In this section, I draw an analogy between the Indian debate on dreamless sleep experience and the contemporary debate on the trustworthiness of dream reports. This analogy provides the resources for overcoming the first two challenges to Thompson’s argument. In particular, it reveals the default view to be inconsistent with the methodological background assumptions of scientific sleep and dream research. Given their own methodological commitments, researchers in these fields should reject the default view.
The first step towards seeing why the default view is inconsistent with scientific dream research is to realize that this field, at least implicitly, relies on the assumption that reports of conscious experience during sleep are trustworthy: at least when they are given under certain (sufficiently) ideal conditions and immediately after awakening from sleep, such reports are taken to reflect what was experienced during the preceding sleep period, and indeed whether anything was experienced at all. What exactly the (sufficiently) ideal conditions for reporting sleep experiences consist in is an empirical question, and in scientific dream research, much work has been dedicated to its investigation (for discussion and further references, see Windt 2013, 2015, chaps. 3 and 4). There is widespread agreement that temporal proximity is a crucial factor: reports given immediately after awakening are commonly taken to be least vulnerable to forgetting. The sleeping environment (at home versus in the laboratory), method of awakening, interaction with experimenters, and precise wording of questions also play an important role (Domhoff 1996, 2003; Hall & Van de Castle 1966; Kramer 2013; Winget 1979). Different reporting techniques may be suitable for different research questions, and aside from being asked for verbal reports, participants may be encouraged to produce a dream drawing or compare the visual imagery in their dream with photographs with varying degrees, for instance, of color saturation or brightness (Rechtschaffen & Buchignani 1992). While there may be uncertainty, in a given case, as to the sincerity of a report, this is a practical matter, not a deep theoretical problem.[5] The key idea is that by improving reporting conditions and tailoring the reporting technique used in a given study to the specific research question, this risk can be minimized. For now, my main point is that this strategy, which is already well established in scientific sleep and dream research, only makes sense against a background of basic trust in at least a subset of dream reports.
This basic idea is very much in keeping with Thompson’s proposal of asking participants to report any feelings or qualitative states experienced prior to awakening, rather than asking them to focus on the contents of conscious thought. By directing participants’ attention to certain aspects of sleep experience or even introducing new experiential categories for their description (an excellent example of this strategy is Lutz et al. 2002), the expressive granularity[6] of individual reports can be increased: types of experiences can be rendered reportable that would otherwise be forgotten. A compelling possibility is that in the case of dreamless sleep experience, such improvements in reporting conditions may not just supplement training, as suggested by Thompson, but may even facilitate the investigation of dreamless sleep experience in participants who lack any particular introspective training.[7]
Admittedly, this approach does not provide a fail-safe method for avoiding or even identifying nonveridical reports. Rather than focusing on the veridicality of individual reports, the strategy is to identify which types of reports are best tailored to a given question and under which conditions they are most likely to be obtained. The problem of identifying individual reports of a certain type for which this strategy has failed is thus not obliterated, but minimized.[8] What is more important is that there is, in this view, a distinction to be drawn between general opinions about experience and reports of individual experiences. Note that reports, in this context, are broadly construed as the product of (verbal or nonverbal) behaviors conducted with the sincere intent of conveying or recording certain relevant information about a specific dream (for details, see Windt 2015, chap. 3.3) Questionnaires asking participants to assess the general frequency with which, for instance, they dream in color do not count as experience reports in this narrow sense. Indeed, there are good reasons for doubting the trustworthiness of responses to such general questionnaires, and in some cases, they have even been shown to be at odds with individual reports (Schwitzgebel 2002, 2011, chap. 1; Windt 2013, 2015, chap. 4.3). At best, such general questionnaires tap into opinions about experience, but whether these opinions match the phenomenal character of the corresponding experiences is a separate question. Importantly, questions about the relative trustworthiness of responses to general questionnaires can be meaningfully investigated only if the trustworthiness of at least a subgroup of dream reports is assumed (Windt 2015, chap. 4.4). This subgroup can then act as a baseline and can be used to determine the relative trustworthiness of answers to general questionnaires, but also of different types of reports. While the exact details continue to be debated (for instance on the laboratory effect), there is widespread agreement in scientific dream research that dream reports gathered immediately upon awakening, as is common in laboratory studies using timed awakenings from different sleep stages, are the gold standard against which other types of dream reports (such as home dream diaries compiled following spontaneous awakening) can be measured (again the debate on dream color is a good example; see Hoss 2010; Murzyn 2008; Schredl et al. 2008).
Importantly, as discussed earlier, the assumption that dream reports are trustworthy translates into a research strategy only if reports of nondreaming are taken to be equally trustworthy as reports of dreaming, at least when they are gathered under the same conditions. If the reporting conditions used in a given study are (sufficiently) ideal, it would, surely, be arbitrary to disqualify a subset of these reports on the basis of their content alone. In order to do so, some independent reasons for attributing reports of nondreaming to disturbing factors would be needed. It does not make sense to trust dream reports, but selectively distrust reports of nondreaming gathered under the same conditions and in the absence of any empirical evidence for distrusting them. Put differently, dreams will have to be regarded as reportable experiences, in the sense that given sufficiently ideal reporting conditions, their presence or absence, respectively, can actually be reported. Importantly, both assumptions are implicit in the scientific investigation of dreams. A brief excursion into the history of philosophical and scientific theorizing about sleep and dreaming illustrates this point.
The beginning of scientific dream research coincided with a new experimental paradigm: the practice of obtaining polysomnographic measurements of EEG activity, muscle tone and eye movements from subjects sleeping in the sleep laboratory and of obtaining mentation reports following timed awakenings. This methodology revealed reports of dreaming to be most frequent following awakenings from REM (rapid eye movement) sleep, whereas awakenings from NREM (non-REM) sleep were typically followed by an inability to recall any dreams. In their groundbreaking paper on the correlation between dreaming and REM sleep, Aserinsky & Kleitman (1953) optimistically claimed that that the method of timed awakenings from REM sleep “furnishes the means of determining the incidence and duration of periods of dreaming” (Aserinsky & Kleitman 1953, p. 274; my emphasis).[9] They very naturally took the reports given by their subjects to reflect conscious experience during the preceding sleep period, noting that “of 27 interrogations during [sic] ocular motility, 20 revealed detailed dreams usually involving visual imagery” (Aserinsky & Kleitman 1953, p. 273; my emphasis). Because the method of obtaining reports following timed awakenings in the laboratory is, arguably, the backbone of scientific dream research, this assumption is not unique to Aserinky and Kleitman’s original study. Instead, scientific dream research generally relies on the assumption that dream reports (at least when gathered under ideal reporting conditions, of which timed awakenings in the laboratory are taken to be a prime example) are epistemically transparent in the sense that they are trustworthy sources of evidence about the occurrence and phenomenal character of experience during sleep. I call this the transparency assumption (Windt 2013, 2015).[10]
It is important to see that on its own, the transparency assumption would be insufficient to establish the presumed correlation between dreaming and REM sleep. Claims about the sleep-stage or neural correlates of dreaming require that reports of dreaming and of nondreaming, when gathered under the same conditions, are equally trustworthy: if only reports of dreaming were trustworthy, but reports of nondreaming were not, then the analysis of dream reports would be insufficient to determine the occurrence and frequency of dreams during different sleep stages. Saying that dream reports are transparent is not quite enough: one will also have to assume that dreams are reportable experiences in the sense that had any dream occurred in a given sleep stage, one would in fact be able to report it, at least under sufficiently ideal reporting conditions. I call this the reportability assumption (for details, see Windt 2015, chap.s 3 and 4). Only this added assumption casts reports of dreaming and of nondreaming as equally trustworthy and thus enables reports to be indicative of the occurrence and frequency of dreaming in different sleep stages. The emerging picture is that scientific dream research not just uses dream reports, under the assumption of transparency, to investigate conscious experience during sleep, but that in doing so, it is also methodologically constrained by the space of reportable dreams. Its implicit commitment to the trustworthiness of reports of dreaming and of nondreaming means that it cannot go beyond what is in fact reported without risking internal inconsistency; it can only strive to render further aspects of dreaming reportable. Metaphorically speaking, the space of reportable dreams can be expanded; it can be broadened to cover more aspects of what characterizes typical dreams, or perhaps also to include more diverse types of dreams; and it can be deepened, by probing the unique aspects of certain types of dreams (such as nightmares) or the dreams of certain subject groups (such as meditators) in more detail (see Solomonova et al. 2014). Importantly, this reliance on dream reports is not a liability, a problem to be overcome: it is built into the very nature of dream research. Conversely, studies relying only on the polysomnographic analysis of sleep stages and/or neuroimaging data gathered independently of dream reports do not form part of dream research proper (Windt 2015, chap. 3.2).
How does this account of dream reporting help address the objections to Thompson’s argument discussed at the end of the last section? The strategy of focusing on reports gathered under (sufficiently) ideal reporting conditions and working towards a continuous improvement of these conditions is clearly relevant to the first objection, according to which the mere possibility of veridical reports is not enough. As soon as we broaden our focus from reports of dreamless sleep experience to reports of sleep experience (including dreams) more generally, it becomes clear that scientific dream research has long been centered on the project of identifying and optimizing the trustworthiness of such reports, as well as on determining the adequacy of different kinds of reports for addressing various research questions. Indeed, the very existence of scientific dream research hinges on the assumption that this can be done. Moreover, we have seen that the assumption that reports of dreaming and of nondreaming are equally trustworthy is implicit in this research strategy. This assumption is directly relevant to the second objection, according to which reports of dreamless sleep experience can be used for the investigation of dreamless sleep experience only if they help detect both its presence and its absence.
Moreover, this proposal is, I think, compatible with Thompson’s own strategy of focusing on reports from certain expert groups and improving the wording of questions. Indeed, this strategy of directing participants’ attention to certain aspects of their experience rather than asking for a free report nicely parallels recent work suggesting that a self-scoring method, where participants respond to a standard questionnaire, for instance, about the emotions experienced in a particular dream, is a better measure of dream emotions than data obtained by external raters scoring free dream reports (Sikka et al. 2014; see footnote 8 for discussion). This suggests that Thompson does not mean to reject, as a matter of principle, the claims that conscious experiences are reportable and that an absence of memory is sufficient to infer an absence of experience. Rather, I think his position involves the weaker claim that we should not easily and uncritically trust just any type of experience report to actually reflect the presence of such experience, nor should we easily and uncritically trust just any failure to remember previous experience as indicating an absence of such experience. But this weaker position is in keeping with the account of dream reporting outlined in this section. The challenge then becomes how to narrow the gap between experiences that are in fact reported and those that could (and would) be reported, given sufficiently ideal conditions. I think this is exactly the problem that large parts of report-based dream research are already trying to address.
Note that nothing I have said so far suggests that the transpareny and reportability assumptions are theoretically justified (but see Windt 2013, 2015); if my analysis is correct, however, both are implicit in and in fact crucial for the entire field of scientific dream research. This shifts the burden of proof: while reports of dreamless sleep experience may seem to be an easy target, if only because of the novelty and alleged remoteness of Thompson’s proposal for investigating dreamless sleep experience, we can now see that the proponent of the default view will in fact have to take on the entire field of (report-based) scientific dream research as well. This raises the bar considerably; but first, more has to be said about how the methodological background assumptions of scientific dream research actually parallel questions asked in the classical Indian debate.
To begin with, note that the transparency assumption is analogous to the Advaitin and Yoga claim that upon awakening from dreamless sleep, we can veridically remember and report that we experienced nothing during sleep. To be sure, this type of report describes an experience marked by the absence of the complex imagery and narrative contents that characterize dreaming. Yet, in the Advaitin view, these are reports of an experiential state: in reporting having slept dreamlessly, we are reporting that we experienced nothing, in the relevant sense, during sleep;[11] we are not reporting the absence of experience. Thompson suggests that in order to turn the Advaitin view into a research strategy, the most reasonable and cautious approach is to assume that dreamless experience exists only intermittently, rather than persisting throughout dreamless sleep. The frequency with which dreamless sleep experience is reported to occur upon awakening will then be regarded as indicative of the actual occurrence of such experience. This is analogous to the reportability assumption. To endorse the stronger claim that dreamless sleep experience persists throughout sleep, at least prior to empirical investigation, would be to legislate an answer to the question of dreamless sleep experience. The weaker claim complements the assumption, implicit in scientific dream research, that periods of dreaming contrast with periods of nondreaming, which is quite different from saying that dreaming persists throughout sleep.
By combining my analysis of the methodological background assumptions of scientific dream research with Thompson’s proposal on the investigation of dreamless sleep experience, we can see that if we were to translate the Yoga and Advaitin view into a research methodology, we would find it to rely on assumptions that run parallel to those of scientific dream research. Dreamless sleep experiences, or so a modern-day, scientifically-minded Advaitin would be forced to admit, are reportable experiences; and if it should happen that (under sufficiently ideal reporting conditions, such as immediately after having awakened from sleep) one were unable to recall any such experience having happened during sleep, this would indicate that no such experience had occurred.
This also tells us that reports of nondreaming should be further qualified: reporting the absence of experience during sleep is not the same as reporting dreamless sleep experience. The former is an instance of reporting an absence of experience, the latter is an instance of reporting a form of experience characterized by the absence of intentional objects; but it is still an experience report. Yet, while this requires terminological adjustments and shows that the concept of reporting a state of nondreaming is ambiguous, this adjustment is consistent with the familiar methodology; indeed, it falls out of the methods already used in dream research, when they are applied to the target of dreamless sleep experience.
From this, we can conclude that the default view of dreamless sleep as being characterized by the absence of subjective experience is intrinsically flawed for two related reasons. The first is that by treating dreamless sleep experience as a conceptual absurdity rather than as an open and empirically tractable question, it misconstrues the nature of the question of dreamless sleep experience. The second is that it stands in outright contradiction to the assumptions implicit in the scientific investigation of conscious experience during sleep. Dream research, understood as the scientific investigation of conscious experience during sleep, should be expanded to include dreamless sleep experience as well. And while this certainly will involve an adjustment of its conceptual resources, the good news is that its existing methodological background assumptions can remain largely intact.
The analogy between the Indian debate on dreamless sleep experience and the background assumptions of scientific dream research not only highlights the inconsistency of the default view. There are also valuable lessons to be learned in the other direction, and considering the historical Indian debate can enrich contemporary debates on the status of dream reports as well. In particular, note that it is one thing to say that scientific dream research is implicitly committed to the transparency and reportability assumptions; but it is another to say that these assumptions are also theoretically justified. Elsewhere, I have defended the view that explanatory considerations justify the transparency and reportability assumptions: construing dream reports as (largely veridical) memory reports provides a better explanation of dream reporting behavior than skeptical alternatives that construe dream reports as the result of inference, misremembering or outright confabulation (Windt 2013, 2015, chap. 4). Here, I want only to point out that similar considerations apply to reports of dreamless sleep experience. In fact, Thompson’s response to the Nyāya argument against dreamless sleep experience shows that casting reports of having slept dreamlessly as based on inference rather than memory is not a proper explanation at all. Instead, it leads to an argument that either results in an infinite regress or is circular. Again, there is a striking similarity to a similarly skeptical account of dream reporting from the 20th century. This time, the analogy with the historical Indian position will lend additional support to anti-skepticism about dream reporting.
To see why, another brief excursion into the history of theorizing about scientific dream research is instructive. Let us consider Norman Malcolm’s (1956, 1959a) skeptical argument against the claims that dreams are conscious experiences occurring during sleep and that dream reports transparently show this to be the case. This argument was a direct reaction to early attempts, following the discovery of REM sleep, to operationalize dreaming as a REM sleep phenomenon. Malcolm’s argument hinges on the conceptual claim that “if a person is in any state of consciousness it logically follows that he is not sound asleep” (Malcolm 1956, p. 21). According to Malcolm, even though we use the same language to describe dreams and waking experiences, dreams (or at least such dreams as occur during sound sleep, which Malcolm, again for conceptual reasons, takes to be representative of dreaming proper) are not experiences, and for the same reason dream thoughts, feelings, and emotions are not real instances of their kind. As Malcolm puts it,
if a man had certain thoughts and feelings in a dream it no more follows that he had those thoughts and feelings while asleep, than it follows from his having climbed a mountain in a dream that he climbed a mountain while asleep. (Malcolm 1959a, pp. 51-52)
Malcolm’s view is complex and a detailed discussion is beyond the scope of this commentary; suffice it to say that one of its more controversial upshots is that dream recall is not a real instance of remembering experience during sleep. Instead, “statements of the form ‘I dreamt so and so’ are always inferential in nature” (Malcolm 1959a, p. 65): one infers that one has dreamt when one realizes, upon awakening, that the events one seems to remember did not in fact occur. This claim struck many of his critics as contradicting both the common-sense understanding and the phenomenology of dream recall (see Dunlop 1977 for a collection of some of the most important critical essays; see Windt 2013, 2015, chap. 1 for discussion). Elsewhere, (Malcolm 1959b) explains that he takes dream recall to be inferential not in the psychological sense of actually drawing this inference when we notice that we have dreamt, but in the sense that we could give grounds for our belief that we dreamt if pressed to do so. However, because he fails to clarify what exactly these grounds are, his account remains sketchy. By applying Thompson’s reconstruction of the Nyāya syllogism to Malcolm’s claim, it quickly becomes clear that even a more complete reconstruction of the inference would be intrinsically flawed. The result would be something like this:
While I was sound asleep, I had no experiences, including sensations, conscious thoughts, feelings, beliefs, or emotions.
This is because (i) I was in a special state (that is, not awake) or (ii) I lacked the necessary means for having experiences, including sensations, conscious thoughts, feelings, beliefs, or emotions (that is, my senses and mental faculties were shut down).
Whenever (i) I am in a special state (that is, whenever I am not awake) or (ii) I lack the necessary means for having experiences, including sensations, conscious thoughts, feelings, beliefs, or emotions (whenever my senses and mental faculties are shut down), I do not have experiences, including sensations, conscious thoughts, feelings, beliefs, or emotions.
As in the case of fainting or a blow to the head.
While I was sound asleep, I had no experiences, including sensations, conscious thoughts, feelings, beliefs, or emotions.
Malcolm concludes that sound sleep is comparable to other states of unconsciousness, and “to a person who is sound asleep, ‘dead to the world,’ things cannot even seem” (Malcolm 1956, p. 26).
If we follow this reasoning, then dream reports cannot ever be veridical experience reports: if we cannot have thoughts, feelings or emotions during sleep, then we also cannot have them during dreams, and we cannot actually remember (or veridically report) having had them after awakening. Rather, we sometimes awaken with the impression of having had such thoughts, feelings and emotions during sleep; and when we realize that they did not in fact occur, we infer that we dreamt.
To be fair, there might well be cases in which dream recall does have such an inferential nature. To use Malcom’s example, it seems possible that I could awaken with the particularly vivid impression of having climbed a mountain and then might realize, from the simple fact that I was lying in bed and nowhere near a mountain, that I had not actually climbed a mountain, but had been asleep. However, even if I was now quite sure that I had merely dreamt that I had climbed a mountain, it would not follow that the thoughts and feelings I remember having in the dream did not really occur. In order to draw this further inference, I would have to know that dreaming is a special state that is devoid of any experiences whatsoever.[12] As is the case for the Nyāya syllogism, this immediately invites the dual threats of circularity and of infinite regress: If I say I was in a special state because the thoughts and feelings I experienced in my dream were not real instances of their kind, I am reasoning in a circle. And if I say that I was in a special state because the mental faculties required for having thoughts and feelings were shut down (or because, as would better befit Malcolm’s argument, I had temporarily lost the capacity for producing the types of behavioral evidence that would enable another person to verify that I had been dreaming), then independent evidence would be needed—and so on. Again, without appealing to memory, no such evidence is available.
At this point it might seem that there is an easy solution: Perhaps, independent evidence for saying that dreaming is a special state has, in the meantime, become available. Malcolm’s analysis of dreaming was a direct reaction to early studies, discussed in section 2_1, on the correlation between REM sleep and dreaming, and his argument made much of the alleged impossibility of acquiring independent evidence, over and above dream reports, for the occurrence of dreams during sleep. Among Malcolm’s critics, there was widespread agreement that he was simply mistaken about this latter point: sleep behavior, (for instance in patients with REM sleep behavior disorder, who are thought to act out their dreams due to a loss of REM sleep-related muscular atonia; see Schenck 2005, Valli et al. 2012) sleep talking, and also polysomnographic measurements were (and continue to be) thought to provide exactly such independent evidence, perhaps even to the point of enabling researchers to verify dream reports (see for instance Ayer 1960; Rosen 2013; signal-verified lucid dreams are another example, as proposed by Revonsuo 2006; see sec. 4_1 for a fuller discussion). Yet, even though the appeal to scientific dream research slightly changes the content of the argument, this merely restates the familiar syllogism, including its problems in a new guise.
To see why, let’s say that rapid eye movements had indeed been found (as stated by the so-called scanning hypothesis; see Dement & Kleitman 1957, to be directly related to visual dream imagery. Could we now analyze these eye movement patterns to diagnose the occurrence (and perhaps even the content) of dreaming even in the absence of (or in contradiction to) dream reports (see Dennett 1976 for the discussion of this possibility)? Note that this is not an abstract philosophical issue: dream researchers have long dreamt the dream of moving beyond dream reports in the study of dreaming altogether. This ranges from science fictional visions of televising dreams (Hall & Van de Castle 1966) or of perhaps modeling them as an immersive, interactive virtual environment, as in Antti Revonsuo’s (2006, pp. 300-303) dream catcher test, to real-world attempts to predict the content of dream reports from behavioral (Leclair-Visonneau et al. 2010) or neuroimaging (Horikawa et al. 2013) data. Again, the idea is that in the future, the analysis of neuroimaging data might be a way to verify dream reports, or even to move beyond their collection and analysis altogether. Elsewhere, I have argued that such attempts are circular: Dream reports, under the assumption of transparency, are used to identify potential sleep-stage and neural correlates of dreaming; but the evidence such potential correlates provide is only as strong as the correlation, and so one cannot then turn around and use such measures as independent evidence to verify dream reports. Now, the Nyāya syllogism and its failure present a nice and crisp illustration of why this is the case. I think this is a nice example of the fruitfulness of a cross-cultural perspective on the methodological and conceptual issues involved in studying the occurrence of consciousness during sleep.
But there is another lesson to be learned. This is that the Nyāya syllogism is not an outdated problem, but one that persists even if we place it in the context of scientific dream research. The question of whether reports of having slept dreamlessly are experience reports or inferential is not of mere theoretical interest, but makes a real difference: assuming such reports, at least when given under ideal reporting conditions, to be veridical memory reports is the condition for a report-based scientific investigation of the relevant experiences in the first place. The historical debate, and Thompson’s reconstruction of it, nicely highlights the need for acknowledging the relevance of first-person reports. Together, they also strengthen the theoretical case against skepticism about the trustworthiness of dream reports. With this anti-skeptical account in place, we can now move forward. In the next section, I sketch the outlines of a conceptual framework for describing dreamless sleep experience and its relation to dreaming.