every kind of phenomenal consciousness; for example, it may not be crucial for the kind of cognitively unaccessed consciousness that Yoga and Vedānta maintain is present in dreamless sleep (though they also maintain, as we shall see, that this kind of consciousness is accessible if one is highly trained in certain types of meditation).
The upshot of this critical assessment of the default view is that neither the subjective report data nor the objective neurophysiological data suffice to rule out the possibility of a subtle mode of phenomenal consciousness occurring in certain phases of dreamless sleep. To put the point another way, the sleep science construct of “dreamless sleep,” defined electrophysiologically as slow-wave sleep, may need phenomenological refinement. We need to allow for the possibility that certain types of slow-wave sleep may have a phenomenal character—a possibility that could in turn lead to refinements in the physiological construct of slow-wave sleep. It follows from these considerations that the standard neuroscientific definition of consciousness as “that which disappears in dreamless sleep and reappears in waking and dreaming states” is not acceptable. At the very least, it needs qualification in light of the present considerations, and it may need to be either substantially revised or abandoned in light of further research.
The case of dreamless sleep suggests that we need to allow at least for the possibility of there being modes of phenomenal consciousness that may not be cognitively accessible in the usual ways. At the same time, Yoga and Vedānta, as well as Indo-Tibetan Buddhism, maintain that aspects of the mind in deep and dreamless sleep can become cognitively accessible through meditative mental training. This is the last topic I wish to discuss. My main point will be that considering sleep from this contemplative angle suggests new experimental questions and protocols for the cognitive neuroscience of sleep and consciousness.