When epistemologists focus on questions or problems directed at the nature of knowledge, particular domains of knowledge, or sources of knowledge, they usually assume that the knowing subjects they discuss are in a particular state of consciousness. They usually assume that the subjects in question are awake. In consequence, the epistemological significance of that background state of consciousness—wakefulness—is rarely discussed explicitly. The most obvious exceptions to this generalization are to be found in certain discussions of dream scepticism. Some address the question of whether and how one can know that one is not dreaming by suggesting that there is an epistemic asymmetry between dreaming and wakefulness that a response to dream scepticism might exploit; and among those who adopt this strategy, proposals are made about the differences between these states of consciousness that would explain that epistemic asymmetry. These hypotheses about the epistemic deficits associated with dreaming prompt us to consider what is so often taken for granted in epistemology—the nature of, and the epistemic significance of, the wakeful state of consciousness.

Ernest Sosa’s paper ‘Dreams and Philosophy’ perhaps best exemplifies the approach to dream scepticism that I have in mind.Footnote 1 In that paper Sosa offers a response to dream scepticism that depends on the claim that when you are dreaming there are impediments to your being able to tell whether or not you are dreaming that are absent when you are awake. Sosa is not the first to adopt that strategy, but what is distinctive in his proposal is his suggestion that the respect in which the dreaming subject is epistemically deprived is to be explained by a metaphysical thesis about the constitution of dreams. Sosa rejects what he takes to be the orthodox conception of dreaming —the view that dream states and waking states are “intrinsically alike, though different in their causes and effects” (2005: p. 7). According to this orthodox conception of dreaming, when you dream you undergo hallucinatory experiences that lead you to acquire false beliefs about what is then occurring and you make judgements and form intentions on that basis. The alternative that Sosa proposes is that “to dream is to imagine” (2005: p. 7). According to this imagination model of dreaming, our dreamt conscious beliefs, experiences, affirmations, decisions and intentions are not “real” insofar as they are all merely imagined beliefs, experiences, affirmations, decisions and intentions. Sosa goes on to argue that this imagination model of the constitution of dreams has significant implications for debates about dream scepticism. If Sosa is right about this, then there may be a metaphysical difference between the respective states of consciousness—dreaming and wakefulness—which can not only provide a novel response to dream scepticism, but which may also provide clues to the nature of, and the distinctive epistemic significance of, the wakeful state of consciousness.

My aim in this paper is to assess the epistemic implications of Sosa’s imagination model of dreaming and consider whether his metaphysical thesis about the constitution of dreams can illuminate our understanding of the epistemological significance of wakefulness. In Sect. 1 I outline and assess the reasons Sosa gives for thinking that his imagination model of dreaming introduces a new dimension to debates about dream scepticism. Sosa argues that his imagination model of dreaming invites a more radical version of dream scepticism and also makes available a novel and more powerful response to dream scepticism. I shall raise objections to both of those claims. This will lead me to consider an epistemically based challenge to Sosa’s imagination model of dreaming. This is the concern that Sosa’s imagination model of dreaming lacks the resources to accommodate the intuition that there is something illusory or misleading about one’s situation when one is dreaming, and as a result his account of dreams fails to accommodate the common intuition that there is a sceptical problem about dreaming but not about dreamless sleep.

In Sect. 2 of the paper, taking my cue from Brian O’Shaughnessy’s (2002) discussions of dreaming, I shall elaborate a version of the imagination model of dreaming that can overcome that challenge. This version of the imagination model of dreaming goes beyond what Sosa explicitly commits to when he outlines his view of dreams. However, it exploits ideas that are integral to a key theme in Sosa’s recent writings on virtue reliabilism—namely his proposal that epistemic agency should be accorded a central place in that approach to knowledge, and his related proposal that agency is exercised in conscious judgement.Footnote 2

In the final section of the paper I shall consider whether this version of the imagination model of dreaming has anything novel to contribute to debates about dream scepticism, and what it implies about the nature and epistemic significance of the wakeful condition. Here I will be suggesting that if this version of the imagination model of dreaming is correct, then central to an account of the nature and epistemic significance of wakeful consciousness should be an elucidation of a connection between the wakeful condition and our capacity to exercise agency over our mental lives.

1 Sosa’s imagination model of dreaming

As I previously mentioned, Sosa is not the first to suggest that there is an epistemic asymmetry between dreaming and wakefulness that a response to dream scepticism might exploit. That is a proposal that is perhaps most famously associated with Bernard Williams’ discussion of dream scepticism in an appendix to his book on Descartes’ Meditations. Footnote 3 In that appendix Williams attributes to Descartes a view that he (Williams) thinks should be rejected as absurd. This is the proposal that “The standpoint of rational enquiry is neutral between dreaming and not dreaming” (1978: p. 312).Footnote 4 Williams claims that we should reject as implausible the following explanation of one’s failure to tell whether one is dreaming when one is dreaming: as one dreams one is able to exercise the ability to rationally decide, and come to rational conclusions about whether one is dreaming, but one’s experience is such as to lead rational decision to the wrong conclusion. According to Williams, it is far more plausible to hold that there are impediments to your being able to tell whether you are awake or dreaming when you are dreaming that are (usually) absent when you are awake, and he likens the dream state to conditions in which one is drugged or severely drunk. In such conditions one may be unable to come to any conclusions, or at least any rational conclusions, about anything. But such epistemic impediments are absent in the sober, wakeful condition. So the standpoint of rational enquiry is not neutral between dreaming and wakefulness.

The relevance of this claim to a potential response to dream scepticism is expressed by Williams in the following passage:

The idea of one’s being able to tell whether S seems to entail the conjunction of two things:

  1. (a)

    One can tell that S when S;

  2. (b)

    One can tell that not-S when not-S.

We may be tempted to think that (a) and (b) must go together. The vital point... is that they do not necessarily go together (1978: p. 310)

Williams points out that for some values of ‘S’, one of these conjuncts can obtain without the other. For example, suppose that ‘S’ is ‘one is dead’, or ‘one is in dreamless sleep’, or ‘one is severely drunk’. You cannot tell whether you are dead when you are dead, but from this it doesn’t follow that you cannot tell whether you are dead when you are alive. You cannot tell whether you are unconscious when you are unconscious, but from this it doesn’t follow that you cannot tell whether you are unconscious when you are conscious, and so on. Williams suggests that dreaming likewise fits this pattern. He suggests that it is plausible to say that “one cannot tell that one is dreaming when one is dreaming—or at least rarely and randomly so, if at all” (1978: p. 312), but from this it doesn’t follow that you are unable tell whether you are awake when you are awake.

This much Sosa agrees with. When comparing Williams’ response to dream scepticism with his own response, Sosa remarks, “We both rely on what dreaming shares with being unconscious or dead: i.e., we both rely on your ability to tell that you avoid such a fate, when you do, despite your inability to tell that the fate befalls you when it does” (2005: p. 17). However, Sosa also remarks that their responses to dream scepticism are, nonetheless, “substantially different and incompatible on the whole” (2005: p. 17). And according to Sosa, this is a consequence of a disagreement between them on the question of the metaphysical constitution of dreams. Sosa remarks, “The crucial respect of difference is that for Williams we do have real conscious beliefs and experiences in dreaming” (2005: p. 17). The alternative that Sosa proposes is that “to dream is to imagine” (2005: p. 7), and according to that alternative model of dreaming, we do not have “real” conscious beliefs when we dream, for as we dream, our dreamt conscious beliefs, experiences, affirmations, decisions and intentions are all merely imagined beliefs, experiences, affirmations, decisions and intentions.

Sosa suggests that this departure from Williams on the question of the metaphysical constitution of dreams bears on the epistemological issue of dream scepticism in at least the following two ways: the imagination model of dreaming invites a more radical version of dream scepticism, and it also makes available a novel and more powerful response to dream scepticism—a response that isn’t available to those who hold, with Williams, that we do have real conscious beliefs in dreaming. He argues that if the imagination model of dreaming is correct, then a more radical version of dream scepticism looms, because according that model, dreamt conscious beliefs, experiences, affirmations, decisions etc. are not “real” insofar as they are all merely imagined. In which case, writes Sosa, “The possibility that we dream now threatens not only our supposed perceptual knowledge but even our supposed introspective knowledge, our supposed takings of the given. It is now in doubt not only whether we see a fire but even whether we think we see a fire, or experience as if we see it” (2005: p. 10). However, Sosa further claims that the imagination model of dreaming also makes available a new and more powerful response to dream scepticism, for it puts one in a position to defend the claim that “<I am hereby awake> shares the special epistemic status of the cogito” (2005: p. 18).

Let us first consider the second of those two proposals—the proposal that the imagination model of dreaming makes available the claim that <I am awake> shares the epistemic status of ‘cogito’ propositions. Sosa argues that in the case of a cogito proposition, such as <I think> or <I am>, one knows by reflection that when faced with the options of believing the proposition, disbelieving the proposition, or suspending judgement, believing the proposition is the best epistemic option. This is because one knows by reflection that disbelieving such a proposition will be defective, since self-defeating; and one also knows by reflection that if one suspends judgement on the question of the truth of the proposition one will thereby fail to avail oneself of a better epistemic option—i.e. the option of acquiring or retaining the correct answer to the question. So in the case of a cogito proposition, he says, “the believing option is the only one about which I know ahead of time that my taking it will obviously imply that I am epistemically right in so doing” (2005: p. 15).

Sosa goes on to argue that on the imagination model of dreaming, the proposition <I am awake> also has this special epistemic status. This is because according to the imagination model, all dreamt conscious beliefs and affirmations are merely imagined conscious beliefs and affirmations, and not real conscious beliefs and affirmations. So in the case of the proposition <I am awake>, I know that believing is the “the only epistemically undefective option” 2005: p. 15). He writes that in the case of the proposition <I am awake>:

Both suspending judgment and disbelieving will share the following feature: that I know ahead of time, as I ponder my question, that I am better off epistemically if I take a particular other option, namely, the belief option, since only about that option is it obvious to me now that if I take it I will be right (2005: p. 15).

Sosa claims that Williams is denied access to this response to dream scepticism. Williams cannot claim that <I am awake> shares the special epistemic status of the cogito, for in granting that subjects may have “real” conscious beliefs when dreaming, Williams thereby allows that it is possible for the conscious belief that one is awake to be false.

However, I think we can question whether Sosa has done enough to secure his distinctive proposal that <I am awake> shares the special epistemic status of cogito propositions, for this proposed response to dream scepticism rests on an assumption that has not been established. It rests not just on the claim that the imagination model of dreaming is in fact correct, but also on the further claim that one knows by reflection that the imagination model of dreaming is obviously correct. It is only by establishing that one knows by reflection that the imagination model of dreaming is obviously correct that Sosa can secure the claim that <I am awake> shares the special epistemic status of the cogito; for only so can one know by reflection, “ahead of time”, that it is “obvious” to one that in the case of the proposition <I am awake> one is bound to be epistemically better off if one takes the belief option.

The general point here is that from the reflective standpoint of a Cartesian meditator who is seeking to escape scepticism by searching for propositions with the special epistemic status of the cogito, the availability of a response to dream scepticism will turn not just on the question of whether a particular account of the nature and constitution of dreams is in fact correct, but also on the question of whether one can know by reflection which particular account of the nature and constitution of dreams is correct.Footnote 5 From that perspective, one will be able to help oneself to Sosa’s imagination model of dreaming in responding to dream scepticism only if one has assured oneself that there are no grounds for doubting that the imagination model of dreaming is correct. However, even if the imagination model of dreaming is correct, one might question whether it is obvious to reflection that it is correct; and Sosa has not done enough to show that it is.Footnote 6

Earlier I said that Sosa suggests that his departure from Williams on the question of the metaphysical constitution of dreams bears on the epistemological issue of dream scepticism in two ways: the imagination model of dreaming invites a more radical version of dream scepticism, and it also makes available a novel and more powerful response to dream scepticism. Having just raised an objection to the second of these claims, I now want to consider an objection to the first. The objection has it that far from introducing a more radical version of dream scepticism, Sosa’s imagination model of dreaming fails to accommodate the intuition that dreaming poses a sceptical threat at all.

To see how this objection might be developed it will help to consider Williams’ ordering of different sorts of explanations that can be given of why a subject may be unable to tell whether S when S, where such explanations do not automatically call into question that subject’s ability to tell whether S when not-S. Williams orders these different forms of explanation from the more drastic to the less drastic. He suggests that a subject A may be unable to tell whether S when S if S is such that:

  1. 1.

    A does not exist to judge anything. (E.g. death).

  2. 2.

    A cannot do anything, or more particularly cannot do anything of which she is conscious. (E.g. dreamless sleep).

  3. 3.

    (a) A cannot tell that anything, come to any conclusions about anything, make any judgement. (E.g. some kinds of drugged condition). (b) A cannot rationally tell that anything, come to any rational conclusion about anything. (E.g. severe drunkenness).Footnote 7

Williams proposes that the most plausible explanations of why a subject is unable to tell whether she is dreaming when she is dreaming will belong to either 3(a) or 3(b). In (3) cases, the subject doesn’t have available to her a rational way of telling anything, but this isn’t because the subject is incapable of doing anything of which she is conscious [as in (2)]. Williams assigns dreaming to (3) cases, because he thinks that an adequate explanation of why a subject is unable to tell whether she is dreaming when she is dreaming should have the resources to accommodate the common intuition that there is a sceptical problem about dreaming but not about dreamless sleep. Assigning dreaming to (3), rather than (2), is what is minimally required if we are to accommodate that intuition. And this may also be why he adds that in (3) cases, “A may have false beliefs” (1978: p. 313).

One way of expressing this thought is as follows. In certain cases a subject may be unable to tell whether S, because her situation is so epistemically impoverished that there is no question of its seeming to her as if S, and also no question of its seeming to her as if not-S, and hence in such cases there is no question of whether mistakes, illusion, or deception can get a grip. For example, during dreamless sleep, there is no question of whether it misleadingly seems to one as if one is awake. During dreamless sleep there is no question of one being deceived or misled about one’s state of consciousness. That makes scepticism that questions one’s grounds for ruling out the possibility that one is now dreamlessly sleeping appear to be obviously impotent. However (so the thought goes), even if dream scepticism can be answered, it is not so obviously impotent. So a condition of adequacy on an explanation of why a subject cannot tell that she is dreaming when she is dreaming is that it should have the resources to accommodate at least something erroneous, misleading or deceptive about one’s situation when one is dreaming. A concern that might be raised for Sosa’s imagination model of dreaming is that it lacks the resources to accommodate that idea, and as a result the account fails to accommodate the common intuition that there is a sceptical problem about dreaming but not about dreamless sleep. That line of objection can be elicited by enquiring further into the nature of the connection between the metaphysics and epistemology of dreaming that is assumed on Sosa’s imagination model.

Sosa proposes that his imagination model of dreaming commits to the idea that waking states and dream states are “constitutively different” (2005: p. 11). However, the claim that “to dream is to imagine” does not by itself entail that there is a “constitutive” difference between waking states and dream states, given that one can imagine things when one is awake. So in what respects are dreaming states and wakeful states constitutively different according to Sosa’s imagination model of dreaming? To that metaphysical question we can add the following epistemological question. When we are awake and imagining, we are not at all deceived or misled by our imaginings, or rarely so. So how does the claim that “to dream is to imagine” explain the epistemic difference between our dreams and our wide-awake imaginings?

On one understanding of Sosa’s imagination model, the second of these two questions (i.e. the question about the epistemic difference between our dreams and our awake imaginings) is addressed by a metaphysical assertion: the assertion that when we are asleep and dreaming we lack “real” conscious beliefs. The thought here is that when we are awake, we can have knowledge of our conscious mental life, including our conscious imaginings, but in sleep we lack this knowledge, because in sleep we lose the ability to acquire conscious beliefs at all. Any candidate mental phenomenon that we, as theorists, might have taken for a real conscious belief is in fact an imagined conscious belief. Our dreams are constituted by imaginings and only imaginings, and this amounts to a “constitutive” difference between dreaming and the wakeful state.

However, a potential problem with that proposal is that it appears to rule out too much knowledge and commit to too much ignorance in the dreaming condition. For it appears to commit to the claim that as we dream we are completely ignorant of concurrent conscious phenomena. In denying that we are capable of acquiring any conscious beliefs at all when we dream, this response would appear to commit to the idea that as we dream we are not only ignorant of the fact that we are imagining, we are also entirely ignorant of what we are imagining. It commits to the proposal that whenever we dream, as we dream, we are completely ignorant of the contents of those dreams. For if we are not, after all, ignorant of the contents of our dreams as we dream—if we have at least some knowledge of the content of what is being dreamt as we dream, even if that knowledge is relatively sparse and fleeting—then we must have ‘real’ beliefs about the contents of those dreams.

Arguably, one will need to avoid committing to the idea that we are completely oblivious to the contents of our dreams as we dream if one is to accommodate the common intuition that there is a sceptical problem about dreaming but not about dreamless sleep. So arguably one ought to concede that as we dream we are capable of having conscious beliefs, where this is understood as the claim that as we dream we are capable of having beliefs about concurrent conscious phenomena—and in particular, beliefs about the contents of our dreams. Would making that concession mean abandoning an imagination model of dreaming and abandoning Sosa’s proposal that there is a “constitutive difference” between dreaming and wakeful states?

Before I address that question in the next section, it will help to clarify the notion of a “constitutive difference” between dreaming and wakeful states. Here is one way of understanding that idea. The metaphysical constitution of dreams is such that the difference between dreaming and wakeful states of consciousness is a difference in kind that isn’t fully captured by Williams’ suggestion that dreaming is akin to being in a drugged condition or severely drunk. For the constitution of the stream of consciousness in the dreaming condition differs from the constitution of the stream of consciousness when one is in a drugged condition or severely drunk, insofar as the latter conditions are consistent with being awake. And this is because the constitution of the stream of consciousness in the wakeful condition (even in degraded forms of the wakeful condition, such as drunkenness) is such that its occurrence is sufficient for wakeful consciousness.

So the question I now want to address is this. Is there a version of the imagination model of dreaming that can accommodate that idea of a constitutive difference between dreaming and wakeful states, without falling prey to the objection that the model fails to accommodate the common intuition that there is a sceptical problem about dreaming but not about dreamless sleep?

2 Dreaming as seeming-to-be-awake

There is something intuitive to the idea that when we wake from a dream, the experience can be somewhat akin to the unmasking of an illusion. That thought is nicely expressed by Brian O’Shaughnessy, who suggests that in such circumstances, wakeful consciousness acts as something like a “corrective”. He writes, “on waking we discover that ‘the world is not like that”’. Wakeful consciousness acts as corrective “of something that was ‘passing itself off as’ precisely that corrective agency—rather as a man might unmask an imposter posing as himself by appearing in person upon the scene” (2002: p. 425).

The idea that O’Shaughnessy articulates here is, I suggest, sufficient to accommodate the point that Williams was concerned to capture in his response to dream scepticism—the intuition that there is a sceptical problem about dreaming but not about dreamless sleep. O’Shaughnessy’s suggestion is that dreaming is akin to seeming-to-be-awake—a seeming state of consciousness that is unmasked for what it is not by the genuine article. The sceptical thought this invites is the following: How do you know that your current state of consciousness is not just such a counterfeit condition? Only waking could show it to be so, and you cannot be certain that waking doesn’t lie ahead of you. Dreamless sleep, by contrast, is no such counterfeit condition. It is not a seeming state of consciousness. When one is in it, it neither seems to be what it is, nor seems to be what it is not. I submit, then, that if the imagination model of dreaming has the resources to accommodate O’Shaughnessy’s characterisation of dreaming as akin to seeming-to-be-awake, then it can meet the challenge I identified earlier—the challenge of accommodating the common intuition that there is a sceptical problem about dreaming but not about dreamless sleep. So how might a proponent of the imagination model of dreaming go about doing this?

Suppose that the proponent of the imagination model of dreaming concedes that the dreaming subject is not entirely ignorant of what she is dreaming when she is dreaming. That concession may mean granting that “real” conscious beliefs can accompany our dreams. However, that concession may yet be consistent with another one of Sosa’s central claims—the proposal that while we dream, all of our dreamt conscious affirmations, all of our dreamt conscious acts of judging, all of our dreamt choices, are merely imagined and not “real”.Footnote 8 The version of an imagination model of dreaming that combines those claims requires some filling out, and I now want to consider one such filling out. According to it, sleep induces the temporary incapacitation of one’s ability to consciously affirm, judge, and decide; but it does not incapacitate one’s ability to consciously imagine (which includes an ability to consciously imagine judging, affirming and deciding). In this condition, when one imagines, one is not entirely ignorant of what one is imagining, but one cannot tell that one is merely imagining.

This particular way of filling out the imagination model of dreaming naturally gives rise to the following questions. (1) What is the connection between (i) the incapacitation of one’s ability to consciously affirm, judge, and decide, and (ii) one’s inability to know that one is merely imagining? And (2) what is the difference between (i) consciously affirming that p and (ii) imagining consciously affirming that p while being unable to know that one is merely imagining? I suggest that we can begin to address these questions by invoking an idea that Sosa emphasises and elaborates in more recent work, and in particular in his recent book Judgement and Agency. This is the proposal that affirming and judging are agential mental acts. Moreover, I want to suggest that by invoking Sosa’s proposal about the agency that is exercised in such mental acts, we can accommodate a respect in which there is a constitutive difference between waking states and dream states, and we can also accommodate O’Shaughnessy’s characterisation of dreaming as akin to seeming-to-be-awake.

Let us start with question (2) above: what is the difference between (i) consciously affirming that p and (ii) imagining consciously affirming that p while being unable to know that one is merely imagining? It is not clear that there would be an obvious phenomenological difference between these conscious episodes, and so given their close phenomenological correspondence, one might wonder: what role does the former play in one’s mental life which cannot be played by the latter? Are we helped in answering this question if we assume that our conscious affirmations are mental actions? When we imagine performing some action, we do not thereby perform that imagined action, and the fact that we do not thereby perform the imagined action does not leave in its wake a failed attempt to perform that imagined action.Footnote 9 Moreover, I take it that the aforementioned line of thought doesn’t only apply if one knows that one is merely imagining performing the action in question. That is to say, one’s ignorance that one is merely imagining does not convert the imagined act of \(\upphi \)-ing into a genuine act of \(\upphi \)-ing, nor does it result in a failed attempt to so act. So if we assume that our conscious affirmations are mental actions, we have a straightforward way of preserving the idea that there is a crucial difference between an act of conscious affirmation on the one hand, and an imagined act of conscious affirmation on the other, even in a case in which one fails to realise that one is merely imagining an act of conscious affirmation.Footnote 10

Now to question (1): What is the connection between (i) the incapacitation of one’s ability to consciously affirm, judge, and decide, and (ii) one’s inability to know that one is merely imagining? If we assume that one’s conscious affirmations, judgements, and decisions all involve the exercise of one’s agency, then the simplest explanation of the incapacitation of one’s ability to consciously affirm, judge, and decide during sleep is the hypothesis that sleep induces the temporary incapacitation of one’s ability to exercise agency over one’s mental life. According to this hypothesis, when one sleeps, one cannot perform mental actions. In consequence, when one sleeps, a particular avenue to self-knowledge is closed off to one—namely the sort of practical self-knowledge that arises through the agency one exercises over one’s conscious mental life. The proposal about the connection between (1)(i) and (1)(ii) that this invites is the following. This particular source of self-knowledge (i.e. the distinctively practical self-knowledge that arises through the agency one exercises over one’s conscious mental life) plays a crucial role in explaining how we know what we are up to when we know that we are merely imagining (or when we realise that we have been imagining, as when we catch ourselves adrift in the midst of a daydream). When our capacity to exercise agency over the course of our mental lives is knocked out, so too is our practical knowledge of what we are up to, and in consequence so too is our ability to know that we are (or have been) merely imagining. On the assumption that not all knowledge of one’s conscious mental life is practical self-knowledge that arises in and through the exercise of one’s agency, this proposal is consistent with the claim that there are forms of self-knowledge that are not knocked out when one dreams—e.g. knowledge of the content of those dreams.Footnote 11

This response to question (1) now suggests a way of cashing out the proposal that there is a constitutive difference between waking states and dream states. According to this cashing out of the proposal, the constitutive difference between these two states of consciousness results from the incapacitation of one’s ability to exercise agency over the course of one’s mental life when asleep. This version of a ‘non-orthodox’ view of dreaming has been argued for by O’Shaughnessy (2000, 2002), and more recently by Thomas Crowther (forthcoming). O’Shaughnessy writes, “Consciousness [i.e. wakeful consciousness] necessitates an overall mental activeness, for the reason that the conscious [i.e. the awake] are in control of the overall movement of their own minds, and the dream is an essentially inactive phenomenon” (2002: p. 426). He suggests that “When one is awake one is ‘master in one’s own mental house’—a fact which is in no serious conflict with the unceasing spontaneity of thought, the continuing unanticipatedness of one’s next thought. After all, to be a master in one’s own mental house is not the same thing as being a tyrant!” (2000: p. 201). According to O’Shaughnessy, although many conscious occurrences in the awake subject are not chosen, the flow of conscious thought and imagination is nonetheless “active as a whole”. So, he suggests, the “overall mental activeness” that is necessitated in wakeful consciousness ensures that the flow of the stream of consciousness in the awake subject is “distinctive and inimitable”. It cannot be reproduced in dreaming, which is essentially inactive. Hence, “the stream of consciousness of the conscious [i.e. the awake] is a sufficiency for [wakeful] consciousness” (2002: p. 426).Footnote 12

This proposal about the constitutive difference between dreaming and wakeful consciousness in turn offers a way of accommodating O’Shaughnessy’s characterisation of dreaming as being akin to seeming-to-be-awake. For we now understand dreaming as an essentially inactive phenomenon that involves seeming-to-be-active. According to that understanding, when one dreams, one undergoes passive episodes of imagining, at least some of which involve imagining oneself to be active. This merely imagined activeness is not the genuine article and in its absence (which is to say, when one is unable to exercise any active agential control over one’s conscious mental life), one cannot tell that one is merely imagining. In consequence, the counterfeit nature of that merely imagined activeness is concealed from one. It can only be revealed for what it is not when one awakes.

Earlier I suggested that if the imagination model of dreaming has the resources to accommodate O’Shaughnessy’s characterisation of dreaming as akin to seeming-to-be-awake, then it can meet the challenge of accommodating the common intuition that there is a sceptical problem about dreaming but not about dreamless sleep. I have now outlined a version of the imagination model that has those resources. In the final section of the paper I shall consider whether this version of the imagination model of dreaming has anything novel to contribute to debates about dream scepticism. Does it invite a more radical version of dream scepticism, or provide a novel response to dream scepticism? And what does it imply about the nature and epistemic significance of wakeful consciousness?

3 Dream scepticism and the epistemic significance of the wakeful condition

Does the imagination model of dreaming that I outlined in the previous section supply us with a more radical and powerful form of dream scepticism? The dream situation that it envisages is the following. One cannot consciously affirm, judge, or decide. One cannot exercise any control over the general direction taken by one’s conscious mental life. One is incapable of exercising any agency over the occupation of one’s conscious attention. In short, one suffers something like a form of mental paralysis—a form of mental paralysis that doesn’t even allow for the possibility of failed attempts to exercise agency over one’s thinking. But it is a form of mental paralysis that is accompanied by the illusion of agency—the illusion of seeming to affirm, seeming to judge, seeming to decide, seeming to be mentally active. A sceptical argument that purports to show that you cannot rule out the possibility that this is your current situation would certainly seem to be more radical than anything that Descartes envisaged. And notice that this sceptical argument need not depend on the claim that this is in fact your situation when you dream. That is to say, it need not depend on the assumption that this version of the imagination model of dreaming is in fact correct. It can deploy the weaker assumption that you cannot rule out by reflection that this imagination model of dreaming is correct. So what response can be given to this form of dream scepticism?

This looks to be a form of scepticism to which a Williams-style response can be given. According to this sceptical line of thought, when you suffer from the envisaged form of mental paralysis, you are unable to know that you are merely imagining; so you are not in a position to know that your imagined conscious affirmations, judgements, and decisions are not “real”. And this is so, because when you suffer from this form of mental paralysis an avenue to self-knowledge is closed to you—the avenue to practical self-knowledge that arises in and through the exercise of your mental agency. The Williams-style response that this sceptical line of thought naturally prompts is the following. From the fact that you are deprived of practical self-knowledge when you are unable to exercise your mental agency it of course doesn’t follow that you are deprived of that source of self-knowledge when you are able to exercise your mental agency. And so from the fact that you are unable to know that you are merely imagining when you imagine during sleep, it doesn’t follow that you are unable to tell whether you are merely imagining when you are awake.

However, what if you were to adopt a higher-order reflective stance on the question of whether this avenue to self-knowledge is currently available to you? Would it be arbitrary of you to assume that this avenue to self-knowledge is available to you? In response one can say the following. If the knowledge in question is practical self-knowledge that arises in and through the exercise of your mental agency, then this isn’t the sort of knowledge that you can successfully distance yourself from and call into question by adopting the reflective stance. For the adoption of that reflective stance requires self-critical reflection, and you cannot engage in self-critical reflection, and withhold assent from doubtful matters, if you are unable to exercise your mental agency. So any attempt to distance yourself from, and call into question, this form of self-knowledge by adopting the reflective stance, is bound to fail; for in doing so you inevitably presuppose, and exploit, the availability of such knowledge.

Although Descartes doesn’t explicitly consider and respond to this more radical form of scepticism, it seems to me that his treatment of scepticism, from the reflective stance, has something in common with the response that I have just outlined. For his response to scepticism consists, in part, in drawing out and exploiting the presuppositions of adopting a self-critical reflective stance and engaging in the Method of Doubt. That one exists, and that one is thinking, are among the more obvious presuppositions that he identifies. But a further presupposition that he identifies is the ability to exercise free agency over one’s thinking. In the Principles of Philosophy, Descartes is explicit that it is the exercise of our free will that allows us to withhold assent in doubtful matters,Footnote 13 and perhaps more significantly, in the Synopsis of the Meditations he is also explicit that in supposing the non-existence of all the things about whose existence one can have even the slightest doubt, “the mind uses its own freedom”. Arguably, the exercise of that mental freedom itself presupposes the exercise of a capacity for knowledge, namely practical self-knowledge that arises in and through the exercise of that mental freedom. So yet a further presupposition of adopting a self-critical reflective stance, and engaging in the Method of Doubt, is the ability to exercise a capacity for practical self-knowledge.

The imagination model of dreaming that I outlined in the previous section offers a way of envisaging the loss of this mental freedom, as well the loss of the practical self-knowledge that accompanies it. It provides a way of conceiving of the mere illusion of the exercise of that mental freedom. But not in a way that gives rise to a form of radical scepticism that can be both posed from the reflective stance, and yet left unanswered by that reflective stance. For you can only adopt the reflective stance if you presuppose and exploit the variety of knowledge that this radical scepticism would have you call into question. In this case at least, the sceptic cannot exploit any common metaphysical (or epistemic) factor that is shared both by the self-knowledge that accompanies the adoption of the reflective stance, and the illusion of merely seeming to adopt that reflective stance.

In this paper I have been attempting to draw out the epistemic significance of a version of the imagination model of dreaming. However, I haven’t provided arguments for thinking that this imagination model of dreaming is correct. So it should be said that any conclusions that can be drawn about the nature and epistemic significance of wakeful consciousness from this proposal about dreaming will, for now, have to remain conditional.Footnote 14 However, what this model of dreaming implies (if it is correct) is the following. The capacity to exercise agency over our mental lives is lost when we dream, and in consequence, when we dream, the fact that we have lost that capacity is also thereby lost to us. When you dream you cannot tell that you are unable to exercise agency, for that insight would require the exercise of your agency. So the background state of consciousness that epistemologists so often assume, and rarely discuss, namely the state of wakeful consciousness, has a vital epistemic function. Central to an account of both the nature of wakeful consciousness, and the epistemic function played by that state of consciousness, should be an elucidation of a connection between the wakeful condition and our capacity to exercise agency over our conscious mental lives.