[1]
In some readings of the term dreamless sleep, the default view in not just obviously false, but it is also unclear that it is actually endorsed by many researchers working on dreaming and sleep. Most would acknowledge, for instance, that hypnagogic imagery during sleep onset or repetitive and non-progressive types of sleep thinking involve phenomenal experience during sleep; yet, because they are also commonly distinguished from full-fledged dreaming, they can be said to occur in dreamless sleep. This, however, is different from the type of dreamless sleep experience that Evan Thompson has in mind and that is the focus of this commentary. As will become clear later, in the narrower reading endorsed by Thompson, dreamless sleep “is that sleep state in which there are no sensory or mental objects of awareness, that is, no images and no thoughts” (p. 14); the question, denied by the default view, is whether this state of sleep can sometimes involve phenomenal experience. Dreamless sleep experience of this type, if it exists, is also distinct from experiences occurring during sleep-wake transitions in that it is thought to occur during deep sleep. In the context of this commentary, I will always, unless explicitly noted otherwise, use the term dreamless sleep experience in this narrow sense. In other readings, the default view may be thought to be trivially true: if one defines dreams as involving any kind of phenomenal experience during sleep (Flanagan 2001), then the occurrence of phenomenal experience during dreamless sleep is indeed ruled out by conceptual considerations. This reading, however, is too permissive in that it fails to acknowledge the distinction between different types of experiences occurring during sleep, ranging from imagistic, narratively complex, and often emotional dreams to thoughtlike activity. For now, this suggests that the default view is too simple: the question is not whether there are experiences during sleep that fall short of full-fledged dreaming in some particular sense but whether there is a further group of experiences—call them dreamless sleep experience in the narrow sense—that is distinct from any of the established forms of conscious experience during sleep, including hypnagogic imagery and sleep-thinking. Thompson acknowledges this issue (p. 14) and I only emphasize it here to avoid misunderstanding.
[2]
Note that throughout this commentary, I will use the terms “experience”, “subjective experience”, and “consciousness” interchangeably to describe states that have phenomenal character, or for which there is something it is like to have them.
[3]

At first sight, there is an inherent ambiguity in the concept of pure subjective temporality in that it can refer to the experiential character of nowness, but also to the experience of duration and of succession. In section 3, it will become clear that in the account defended here, the two aspects of nowness and duration are not strictly dissociable: the simplest forms of temporal experience are characterized by both a phenomenal now and the experience of duration, because the phenomenal now itself is temporally stretched. Though for reasons of space, I cannot discuss this any further here, note that once the distinction between the phenomenal now and the experience of duration collapses, the experience of seriality or of succession disappears as well: if the phenomenal now is no longer embedded within a larger temporal reference frame, then there will be no separate events that can be experienced as succeeding each other. 

[4]
It remains controversial whether different forms of meditation actually enhance introspective accuracy. While there is some evidence in support of this claim (Fox et al. 2012; Sze et al. 2010), at least one study has suggested that meditators may feel more confident than controls about their ability to successfully perform interoceptive tasks (such as heartbeat detection), but that this confidence is not paralleled by an actual improvement in task performance (Khalsa et al. 2008).
[5]
Researchers occasionally worry, for instance, that participants may underreport embarrassing dream content; censorship of this type may be why sexual dream content is only rarely reported in laboratory studies (Hobson 1988); see also Rosen’s (2013) discussion of willful narrative fabrication of dream reports. For the investigation of dreamless sleep experience, which is, after all, thought to be devoid of such content, such worries about censorship do not seem to apply.
[6]
I owe this term to Sascha Fink; see for instance Fink 2015, p. 23; for discussion, see Windt 2015, p. 92.
[7]
As Solomonova et al. (2014) note, it is important to distinguish questions about the range of possible experiences in dreams (or the “depth” of dreaming) from those about their typical characteristics in the general population (or the “breadth” of dreaming), and what counts as the ideal reporting conditions in the context of a given study depends on which of these questions is being addressed. For now, note that because expertise is likely most useful for answering questions about the depth of experience, and because expert reports may not be representative of the breadth of the target phenomenon, broadening the investigation of dreamless sleep experience beyond expert groups is an important goal for future research.
[8]
Strictly speaking, it cannot be ruled out that even for reports obtained under seemingly ideal conditions—for instance immediately after awakening, and using appropriately worded questions—certain subject groups are particularly prone to memory failure or confabulation (Rosen 2013), or that results are distorted because of further disturbing factors that have so-far been overlooked. The challenge will then be to identify such potentially disturbing factors, manipulate them experimentally, and derive certain predictions on how they will affect data obtained from the analysis of dream reports. These factors can then be integrated into a future, improved and more empirically plausible account of the ideal conditions of dream reporting. For now, my main point is that this strategy only makes sense if one already assumes that some subset of dream reports can be used as a baseline against which other, less trustworthy ones can be measured.
The study of dream emotions is a nice example of how this strategy has been put to work in dream research. Views on the both the frequency and the types of emotions experienced in dreams have changed quite dramatically as new methods of collecting and scoring dream reports have been developed. Whereas older studies using classical dream content analysis suggested that dream emotions are relatively rare (Hall & Van de Castle 1966), the frequency of reported dream emotions increases tenfold when subjects are specifically asked to report their emotions on a line-by-line basis (Merritt et al. 1994). Affirmative probes of this sort suggest that dreams are “hyperemotional”, with emotions being mentioned in 95 percent of dream reports and the average dream report containing several different types of emotions. A plausible explanation is that dream emotions are underreported in free dream reports of the type used in older studies; free dream reports, in this view, are insufficient to capture the actual frequency of dream emotions. Until very recently, the accepted view was that the types of emotions experienced in dreams differ from those experienced in wakefulness in that dream emotions are predominately negative (Hobson et al. 2000). However, a recent study compared external ratings of emotions in dream reports to scores obtained when participants answered a standard emotion questionnaire themselves. Sikka et al. (2014) found that external ratings underestimate not only the frequency but also the types of emotions experienced in dreams. A particularly surprising result was that self-ratings showed positive dream emotions to be six times more frequent than negative ones. The systematicity of the differences is compelling and the same pattern was found in a number of follow-up studies (Sikka et al. 2014), suggesting that the use of self-ratings is a more reliable method for capturing the frequency and types of dream emotions than the use of external raters. This is not so say that the conditions for reporting and scoring dream emotions cannot be further improved. But this example does illustrate that theoretical views on dream emotions changed in tandem with changed and likely improved reporting and scoring conditions. Again, the idea is that methodological adjustments can obscure or render visible different aspects of the phenomenology of dreaming.
[9]
Today, it is widely recognized that dreams can occur in all stages of sleep and are not exclusively a REM sleep phenomenon. Incidentally, this recognition may also lead to refined sleep-stage scoring systems and a blurring of the borders between REM and NREM sleep (Nielsen 2000; see also Windt 2015, chap. 2).
[10]
Here, I use the concept of epistemic transparency in a non-technical and metaphorical sense, intending to capture the intuition that dream reports are the closest researchers can come to “watching the sleeping mind” (Cartwright 2010, p. 17). The choice of terminology also reflects the fact that dream reports are not identical with, but better conceived of as separate from dreaming. Finally, transparency is a nod to the historical situation that the theoretical problems raised by dream reporting were nearly invisible throughout most of the history of philosophical theorizing about dreaming.
[11]
At this point, it might be objected that this formulation rides on a reification of the word “nothing”, as if “nothing” itself could be turned into an object of experience. I return to this problem in section 3; as will hopefully become clear, my own positive model of dreamless sleep experience avoids this problem by introducing a qualified reading of what is described, in the Advaitin view, as experiencing or knowing nothing.
[12]
Incidentally, note that if it were the case that dreams are devoid of any experiences whatsoever, it would be utterly mysterious why we should awake with the vivid impression of having had such experiences in the first place. Indeed, Malcolm provides no explanation of why this happens. By contrast, my erroneous impression of having climbed a mountain during sleep is nicely explained by saying that during sleep, I had experiences that were sufficiently similar to their waking counterparts to create this impression. Again, this comes back to the idea that explanatory considerations favor the view that dream reports are actual memory reports, and not inferential. Perhaps, the difference between dreams that are belief-inviting beyond the borders of sleep, for instance by making us actually believe, if only for a moment, that the corresponding events actually occurred, and more commonplace dreams that do not induce such false beliefs can even be described in phenomenological terms (for a first proposal of how this might be done, see Windt 2015, chap. 10).
[13]
Note that this way of thinking about phenomenal selfhood is quite different from the way the term “self” is used in the classical Indian literature. In his reconstruction of the Advaita Vedānta concept of witnessing, Fasching (2010) notes that the “‘witness’ (sāksin) is not understood as an observing entity standing opposed to what it observes, but as the very taking place of ‘witnessing’ itself, and ‘witnessing’ is nothing other than the taking place of the experiential presence of the experiences, in which the experiences have their very being-experienced and thereby their existence.” (p. 204)
In this conception, “the ‘self’ is nothing other than becoming aware of experiential presence (consciousness) as such” (p. 207); it is “not a structural moment of what is given, but is the very taking place of givenness itself” (p. 210). Recall that one of the points of agreement between the Advaitins and the Nyāyas was that the self persists throughout sleep. But this is not the reading of the concept of “self” that Thompson (2014; see for instance chap. 10) has in mind when he says that in dreamless sleep experience, there is no longer an awareness of the “I”, or what I mean when I speak of phenomenal selfhood.
[14]
In fact, if we conceive of temporal experience as involving a specious present, we might say that the phenomenal now simply is identical with a rudimentary form of a temporal reference frame. I return to this point later. Alternatively, if we conceive of temporal experience as consisting of a series of unconnected moments that themselves have no temporal extension, then again it would seem that each of these could occur in isolation and without being embedded in a larger temporal reference frame.
[15]
This phenomenological observation is reflected in the classical idea that the mind cannot be spatially located in the physical world. Mental states persist over time, but they do not have spatial characteristics such as expansion or separable parts. Perhaps, this phenomenological observation lies at the root of metaphysical claims about the relationship between mind and body.
[16]
This is not to deny that experiences (or qualitative aspects of experiences) could exist that are beneath the cutoff line of memorability and reportability. Certain subtle aspects of phenomenal experience, such as hues of color, do seem to outrun our ability to categorize and reidentify them over time (Raffman 1995). Here, I am only claiming that such subtle aspects of experience are not candidates for the report-based type of scientific investigation I am interested in here.
[17]
This is a prediction, and different subjects may mean different things when they describe an experience as selfless. For some this may mean an experience characterized by spatiotemporal self-location, but in which they had the experience of being a disembodied entity (cf. Windt 2015, chap. 7); others may describe episodes characterized only by their temporal features as involving a self. There is also the familiar problem that reports of selfless experiences easily slip into a performative self-contradiction, of the type “I had a dream in which I was not present”; such episodes are clearly remembered and reported by someone. But we should not expect the folk-psychological use of terms such as “I” or “self” to align perfectly with a particular technical definition. This is a good example of where specific interview questions might increase the expressive granularity of retrospective reports.
[18]
Note that this is related to a terminological difficulty that is implicit in the Indian debate, as well as in Thompson’s reconstruction of it. As noted earlier, both sides in the Indian debate assumed the self to persist throughout sleep; they merely disagreed whether the self is necessarily conscious. My proposal that we redescribe dreamless sleep experience in terms of pure subjective temporality captures this idea that the self persists in a thin sense even when awareness of any intentional contents is lost. At the same time, recall that dreamless sleep experience is thought to be characterized by a collapse of subject-object duality and by an absence of any intentional objects of awareness. In this state, nothing, including the self, is thought to be known or cognized. There is no longer an individual, consciously experienced first-person perspective. It is this thicker and more substantial notion of a self experienced as distinct from other objects or persons that I propose is lost in dreamless sleep experience; the persistence of such a self would mean that there would still be an intentional object of awareness, and thus would indicate a more complex state than that characterized in the Indian debate as dreamless sleep experience.
[19]
This could, of course, turn out to be false. Even if the underlying neural representations are temporally extended, the same may not be true of conscious states themselves; these may still be conceived of as elementary and momentary events lacking spatial or temporal structure. For a recent defense of such a view, inspired by the Abhidharma doctrine of momentariness, see Chadha (forthcoming). Yet, even if the experience of continuity and persistence over time turned out to be an illusion, this would still be an interesting structural feature of phenomenal experience. For present purposes, the basic phenomenological claim, according to which the phenomenal now is temporally stretched rather than momentary and discrete, is enough.
[20]
Again, there are subtle terminological differences. For instance, Williford (2015a, pp. 10-11; see also Williford 2015b) writes that reflexivity or self-acquaintance is “an essential structural feature of all consciousness; and I take it to be a phenomenological datum. All streams of consciousness are immediately aware of themselves, and that is the foundation of all other forms of self-representation, autobiographical cognition, and so on. This reflexivity is subjective character (for-me-ness), but it is a mistake to turn this structural feature into a kind of entity or homunculus.”
My account is compatible with much of what Williford says here; I agree that we are considering a basic and essential feature of conscious experience, and one that should not lead us to posit an independent entity that is identified as the self. Yet, I think there is room for phenomenal selfhood as a structural feature of experience over and above the reflexivity of even the simplest kinds of phenomenal states. Even readers who disagree with my description of this as a form of phenomenal selfhood might still agree that the target property of spatiotemporal self-location is distinct from the more basic reflexivity of consciousness. Adopting the conceptual convention of describing this as a form of self-experience does not, I take it, require us to reify the self or to slip into a homuncular view, but simply offers a conceptual tool for describing the way we experience ourselves as being or having a self.
[21]
A prediction that seems implicit in Proust’s observation that if we are suddenly overcome by sleep, we no longer know what time it is upon awakening, is that dreamless sleep experience may bear an interesting relation to the ability to estimate how long one has slept. Perhaps, intermittent periods of dreamless sleep experience even ground our awareness that some time has passed or are responsible for the ability, which may be more pronounced in certain subjects, to awaken just before the alarm clock goes off (thanks to Valdas Noreika for pointing this out). By contrast, if we awaken from a state lacking any form of phenomenal experience whatsoever—as in some forms of anesthesia—there may be no sense of a preceding temporal gap and a more profound sense of temporal disorientation. At present, this is, of course, entirely speculative, but it might be a question worth asking.
[22]

The temporally dynamic nature of experience is also of central importance for understanding the neural correlates of conscious experience. As Melloni (2015) points out, while the mechanisms for updating the contents of consciousness have been investigated by numerous studies, the mechanisms underlying the maintenance or the flow of conscious experience fall outside the scope of most existing paradigms. She also proposes that the temporal flow of consciousness is a fundamental property of experience and an important target—perhaps the most important target—for future research on the neural correlates of consciousness. Similarly, Noreika (2015) suggests that focusing on the analysis of individual contents of consciousness, as is standardly done in mainstream research on the neural correlates of consciousness, misses the temporality of consciousness; instead, to make progress toward understanding this more fundamental property, he proposes contrasting conscious and nonconsciousness states.

[23]
A similar link between the dissolution of the self and the experience of timelessness, or of an indefinite duration, may exist in deep meditative states (Berkovich-Ohana et al. 2013), but also, for instance, in drug-induced altered states of consciousness (Wittmann in press).
[24]
Similarly, in the tradition of dream and sleep yoga, dream lucidity is sometimes described as a preliminary stage of becoming aware of sleep; again, realizing that one is dreaming precedes the dissolution of dream imagery while maintaining awareness of dreamless sleep. See for instance Wangyal & Dahlby (1998).
[25]
This is also why such examples of prelucid dreams or of sleep-state misperception do not threaten the transparency of retrospective reports. In the present framework, reports are transparent with respect to the occurrence and phenomenal character of experience only; but we should not expect them to accurately reflect the sleep state in which the respective experiences occurred (or indeed whether they occurred in sleep at all), just as we should not expect them to accurately identify the underlying changes in neural activation patterns. Perhaps training, as Thompson (2015) suggests, can indeed improve the match between subjective experience reports and objective measures of sleep states or of brain activation; but this is in no way guaranteed. Or perhaps, objective measures of sleep should be informed by the conditions under which different subject groups experience themselves as being asleep. A mismatch between subjective and objective measures need not indicate a flaw in subjective reports; it might also indicate that objective sleep measures are poorly suited to capture what normal, healthy subjects mean when they say they have been asleep. Indeed, this latter suggestion is in keeping with Thompson’s proposal of a phenomenologically enriched taxonomy of sleep states.
[26]

Clearly, this is just the very beginning of the conversation on how to refine sleep-state taxonomy. Ultimately, the investigation of dreamless sleep experience, and the addition of dreamless sleep experience to the conceptual tool kit used for the description of sleep and wakefulness, may prove to be no more than a first step in this direction. And while the reconstruction of the Indian debate and its contrast with contemporary views of sleep is a rich and valuable project, important but easily forgotten lessons might be found closer to home as well. The monophasic sleep pattern currently investigated in Western sleep laboratories and taken to be the biological norm may be only a few generations old (cf. Greene 2008, pp. 238-240) and is likely an artifact of a profound change in sleep behavior brought on, to a considerable extent, by electrical lighting. In preindustrial times, sleep was biphasic—two periods of sleep, called the first and the second sleep, were structured around a period of wakefulness that was made up of quiet rest, perhaps even resembling certain meditative states and often involving the contemplation of dreams (Ekirch 2001; Ekirch 2006). Research suggests that under appropriate conditions—in an environment without artificial, electrical lighting and without various nighttime activities that become possible in such an environment, that compete for our attention and increase the pressure for and attraction of staying awake rather than going to bed—we naturally return to this biphasic sleep pattern (Wehr 1992). It does not seem unreasonable to think that the transition to a monophasic sleep pattern, alongside factors such as increased electrical lighting, traffic noise, and time constraints—will have changed not just the structure of sleep, but our experience of sleep as well. With less and less time allotted to sleep, the temptation to simply black out during sleep (or to view sleep as involving such a blackout) may have increased; yet, current sleep behavior in rich, Western societies may be a highly artificial and learned behavior. If we want a taxonomy of sleep states to reflect universal features of sleep, rather than our culturally specific, contemporary sleep habits, we would do well to remember this.