Not a HOT Dream

Presenter: Miguel Sebastian, LOGOS University of Barcelona


Commentator 1: Matthew Ivanowich, University of Western Ontario


commentator 2: Josh Weisberg, University of Houston

Advertisements

86 Comments

  1. Welcome to the discussion!
    Before starting, let me thank Josh Weisberg and Matthew Ivanowich for their great and really useful comments.
    Let me start with a reply to Matthew Ivanowich.
    He calls my attention to the fact that I have not consider the possibility of a misinterpretation of the result of the experiment.
    If I have properly understood it, his proposal is the following: we can distinguish two different kinds of higher-order thoughts:
    1.A HOT1 to the effect that one is having a particular Visually Phenomenally Conscious Experience (VPCE).
    2.A HOT2 to the effect that one’s visual experience is of a square or of a diamond.

    HOT theory maintains that a mental state M is phenomenally conscious iff it is accompanied by the right kind of HOT. According to Ivanowich HOT1 but not HOT2 is the right kind of HOT and Lau and Passinham’s (L&P) experiment only shows that dlPFC is the neural correlate of HOT2; L&P only show that the stimulus was perceived as square or as a diamond not that it was phenomenally consciously perceived. The reports during the long SOA condition shows cases of perceptual certainty but not of VPCEs.

    Ivanowich agrees with L&P that the function of the dlPFC is to work as a decision mechanism that allow us to make categorical decision in the presence of stimuli, but denies that this mechanism is required for having the right kind of HOT.
    But that seems to be wrong, how could we entertain the right kind of HOT without the corresponding categorization? If Ivanowich objection is to succeed, it should be possible to have a HOT1 without a HOT2, but I fail to see how we can have the right kind of HOT, HOT1, without such a categorical decision. The right kind of HOT requires that we deploy the right kind of concepts and that requires categorization abilities. For that purpose the activity of the dlPFC is, as Ivanowich concedes, required.
    Did I correctly understood your proposal?
    I am very curious about the opinion of some of the higher-order experts attending the conference. Interestingly, this alternative (at least I think that it is a different one) HOT theory that Ivanowich is proposing would be a direct opponent of Lau’s theory, if Lau is right then HOT is wrong and vice-versa.
    Leaving this issue aside, I don’t understand very well the objection Ivanowich presents to L&P (http://www.fil.ion.ucl.ac.uk/~hclau/Lau_IP_PNAS.pdf; p.16) to support his interpretation. During the fMRI measurements they adjusted the SOA and the size of the stimuli to get a condition in which the subject performs the discrimination task correctly in 71% of the cases in both, short SOA and long SOA condition. Could you please elaborate your critic? What is missing in the set up of the experiment? How would you explain the differences in the performance capacity?

    In a second part of his comments, Ivanowich concedes that dlPFC is required for the right kind of HOT and objects that we cannot report on VCPE during sleep unless there is an activation of the dlPFC.
    I think that Ivanowich has misunderstood my argument. I think that L&P show that dlPFC is necessary for having a HOT (this is related to the first objection), the question now is whether HOTs are required for having a dream. Let me explain it more carefully. I will use ‘dreams’ to refer exclusively to the phenomenally conscious experiences we have during sleep. The question is whether there are dreams or not. I take it for granted, that IF there are dreams no one will try to resist the argument by claiming that they do not include VPCEs.
    LaBerge et al. experiments suggest that there are dreams: there are phenomenally conscious experiences, at least during lucid dreams. The question is: what happens with the dlPFC in these cases?
    Let’s consider two possibilities: dlPFC is activated during lucid dreams (the most plausible option) or it is not.

    1.If it is not activated then the subject might be reporting having phenomenally conscious experiences but not a visual one. In order to decide who is right, in an experiment they could be asked to move their gaze when they would “see something” during a lucid dreams. If there were no activation of the dlPFC in this condition, then HOT is save and my argument fails because that would show that an increase of the activity in the dlPFC is not necessary for having a VPCE.
    2.For different reasons, most scientist expect an activation of the dlPFC during lucid dreams. If this is the case, then the only option available for HOT theories is to maintain that there are only lucid dreams. This seems to be a “half baked reply”.

    A colleague, Manolo Martínez, has suggested to me an interesting case: sleep talkers. A sleep talker is someone who talk during sleep. I do not know any experiment with sleep talkers so what I will say in what follows is not scientifically supported. There is a coherency between what sleep talkers say while they are sleeping and what they report having being dreaming about. That suggests, that the sleep talker was having a dream during sleep. I think that in this cases the dlPFC will be activated. Again the defender of HOT could maintain that the sleep talker is having dreams but not a non-sleep talker; this position is, I think, really unsustainable.

  2. Thanks again to Miguel Sebastian for a great paper! Let me try to respond to your comments. In particular:

    “… how could we entertain the right kind of HOT without the corresponding categorization? If Ivanowich objection is to succeed, it should be possible to have a HOT1 without a HOT2, but I fail to see how we can have the right kind of HOT, HOT1, without such a categorical decision. The right kind of HOT requires that we deploy the right kind of concepts and that requires categorization abilities. For that purpose the activity of the dlPFC is, as Ivanowich concedes, required.
    Did I correctly understood your proposal?”

    Yes, you seem to have correctly understood my proposal. So the question is whether it is possible to have a HOT1 (the kind required to report that one is having a visual experience) without a HOT2 (the kind required to report what one’s visual experience is of).

    I take it as clearly possible that that one could be aware that one is having a visual experience without being able to tell whether it is a visual experience of a pentacontagon (a 50-sided figure) or of a circle. Likewise, it seems possible that one could be aware that one is having a visual experience without being able to tell whether it is of an icosagon (20-sided figure) or of a circle. What about an octagon (8-sided figure) or a circle? Can one be aware that one is having a visual experience without deciding whether it is of one or the other in this case? Whatever your intuitions, I take it that—in order to reject my proposal—Sebastian would have to deny that one can be aware that one is having a visual experience without deciding whether it is of square or of a circle. And that seems incredible to me if we readily allow for visual experiences in the cases where the subject can’t tell if their visual experience is of a pentacontagon or of a circle, or of an icosagon or of a circle, or even of an octagon or of a circle.

    In other words, Sebastian is right to say that “The right kind of HOT requires that we deploy the right kind of concepts and that requires categorization abilities”. However, it’s an open question what kinds of concepts are required. And it’s at least conceivable that the HOT to the effect that one is having a visual experience need not deploy concepts like “square” or “circle” (just as it need not deploy concepts like “pentacontagon” or “icosagon”).

    More generally, it doesn’t seem particularly strange to me to say that one could have a visual experience of, say, a picture presented on a screen (and be able to report this fact) without being able to say what the picture was of (because, e.g., it was presented to briefly). However, Sebastian seems to deny that it’s possible for one to be aware that one has had a visual experience without being aware of what that visual experience was of. Once again, to me this idea seems nearly inconceivable.

    As a finaly analogy, consider prosopagnosia. People who suffer from this condition are unable to recognize faces, even though they are fully aware of all the visual features of a face that they are looking at (e.g., that it has blue eyes, a wide nose, thin lips, etc.). So clearly activation of whatever perceptual decision-making mechanism that underlies face recognition isn’t necessary in order to have a visual experience of faces. Likewise, my claim is that the perceptual decision-making mechanism that underlies square and diamond recognition (presumably the dlPFC) may not be necessary in order to have a visual experience of squares or diamonds. (Note that I’m not claiming that prosopagnosia is in some way related to a deficit in the dlPFC; rather, simply that it’s an example of an experience where one seemingly has a HOT1 without having a HOT2.)

    As for the second part of my comments, I think I must have misunderstood your argument. I thought that you believed that the dlPFC was likely to be inactive during even lucid dreams. So, assuming that it is in fact activated, then there is no special problem about how we are able to report lucid dreams (and there’s no contradiction in your argument). However, (as you point out), there is no evidence (at least, none that I’m aware of) to show that the dlPFC is in fact activated during lucid dreams. (Some experimental work clearly needs to be done to settle this.) Furthermore, (as you also point out), the defender of HOT theory still has a way out (namely, the “half-baked” way out): she can hold that we are only phenomenally conscious during lucid dreams but not during non-lucid ones. Now, I agree with you that this reply is somewhat of a stretch; however, that’s not to say that it’s impossible.

  3. Here is my reply to Josh Weisberg comments (thanks agains Josh).
    The first thing to be noted is that I am not presenting an argument against higher-order theories in general. I am arguing exclusively against theories that maintain that the cognitive accessibility that underlies reports is constitutive of phenomenal consciousness. Rosenthal’s HOT theory endorses this claim, so it is targeted by my argument.
    It is important to note that my claim is not based at all, on a phrenologist view about the brain, the dlPFC was the area identified by L&P, that’s why I keep talking about it.

    I will precede premise by premise as Josh does:
    Premise one: Josh disregard Malcom, thank you, he is one of the few that have opposed the common sense position and maintained that there are no dreams (I keep on the convention I used above that dreams refer exclusively to phenomenally conscious mental states).
    I didn’t merely rest on a common sense position. Josh mentions Dennett, as I do, as an example the skeptical position and ask how we can rule out this possibility. That’s precisely what I do in section 5.2 and given that this evidence is not challenged, I still think that this is a reply to Dennett’s like skepticism.
    Josh mentions the emotional salience of dreams. Emotions are phenomenally conscious states (we could maybe appeal to a distinction between the emotion and the feeling of the emotion as Damasio does, but Josh seems to be talking about the second one)
    Premise two:
    I didn’t intend to say that there is NO activity in the dlPFC during REM phase. There is much less activity (I used the expression ‘deactivation from phase to phase’ what may be confusing) in the dlPFC during NREM phase and even less during REM phase.
    If there were HOTs during sleep we would expect an increase in the activity of the dlPFC, but what happens is the opposite. Contrary to what Josh considers, Lau may be able to offer an explanation of the fact that we have dreams despite the lower degree of activity. But I do not see how Rosenthal can explain such a significant decrease.
    The question is not whether this low level of activity suffices for having a HOT, the question is how can we justify such a significant decrease when what we would expect is precisely the opposite. Maybe it can be done, but I am still waiting for such an explanation, and the commitments of such a reply to judge the plausibility of the theory.
    Premise 3: the premise is, I think, completely unchallenged. In his reply Josh offer alternative higher order views. But these views are not targeted by my argument.
    Only theories that maintain that the cognitive accessibility underlying reports is constitutive of phenomenal consciousness are jeopardized.
    For instance, Carruthers’ theory explains consciousness by appealing to the availability of the mental state to a theory of mind. It is a higher-order theory but it is not jeopardized by my argument.
    With regard to Damasio’s theory I think that it is important to get clear about how I understand what is a higher-order theory, because I think that Damasio’s is not one them (or at least doesn’t have to be).
    I take it to be the difference between higher-oder and first-order theories that the former and not latter requires a further representation of a mental state for it to be conscious. Interestingly, the relation of representation requires that there can be cases of misrepresentation (a very well known problem for higher-order theories with a bunch of replies). Imagine, just for illustration that we hold on a very simplistic teleological theory of representation and very simple mental states (ON/OFF): a mental state A being ON represents a mental state B iff A being ON has the function of indicating B being there (your plug in your favorite theory as an alternative, here is a simplistic one just for illustration). Normally when A is on, then B is there. However, there might be cases in which A is ON but B is not there. A being ON still has the content that B is there.

    Damasio called structures that map the relation between the first-order representation of in our, say, visual system and the proto-self higher-order representations. I think that Damasio, who is not a philosopher, didn’t intend to use higher-order representation to this higher-order structure. Because if the relation that holds between higher-order maps and the proto-self for instance were a relation of representation, then there could be cases of phenomenally conscious states without the activation of the nuclei of the brains stem that constitute the proto-self and I am pretty sure that Damasio would deny that. It is therefore a first-order theory (there is not representational relation between first and second order maps) Maybe someone has called his attention to this fact and in his last book and refers to these structures as “neural coordinators” (there is also one article by Damasio that I don’t remember now in a Kriegel’s compilation of same-order theories).

    Finally one last remark, Josh maintains that my theory can also jeopardize other SOR theories. I think it doesn’t.
    Consider Kriegel’s theory, for Kriegel, roughly, a mental state is phenomenally conscious iff it is a complex of two parts, where one of them represent the other directly and thereby the whole complex indirectly. He appeals to the dlPFC to implement the relation between the two parts but it doesn’t have to be the only way they can be connected, during sleep the relation between the two parts could be implemented by the ACC, for instance (Kriegel himself consider that the dlPFC is deactivated during dreams) or alternatively (I think that a worst one) somehow maintain that the remaining activity suffices for the interconnection of the parts.
    Of course, he maintains that phenomenal consciousness is the categorical basis of the cognitive accessibility that underlies reporting but he could, if he were convinced by the argument, reject this relation between access and phenomenal consciousness.

  4. “More generally, it doesn’t seem particularly strange to me to say that one could have a visual experience of, say, a picture presented on a screen without being able to say what the picture was of .However, Sebastian seems to deny that it’s possible for one to be aware that one has had a visual experience without being aware of what that visual experience was of. Once again, to me this idea seems nearly inconceivable”
    Same to me!! I didn’t say that. In fact, I think that phenomenally conscious experiences and “awareness” thereby, do not depend on concepts, and therefore not on HOTs.(but that’s irrelevant for the argument). Obviously, as you point out I can have a PCVE as of a pentacontagon without having such a concept (I didn’t have this concept two minutes ago when I looked at the picture in the Wikipedia ).
    I just pointed out to the fact that concepts are need according to HOT and that these concepts determine the phenomenal character of the experience.
    Rosenthal’s theory for instance is not committed to such a crazy view in the case of the pentacontagon (I love this word!). But he would maintain, I think, that other concepts like FIGURE or COLOR are deployed in the HOT and obviously we require the categorization for deploying these concepts. For instance, according to Rosenthal I do not need to have the concept RED34 to have a PCVE as of RED34, suffice it to have the concept RED and for example DARKER and LIGHTER and some others (I am not evaluating the merits of this view, and obviously not endorsing it)
    My objection to your position is that whatever concepts you require you need such a mechanism, or I fail to see how you can avoid it.
    In the example you can claim that the concepts deployed are not SQUARE or DIAMOND, but obviously it wasn’t DRAGON. The point is that if HOT1s are required they have to be as something and not as some other things; that entails a HOT2 (maybe not including SQUARE or DIAMOND). No matter how imprecise the visual experience was, it wasn’t as of an A played by a viola.
    According to your proposal that can be done without categorization, I cannot see how. Maybe I am missing your point, Rosenthal provides an excellent characterization of how HOTs work, but I fail to see yours. An example of the concrete content in the case of the experiment and a clarification of the L&P’s set up may help.
    >However, it’s an open question what kinds of concepts are required.
    I thought it wasn’t at least in simple cases.
    One final remark, your example with prosopagnosia merely show that the decision-making mechanism is not sufficient for having a PCVE as of a face, not that it is not necessary, right?. In fact, these patients (similarly to blindsight ones) can make force-choices above 50%, about the persons face, suggesting that there is a decision-making mechanism doing its job.

  5. 1. An issue that bears mentioning: Rosenthal appeals to reporting in order to argue for the presence of HOTs, but he does not think that reporting is necessary for HOTs. So it might be that HOTs usually connect to the reporting system through dlPFC, but HOTs are independent from dlPFC. HOTs may still instantiate the “cognitive accessibility that underlies reporting,” even though the normal channels between HOTs and the actual report/language mechanisms are inactive.

    2. A caveat about emotions: they can occur nonconsciously. So the presence of emotions alone is not enough to entail consciousness. But the key point, I think, in my challenge to premise 1 (dreams are conscious) is that dreams may be far “thinner” in content than common sense takes them to be. The emotional salience of dreams accounts for our memories of them (emotional salience seems to increase the likelihood of remembering. See, e.g. LeDoux), but their sensory content may be much less than we think. I drew a possible analogy with the mistaken impression that the conscious visual field in waking life is full of detail out to the edge. Maybe dreams are like that, only more so. Can you really tell how detailed your dreams were in sensory content? And if they are thin in this way, there is that much less for dlPFC to do in dreams.

    3. On Premise 2 (dlPFC is less active in REM dreams). When compared to ordinary waking consciousness, Rosenthal can say that the details we’re conscious of with waking sensory states, as they change due to constant impact from the environment, is far greater than the detail we’re conscious of with dreaming sensory states. So dlPFC activity should be less, in accordance with the sleep data. The questions than becomes, why the decrease from NREM sleep to REM sleep (how significant is this change, by the way? I’m sorry that haven’t looked more closely)? Here, Rosenthal can claim that dlPFC alone is not sufficient for consciousness and so in NREM sleep dlPFC is engaged in something other than connecting HOTs and verbal reports. That requires some story about the non-dlPFC aspects of HOT, but I’ve tried to give that in my attack on premise 3. So, decreased dlPFC activity from wakefulness to conscious REM dreams is to be expected on the HOT view, because of decreased detail in conscious content (this is close to what Mathew is saying in his commentary, I think). And the increase in dlPFC activity in NREM compared to REM has nothing to do with HOTs or reporting.

    4. About ToM — I mentioned Carruthers only to cite his work on ToM, not to endorses his alternative dispositional HO view. The point, rather, is that HOT is plausibly constitutively connected to ToM. ToM lets us think about mental states. HOTs are thoughts about mental states. Indeed, this could be the system connected to reporting via dlPFC in waking experience. Medial PFC is implicated in ToM tasks, and it shows greater activation in REM dreaming. So, if HOTs are ToM states, and MPFC instantiates ToM, HOTs are active in dreams in just the way you demand. Hobson thinks ToM is active in dreams, as a bonus.

    5. Damasio explicitly labels his view a “higher-order sensing” view in a footnote in his 1999 book, and he cites Rosenthal there, for what it’s worth. About misrepresentation, I’ve argued elsewhere that same-order theories either don’t escape the misrepresentation problem or they leave phenomenal consciousness unexplained (that is, the do not reduce phenomenal consciousness to representation, and hence to access). But I think the key points here can be made independently of this issue.

    The question for Damasio is what do the HO maps do? What is their function? If their job is to carry information about other brain areas, areas instantiating the proto-self and sensory content, then it looks to me like they are representations of representations, whatever Damasio wishes to call them. Also, Damasio, as a brain scientist, is more worried about how actual brains function than about odd possible cases. The brain is full of feed-forward and feedback circuitry, and that suggests that one region firing without another is generally not going to happen. So he’s probably not thinking too much about the “pure” misrepresentation case that is supposed to separate HO and SO theories.

    But forget what Damasio says here, for a moment. Can’t the HOT theorist simply take what Damasio has posited to be the realization of HOT? HOT theory says HO representation accounts for phenomenal consciousness, Damasio has discovered HO representations in the brain in ACC (even if he doesn’t see it that way), and so HO theory avoids the REM dream argument.

    6. On Kriegel and SOR in general. My feeling is that there’s very little separating Uriah’s view from HOT (he, unsurprisingly, disagrees!). So the HOT theory can appropriate Uriah’s move here, I believe. There are parts and connections on both views. The HO part on SO theory is the HOT on HOT theory. The FO part on the SO theory is the target of the HOT. The dlPFC is the connection between the two on SO theory, binding the complex into one state. The dlPFC is the connection between the two on the HOT theory, serving as the ordinary causal connection between the two. Now run the same story for dreams, on either of Uriah’s alternatives. (Again, I think this has affinities with Mathew’s line above, though I don’t want to rope him in if he disagrees!).

    7. Final points: what about workspace views? Some workspace views hold that dlPFC is important to consciousness because of its function in working memory. (Also, is this a “first-order” view on you taxonomy? I assume it is.) I would think that your argument hits these views quite directly. Also, what about the fact (which I take from Damasio) that dlPFC damage does not undermine conscious experience? Nor, incidentally, does it undermine reporting ability, and hence the cognitive access underwriting reportability. I’d like to hear what you think of these issues.

    Thanks again for the great paper!

  6. 1.Completely agree, I hope I haven’t say anything to the opposite
    2. That’s of course a plausible possibility but I don’t think that this will undermine my argument. In the L&P experiement the sensory complexity is minimal.
    3. One question: do we agree that the subject has a Rosenthal’s HOT in the long SOA condition and not in the short SOA condition in L&P experiment? I fail to understand in Mathew’s reply what’s wrong with L&P experiment. Would you agree (better, would you agree that someone that holds on a theory like Rosenthal’s would agree) with L&P that the subject is having an experience as of a square in the long SOA condition and not in the short SOA condition?
    In such a case, the problem is not the decrease in the activity from wakefulness to NREM but from REM to NREM. Lau maybe can explain it, I fail to see how Rosenthal can. Maybe other areas are involved but in this case, wouldn’t we have expected to see them in the fMRI? I think that if Lau is right, Rosenthal is wrong and the other way around.
    4. Sure, this theory is not targeted by my argument, being accesses (or accesible) to a ToM is what determines that a state is conscious. The cognitive accessibility that underlies reports is not constitutive of phenomenal consciousness and therefore is immune to my argument.
    5. Of course, and these theory is not targeted by my argument. There are many higher-order theories that are save if my argument is right, you mentioned some of them and I agree with you.
    6 and 7: I think that my way of presenting the argument is confusing because it makes you think that I am arguing against higher-order theories.
    A first-order theory that maintains that the kind of “access” necessary for consciousness is the one that underlies reports is also a plausible targeted by my argument. I chose Rosenthal’s HOT theory because he explicitly endorses the relation between reports and phenomenal consciousness (qualified as you point out in 1) and the connection with L&P result is pretty immediate (maybe not as Matthew point out, but I don’t understand what is wrong with L&P experiment)
    Which theories are targeted by the argument? Only some of the theories that maintain that the cognitive accessibility that underlies reporting is necessary for consciousness. Rosenthal’s is one of them.
    Arguably, the GWS is anther one, let’s assume that the access to the GWS depends on the dlPFC. We can think of the GWS as the categorical basis of cognitive accessibility, the states there are freely available to reports, etc. The access to the GWS depends on the salience of the process, processes form feed-forward loops increasing the salience of the information, and recruiting other areas. I do not know the neurophysiological details but I think I do not make a big mistake if we think of the salience of a process in terms of signal to noise ratio and stability. If there are dreams then the dlPFC is not contributing significantly to the salience of a process; that suggests that phenomenally conscious states during dreams are not implemented in the GWS, if this is true, being encoded in the GWS not the kind of access we are after.
    If Lau offers a theory of the Bayesian decision-mechanism that is compatible with the decrease of the activity between NREM and REM (granting that there are dreams, he accepts it) I think that his theory has one point more than GWS.
    Probably the GWS theory has a reply to the former question, my point is that a HOT theory like Rosenthal (and not any higher-order theory) has problems for accommodating the data.
    If the cognitive accessibility underlying reports is not constitutive of phenomenal consciousness and such accessibility depends on the dlPFC then damage in the dlPFC doesn’t undermine conscious experiences. I think he is right, Lau and Rosenthal would disagree. Damasio is not at all committed to the view that the cognitive accessibility underlying reports is essential to phenomenal consciousness, so it is not surprising that he thinks that dlPFC is not required. I agree with him.
    The question os whether the “access” required for consciousness is the same one that underlies reports. Rosenthal says yes and I argue that this is problematic for his view.
    I do not maintain that dlPFC is necessary for PCVE but that it is necessary for reporting a PCVE (and not with the doing the report, like muscle control, etc). I, contrary to Rosenthal, think that there are conscious states on which we cannot report (they are not in the GWS). Rosenthal concedes the connection between such a report abilities and phenomenal consciousness and give reasons in favor of his claim, here is one that intends to be against, dreams!

  7. Thanks by the way, this is a lot of fun.
    If I have time (finishing my dissertation is killing me) I will check Damasio’s and Laurey’s book and tell you a bit more about what I think (I have forgotten the details and Damasio’s reason for rejecting this -I think that is related to Balint syndrom and the two visual paths story that Hakwan objects to -see 2nd conference online)

  8. Hi Miguel. Neat talk! And this is a great discussion, guys.

    Picking up on Josh’s arguments against premise two (that dlPFC is inhibited), I know that you acknowledge that the studies don’t show that there isn’t any dlPFC activity, but I wonder why you think that they even show that “there is much less activity” (I think this is really just echoing what Josh said).

    Let me first admit that I haven’t looked at those studies, but from what little I know about fMRI research in general, it’s often hard to conclude much from the fact that there was some deactivation. For one thing, fMRI is a very rough tool and it’s not obvious what to count as great amounts of deactivation vs. modest amounts. And, perhaps more importantly, (de)activation is always measured against a baseline, and it’s often not clear how that baseline is generated. And selecting different baselines will give different results.

    David Carmel told me these things about fMRI when discussing another putative argument against HOT theory, that I think is reminiscent of your argument. In his reply to comments on his “Mesh” paper, Block cites the Malach study (Goldberg et al 2006) wherein subjects categorized the same stimuli during an introspective task (where they are asked to report on their emotional reactions to the stimuli) and a rapid categorization task (where the stimuli are presented in quick succession). Unsurprisingly, frontal areas associated with self-awareness were activated during the introspective task, but deactivated during the categorization task. Block concludes, playing on the title of the Goldberg paper, “that the self is lost in intense perceptual activity” (2007, 538), which is supposed to be a challenge to HOT theory if those perceptions are conscious.

    David’s point, if I understood him correctly, was that simply finding deactivation relative to the baseline doesn’t mean that there wasn’t substantial dlPFC activity. As I said, I’m not familiar with the dream studies you found, but I think all fMRI studies work this way, so I’m thinking the same worries about what to conclude apply.

  9. Hi Jake, thanks for joining the discussion!!

    The Braun study is not a fMRI study, but a PET study. This is not very relevant, for I think that you critic also applies to PET studies (you can just open the paper and look for the big tables to have an idea if you are interested); what is important is that there are independent studies with independent techniques which results are consistent with each other.

    During sleep some areas are more active and others less, Braun study shows that one of most deactivated areas when we compare the activity just before going to sleep and slow wave sleep is the dlPFC (that’s what I mean by much less activity). When we enter the REM phase the activity of some areas increases. This suggests that the dlPFC is not doing what it usually does. But, as you point out, it might be that it still does many things.

    I completely agree with you that deactivation relative to the baseline doesn’t mean that there wasn’t substantial dlPFC activity. You can try to use that to criticize Block when he quotes Malach, but it is also fair to say that data play in his favour.

    There is something methodological underlying the discussion, the measurement techniques that we have merely suggest one possibility of other (and the kind of replies that you offer to the experiment that Block mentions will always be available). We can make predictions and see the results, but it is very hard that with the techniques we have we can rule out a theory, but can make an alternative theory more plausible. I intend to offer an argument that suggest that the prediction that certain theories should make fail.

    Let me focus exclusively in the point you are making.

    L&P show that when I report that I have see the stimuli there is more activity in the dlPFC.

    L&P also show, I do not mention it (and I should), if we TMS over the dlPFC we decrease the perceptual certainty (assotiated I think with HOTs)

    Let’s assume the following (which is not totally true): we do not dream during NREM and we dream during REM. According to Rosenthal’s theory (assuming that there are dreams), there are HOTs during REM and not during NREM. If the dlPFC is necessary for HOTs (you can try to deny that and reply to Hakwan’s experiment, Matthew is pressing in this direction) we would expect this are to work in an activity it wasn’t working before and therefore to increase its metabolic activity, but what happens is precisely the opposite as it has been measured with different techniques. I am not claiming that this level of activity is insufficient for thinking that we have HOTs, I am claiming that we would expect an increase in the activity of this area and we get exactly the opposite. Hakwan can probably offer an explanation, I am not sure that David can.

    Possible alternatives that I see to reply the argument, specifically for HOT, thanks to all your comments (assuming that there are dreams):

    1. Reply to L&P experiment. I think that I have a HOT of the kind ‘I see a square’ in the long SOA condition and not in the short SOA condition. Matthew has not convinced me to the opposite, because I do not understand how he replies to the experiment, but there is still something I must be missing from his reply.

    2. Explain how this result can be compatible an increase of the activity exclusively in the dlPFC when we have a HOT in the L&P experiment, with a decrease of the activity in the dlPFC from NREM to REM, asumming that there are HOTs in REM and not in REM. I fail to see how.

    Of course one can try to speculate that this are is involved in some other activity during NREM and not in REM what compensates the difference, but I haven’t read of any hypothesis in this direction. Tononi in an article in “The neurology of consciousness” gives a nice account of the activation-deactivation of areas in the brain during sleep and its consistency with dream reports.

    P.S yesterday my memory played a dirty trick on me, balint syndrome is not associated neither with the dlPFC but the parietal cortex and plausibly not with consciousness. Damasio denies that the dlPFC is associated with consciousness because damage in this areas seems not to affect consciousness (according to him).

    P.S.2 one worry (that I mention in the paper) is that, in Braun’s experiment contrary to the one by Maquet, they didn’t control for reports of dreams, but we can make use of the result of other experiments that show that if subject are awaken in the REM phase they report having being dreaming on more that 80% of the cases (given that we are not discussing here whether there are dreams or not; we can assume here that these reports indicate that the subjects dream, for the sake of this discussion)

  10. Hi all,

    First, thanks to Miguel and the commentators for an interesting paper and discussion.

    I find myself in almost total agreement with Josh Weisberg about the main argument of Miguel’s paper.

    I wish we could pay more attention to Premise 1, i.e., REM dreams are conscious.

    The methodological issue raised by Dennett’s argument against this premise is the elephant in the room, at least for me.

    We are all supposedly arguing about consciousness. And the only data that we have to delineate the topic of conversation is conscious appearance and reports (usually verbal) about it. I assume we all agree that a theory of consciousness is supposed to be the explanans for those appearances and reports.

    I presume we can also agree that that is a handwavy way of defining the topic of conversation, but it is unfortunately all we’ve got.

    But are dreams really a part of the explenanda that a theory of consciousness is supposed to explain?

    I guess it depends on how handwavy we want to be, but it isn’t obvious that they are, given Dennett-style worries. This raises a difficulty for Miguel’s whole enterprise.

    In his paper, Miguel cites some studies and concludes “the experiments on lucid dreams provide evidence that we have conscious experiences during sleeps [sic], and give us the opportunity to record reports to that effect” (p. 10).

    I am not sure that any of these studies mention real reports on conscious appearance. It seems to me that eye movements do not count as subjective reports. In order to count as such the participants would also have to be disposed to make verbal reports with, roughly, the content that they were making eye movements with the intention to inform the experimenter of the direction they were looking at in their dream. Were the participants disposed to verbally report this way? I think we will not know until the explenandum gets less handwavy. There are some directions in which things can go such that we might have a criteria independent of verbal reports, but we are not there yet (check out: “Criteria for consciousness in humans and other mammals” by Anil K. Seth, Bernard J. Baars, David B. Edelman in Consciousness and Cognition 14 2005).

    So, with Dennett, I think that the relevant first-person data, i.e., data about conscious appearance, always comes *after* REM sleep. So the relevant reports never correlate with any brain activity during REM sleep. They correlate with brain activity after we wake up.

    Again, thanks for the wonderful paper and discussion. I hope to hear back from you on this.

    Michal

  11. Hi Michal, nice to see you here!!
    You said that you agree with Josh, but that seems to left the argument untouched. Am I missing something?
    With regard to Dennett skepticism: the subject say that he has a dream; we cannot trust her because it may be a case of false memory and there is no way to test it because we cannot ask the subject whether he is having a dream. But there are lucid dreams. So we ask the subjects to make an extreme eye movement when they realize they are dreaming; they do and after that they say that they were dreaming. What is your explanation of lucid dreams? Dennett said that lucid dreams was compatible with the subject dreaming* (dreamind* is an unconscious proccess) that he was dreaming*, but the evidence shows that this is not the case.
    There is no much room for skepticism from my point of view but if there are people skeptic about the external world I understand that there might be people skeptic about lucid dreams. If the evidence from lucid dreams does not convince you I do not have any other argument.

    The case of the sleep talkers is also interesting but if you remain skeptic with regard to lucid dreams, I do not think that this is going to be of any help, because I have any empirical evidence beyond what friends and their partners have told me.

    The proposal of Baars and Seth is…well a proposal. If these conditions are necessary conditions, then they are too controversial (from the philosophical point of view and from the neuroscientific point of view) to be useful in any sense as an start point, don’t you think?

    I agree that some form of “acess” is necessary for phenomenal consciousness. Some people think that this access is the same cognitive acces that underlies our ability to report (Baars is one of them so taking his definition as an start point sounds a bit like cheating, doesn’t it?). I present an objection against these theories.
    But you proposal is interesting, why not starting with something as less demanding as possible and discuss about what we need to add? consider cases of minimally conscious states (like the cases of vegetative or coma in Laurey’s patients) and discuss whether there is something it is like for the subject to be in these states.

  12. Miguel,

    It is very nice to be here at CO3 (thanks to St. Richard Brown and company). It’s nice to see you too!

    I said I agree with Josh, which meant that I think his criticism of your argument is right on. Every one of your premises can be contested, but I also think that the way you go about constructing your argument is good.

    My only disagreement with Josh is, I think, on how much emphasis to put on Dennett’s point. From what I understand, Josh doesn’t think it’s that big of a deal. I think it is–hence my post.

    You said:

    “But there are lucid dreams. So we ask the subjects to make an extreme eye movement when they realize they are dreaming; they do and after that they say that they were dreaming.”

    Right, but the point is that these people say that they were dreaming after they wake up. So their reports are based on the mental states that they are having then, not the mental states they had when they were dreaming. You are simply *assuming* that they had conscious mental states when they were dreaming.

    I don’t see any non-question begging evidence for the claim that we are conscious of our mental states when we dream. You say you have some. Is the eye movement thing it?

    As to my proposal, it really wasn’t one–just a pointer to how consciousness can be operationalised so we can talk about the same thing.

    I guess you disagree with me about what the topic of our conversation is. You have your conception of what delineates the explenanda and I have mine. Mine is grounded in subjective reports on conscious appearance. I think that yours has something to do with conscious appearance, but I am not sure what.

    And “minimally conscious states” sounds question-begging to me. How do you know that those states are conscious at all if you do not have an independent way of getting at consciousness?

    Further, I’ve no idea whether there is anything that it is like to be in a vegetative coma. But I know that vegetative coma is different from locked-in syndrome. In locked-in syndrome the patient has conscious mental states, but can’t report on them because they are paralysed. So they have a disposition to report on their conscious states, but can’t.

    That contrast leads me to believe that people in vegetative comas are not aware of any mental states they might have–if they have any at all. Some comas are, after all, just a form of deep sleep. 😉

    Michal

  13. Miguel–

    It seems to me that there are two key points here between us.

    1. Your whole argument boils down to the claim that HOT theory does not predict increased dlPFC activity in NREM sleep as opposed to REM sleep. If the HOT theory can provide an answer to this worry, it can then appeal to the line Jake has laid out (about residual dlPFC activity, difficulties with measurement, etc.) and avoid the argument. Would you agree with that?

    So, back to my response to REM<NREM worry. HOT theory (assuming your reading of L&P) is committed to dlPFC activity being necessary for consciousness. I’ll accept that for the sake of argument (see point 2 below, however). But that does not entail that dlPFC activity is sufficient for HOT. Why think that it is? Further, what is the function of the brain activity we see in NREM? Is it so clear that all brain regions are functioning as they do in REM or in waking experience? Here, perhaps, differences in global oscillations might mean something (see Llinas and Ribary, for example). Or maybe there are deep metabolic differences in the brain during NREM that matter for the function of the dlPFC. Or perhaps something else entirely. But since I see no reason why the HOT theory is committed to the dlPFC sufficiency thesis, I think your argument does not work. To fill in the gap in your argument here, you need to establish a) that the dlPFC activity in NREM sleep is relevantly similar to activity at other times, and b) that the activity in dlPFC is sufficient for HOT.

    And even if you establish a & b, there’s still an open response for the HOT theory, one that should be acceptable to you, in particular. The theory can hold that there are unreportable conscious experiences in NREM sleep. Given that you acknowledge that reporting is not necessary for HOT theory, this is an open move. I suspect, a bit, that you are reading the HOT theory as Dennett’s “Cognitive Theory of Consciousness” from Brainstorms. He pretty much collapses consciousness and reporting, and he’s pretty verificationist about dreams. But that’s not the HOT view. So, given that there’s enough dlPFC activity to account for HOTs and consciousness in REM sleep, there’s certainly enough to account for HOTs and consciousness in NREM sleep. There’s no reportability, but that’s not necessary for consciousness, on all views here.

    2. Several times you’ve said you accept my counterarguments (5, 6, & 7), but you claim that they are not about the HOT theory, which is the explicit target of your attack. But I read the dialectic differently. What my counterarguments show is that your argument is not about HOT theory but about some more restricted theory, let’s call it dlPFC-HOT, or d-HOT for short. As I see it, HOT theory is committed to the idea that conscious states are states we’re aware of ourselves as being in, and that this awareness is instantiated by noninferential, non-sensory representations. d-HOT theory is further committed to HOTs being realized in dlPFC. Now, I accept that one possible realization of HOT is dlPFC activity (see point 1 above). But that’s certainly not the only possibility, and to say that this is a settled issue strikes me as premature at best.

    So, does HOT theory logically entail d-HOT theory? Clearly no—d-HOT adds an empirical claim about brain regions. So, does L&P show that HOTs are realized in dlPFC? I’m not convinced, personally. dlPFC seems to be realizing confidence judgments about whether subjects are seeing a particular stimuli or not. Note also, as Mathew points out, that subject are in conscious visual states in the short SOA condition as well. They consciously see the computer screen, the mask, probably the wall behind them, etc. Is there increased dlPFC activity correlating with this? I take it no—it only correlates with confidence judgments about seeing the target stimuli. Most likely, dlPFC activity is functionally associated with HOT (and perhaps with reporting, at least about confidence) in specific conditions, but is not the realization of HOT. Mere association does not entail realization—there is not a necessary connection here. And so, if dlPFC is not necessary for HOT, then its absence in dreams does not bother the HOT theory at all, assuming dreams are indeed conscious (I’m glad Michal has stressed this worry!). Your argument refutes only d-HOT, not HOT. (And it doesn’t even refute d-HOT—see point 1 above.)

    (Hi Jake and Michal! Thanks for joining in.)

  14. Hi all,

    Thanks, Miguel, for an interesting paper! And thanks to everyone here for this terrific discussion.

    I think that most of my reactions to Miguel’s paper have been covered by what others have said here. But, since a lot seems to hinge on the lucid dreaming case presented near the end of the paper, I thought I’d raise a couple of further questions about that, stemming from concerns similar to the ones that Michal expresses:

    1. You say in your paper that “LaBerge (1981) provided evidence in favor of lucid dreams by asking the subjects to make distinctive patterns of voluntary eye movements when they realized they were dreaming” (10).

    In order for this to count as evidence of lucid dreaming, you would need to show, at least, that the eye movements the subjects made were performed with the intention of showing the experimenters that they were aware of themselves as dreaming. Perhaps, if the eye movements were “distinctive” enough, that’s evidence that they were performed with this intention. But I wonder what “distinctive” amounts to here? Specifically, I wonder whether it can do the work required to rule out the possibilities that the eye movements were performed either unintentionally, or in the service of some other intention, neither of which would provide any evidence for lucid dreaming.

    2. Suppose one accepts that these eye movements are indeed evidence of conscious visual experience during lucid dreaming. Suppose one accepts further that dlPFC activation is necessary for HOTs, and that there is dlPFC activation during lucid dreaming, as you say is most plausible. Now, you call the reply that we have conscious visual experiences during lucid dreaming, but not during ordinary dreams “half baked”, but I’m still not sure I see why. Would you mind elaborating a bit? After all, lucid dreams are often set apart from ordinary dreams, both in commonsense and in the scientific literature on the basis of their distinguishing features. Why not think that this is one of them?

  15. Following up on Myrto on lucid dreams, concern over which Miguel regards as half baked:

    Even if lucid dreamers’ eye-movement signals are intentional, it’s not always obvious that nonverbal movements are the psychological equivalent of a verbal report; it might be a behavioral sign that doesn’t require any awareness of the state it’s a sign of. This is not mere skepticism, which Miguel charges in reply to methodological concerns generally. Subjects’ in Marcel’s famous 1993 study were far more sensitive when they responded with eye blinks to stimuli that weren’t consciously visible than when they responded with verbal guesses. The stimuli were presented to hemianopic subjects; so the eye blinks did not indicate presence of any conscious states.

    Marcel, Anthony J., ‘‘Slippage in the Unity of Consciousness,’’ in Gregory R. Bock and Joan Marsh, eds., Experimental and Theoretical Studies of Consciousness, Ciba Foundation Symposium No. 174, Chichester: John Wiley & Sons, 1993, pp. 168–186.:

  16. Thank you Myrto and thank you David for your comments, it is very nice to discuss with you.

    1. They would have been too naïve to make this mistake and there would have been many replies to the experiment. However the experiment has been replicated in several laboratories.
    The signaling was prearranged before the subject when to sleep, so it is fair to assume that they chose a distinctive signal, I also assume that this is the first thing that the referee of the journal would have checked. If it is not distinctive enough (by the corresponding standards) the result is just nonsense. I haven’t read any critique in this direction and it is the most obvious one. Don’t you think so?
    The subjects have to report that they are aware of being dreaming and that is what they seem to do.
    Is there any controversial issue to criticize what the scientific community accepts? Do you know any study that questions LaBerge and other laboratories’ result?
    David presents a worry along this lines and Richard seems to agree with him. Thank you very much for the reference, I will check it, but I do not have access to it (I need to request the book to other library)I do not understand how this study is of any help, but I am replying just from what I have learned from David’s comment. I think that what I said to Myrto also applies in this case, but maybe I go too fast, and there is room for critique in this direction to LaBerge’s results. David could you elaborate a bit (tell me what do you think that the subject is reporting) or can someone send me the paper? Do you have in mind that maybe the subject is aware of being dreaming but doesn’t have visually conscious experiences? Or exactly the same as Myrto was pointing out.

    2.We have good reasons for believing that there are dreams, many of us remember our nice dreams and nightmares. Skepticism about dreams is based on the fact that we report only when we wake up and they could be cases of false memories. Now, we have found a way to test whether these reports can be trusted. The case of lucid dreams shows (or so is usually accepted) that there are conscious experiences during sleep.
    My recall about lucid dreams is not that different from my ordinary dreams and they usually mix (especially in the morning with my alarm clock). Claiming that ones are false memories but not the other seems to me to be something of a reach. Especially, when what we know about the activity in the brain during sleep is consistent with the reports we give about these episodes.
    But this is as far as my argument goes, if someone finds the position that holds that only lucid dreams are phenomenally conscious comfortable I do not have much to add. If the argument is right this is one commitment of the theory, let see whether the opponents can do it better and compare theories, not only in this respect.
    What do you think about sleep talkers? If there were evidence that there is a correspondence between what they say during sleep and what they report having being dreaming would you hold that only sleep talker have dreams?

  17. Hi Josh,
    The reply to the first question is that this will weak the argument.

    When I replied to Jake I considered the possibility of there being a higher activity in the dlPFC during NREM that decreases during REM. If you want to go this way good luck, I don’t know of any function required for sleep that demand an such an amount of the metabolic activity in this area during NREM (just have a look to the tables and do the maths), nor any theory that postulates anything in this direction. But if you find something then you will have a reply to the argument.

    >To fill in the gap in your argument here, >you need to establish a) that the dlPFC >activity in NREM sleep is relevantly >similar to activity at other times, and >b) that the activity in dlPFC is >sufficient for HOT.
    A: The activity of the of the dlPFC during NREM is not relevantly similar to activity at other times, it is relevantly lower. (as I show in the paper!)
    B:I do not need it to be sufficient, only necessary for HOTs.
    Josh, I have told it many times but I think that I am not explaining myself. I am only interested in theories that maintain that access consciousness=phenomenal consciousness or that phenomenal consciousness is the categorical base of acces consciousness.
    Access consciousness: a mental state is A-conscious iff it is available for reports, believes and rational control of action or if you prefer for global control.

    I think that David maintains that a mental state is A-cosncious iff it is accompanied by a higher-order thought. A mental state is available for reporting iff it is accesed by a HOT (of the right kind of course) and that A-conscious states are P-conscious states.
    I agree that phenomenal consciousness requires “awareness”. The question is what is “awareness”. A theory that maintains that “awareness” is access consciousness (properly speaking the categorical basis of acces consciousness)is the target of this argument.
    L&P experiment:
    Short SOA no availability for report
    Long SOA availability for report.
    That suggests that I am access conscious of the latter but not of the former. What is the difference of this availability? The dlPFC
    Let’s consider one particular Higher-order representational theory, David’s HOT theory. What is the difference for David’s theory: in both (long and short SOA) I have a state with the content square, that explains that the performance is identical. But in the long SOA and not in the short SOA there is a HOT to the effect that I see a square that explain that I can report on it. David endorses, I think, the connection between access and phenomenal consciousness. That is the only thing I try to present an argument against.

    If the dlPFC is not related to HOT then it seems that in such a theory HOTs are not directly related to access consciousness, because L&P experiment suggest that there is a connection between dlPFC and the posibility of reporting.

    Obviously not any higher-order theory entails access consciousness, I agree with you in that from the very beginning.

    The dlPFC is necessary for reporting that is what L&P show. If the dlPFC doesn’t do its job the content is not available for reporting, it is not access-conscious.
    >So, does L&P show that HOTs are realized >in dlPFC? I’m not convinced, personally. >dlPFC seems to be realizing confidence >judgments about whether subjects are >seeing a particular stimuli or not. Note >also, as Mathew points out, that subject >are in conscious visual states in the >short SOA condition as well. They >consciously see the computer screen, the >mask, probably the wall behind them, etc. >Is there increased dlPFC activity >correlating with this?
    That cannot be shown byt the experiment because this is the base of the comparison. fMRI is a differential measurement, the situation you describe is the base of the comparison (something that remains constant in the two conditions we want to compare), that’s the reason why L&P have to adjust the time for the long SOA and the long SOA so that they get the same performance capacity and they can have a base for the comparison. They are not cheating as Matthew suggests (I am joking)they are doing what they have to do for measuring what distinguish these to conditions.
    > I take it no—it only correlates with >confidence judgments about seeing the >target stimuli.
    This is precisely the reply of the first-order proponent. In the long SOA there is a higher-order judgement about a first-order perceptual state (one that if you want is already a higher-order one). But isn’t access consciousness precisely this?

    The subject says that she showed the stimuli in one condition and not in the other and the difference between this two conditions is the dlPFC. Where would you locate the curve of phenomenal consciousness in the L&P graphic, in the same place as Block located it last year (the red line in my presentation)?

  18. You write that “[t]hey would have been too naïve to make this mistake.” I’m not sure what mistake you have in mind, especially since later on you ask me to elaborate. But whatever the case about that, uncritical faith in authors and journal referees seems a slender reply to substantive methodological concerns. Lack of published worries, especially in a relatively marginal area of research, doesn’t say much if anything; people knowledgeable in psychology and neuropsychology often have serious concerns about major research that they don’t bother to publish. And the question isn’t whether LaBerge et al (1981) has been replicated; it’s how to understand what the results actually show.

    You go on to write that “subjects have to report that they are aware of being dreaming and that is what they seem to do.” Well, that’s apparently how it seems to you. But it’s not at all obvious. Reports are speech acts, not nonverbal signals. It is a relatively standard, undefended line in much–but by no means all–writing in psychology and neuropsychology that any prearranged signal, e.g., a button press, functions as the equivalent of a report. Some have seen that this isn’t always so, and that more delicate experimental design is sometimes required. Alleged reports of dreams seems a likely case. Additional evidence is needed here to construe an eye movement as a report.

    I assume you think it’s enough that subjects agreed in advance to signal the occurrence of a dream. But as I said before, there’s reason to think this doesn’t settle the question. The Marcel study shows that eye blinks (and button presses as well) can function differently from verbal reports as responses to stimuli that were not consciously visible. They were far more accurate than verbal guesses, for one thing. Evidently subjects often (not always) can on request set up in advance a connection in virtue of which a particular occurrence will trigger a particular movement–even when subjects are not consciously aware of the relevant occurrence.

    We seldom if ever know what the triggering mechanism is in such cases; so unless there’s something quite special in LaBerge et al (1981) that you haven’t reported, it’s overwhelming likely that we don’t know what the triggering mechanism is in that case. So it’s unwarranted simply to assume that the LaBerge eye-movement signal plays the role of a report that indicates conscious awareness of the dream. We need an independent consideration that points in that direction, as against some alternative triggering mechanism.

    On another matter you write: “We have good reasons for believing that there are dreams, many of us remember our nice dreams and nightmares. Skepticism about dreams is based on the fact that we report only when we wake up and they could be cases of false memories.”

    I don’t know who in this exchange you think is expg doubts about whether dreams occur. Your discussion exhibits a odd tendency automatically to assimilate doubts about whether dreams are conscious states to doubts, such as Malcolm’s, about whether they occur at all. Somebody raises the question of whether dreams occur consciously and you charge the person with skepticism on a par with doubts about the external world. Let’s accept that doubts about whether dreams do occur are skeptical in some relevant way. That simply has no bearing on whether the dreams that do occur are conscious mental states.

    My guess is that you may conflate these two questions because of a particular understanding you have of what you call phenomenal consciousness. I’ll say more about that in a moment. But let me first stress, vis-à-vis your recent reply to Josh Weisberg, that your simply assuming Ned Block’s framework of phenomenal and access consciousness is unwarranted here, and may well cause difficulties. There are issues about both of Block’s notions that prevent them from capturing what higher-order theories, my own included, are doing. So I do not, as you say you think, “[maintain] that a mental state is A-cosncious iff it is accompanied by a higher-order thought”; Block’s notion of access consciousness simply does not capture the relevant higher-order awareness.

    I have written about these matters, so I won’t further pursue them here except for one thing. In your reply to Josh you conceded “that phenomenal consciousness requires ‘awareness’,” but immediately went on to say that “[t]he question is what is ‘awareness’.” But all you then went on to say was that my answer to that question–which you miscast in terms of access consciousness–was faulty; you did not yourself offer any alternative answer to your own question. Nor to my knowledge has Block given any explicit, positive account of what the awareness is that figures in phenomenal consciousness. Until one of you does, it’s unclear that there is any serious alternative to a higher-order theory.

    I’ll now return to why I think that your invoking a notion of phenomenal consciousness like Block’s leads you to conflate the two issues about dreams that I distinguished three paragraphs back. When you speak of phenomenal consciousness, it sounds as though it’s just a courtesy use of ‘consciousness’, which applies automatically to any mental state simply in virtue of that state’s exhibiting some mental quality. I say a courtesy use because states with mental quality need not be conscious as most people use the term ‘conscious’. Indeed, they often aren’t conscious, as we know from robust experimental work on masking and other forms of subliminal perception.

    A state is conscious only if one is in some way subjectively aware of it. It’s quixotic at best to regard as conscious a state that somebody sincerely denies being in and has no subjective awareness of whatsoever. But states of which the individual evidently has no subjective awareness can and often do exhibit mental quality.

    So if, as I assume, you regard any state that exhibits mental quality as automatically conscious in this courtesy way, that would explain your eliding the difference between doubts about whether dreams are conscious and doubts about whether dreams even occur. Dreams do presumably involve mental qualities. So if qualitative states cannot occur without being conscious, somebody who doubts whether dreams are conscious states is in effect doubting whether they occur at all.

    But we actually have little reason to believe that dreams themselves–as against our lighly vivid memories of dreams on awakening–are conscious states. So it’s an unwarranted caricature to describe concern over that question as the result of skeptical doubts. The question whether dreams are conscious states naturally arises as soon as we acknowledge the occurrence of qualitative states that aren’t conscious.

    I expect you may be uncompromising about that, and simply deny that states with mental quality can occur without being conscious. States that occur in subliminal cases, you might insist, are subpersonal or neural states that lack mental quality.

    But these states function psychologically in the many of the ways that conscious qualitative states function and differences in function from one nonconscious state to another correspond to differences in mental quality in the conscious cases. So denying qualitative states that fail to be conscious is in this context the true skeptism; as with standard cases in the traditional philosophical literature, it’s simply reinterpreting of empirical findings to fit arguably groundness a priori convictions.

  19. Hi Miguel,

    Thanks for your reply. Just a quick follow up.

    I was wondering about the specific methodology that LaBerge used in the studies you mention in your paper. I don’t have convenient access to either LaBerge paper you cite through my library, unfortunately. But I was able to locate LaBerge (1990), and learn more about how they did it.

    It seems that the eye movements they asked the subjects to perform were not all that distinctive. They were asked to perform a pair of left-right eye movements to signal the onset of lucid dreaming, and since lucid dreaming occurs during REM sleep, where presumably there are plenty of consecutive left-right eye movements, it doesn’t seem to me that the signals would stand out.

    What’s doing most of the work with respect to the movements being intentional is the fact that the subjects actually reported how many times (and supposedly when, although I’m still not clear about how that’s supposed to work) they made the signals upon waking up. Their awake reports were then verified by judges looking at their polysomnograms. Given all that, I’m fine with saying that the movements were intentional.

    (But the question of whether they should also be treated as reports is a different question—as David and Michal have both stressed in their comments.)

    As for sleep talkers, if we had the kind of result you imagine, and the sleep talking in question occurred during ordinary dreams, I’d probably say that was evidence that ordinary dreams are conscious. But I’m not sure how that relates back to what I said about lucid dreams. Lucid dreams have relevant psychological and neurological differences from ordinary dreams—even if they subjectively seem to you to sometimes be similar, or to mix together—so I still think it’s reasonable to be cautious about making any generalizations from their features to the features of ordinary dreams. That’s all I wanted to highlight.

    Anyway, thanks again for the interesting paper and discussion!

  20. Ok, here’s the dialectic right concerning dlPFC, REM, and NREM as I see it, plus a new step:

    1. Miguel: HOT is associated with dlPFC activity. dlPFC activity is reduced/not present in dreams. So no (less) HOTs in dreams. But dreams are conscious, so consciousness without (with less) HOT.

    2. Josh & Jake: There still could be dlPFC activity in REM sleep—it’s just missed by fMRI, PET, etc. And that’s enough for HOT.

    3. Miguel: But, “In such a case, the problem is not the decrease in the activity from wakefulness to NREM but from REM to NREM.” And: “According to Rosenthal’s theory (assuming that there are dreams), there are HOTs during REM and not during NREM. If the dlPFC is necessary for HOTs… we would expect this are to work in an activity it wasn’t working before and therefore to increase its metabolic activity, but what happens is precisely the opposite as it has been measured with different techniques. I am not claiming that this level of activity is insufficient for thinking that we have HOTs, I am claiming that we would expect an increase in the activity of this area and we get exactly the opposite.”

    4. Josh: NREM activity is different, and just because there’s dlPFC activity does not entail HOT. dlPFC may do more than just HOTs, it may do something other than HOTs in NREM, etc.—that’s the anti-sufficiency claim.

    5. Miguel: “If you want to go this way good luck, I don’t know of any function required for sleep that demand an such an amount of the metabolic activity in this area during NREM (just have a look to the tables and do the maths), nor any theory that postulates anything in this direction. But if you find something then you will have a reply to the argument.”

    Now a new step:

    6. Josh: In NREM (but not REM), we see slow (delta) waves across the brain. As it turns out, these waves are “traveling waves” and their points of origin can be mapped. This has been done by Tononi’s group in 2004 (Massimini M, Huber R, Ferrarelli F, Hill S, Tononi G (2004) The sleep slow oscillation as a traveling wave. J Neurosci 24:6862–6870). They find that the slow waves generally propagate from front to back, and the usually originate at the border area between orbital PFC and dlPFC. So dlPFC is involved in propagating delta waves in NREM (SWS 3 & 4). Why might this be? Tononi and co. speculate that “this brain area appears to have a stronger need for sleep” and “the orderly propagation of correlated activity along connected pathways during sleep may play a role in spike timing-dependent synaptic plasticity …and lead to synaptic consolidation… or downscaling” (6869).

    So that’s what dlPFC is doing in NREM. It’s not playing its usual function of underwriting confidence judgments—it’s kicking off delta waves to achieve synaptic consolidation or downscaling. Then in REM sleep, delta wave activity stops, so dlPFC (as well as orbital PFC) is no longer active in the task of propagating the waves and its activity decreases, though it maintains the residual activity needed to do the thin HOTs of REM dreaming (as per Jake). No need to think that anything involving HOT is going on in NREM, unless one is committed to the claim that ANY dlPFC activity is sufficient for HOT, which you are not. So, I have replied to your argument.

    More about access, dlPFC, and HOT to follow.

  21. Hi everyone, excellent discussion going on in here!

    I think I am in agreement with Josh, Jake, Myrto, Michal and David on the responses given. Miguel’s argument doesn’t threaten HOT, or even PFC HOT (d-HOT) for the reasons given. But I want to suggest that fMRI data isn’t in principle the kind of data that could falsify HOT. All that fMRI data can do is to provide likely locations for where HOTs are to be found. On PFC HOT the location is dlPFC, but there is absolutely no reason to think that more activity, or less activity is predicted by PFC HOT when dreams are phenomenally conscious. This is because we don’t yet know what that activity is, as Josh has pointed out. What is needed is data about the electrical and chemical activity of the brain and a theory about how that activity instantiates brain functions which in turn are instantiations of metal theories. fMRI can’t give us that.

    This isn’t to say that fMRI data is useless. I think that the recent use of fMRI in so-called ‘brain decoding’ is very exciting, but ultimately brain decoding will only be truly successful when we start to use more fine-grained measures of what neurons are doing (brain decoding, which as Colin Clifford showed us last year is just multivariate pattern analysis, can be done on any kind of brain imaging technology.

  22. Myrto: I take your point, I do not know anything about what kind of eye movements usually happen during REM, I assume that they are characteristic (that’s why the phase is called rapid eye movements)

    In another article in 1982, subjects control their breath (stop breathing) when they realize they were dreaming, do you think that this is a better example, do you think that I should collect and detail this measurements?

  23. Josh:
    4. I don’t understand what 4 has to do with the argument. I never made a claim about what happens in NREM beyond saying that there is a decrease in the activity of the dlPFC.

    5 is not a reply to 4 in any sense.

    6 You reply doesn’t show that you have replied the argument in any sense.
    It is not a reply because saying that there are slow-wave signals during the slow-wave signal phase is not making a lot of work.

    Furthermore, take into account that slow-wave signals are not present during wakefulness, and they are not exclusive from this area.
    If you look at the tables and to the function of different areas and how they are associated with the phenomenology of dreams (here we are assuming that there are dreams). Of course, the metabolic consumption will depend on the kind of neurons involved that depend on the cortex layer.

    Tononi (the same Tononi that you quote and who know about all this facts), as I mention in the paper, and explains how the activity in the brain is consistent with the reports on dreams. For instance, I mentioned in the paper that he would expect activity in the dlPFC because according to him it is related to volitional actions and reports on lucid dreams are volitional.

  24. David.

    I appealed to access consciousness because I was trying to explain Josh the target of the paper and how not all higher-order theories are committed to what you say (see the quote below).
    I apologize for over simplifying the issue and come back to the claim that a theory might be jeopardize by the argument only if it holds that the cognitive access underlying report is essential to phenomenal consciousness.

    “Nor to my knowledge has Block given any explicit, positive account of what the awareness is that figures in phenomenal consciousness. Until one of you does, it’s unclear that there is any serious alternative to a higher-order theory.” The point is fair, that is what I said above “let see whether the opponents can do it better and compare theories, not only in this respect.”
    Kriegel in philosophy or Damasio in neuroscience, to cite two examples, have offered such theories (but this is not the place to evaluate their merits). This is the main topic of my dissertation, I could send you the chapter related to this topic and I would really and strongly appreciate your opinion about it.

    Let’s focus:

    To avoid leaving the topic I will accept that there are states with quality that are not phenomenally conscious. The question is whether this awareness is the cognitive accessibility that underlies reports. You endorsed this view in your book “The ability to report being in a particular mental state therefore correspond to what we intuitively think of as that state’s being in our stream of consciousness.”

    The quote above allow us to establish a link between phenomenal consciousness and reportability. Phenomenally conscious mental states are those available for report.

    “The Marcel study shows that eye blinks (and button presses as well) can function differently from verbal reports as responses to stimuli that were not consciously visible. They were far more accurate than verbal guesses, for one thing. Evidently subjects often (not always) can on request set up in advance a connection in virtue of which a particular occurrence will trigger a particular movement–even when subjects are not consciously aware of the relevant occurrence.”

    This differences can be explained at two levels, either there is a different mechanism underlying the accesibility or the difference reflect the differences in the system that accesses the information. The mechanisms underlying the motor response are more simple that those involved in verbal report what interferes with the task. Are you suggesting that they are different mechanisms?
    The cognitive accesibility that underlies reporting, is according to the GWS (Hakwan is silent on this in I don’t know any other theory) postulate a unique mechanism underlying both verbal report and rational control of action, are you suggesting that they are different mechanisms? Or that the subject is not rationally controlling his action.

    See also my reply to Myrto for different examples.

  25. Richard:
    “All that fMRI data can do is to provide likely locations for where HOTs are to be found. On PFC HOT the location is dlPFC, but there is absolutely no reason to think that more activity, or less activity is predicted by PFC HOT when dreams are phenomenally conscious.”

    Of course that there is reasonable room from raising doubts about our current measurements of the brain activity, but not only during sleep. However the empirical evidence we have can suggest that one theory seems more plausible than the others
    If there is a connection between L&P result and certain HOT theories, they will predicts an increase in the activity during REM phase (if there are dreams).

    It is not going to falsify the theory but raise serious doubts about its plausibility.

    The levels of activity are consistent with what happens in dreams (here is the Tononi’s quote, I mentioned in my reply to Josh)
    “Despite the remarkable similarities between waking
    and dreaming consciousness, dreaming consciousness
    often presents some distinctive features. These
    include:
    (1) disconnection from the environment->reduced activation of primary sensory and motor areas in PET studies of REM sleep.
    (2)internal generation of a world-analogue->strong activation of
    temporo-occipital and parahippocampal association
    areas that is observed in REM sleep
    (3) reduction of voluntary control and reflective thought->dorsolateral prefrontal cortex, which is involved
    in volitional control and self-monitoring, is especially
    deactivated during REM sleep
    (4)amnesia->It is not clear why the dreaming brain is so profoundly
    amnestic since, for example, parahippocampal
    and limbic circuits are highly active during REM sleep although prefrontal cortex, which also plays a
    role in episodic memory, is deactivated.[he seems worried about this]
    (5) high emotional involvement->REM sleep is in fact associated
    with a marked activation of limbic and paralimbic
    structures such as the amygdala, the anterior cingulate
    cortex, the insula, and medial orbitofrontal cortex.” The Neurology of Consciousness (Laurey and Tononi, p.100-101)

  26. Miguel—

    1. You agree that there is enough activity, as far as we can tell, in dlPFC during REM to account for the presence of HOT. (Jake’s point)
    2. That amount of activity is low in dlPFC in REM.
    3. The amount of activity in dlPFC in NREM is a bit higher than in REM (z-score of +2.90, Braun et al 1997).
    4. Activity in dlPFC is considerably lower in NREM than it is in wakefulness (-4.0 or so, op.cit.).
    5. So, we don’t need much work here—enough to account for the difference between NREM and REM only.
    6. The point wasn’t that delta waves don’t propagate from other regions in NREM; rather, it’s that they often propagate from dlPFC.
    7. How do you know that the work being of propagating delta waves isn’t work enough to account for the added work in NREM? Or perhaps the work is related to synaptic consolidation or downscaling, as Tononi et al mention?
    8. Tononi’s stuff about dream reports and brains is not relevant to this claim, because we’re talking about NREM, not REM dreams or lucid dreams. The HOT claim is that in NREM, the dlPFC has a different function, one requiring some, though not a lot, of work. This accounts for the decrease between NREM and REM.

    Now, about access, HOT, and the dlPFC.

    You say: “If the dlPFC is not related to HOT then it seems that in such a theory HOTs are not directly related to access consciousness, because L&P experiment suggest that there is a connection between dlPFC and the possibility of reporting.”

    And also: “The dlPFC is necessary for reporting that is what L&P show. If the dlPFC doesn’t do its job the content is not available for reporting, it is not access-conscious.”

    Why think that dlPFC is necessary for reporting the awareness of the diamond or square? All L&P show is that dlPFC tracked confidence judgments. That is, when we are more confident, dlPFC activity increased. That isn’t a judgment about the presence of a diamond or square. It’s a judgment about how confident you feel that you’ve seen the target or if you are just guessing.

    If the dlPFC does not do its job, you can’t do confidence judgments. And that fits with a fair bit about what we know of the PFC. PFC damage messes up things like the ability to correctly track how accurate your beliefs and perceptions are, etc. (see Damasio’s Descartes’ Error, e.g.).

    HOTs are “directly related to access consciousness” by making subjects aware of the states they’re in. If you have to guess what state you’re in, you’re not aware of that state in a way needed for consciousness (and hence access). So dlPFC TRACKS consciousness (because it underwrites confidence judgments, which reliably indicate consciousness), but there’s no reason to think it underwrites consciousness, instantiates HOTs, underwrites the cognitive processes needed for access, etc.

    The HOT theory is not committed to dlPFC = the access underwriting reporting. They accept the association (which is the term you’ve been using), because if we’re just guessing, we have no access, and increased dlPFC activity tracks the move from guessing to confidence.

    Further, you seem committed to the idea that damage to dlPFC should make it impossible to report specific conscious visual contents. This does not seem to be the case, as far as I can tell (see Malach 2005, 62, for example; Damasio 1999 reports similar things).

    HOT theory is indeed a theory holding that phenomenal=access (in the sense you’re using the terms—not, NB, the way David ever says things!). States are phenomenally conscious (there’s something it’s like for the subject) when we are aware of the states by way of HOT. And when we’re aware of the states by way of HOT, they are available for reporting, barring unusual circumstances. So, phenomenal= access. But none of this requires the activation of dlPFC, which is only a reliable (though contingent) sign that conscious experience is occurring, due to it’s tracking of confidence judgments.

    Your argument is not targeting HOT theory. It’s targeting d-HOT theory. HOT survives your challenge.

  27. Wow, I’ve missed a lot of great discussion in the last few days! Time to throw my own response back into the mix…

    First of all, I really like what Michal, Myrto, and David have said about the question of whether we actually have conscious visual experiences while dreaming. However, for now I’m just going to assume that we do have such experiences (though I’ll have something to say about that later) and focus on some of the ideas I was trying to express earlier.

    I think my comments are very much in line with what Josh has just said above, but here’s my take anyway…

    Let me start by trying to characterize the different between my view and Miguel’s. (Hopefully, I’m getting your views right here!) Miguel seem to be suggesting that the L&P experiments show that the dlPFC is necessary to ACCESS the lower-level sensory information that it was, say, a square that was seen. For example, when Miguel says:

    “Short SOA – no availability for report
    Long SOA – availability for report.
    That suggests that I am access conscious of the latter but not of the former. What is the difference of this availability? The dlPFC”

    Miguel seems to be suggesting that in the long SOA condition we have access to specific lower-level representations that are inaccessible in the short SOA condition (and that this difference in access is because of the activation of the dlPFC). That is, the idea seems to be that the HOT theorist should view the dlPFC as a kind of ‘bridge’ which allows low-level sensory information to be presented to consciousness. However, I think what’s really going on is that in the short SOA condition no decision was made about what was seen and therefore there’s nothing to report.

    That is, rather than functioning to ‘grease the wheels’ of accessibility (so to speak) so as make possible the reporting of low-level sensory information, I think that, prior to the contribution of the dlPFC, there is no definitive representation-of-a-square to access.

    Now, let me qualify what I just said: clearly, there is low-level sensory information about the stimulus, and that sensory information can affect behavior (e.g., blindsight). However, that’s compatible with the claim that that kind of low-level sensory information never itself becomes directly conscious, not even in the long SOA condition. Rather, it’s some kind of ‘downstream’ representation (not identical to the low-level content) that eventually reaches consciousness in the long SOA condition – possibly even the output of the dlPFC itself.

    In this sense, the outcome of the “decision” made by the dlPFC-state it itself the target of a higher-order thought. (This is essentially what I meant when I said that the dlPFC isn’t the right kind of HOT.)

    Now, assuming that something like what I just said can be reasonably held, I think that Miguel’s best critique of my objection comes from this line of reasoning:

    “if HOT1s are required they have to be as something and not as some other things; that entails a HOT2 (maybe not including SQUARE or DIAMOND). No matter how imprecise the visual experience was, it wasn’t as of an A played by a viola. According to your proposal that can be done without categorization, I cannot see how.”

    That’s not quite my view. In fact, I agree that HOT theory definitely seems to require that there be some kind of categorization involved. But I think Miguel’s main objection to my view is that, as he puts it, “whatever concepts you require you need such a mechanism, or I fail to see how you can avoid it.” I take it that the main point is that since application of some concept or other is necessary for a HOT, it follows that no matter what kinds of concepts need to be deployed in order to make a visual experience phenomenally conscious, you need some kind of mechanism to decide which concepts to apply (i.e., whether to apply concept A or concept B).

    However, the kind of categorization required to make a visual experience phenomenally conscious might be realized by a different neural mechanism than the dlPFC

    Indeed, I think Josh is exactly right when he points out that Miguel’s argument only targets a very specific version of HOT theory (“d-HOT”), and that for his argument to be successful, he’d have to hold that the defender of HOT is committed to the claim that that mechanism must (in all cases) be the dlPFC. However, as Josh points out, HOT clearly doesn’t entail d-HOT, and, as I’ve tried to show, L&P’s experiments in no way show that HOTs must be realized in the dlPFC.

    Now, it may be true that no matter whenever concepts are applied, there’s must some mechanism which decides between which of multiple competing concepts should be applied. However, assuming that is true, all I’m claiming is that the defender of HOT theory should try to hold that the neural mechanism that applies whatever concepts make a visual experience (of say, an indeterminate shape) phenomenally conscious that is not the dlPFC.

    So let me ask Miguel: do you think that the defender of HOT theory can reasonably hold that one can have a visual experience of an indeterminate shape that is neither a square nor a diamond? It seems to me that if that turns out to be a defensible position, then it seems your argument against HOT theory doesn’t work.

    (By the way, you’re right that I (unintentionally) mischaracterized your views on this question above. What I meant to say was: “Sebastian would have to deny that [the defender of HOT theory can reasonably hold that] one can be aware that one is having a visual experience without deciding whether it is of square or of a circle.”)

    Obviously, I think that the defender of HOT theory can in fact reasonably hold this. (Furthermore, I think that Rosenthal does hold something like this, for he says, e.g., that one can be “aware of one’s perception of an ‘A’ as a perception of some alphanumeric character or other, but not as a perception of an ‘A'” (https://wfs.gc.cuny.edu/DRosenthal/www/DR-Lau-NYU.pdf)

    Anyway, catching up on all the comments have been great, and I’ll have

  28. Miguel, you say

    If there is a connection between L&P result and certain HOT theories, they will predicts an increase in the activity during REM phase (if there are dreams)

    but this is exactly the point I was meaning to criticize. Given that we have no idea what the activity in dlPFC is actually doing we have NO reason to predict that there would be an increase. So let me make teh point differently: what reason/argument do you have for accepting the above conditional?

  29. Josh,
    In my reply to Jake I said that we don’t know whether this activity suffices for HOT. The data we have suggest that this area is much less active.
    As Richard point out with our current techniques we are going to hardly rule out any theory, because suggestions like the one Jake made are always available. Neuroscientist make hypothesis based on the data they have and those data rarely amount to no activity at all in an area. We should evaluate theories according to the empirical evidence we have, a theory that does not require this area fits better the empirical data.
    With regard to 5, 6, 7 and 8. I think I haven’t explain myself or you don’t want to understand the point.
    If you postulate that the dlPFC has certain activity that is also present in other areas, you should consistently apply the same kind of reasoning with regard to other areas, if there is this activity in area A, you should interpret the data in REM with regard to A as you interpret the data about dlPFC, that is why Tononi’s comments on the function of other brain areas during dream are relevant.
    If you want to ignore the other areas then you require something exclusive of the dlPFC, and something that people like Tononi, Damasio, Lau, etc. don’t know, because they consider that there is a significant deactivation of this area during dreams as we have seen and they see problems in their theories when they have to appeal to the activity of this area with this levels (see Tononi, or see Hakwan’s paper) Taking one data out of context and without considering many other things that require consideration is not taking the work of this people very seriously.
    Of course, as David points out, we can reinterpret empirical findings “to fit arguably groundness a priori convictions.” But then one have to face questions due to you accomodation of the data, one could maintain that during dreams more of the states are unconscious qualitative states (whatever that might be), but then you have to explain why we can remember qualitative states and we usually do not remember sub-personal states, for instance.

    On the second part:
    I agree with Damasio that dlPFC is not necessary for consciousness. As you point out damage in this area doesn’t seem to compromise consciousness. The question is whether HOTs are compromised in this case, as you point out, reports of the subject with damage in this area seem to support this claim.
    The interpretation I offer of the data I make is shared by David, Kriegel, Dehaene, etc.
    Your interpretation of L&P is the following:
    “Why think that dlPFC is necessary for reporting the awareness of the diamond or square? All L&P show is that dlPFC tracked confidence judgments. That is, when we are more confident, dlPFC activity increased. That isn’t a judgment about the presence of a diamond or square. It’s a judgment about how confident you feel that you’ve seen the target or if you are just guessing. “
    When you ask me whether I have seen the square or I am just guessing I need a higher-higher-order thought (assuming that a HOT is needed for seeing the square) that is a judgement about whether I have seen the stimuli or not. Do I need a higher-higher-order thought, why a higher-order thought is not sufficient?
    Anyway, you seem to object to the experimental set-up and calibration of the measurement.
    Given the time resolution of the fMRI, I hope that the measurements they are doing are during the presentation of the stimuli not during the reply to the questions!!!!
    I was trusting that all the time, because the details of this details of the set-up are not presented and I have no idea how to interpret the set-up for fMRI measurements either (I have been told that the set up of fMRI measurement is quite complicated), that they have done a good measurement and that the data the present are measured during the presentation of the stimuli not during the reply that is when the judgement you mention is required.

  30. Quick to Richard:

    The target of the paper are theories that maintain that the cognitive accesibility that underlies reporting is essential to consciousness. Some higher-order theories (and also first-order ones, Tye’s PANIC for instance) hold this idea.

    The subject report having seen the stimuli in the long SOA and not in the short SOA, there you have the connection. The content of the mental state is available for report in the long SOA condition but not in the short SOA and this conditions only differ in the dlPFC. What is your question?

    I will try to reply to Matthew tomorrow (thanks for coming back)

  31. Hi Miguel, yes we are in agreement about who you take to be the target of the argument but my point is that what you point to shows that when subjects are awake and also when subjects make certain kinds of judgments there is more activity in a certain area of the brain. None of this suggests that we would expect to see an increase in activity in that area when the subject is asleep and transitioning from a completely different state (NREM). So, the baseline in Lau et al is resting wakefulness and with respect to that you see an increase. In your argument the baseline is NREM sleep and with respect to that you see a decrease but what you haven’t shown is that there is any reason to think that this is at odds with the theory, or as you put it that the theory predicts an increase. What reason is there to think this? That there is an increase in the other case? That doesn’t work, largely for the kinds of reasons that have been appealed to here.

  32. Which part is supposed to be an answer?

    Look, Josh is in the business of trying to explain why there is a decrease; as he says it might be because the dlPFC is doing something besides what it does usually, and then when the dreaming starts the level of activity decreases relative to that. My point is related to this, namely that we don;t have any reason to think that it should increase…that is unless the dreaming people were engaged in reporting, which they aren’t. This has been brought up by others, but the point is that you may need to have a HOT to report but just because you have a HOT doesn’t mean that you are reporting it.

    Again ‘activity increase relative to this or that baseline’ is not the right kind of information that we need. What we need to find out what the activity is doing.

  33. That’s what I explained above (second part related to the measurement) the dlPFC is involved in the reportability not in the reporting.
    Where is the neural correlate of the the concepts deployed fot the HOT in L&P?
    If the dlPFC is part of the neural correlate of HOTs then in dreams we should expect this activity.
    Sorry if I am not understanding the point you try to make.

  34. “If the dlPFC is part of the neural correlate of HOTs then in dreams we should expect this activity”

    Yes that is right. We should expect activity in that area when there is conscious experience. That is ALL that the Lau et al results show. That is what we find. What you need for your argument to work is to show that the dhot theory predicts that there would be more activity when dreams are conscious and you haven’t shown that.

  35. You expect an activity during REM phase that you don’t find.
    Feartures in dreams are consistent with the activity of different areas, we expect the activity of the dlPFC also to be consistent with that. Just look at Tononi’s considerations…

  36. Let me try to come at this in a different way.

    Let’s grant all of your assumptions. It would only follow that the activity during REM sleep was on an order of that seen during waking if the dream experience were super vivid and in fact indistinguishable from waking experience. But you yourself say that you are not going to assume that dream experiences are indistinguishable from waking experience. You say that you will only insist that dreams are phenomenally conscious, not that they are the same as ordinary conscious experiences. But then you have no reason to expect that the activity during REM should be the very same as that during waking. If dream phenomenology is sparse in the way that Josh suggested (i.e. the content of the HOTs is general or vague) then the lowered activity would be expected. But, importantly, dreams would still be phenomenally conscious with just thiner content.

    This is the point I was trying to make above. What matters isn’t increase/decrease with respect to various baseline but what the phenomenology is like and what the actual content of the relevant states is. Knowing that dlPFC is active less in REM sleep is interesting but as you note what matters is not that but what the dhot theory predicts and you haven’t shown that it predicts more activity. Even if we accept all of your (controversial) assumptions all that it shows is that we expect more activity when dreams are very vivid and realistic, and that is just what we find.

  37. Miguel–

    On the first issue (decrease dlPFC still might account for HOT):

    I think you are missing my point. I’m not saying that dlPFC has exclusively one function in waking experience and REM, and a completely different exclusive function in NREM. The point is that dlPFC might make us aware of our states in waking and REM, but be active in consolidation activities in NREM. The deactivation of dlPFC from waking to REM is due to reduced detail in HOT content (as Richard notes). The more minor deactivation of dlPFC from REM to NREM is due to cessation of consolidating activity. All this I take to be consistent with the work of Tononi, Lau, Damasio, etc. (Indeed, it was Tononi’s work I cited in terms of what the dlPFC might be doing in NREM). Just because you disagree with my interpretation of the data doesn’t mean I’m not taking the work of these people seriously! 😉

    Here are 2 other papers that seem relevant on this issue:
    Tian et al (2006). Assessing functioning of the prefrontal cortical subregions with auditory evoked potentials in sleep–wake cycle. Neuroscience Letters 393, 7–11.
    Ioannides et al. (2009) MEG identifies dorsal medial brain activations during sleep. NeuroImage 44, 455–468.

    An interesting finding from Ioannides et al.: “However, we found clear asymmetries in the DLPFC when REM was contrasted with the awake state: The L-DLPFC showed significant increases in slow wave and gamma band (Fig. 4D) while on the right no significant differences were present (Fig. 4E). It would then be possible to speculate that the L-DLPFC and the L-DMPFC are two areas which maintain some frontal executive and attention functions when the body is paralysed in tonic REM sleep” (466).

    Perhaps this is the activity Jake and I were hoping for? Anyway, be careful of the lure of phrenology. dlPFC may be active in one kind of circuit or network in waking and REM and another in NREM. Its general executive-type function may be the same, but it may be doing very different things nonetheless. For example, it might monitor lower-order visual states in waking and REM and help consolidate goal-directed motor movement (see Robertson (2009). From Creation to Consolidation: A Novel Framework for Memory Processing. PLoS Biology 7, 1, 11-19.)

    On the second issue (dlPFC may not realize HOT):

    You say, “I agree with Damasio that dlPFC is not necessary for consciousness. As you point out damage in this area doesn’t seem to compromise consciousness. The question is whether HOTs are compromised in this case, as you point out, reports of the subject with damage in this area seem to support this claim. The interpretation I offer of the data I make is shared by David, Kriegel, Dehaene, etc.”

    I’m not sure I get what you’re saying here. I didn’t mean to suggest that damage to dlPFC must compromises HOTs—given that consciousness goes on without dlPFC and consciousness is (ex hypothesi) underwritten by HOTs, damage to dlPFC does NOT compromise HOTs. It may mess with reporting—there are some aphasic syndromes associated with dlPFC damage—but even that does not undermine the reporting underwritten by HOT. At best, it seems that some dlPFC damage may induce aphasia for complex actions, but not for objects (see Alexander, Journal of the International Neuropsychological Society (2006), 12, 236–247). And what is it that you, David, Uriah, etc. agree on here?

    You then say: “When you ask me whether I have seen the square or I am just guessing I need a higher-higher-order thought (assuming that a HOT is needed for seeing the square) that is a judgement about whether I have seen the stimuli or not. Do I need a higher-higher-order thought, why a higher-order thought is not sufficient?”

    You need a HOT to be aware of yourself as seeing the square. But I don’t see why that by itself is enough to judge confidence. What is being asked of subjects here? They have a forced choice between “seen” or “guessed.” Now, how do you know if you’ve seen or guessed? You don’t consider whether there is a square or a diamond. You consider what your experience was like. That is, you think about the conscious experience you are having (or just had). That, on the HOT theory, is what Rosenthal calls ‘introspection.’ You are consciously thinking about your conscious state. That requires a 3rd-order state, on the HOT view. As L&P put it, these are “reports of the experiences themselves” rather than of the stimuli. That suggests conscious metacognition (rather than mere conscious cognition) to me.

    Alternatively, dlPFC is contributing to the HOT—it’s telling the HOT system, “Yes, that is clearly a square, no doubt about it.” The HOT then represents to the effect that “I, myself, am in a state of clearly seeing a square/diamond.” But dlPFC needn’t actually BE the HOT. It might be a mechanism dealing with conflict/accuracy/deliberate search. It then outputs its result to the HOT system, leading to the requisite HOT.

    I in no way object to L&P’s setup or calibration. For what it’s worth, the key activity in dlPFC occurs in SOA trials at 3.9 seconds after the stimulus—it may be a bit late for the experience itself. I doubt it takes subjects 4 seconds to consciously SEE the square, though it wouldn’t be so surprising to think that introspection might take that long. But maybe I’m not reading that graph right—L&P’s figure 3. And as you say, fMRI timing may be quite complicated.

  38. getting in late for this discussion turns out to be a nightmare. so much stuff here. i’m gonna just pick a few things to address. let me first say the debate has gone to quite some details and level of complexity that is sometimes beyond me – good work guys. i’m just gonna chip in my two pennies worth as a scientist who does related things, to provide some alternative perspectives.

    – richard’s take is that the level of fMRI activity in dlpfc doesn’t really say exactly what it does, so we may as well ignore it (kind of). we had this argument last year, and it probably isn’t useful to rehash the whole thing. yeah fMRI sucks. to really figure out what the computation / representation is, you need more fancy / invasive tools. thing is, even for V1, where so much has been done, people still argue exactly what it codes. so if you have the patience, you can wait. but for me who worry i may not live that long, instead of hoping for an exact understanding right now, one can make some vague and plausible predictions on levels of activity. for instance, if we change a stimulus 2 times a second vs once every 2s, we expect more activity in V1. if it is motion, we expect it to be in MT as well – because MT codes motions. this kind of prediction regarding higher and lower gross activity in a region is quite standard, and has generated useful info, it seems.

    – perhaps not too surprisingly, i don’t agree at all with matthew’s argument against the interpretation of L&P. ” I’ve tried to show, L&P’s experiments in no way show that HOTs must be realized in the dlPFC.” again, it’s not like a bullet proof kind of thing. no we don’t do that science, sorry. we make plausible inferences based on available data. matthew said that dlPFC, instead of reflecting HOT, reflects “categorization” a la heekeren. first of all, there’s a fair bit of recent debate as to whether heekeren’s interpretation itself is right. but i like hauke so let’s grant it is right. but it still can’t explain the difference in dlPFC between the long and short SOA, because stimulus categorization performance was matched. matthew said something like “but it’s matched only because we contrived to match it by titrating the stimulus”. i was hoping some philosophers will set this straight because it’s normally not my job to do this logic thing. but here’s a try: yeah the reason that they were matched was indeed because we did something funky to intentional make them matched (not that funky actually; it’s called staircasing), but no matter how, fact of the matter is it was matched. so to the extent that it is a matter of fact, no matter how, that it was matched, i.e. there’s no difference in categorization difficulty / performance, how do you explain the difference in activity there? one very plausible candidate is the difference in subjective report.

    – somebody says “As you point out damage in this area doesn’t seem to compromise consciousness.” it’s a myth, popularized by people like dan pollen. it’s unfortunate. we’ve shown that TMS to dlPFC changes subjective reports of visibility. del cul and dehaene (brain, 2009) showed that PFC lesioned patients has both disturbed subjective reports and objective task performance (discrimination), though the effect seemed stronger on subjective. bob knight has a paper in nature neuroscience around 2000 that showed that PFC damage weakened visual activity measured by ERP/EEG and impaired task performance in a visual detection task. we have recently gave a dopaminergic drug to subjects which supposedly targets PFC and striatum and it changed visual functions, both subjective and objective. i can go on and on…. these may not be a complete knock out of C. but if you tamper with the CPU you don’t get as much knock out to what’s on your screen as cutting the monitor cable. the dlpfc is a central system. if you lesion it bilaterally, people behave a lot like zombies in b-movies actually (bob knight has movie clips of them). these are not reported in cognitive neuro journals because there just isn’t much you can do with these patients anymore.

    – josh wrote “[the subjective report in L&P] suggests conscious metacognition (rather than mere conscious cognition) to me.” yeah i kind of agree. the kind of measures associated with pfc are quite metacognitive -confidence ratings, visibility ratings, post-decision wagering, kind of thing, i.e. “subjective” response, rather than a relatively “objective” response about the stimulus itself. the debate about how to relate these subjective response to HOT is quite complex, and i don’t wanna get into that. but all i can say is, these subjective responses track conscious phenomenology a lot better than objective responses, as has been shown by a whole lot of experiments. e.g. blindsight. or the fact that unattended peripheral vision seem subjective clear but actually doesn’t give much objective processing power. so to the extent that subjective and objective are really only the two main kind of things we can use in a cognitive experiment, subjective, i.e. metacognitive, responses, are your best bet.

    – josh wrote: : “For what it’s worth, the key activity in dlPFC occurs in SOA trials at 3.9 seconds after the stimulus—it may be a bit late for the experience itself.” yeah but that’s only because the blood-oxygenation-level-dependent signal used in fMRI is typically delayed for about 4-8 s. i.e. v1 response onset is around that range too.

    – for those who think REM PFC activity is relatively low maybe only because NREM PFC activity is high. yeah could be, kind of in principle. but it isn’t so plausible. mind you, even in anesthetized cats, in V1, there may be some baseline activity. you’re not gonna expect zero in PFC. but i recall these are PET studies, which unlike fMRI has meaningful physiological units attached. the general consensus is PFC activity in these meaningful physiological terms in REM is really pretty low. is it so low that logically it couldn’t have done anything? no, we don’t know about that. but if you take reported dream phenomenology at face value (i.e. it’s not generated at the moment of awakening), PFC activity does seem too low to do much for it in a direct, positive fashion. i think most scientists would find this a plausible conclusion. again, like i said, we don’t do proofs. but it looks that way.

    – generally, if you guys don’t think dlPFC codes HOT, where else do you think it does? josh made some suggestions. one can imagine, we can go through these one by one, and then richard can shoot them down by saying oh the evidence depends on fMRI only which is rubbish (sorry bro…), and that will be a long debate. but one approach is to think, amongst all these regions, given all the available and necessarily incomplete evidence, which are the *relatively* more likely candidates. it is in this sense that i would pick dlpfc over the others. probably it isn’t the only region involved either; the brain doesn’t quite work like that. but if there’s such as thing as HOTs in the brain, the dlPFC is most likely, compared to other regions, to do something relevant.

    – may i open another can of worms for you guys? there’re some other cases of supposed high level of visual phenomenology associated with lack of (substantial) PFC activity. there’s one paper in neuron by goldberg and malach, often discussed and i don’t think it’s so relevant because there they present things very fast, vs slow, and say when it’s fast there’s less PFC activity (not dlpfc; kind of more medial and dorsal). but obviously, when it’s very fast, you may not see it as much. but two quite cool studies are Tse et al et Macknik’s PNAS 2005 paper, and Kouider, Dehaene et al’s Cerebral Cortex 2007 paper. there they show that when attention was directed away, a difference between visible stimuli and invisible stimuli led to a difference in early visual, but NOT PFC activity. so if you believe that focused attention is not necessary for you to consciously see the difference between the visible and invisible stuff, there you have a case of conscious phenomenology not positively reflected in PFC. i kind of like these cases more because it seem easier to introspect and think what it is like than for dreams…

  39. Hi Hakwan, thanks for joining in!

    “richard’s take is that the level of fMRI activity in dlpfc doesn’t really say exactly what it does, so we may as well ignore it (kind of)”

    Just a quick response to this. In this discussion my point was only that we have no reason to expect that there would be an increase in dlPFC activity during dreaming. I completely agree with the rest of what you say in this paragraph…

  40. hey richard – yeah perhaps i wasn’t really fair to your point. so let me try to be more precise: usually if you have more frequent and salient changes of content of conscious phenomenology, you would expect more frequent changes /updating in HOT representations – assuming HOT is right. and if dlPFC codes HOT, it should show higher activity, because usually we assume that if area codes X, and X changes more frequently (within limits), then fMRI activity in that area goes up.

    for example, MT codes motion. if you present dot motion that changes in direction every second, vs motion that stays the same, you’ll see more fMRI activity there.

    so is it not quite reasonable (though not bullet proof) to expect dlPFC would show higher activity during REM, given all the assumptions e.g. in dreams content of conscious phenomenology changes more frequently than in dreamless sleep, and REM is associated with dreams, and that HOT is right , and HOT is coded in dlPFC, etc? i’m not saying these assumptions have to be right. i’m just saying one would expect an increase in DLPFC fMRI activity if these assumptions were true. not that one *has to* expect that in th logic-chopping kind of way, perhaps, but to the extent that you don’t see the increase, it renders the assumptions less plausible. so fMRI isn’t irrelevant / impotent in contributing to this debate.

  41. Thank you very much for your very interesting comment and for the references Hakwan!
    I have one question with no hidden intention.
    When I see a glass full of wine close to my computer, the bayesian decision mechanism makes a decision judgment about the input corresponding to the glass of wine and a judgment about the inputs of the computer or a judgment about (the glass of wine and the computer)?
    In other words, the more complex the visual input is the more work for the decision mechanism?

    Matthew:
    “do you think that the defender of HOT theory can reasonably hold that one can have a visual experience of an indeterminate shape that is neither a square nor a diamond? ”
    I see no reason for denying that.
    I think that above chance correct judgements about the stimulus indicate that the subject has seen the stimulus: there is no significant difference in the quality of perception (same performance capacity) and there is a difference in their reply to the second question. I see no reason for denying that the subject is not conscious in the long SOA.
    If I have understood you properly you claim that they report having seen a figure, but not a square or a diamond, in this case why they don’t claim having guessed their reply?

    Josh,
    first of all I want to apologize if the comment sounded rude and thank you for the emoticon.
    Your first interpretation of the L&P experiment seems to be the one I offered in the presentation. The second question reflect a higher-order judgement about the conscious state. In this case the result by L&P goes directly against HOT because no extra activity corresponding to the concepts deployed by the HOT is recorded by the experimenters.
    In the second alternative are you suggesting, with Matthew that the subject is not conscious of the square in the long SOA?

  42. hi miguel,

    “In other words, the more complex the visual input is the more work for the decision mechanism?”

    yes certainly, if the perception is to be veridical. you said you have no hidden intention, but let me pretend that you’re trying to challenge my view, so i can hammer this point home.

    in dreams, those are not veridical perceptions. technically, they are hallucinations / false alarms. one can make false alarms for two reasons, 1)the sensory signal is unreasonably high despite the lack of real inputs, 2)the decision system fails to work properly and mistakenly takes noisy sensory signal as real stuff. (e.g. using too low a criterion for detection). the REM PET data actually support a bit of both, so all is good for me.

    note that on my view i don’t think the DLPFC represents HOTs. i think it carries out the *process* of higher order inference (deciding whether a FO signal is real or noise). whereas usually more representations = more fMRI activity, in my case badly done processes can be associated with weak fMRI activity (as observed in many other studies).

  43. Thanks for this response Hakwan.

    I think maybe bringing up my axe to grind with fMRI was a mistake. I was reacting to a specific claim that Miguel made. He has consistently said that the dHOT view makes a certain very specific prediction. The prediction is that there would be an increase in the activity of the dlPFC during REM sleep when compared to NREM. I claim that the dHOT theory doesn’t make that prediction. This is the real bone of contention in Miguel’s talk. It is not just that there is low activity, as that is consistent with HOT (less representations doesn’t mean no phenomenology it just means less).

    I think it is clear why Miguel thinks this. He says in his paper that a decrease in activity here indicates a decrease in the amount of HOTs and so a decrease in phenomenology. But since he think that there is precisely an increase in phenomenology this refutes the dHOT theory. He seems to have something like the following in mind. In L&P when there is no conscious experience there is X activity. When the subjects consciously see the stimulus there is an increase in activity. In sleep the NREM state is like the condition where the subjects don’t see the stimulus, and REM sleep is like when the do consciously see it. If this was right then we would expect to see an increase in the activity of the dlPFC. But my point was that this isn’t right. You can’t infer that just because there was no increase in activity from NREM to REM that there is no phenomenology in dreams on the dHOT view.

  44. I just saw Hakwan’s latest post…

    Hakwan, you seem to be giving an explanation for why there wold be low activity in dlPFC during dreams, but do you agree with Miguel that we should expect an increase in activity from NREM? Does your account predict that?

  45. richard, my unfulfilled due diligence is probably showing. i haven’t read up the whole correspondence above which is probably why you needed to recap for me. sorry and thanks!

    yeah i think if his argument is about REM<=NREM, it isn't so hot (pardon the pun). but see my point above about the fact that these are PET studies with meaningful physiological units. my impression is REM is not just low relative to NREM. it's just pretty low in absolute terms. but in dreams, the phenomenology shouldn't be so low. it's not exactly like awake real life, but it's probably closer to that than to dreamless sleep

    so perhaps it isn't so good as a logically tight argument. because it seems all pretty much depending on subjective guesswork. how much do we dream in REM exactly? how much phenomenology there is in dreams exactly? and how low of PET activity is too low, really? but for what it's worth, i guess most scientists working on this tend to say it *looks as if* DLPFC activity is too low for it to contribute positively to dream phenomenology, i.e. it looks as if the two don't even have a monotonic relationship, if we are to make an educated guess given all of what we know.

    my view, sketched in my reply to miguel before, says that we don't need a monotonic relationship to begin with, so i'm happy. 🙂

  46. Miguel—

    No apologies necessary! I find that discussion of HOT can get, well, heated. (Sorry, awful pun.) I took your comments to be in the spirit of good philosophical debate.

    I disagree with Matthew that subjects are not conscious of the square in long SOA (assuming that’s Matthew’s position). I just question whether the HOT theory must hold that dlPFC itself codes HOT. I do think that L&P shows that dlPFC tracks subjective reports, but that may be due to it’s function in (3rd-order) introspection or in feeding confidence info to the HOT, located elsewhere.

    Another possibility is that dlPFC is sometimes a component of HOT and sometimes not. HOTs might be complex states made up of a number of distributed elements, bound by something like the process described by Hans Flohr. dlPFC may not be a necessary component of HOT, but it might be an ordinary component of HOT in waking experience, due to its role in “deciding whether the FO signal is real or not” as Hakwan put it. That was close to my “alternate” suggestion above.

  47. Hakwan the scientist, sometimes (welcome!)—

    You ask “if you guys don’t think dlPFC codes HOT, where else do you think it does?”

    Again (this reiterates my recent comments to Miguel), for HOT theory, it might be that the dlPFC activity you guys found is more “3rd-order” than ordinary higher-order. dlPFC may track deliberate, conscious focus on one’s conscious state, rather than a process of making us aware of previously unconscious first-order processes. Your task in L&P explicitly asks subjects to deliberately focus on their experiences (rather than the world), so it wouldn’t surprise me if the reports track introspective processes.

    I wonder if the dlPFC activity is as strong (or present at all) if you don’t ask the “seen or guessed” question? That might be informative about this issue. If the activation is less or absent, but we still assume that subjects consciously see the stimuli in the long SOA trials, we’d have reason to think that dlPFC activation is a function of the additional question (and its introspective demands) rather than ordinary conscious experience of the square or diamond. Did you do anything like that with your subjects?

    Another possibility (and one I’ve urged throughout this discussion) is that HOTs may be much more distributed than we’ve been taking them in this debate. As you say about dlPFC, “probably it isn’t the only region involved either; the brain doesn’t quite work like that.” Perhaps the dlPFC is a part of HOT sometimes, but not a necessary part of all HOTs. When there is a conflict or a demand for accuracy, dlPFC may kick in and instantiate HOT content about confidence, clarity, etc. It might be bound with other brain regions into a circuit which altogether codes the HOT. In dreams, the role of dlPFC may not be needed in HOTs, or it may be actively suppressed for some reason. Other regions necessary for HOT would still need to function in conscious dreams, but the HOT theory wouldn’t be so closely linked to dlPFC activity.

    About the data on dlPFC damage and consciousness, I’ll look at those papers. I bet the deficits may be “introspective” rather than ordinary consciousness (including reporting tasks requiring active introspection), but I haven’t looked yet. The challenge here is experimentally separating introspection from HOT consciousness, and given that both seem to track subjective reporting, I’m not sure your methods have done so. (Incidentally, the experimental separation of subjective reporting and performance is just awesome—don’t get me wrong!)

  48. The argument is based on the idea that dlPFC cannot implement HOTs, given its degree of inactivity during REM phase.
    All neuroscientist I have read consider that this degree of activity as indication that it is not performing its function. In support of this conclusion they present PET and fMRI studies. Some of the results of PET are presented as delta(rCBF) and for that reason I thought that it was a comparative measurement and offered a possible reply for HOT theories what originated the discussion with Josh. But as Hakwan has explained they are absolute measurements (in the table the delta represent increase with respect to a normalized value and not a comparison to the previous phase) Thanks you, the argument is the same.

    As he showed, his theory can provide an explanation of dreams contrary to HOT.

    During REM the Bayesian decision mechanism (BDM) malfunctions and when it malfunction it lowers its criterion for detection. No matter how many detections BDM has to do we do not expect a correlation between its level of activity and the phenomenology. This doesn’t seem to be plausible in the case of HOT.
    Only those theories that predict the previous correlation are targeted by the argument.

    One question: dreams are hallucinations due to the BDM working incorrectly. Why does the brain keep consuming resources for a system, the BDM, that is not required, is not working properly and can be dangerous? (dreams seem to be exclusive of marsupials and placental mammals for only they have REM sleep)

  49. “The argument is based on the idea that dlPFC cannot implement HOTs, given its degree of inactivity during REM phase.”

    This is not the argument that you present in the paper. There you say that the relevant fact is that there is no increase in activity from NREM to REM but rather a decrease. My point was only that the theory was not committed to predicting an increase in activity from NREM to REM. As long as there is enough activity to support the phenomenology that is all that is required, and that is compatible with spasre HOTs Miguel seemed to me to be assuming that dream phenomenology is just the same as waking phenomenology. Then maybe we would expect an increase).

    But I do agree with the reformulated argument in the way Hakwan put it. The conclusion of that is that the dlPFC probably doesn’t play a role in generating the phenomenology of dreams.

    Then the argument hinges on whether or not L&P 2006 shows that HOTs live in dlPFC and here I find I am in agreement with Josh here. There are many prefrontal areas that do get ‘reactivated’ during dreaming and here is a recent study that suggests that mid-dlpfc is involved in monitoring order information, and especially seems to support the idea that this is more 3rd order.

    On Josh’s other idea I think it is suggestive that the areas of pfc that do get reactivated are discussed as fitting nicely with dream phenomenlogy. So perhaps as Josh notes we should actually expect HOts to be realized across the pfc each contributing some content and perhaps being bound into a coherent state via synchrony.

  50. Richard, the argument in pg. 7 is the same one (you showed a previous presentation!!). In my support of the claim that neuroscientist does that the dlPFC is deactivated I thought that the PET measurement was differential, as it is the fMRI study by Maquet, that’s why I had to appeal to NREM.
    I am very grateful to Hakwan for correcting my mistake in reading the PET paper and show me that it is not required and to Josh for pressing in this point.

    With regard to your point, the question is not whether HOT live in the dlPFC, the question is that if i) HOTs are required for having a VPCE and ii)the subject in the long SOA condition is conscious and not in the short SOA, this is the most plausible candidate for them to live.

  51. Miguel: yes the arg. on p 7 is the same, but I am talking about, for instance page 6 (see especially the last paragraph) where you say that the key point is that there is no increase in activity from NREM to REM and also in this discussion thread. My point was, and is, that L&P 2006 results don’t give us any reason to think that we should expect MORE activity in dlpfc during dreaming when compared to NREM. Just as you say, the question is where do they live and once we settle that the question is “is there enough activity going on to support conscious dreams’ NOT ‘is there an increase from.

    You suggests that there isn’t but that is a different consideration than the other one.

  52. One more try: The explanation for the low activity in dlpfc has got to be that it is not doing its job very well (this is Hawan’s explanation). You say that the d-HOT theory can’t rely on that because we would expect more activity since there should be more representations. But the dhot theory could well respond that this level of activation is consistent with the phenomenology and with dlpfc not doing its job very well. Perhaps it is only generating occasional very vague HOTs. Then we would have this very strange disconnected phenomenology, much like what usually happens. Now even this may not be compatible with the level of activation that actually occurs in the dlpfc but I can’t see why we should rule it out a priori.

  53. Josh:”I do think that L&P shows that dlPFC tracks subjective reports, but that may be due to it’s function in (3rd-order) introspection or in feeding confidence info to the HOT, located elsewhere.”
    But this elsewhere was not recorded by L&P.
    You suggest that it might be a distributed process. Even if it were a distributed process these areas should appear in the fMRI. You can reply that the fMRI doesn’t have the required resolution, and that might be right, but is something that can be objected to any study.
    Given the evidence that we have the most plausible candidate to be the realizer of HOTs seems to be the dlPFC.

  54. Terrific, I see the point.Josh make this suggestion also in his comments.
    According to HOT you expect the activity of the realizer to increase with the richness of the experience (more HOTs).
    Reports about dreams seems to suggest that phenomenology is not rare in dreams. Hobson who has recorded these reports writes that they are highly visual, with rich colors, movements, sounds, smells, tastes, tactile sensations, emotions, pain, pleasure, etc.
    I am not ruling out a priori that the possibility you present.

  55. Right, I agree that Josh made a similar point. My point was only ever that because of this it is not automatic that we should expect an increase in activity. The response you are giving me now is why I pointed out the passage where you say you are not going to assume that dream experience is as realistic as waking-experience, but here you seem to be relying on something like that to make your argument. If there is sparse phenomenology in dreams which gets misremembered as rich then we wouldn’t expect an increase…

  56. I told that I was not going to assumme that they are idistinguishable, as Descartes thought. In fact I comment on some differences.

    The skeptical position rejects reports and holds that dreams are not conscious, so I focused on this view.
    If the evidence on lucid dreams shows that there are dreams I see no reason for doubting reports to the point that they are just rare and sparse episodes in sleep.
    In fact reports on dreams match the activity of other areas in the brain.

  57. You seem to be missing the point here. In lucid dreams you probably do have very rich phenomenology…we would then need to see the activity of the dlpfc. The point is that in normal dreams this may be how it is, and that is consistent with reports on dreams and activity in other ares.

  58. Right, and those reports are consistent with the sparse phenomenology misremembered as vivid so therefore low activity in dlpfc view with activation in other areas. On the dHOT view most of the visual and emotional states will be unconscious, though there will be some consciousness. So what reason do we have to really trust the reports? Why is it so skeptical to say that they had some sparse consciousness that they misremember as vivid? People used to say that they dreamt in black and white. Did they really or were they just calling up conscious visual images of things they had seen in their dreams that were only sparsely conscious?

  59. Miguel: “The skeptical position rejects reports and holds that dreams are not conscious, so I focused on this view.”

    I’m not clear why that’s a skeptical view. Is ‘skeptical’ just something a person disagrees with? So you, e.g., adopt skepticism with respect to HOTs?

    As I remarked–and as with much I said you ignored–‘skepticism’ has little serious force unless compared, as you do, to such views as those that doubts about the existence of physical objects.

    But nobody’s denying that dreams occur. There’s simply a question about their properties.

    Perhaps it’s still your apparently uncritical use of Block’s notion of phenomenal consciousness, on which all qualitative states are phenomenally conscious. But then subliminal vision either and masked cases (as in Lau & Passingham) are phenomenally conscious despite there being nothing it’s like for the individual–or they’re simply not psychological at all.

    You pick and choose a lot among the things others say as to which you’ll simply ignore.

  60. Richard:
    You are right, we DON’T HAVE to.
    When we wake up someone in the middle of the REM phase and ask her to report her dreams: is she confabulating or is she roughly reporting the content of her dreams?

    The option that you mention is always available, we don’t have to admit that subjects in long SOA is conscious.
    Thought it is completely unfair to put all replies along these lines in the same boat.

    Based on the data available the hypothesis that the dlPFC is not required or is malfunctioning a là Hakwan is more plausible that disregard the subjects report to the point required by the theory.

    P.S. Maquet is not fMRI as I wrote above but PET.

  61. David:
    I used the term skeptical because most people think that these reports track conscious experiences, among them people working in sleep studies.

    I think I gave a reply to this consideration.
    Let’s assume that mental states are qualitative states but not conscious.
    When I wake up someone in the laboratory during REM sleep he give us a report of dreams. You consider the possibility that these states are qualitative states but not conscious states.
    This is of course a logical possibility, but, I think, not very plausible because normally we cannot report on states that are merely qualitative.

  62. “normally we cannot report on states that are merely qualitative.”

    And dream reports on awakening are pretty good cases of what’s normal in reporting qualitative states?

    I find that your rhetorical flourishes (“This is of course a logical possibility”) don’t contribute to a potentially productive exchange. You must of course recognize that it’s far from merely a logical possibility.

  63. Hi David!
    I just wanted to emphasize that I do not want to ruled out this possibility a priori (a few comments above noted that certain possibilities cannot be ruled out a priori and I wanted to stress that I was not intending to do that). I am very sorry if it sounded as an unrespectful flourishment.
    I am the really interested in a productive exchange (actually, for me, VERY productive and I am learning a lot thanks to the discussion) and thankful for your comments
    Please, if you think that I am not answering a question, just ask me, do not read it as a lack of intellectual honesty, it may be that I have forgotten it, that I am thinking about it, that I am replying other questions or that I have wrongly thought that I had already replied or that I didn’t understood it as a question, etc.

    I do not have a good grasp on qualitative states that are not conscious and how to distinguish them from non-qualitative states. So I apologize if I make some mistakes in the following considerations.

    It seems that the content of dreams is systematically stored in the memory buffer that is reachable for report and that this memory vanishes very quickly (there a some theories that explain that). This doen’t seem to be true for qualitative states that are not conscious. I don’t know of cases in which we can make explicit the content of a state that is not conscious. Given that there are billions of things I don’t know, I used normally, meaning that this doesn’t seem to happen to me, that I do not know of cases of states that are qualitative but not conscious on which I can report.
    Do you have in mind any example of a report about a qualitative state that is non-conscious?

  64. I think the strongest thing going for the hypothesis–and I regard it as no more than a hypothesis, despite your apparent certainty–that dreams are conscious states is indeed that we can consciously remember term, and we can seldom in waking life if ever consciously remember mental states that are not themselves conscious states.

    But I am more impressed, it seems, than you (and indeed many dream researchers who seem not to have considered the methodological issues) about the difficulty of extrapolating from ordinary waking life to either mental functioning while one is asleep or mental functioning across the line between sleeping and waking life.

    There are in any case qualitative mental states that occur without being conscious that are consciously recalled–with the help of hypnosis (not during a trance but due to post-hypnotic suggestion).

    So we know that there are unusual cases in which nonconscious qualitative states are consciously remembered. Are the unusual cases that occur with the aid of posthypnotic suggestion too unusual to apply to theorizing about conscious recall of dreams? I see no reason to think so. Indeed, there is reason to think it’s a good analogy.

  65. “In my support of the claim that neuroscientist does that the dlPFC is deactivated I thought that the PET measurement was differential, as it is the fMRI study by Maquet, that’s why I had to appeal to NREM.”

    Here is a quote from Muzur, Pace-Schott, and Hobson’s 2002 TCS paper “The Prefrontal Cortex in Sleep,”

    We are aware that increased delta activity
    does not always mean (complete) inactivity (see,
    for example, Ref. [14]). Rather than evaluating the
    absolute metabolism of the prefrontal cortex, we
    consider ‘deactivation’of the prefrontal cortex in
    terms of relative activity

    So the claim that Hobson is making is that the dlpfc is relatively deactivated in REM. This is why I was saying that, just as Hakwan does, it is open for a dHOT defender to argue that the relative deactivation means that the dlpfc is not doing its job well. Perhaps it is generating few or perhaps sparse representations. Thsi would mean that we do have phenomenally conscious visual experiences when we dream but they are not as vivid as we take them to be on waking.

    This also helps with the worry that you and David are discussing about reporting. if I have conscious visual experience in dreams I can recall it and so report it. But I am apt to misreport it as being more like ordinary vivid waking conscious experience. Notice this is perfectly consistent with everything that the neuroscience finds as well as the relevant phenomenology appealed to. In Hobson’s William James Lectures given in 2009 he documents the aspects of dreams that he is trying to account for and he notes that these cognitive functions are impaired in dreams but not absent.

  66. Miguel,

    I think that if we are a bit more careful about the cases, the intuitions (whatever that’s worth) about dreams (lucid or otherwise) suggest that they are not conscious experiences. This is more than a methodological worry about not being able to show that they are conscious.

    One peculiar feature of dreams is that even if we remember them after we wake up, we often forget them soon after. I assume you’ve had this experience and it’s well documented in sleep studies.

    This suggests, at least to me, that what goes on when people dream is different from what goes on when they have waking conscious experiences. We tend not to forget the kind of things that appear to happen to us outside of dreams.

    I don’t know about you, but mine include all sorts of crazy stuff like flying around, shootouts with mobsters, and so on. If those things happened in real life, I am sure I would remember them vividly for the rest of my life. But I do not remember them. What I remember is that I talked with someone about flying or shootouts after waking up. Maybe there this is not always the case, but it seems to be predominantly so.

    So what is the difference such that I tend to forget my dreams?

    One possibility is that when I realize that I dreamt something it is erased, so to speak, from my memory. But the eraser story seems shaky.

    This is because realizing that a conscious experience is non-veridical doesn’t usually erase the memory of having the experience. For example, if I were to hallucinate that I am flying while awake and then realize that I was hallucinating after the hallucination is over, I suspect I wouldn’t forget what it was like to fly in my hallucination. Not that that ever happened to me—it’s just an intuition.

    Another possible explanation for why we fail to remember dreams is that they are not like hallucinations and not like conscious experiences. They are more like the kind of experiences we tend to forget. And what are those? Here are some I can think of (I am sure there are more):

    1) Lies. As you probably know, it’s easy to spot a habitual liar, because their stories change over time. This happens to them, presumably, because it’s really hard to keep track of conscious experiences that one didn’t actually have.

    2) Peripheral experiences. Was Julie at the party? Well, it seems to you that she might have been—you don’t remember! Why? You did see the other guests! It seems totally plausible to me to think that what happened is that you forgot all the peripheral stuff because you were so focused on someone else that most of your conscious experiences had to do with that person but, sadly, not with Julie.

    3) Boring stuff. There you are looking out the window of a train going from New York to Philadelphia and lots of stuff is passing by. Abandoned factories, graffiti, dilapidated cars, polluted rivers, etc. Will you remember what all that stuff looks like? My intuition is that you won’t remember much. Why? You did see it! It seems to me that you (perhaps intentionally) forget all those things right after you see them, because you don’t really care to remember them; maybe you are “zoning out.” You need something extra to make those experiences worth the real estate in your brain.

    What (1)-(3) have in common is that they involve experiences that we tend to forget. What else? Arguably, (2) and (3) involve first order mental states such as visual sensations of Newark, or Julie, but no higher-order awareness of those first-order mental states. If they do involve higher-order awareness, that awareness isn’t very detailed. You might have a conscious experience of the party as such or New Jersey as such, but not anything more than that. And (1), arguably, involves no first-order or higher-order states at all—at least not of the kind that one asserts to have had when one lies.

    The moral of all of these anecdotal observations, it seems, is that something is missing in dreams that makes them more like (1)-(3) and less like normal waking conscious experiences or hallucinations; things we tend not to forget. And, arguably, what is missing is the higher-order awareness.

    Now this is not to say that higher-order awareness has the function of making us remember our experiences. It’s just that having two mental states involved with the same thing carries more ‘oomph’ in one’s mental economy than having just one. Or so it seems to me.

    If I am right, then we have gone beyond a methodological worry about dreams (lucid or not) involving conscious experiences. Even if they are stored in a buffer, as you say. We have an argument (or a bunch of intuitions) that dreams are, in fact, NOT conscious experiences.

    Michal

  67. David: you mentioned hypnosis, I have asked for some study that support what you said (to get an idea) but I haven’t found it. Probably because I haven’t understood what you have in mind.
    I have heard of some studies about induced states where the subject is more easily influenced or bring someone to a situation closer to a past episode so that it is easier to remember forgotten things, but not of any study where the subject remembers states that were not conscious but qualitative. Could you please recommend me anyone?

    To get a better grasp of what a qualitative state that is not conscious might be: is the state of the subject in the short-SOA a non-conscious qualitative state? Do you have in mind any criterion for distinguishing unconscious qualitative and non-qualitative states?

    Richard: You reply is different from David’s,but of course they may be combined, right? David suggests that we are reporting non-conscious qualitative states and you that the dreams are conscious but that their content might be more sparse that what we think.
    There is no evidence that the activity in the dlPFC is absent. It is plausible that the dlPFC can produces some few or sparse HOTs. I have already conceded that this is an opened possibility, in this case subject confabulate on awakening about their dreams. But again, based on the data available the hypothesis that the dlPFC is not required or malfunctioning in a way (like Hakwan’s) that doesn’t require to appeal to confabulation is more plausible.

    In the lectures you mention. Hobson writes:
    “Dreaming is a state of consciousness that proceeds in sleep without awareness of the outside world; its content is difficult to remember but rich; and when it is remembered it seems strangely different from waking consciousness.”
    His claim is based on reports of subjects awakened during the REM phase so you can argue that they think their dreams were rich but they really wasn’t.

    Michal: The intuition was always that dreams are conscious experiences, that is why Malcolm, who opposes this idea calls it the “common sense view”(the name is not mine, comes from someone that denied that dreams are conscious experiences). In our field that has been taken for granted by Aristotle, Descartes, Kant, Russell, Moore, etc. But of course, they might all be completely wrong (please don’t read any irony in this), I just mentioned this to call your attention to the fact that what seems to be the shared intuition is quite the opposite than the one you suggest.

    You are completely right about the fact that we quickly forget dreams. Hobson for instance writes: “memory for even very vivid dreams is evanescent and tends to
    fade quickly upon awakening unless special steps are taken to retain it”

    I wasn’t saying anything controversial when I appealed to buffers, that’s the usual way of thinking of memory. I used buffer to refer to a memory system, the long-term memory, the short-term memory, the working memory. I am not thinking that they are located in any place, the are just “virtual buffers”.
    The content of dreams do not get consolidated and “stored in a long-term memory buffer” probably due to the low level of certain kinds of enzymes (that’s what I have read)

    I have no idea of what mechanisms undelie the processes you mention (of course I have my views about peripheral experiences), nor whether we have any reason for thinking that they are related to each other, nor in what sense they are related to dreams.
    I am understanding that you are suggesting that if I forget an episodes quickly, this is an indication that it was not conscious, but this cannot be what you have in mind, right?

  68. WHY NOT HOT (Comment on Miguel Sebastian)

    I’m sorry I’ve not commented on your interesting paper but it’s because I think I agree with it.

    HOT is a higher-order theory of consciousness. It is predicated on one process taking another process as its target or input. It more or less defines consciousness as the ability to report on an “experience” (one does not ask whether the experience itself is conscious, because one senses that regress and perhaps even incoherence lurk there!).

    This higher-order activity seems to be correlated with activity in a certain part of the brain. You note that during dreams that part of the brain is not active. Nor are dreams reportable. Yet surely dreams are conscious. So it looks like both the HO theory of consciousness and the neural correlate of consciousness are refuted.

    I think you’re right. I would simply have expressed it differently: I think one could not even have formulated the HOT as a theory of consciousness, if for “consciousness” one substituted, as one ought, “feeling.”

    Not only does a HOT for feeling not make sense to begin with — what could it mean to say that I feel a toothache because a higher-order process takes my toothache as its target or input? — but it is implausible even to suppose that reportability is a necessary condition for feeling (let alone a sufficient one).

    Feeling is feeling. And dwelling upon, pondering, remembering, theorizing about or reporting a feeling is something else (though itself likewise felt). One could have a flicker of a feeling for an instant, and then immediately forget and be unable to report on it. Yet explaining feeling is the problem we call the mind-body problem. The HOT does not even address that problem.

    (With scopolamine, which was used as a pain-killer for women during child-birth, the feeling can be of even longer duration without being reportable, for there are never any complaints about pain during childbirth under scopolamine: The mothers scream even more shrilly than unmedicated mothers, but the retrograde amnesic effects of scopolamine make them forget afterward. Telescope that down to two successive instants. Similar effects occur with electroconvulsive therapy. None of this paramedical melodrama is needed to make the obvious conceptual point that feeling something and being able to report it are not the same thing. The same holds for dreams: Wake a subject during the dream, and he will remember the prior instants of his dream, but wait till the REM episode is over and it will be forgotten without a trace.)

    The error of HOT is, as usual, the conflation of “access” with “consciousness” (a word that is already equivocal on access vs feeling) along with the spurious distinction between “phenomenal consciousness” and “access consciousness.”

    Substitute feeling for “consciousness” and try to make sense of the idea that something is felt if we can access the feeling: Well, felt access is certainly felt. But then all feeling is felt access. And what isn’t felt access is unfelt access. Or, to be more economical, unfelt.

    “Phenomenal consciousness” is of course redundant: “felt feeling.” And “access consciousness” is just an arbitrary kind of feeling, namely, what it feels like to think about what you are feeling. Interesting, cognitively, because if you can think about it, you can do more about it, even talk about it, and that means a lot for our mind-dumping. mind-reading, linguistic species.

    All those terminological degrees of freedom for verbally fooling oneself into thinking that one has made some progress on the mind/body problem (of which there is only one — not two, one each for AC and PC!), and feeling that one has explained something, when one has just been spinning wheels…

    This was not about your paper, however, with which I agree, but about HOT, about which I now know enough to know that I need know no more!

  69. Just something for the record:

    Miguel says that I’m “suggest[ing] that we are reporting non-conscious qualitative states.” No; I want to stress that it’s not obvious that they are conscious, and that the hypothesis that they are needs support. It can’t be assumed for free.

    In that connection, I urged also that Miguel might stop his reductio-style appeal to Malcolm, as in replying to Michal: “The intuition was always that dreams are conscious experiences, that is why Malcolm, who opposes this idea calls it the ‘common sense view'(the name is not mine, comes from someone that denied that dreams are conscious experiences).”

    Writing as though there is no serious issue about something doesn’t show that there is no serious issue about it.

  70. David: I am not using a “reductio-style”. Michal’s intuition with regard to dreams is different from mine and he suggests that this is the common intuition. I only mentioned Malcolm as an example of someone who denied that dreams are conscious but accepted that the fact that dreams are conscious is the common view.

    “…that the hypothesis that they are needs support.”
    That’s why I appealed to lucid dreams to provide support for this claim.
    Furthermore, subjects awaken in the middle of the REM phase report having being dreaming.
    The most plausible explanation is that dreams are conscious, but maybe this evidence is not enough for you.

    In this case, do you think that they are confabulating or that there is such a thing as non-conscious qualitative states (I am still not clear about what they are) that are accessible for report after waking up? (or maybe some other alternative)

  71. WHY NOT NOT HOT (comment on Steven Harnad)

    (First, thanks for joining in!)

    Prof. Harnad.

    I take it that you do not accept the idea that there can be nonconscious sensations. If that is the case, yes, all this HOT stuff is going to seem badly misguided.

    But why think that? It seems to me that both commonsense and science accept talk of nonconscious mental states, and even nonconscious sensory states. The best explanation of masked priming, inattentional blindness, subliminal perception, etc. seems to be that we have nonconscious sensory states. Why think that such states are conscious? And why think they are nonmental if they are nonconscious? Descartes rejected nonconscious mentality, but beyond his considerations, why do so?

    Now, if one accepts that there are nonconscious sensations, the idea of “unfelt feelings” is not incoherent. Unfelt feelings are those sensory states that register unconsciously. When they become conscious, we feel our feelings. Better, when they become conscious, we become aware of ourselves as being in those sensory states. They are feelings because they play the right perceptual role. They are conscious feelings because we are aware of ourselves as being in those states. All this sounds fine to me. Sure, one can argue against this way of speaking, but to simply say that this is “verbally fooling ourselves” or something strikes me as question begging in this context. How do you know that feelings must always be conscious? Or if you reject the Cartesian idea, what’s wrong with this way of parsing the idea of “unfelt feelings”? It may sound a bit jarring to say it that way, but that happens sometimes in science.

    Further, you incorrect that HOT equates consciousness with reporting. The theory equates consciousness (phenomenal consciousness) with being aware of yourself as being in a state. The awareness is explained in terms of HOT. And yes, you can have awareness that you are unaware of! (See above.) So the scopolamine case and others you mention can be handled by the HOT theory just fine.

    I agree that if consciousness is as you define it, so that there are no nonconscious feelings (or no nonconscious mental states at all), then the mind-body problem stands untouched by the HOT theory. But there seems to me no good reason at all to embrace this Cartesian way of thinking. If we do not, the HOT theory allows us (if it works) to reduce consciousness to representation, and then representation to brain activity. That is indeed progress on the mind-body problem, as I see it! Miguel’s challenge still stands even if we reject the Cartesian approach, by the way. But we’ve clearly solved that problem (j/k!!! 😉 )

  72. Hakwan: Of course I realize that science isn’t in the business of giving absolute, “bullet proof” theories, and yes, I obviously realize that science “make[s] plausible inferences based on available data”. I’m pretty darn certain that I didn’t say anything that implied otherwise. (Yes, I did question the specific inferences you made, but I’m not clear why THAT is a problem….) Anyway, as to your claim that my view that the dlPFC is involved in categorization “can’t explain the difference in dlPFC between the long and short SOA, because stimulus categorization performance was matched”, I think you may have misunderstood me. If I’m not mistaken, you yourself hold that the the dlPFC is involved in decisions about perceptual certainty. I’m saying pretty much the same thing, except that I deny that perceptual certainty is required in order to have a visual experience (I think that one can have a visual experience of an indeterminate shape.) So in the long SOA, the dlPFC ‘made a descision’ that it was a square that was seen, so the HOT included the content ‘SQUARE’ (and therefore the visual experience was of a square). In the short SOA condition, the HOT didn’t include a decision about what was seen so there was no visual experience of squareness, even though there might have been a visual experience of an indeterminate shape. Furthermore, in the short SOA condition, the low-level information that the stimulus was a square obviously had some impact on the outcome of forced-choice decisions.

    Miguel (and Josh): I certainly didn’t intend to imply that the subjected wasn’t conscious of a square in the long SOA condition!! All I was suggesting was that in the short SOA condition the subjects could have a visual experience of an indeterminate shape. As I understand it, on Miguel’s (and Hakwan’s) views, the defender of HOT theory can’t hold that this is possible.

  73. WEASEL-WORD WATCH (Reply to Josh Weisberg)

    J.WEISBERG: “I take it that you do not accept the idea that there can be nonconscious sensations”

    Correct. Suitably transcribed: I do not accept the idea that there can be unfelt feelings. (I can’t even make sense of the idea.)

    J.WEISBERG: “But why think that? It seems to me that both commonsense and science accept talk of nonconscious mental states, and even nonconscious sensory states.”

    That’s unfelt internal states. (And the only reason we call them mental is that they are occurring inside a feeling brain; if they were occurring inside a nonfeeling brain — or inside a toaster — we would not call them “mental” states at all.)

    J.WEISBERG: “The best explanation of masked priming, inattentional blindness, subliminal perception, etc. seems to be that we have nonconscious sensory states.”

    Unfelt internal states. But I will go much farther. Most of the internal states of the brain are unfelt. Hence it’s rather a mystery why (and how) any of them are felt. And that’s what the question is: How and why are (some) internal states felt states.

    J.WEISBERG: “Why think that such states are conscious?”

    I don’t. I’m asking how and why those that are, are.

    (And as I’ve explained in my own thread, and my paper, I reject all the weasel-words — synonyms, paranyms, ephemisms that proliferate and do multiple duty here — consciousness, awareness, mind, mental, subjectivity, intentionality, “access consciousness,” “phenomenal consciousness,” etc. etc. — in favor of feel, felt, feeling. If we can explain how and why we feel, we’ve solved the mind/body problem, and explained consciousness. But if we keep trading in the weasel-words, we are just compounding confusion and equivocation.

    So: I don’t think unfelt states are felt. That’s just the point. And the existence of unfelt states certainly does not help explain the existence of felt ones!

    J.WEISBERG: “And why think they are nonmental if they are nonconscious?”

    That’s just words, and I’m happy to call brain-internal processes, loosely, “mental,” but the mark of the mental is feeling, and in a head that did not feel, internal processes could not be called mental, even loosely. And, to repeat, the mystery is not unfelt internal processes but felt ones.

    J.WEISBERG: “Descartes rejected nonconscious mentality, but beyond his considerations, why do so?”

    I’m not sure what Descartes’ considerations were; I reject it because the only thing that makes internal processes “mental” is feeling. An unfeeling system has no mental processes. But I don’t mind calling the unfelt internal processes of a feeling system “mental,” if you like, as long as no one tries to get any further illicit mileage out of that weasel-word…

    J.WEISBERG: “Now, if one accepts that there are nonconscious sensations, the idea of “unfelt feelings” is not incoherent.”

    Now how on earth did you draw that conclusion? I’ve allowed that there are unfelt processes aplenty in the brain, of course. How have I committed myself to unfelt feeling? Was it the weasel-word “sensation”? Then please correct that to “unfelt sensory system activation” so that there should be no mistake about what I have and have not committed myself to. No one is feeling an unfelt sensation, or sensing an unfelt feeling…

    J.WEISBERG: “Unfelt feelings are those sensory states that register unconsciously.”

    Translation: Unfelt sensory system activity is unfelt. No feelings at all. Just sensory system “functing”…

    J.WEISBERG: “When they become conscious, we feel our feelings.”

    Weasel word. When unfelt states become felt states, they are felt. Where there is only an unfelt state, there is no feeling.

    J.WEISBERG: “Better, when they become conscious, we become aware of ourselves as being in those sensory states.”

    Worse, because more weasel-words! When internal states are felt, we feel them. If what we are feeling is the sensory state of being touched, then we are feeling touched. If we are also reflecting on the fact that we are being touched, then we are feeling what it feels like to be touched, and also feeling what it feels like to be reflecting on what it feels like to be touched…

    J.WEISBERG: “They are feelings because they play the right perceptual role.”

    What does that mean? (And is “perceptual” another weasel-word”?)
    If I feel pain, and I withdraw my hand from the scalding pot, the nociceptive activity is felt, and it is playing the right perceptual role: I detected tissue injury, and acted to minimize it.

    J.WEISBERG: “They are conscious feelings because we are aware of ourselves as being in those states.”

    Feelings are felt because we can reflect on the fact that we are feeling something, and that feels like something too… (Higher-orderism is a bootstrap or a skyhook if it does not first account for 0-order feeling itself. That’s the real problem. Solve that and the rest is a piece of cake.)

    J.WEISBERG: “All this sounds fine to me. Sure, one can argue against this way of speaking, but to simply say that this is ‘verbally fooling ourselves’ or something strikes me as question begging in this context.”

    Whereas I think I am exposing question-begging by routing the superfluous weasel-words and calling a spade a spade…

    J.WEISBERG: “How do you know that feelings must always be conscious?”

    How do I know that feelings must always be felt? Let me try a koan: An unfelt feeling is like the sound of one hand clapping…

    J.WEISBERG: “[You are wrong] that HOT equates consciousness with reporting. The theory equates consciousness (phenomenal consciousness) with being aware of yourself as being in a state. The awareness is explained in terms of HOT.”

    Demustelation: HOT equates feeling with what it feels like to think about yourself feeling. The feeling is explained in terms of HOT.

    Well, good. And now can we take on the real problem, which is explaining how HOT explains how and why we feel anything at all?

    J.WEISBERG: “And yes, you can have awareness that you are unaware of! (See above.)”

    Feelings I don’t feel (see above)…

    J.WEISBERG: “So the scopolamine case and others you mention can be handled by the HOT theory just fine.”

    Getting to know the HOT theory better and better, I don’t doubt it!…

    J.WEISBERG: “I agree that if consciousness is as you define it, so that there are no nonconscious feelings (or no nonconscious mental states at all), then the mind-body problem stands untouched by the HOT theory.”

    “If feeling is as you define it, so that there are no unfelt feelings (or no unfelt mental states at all) then the mind/body [feeling/functing] problem stands untouched by the HOT theory.”

    Well, we’re in agreement then (except it’s not just a simple matter of defining feeling any way we like…)

    J.WEISBERG: “the HOT theory allows us (if it works) to reduce consciousness to representation, and then representation to brain activity.”

    Representation is another weasel-word, unless you distinguish felt and unfelt representation…

  74. Miguel,

    I’ve been running a 102 degrees fever last coule of days, so I apologize in advance if this post is a litte demented.

    So, to get right to it, I wish you read what I wrote in my last post with a little more attention to the forest as opposed to the trees. I am afraid you missed the big picture.

    My reasons for this: you spend a lot of space discussing buffers, which I mentioned in an offhand remark. And also commit a great deal of your response to mechanisms that underlie forgetfulness–nothing I said pertains to that. And to Malcolm’s skepticism, which is, as David Rosenthal pointed out above, besides the point.

    And I didn’t simply give you an intuition that dreams are unconscious. I also didn’t simply assert that dreams are unconscious. I gave an argument. I understand it’s roughshod (hence my mention of intuitions), but I bet there are studies I could dredge up to back up all the premises.

    Anyway, to engage with my post, if you are so inclined, I suggest you address the argument (the forest).

    Here it is in rough and dirty form:

    (P1) We tend to forget some of our experiences, such as boring stuff, lies, peripheral stuff, etc.

    (P2) We tend to NOT forget experiences of which we are distinctly aware.

    (P3) We tend to forget our dreams.

    (C1) Dreams are more like experiences in (P1) than in (P2)

    (C2) Dreams do not involve us being distinctly aware of them.

    Your mention of Aristotle, Descartes, Kant and Russell (!) all in one breath is breathtaking. They all share the same intuition about dreams? I bet they don’t share few intuitions about anything.

    It’s still controversial in some circles whether Aristotle even had a notion of consciousness that resembles our own, for example. And Descartes thought that dreams and wakefulness were indistinguishable. That is something anyone besides him thinks is true?

    Toodles,

    Michal

  75. I AM Weasel (reply to Stevan Harnad)

    Lol.

    Right. We differ on the following, clearly:

    You say:

    “That’s just words, and I’m happy to call brain-internal processes, loosely, “mental,” but the mark of the mental is feeling, and in a head that did not feel, internal processes could not be called mental, even loosely.”

    What is the justification for this claim? Not, as far as I can tell, folk psychology. Not (much of) cognitive science. Philosopher’s intuition?

    All my “weasel words” are labeled by you as such because (roughly) you read them as smuggling in feeling, and so rendering all attempt at explanation circular. But (weasel though I may otherwise be) I was not smuggling in this way. I don’t think the words entail feeling. I’ve been thinking of those terms as per Lewis and Armstrong—they seem to provide a clear way to define these terms without reference to feeling. (Argument from authoritative weasels?)

    Of course, this will cut no ice with you! You already know that their analyses are incomplete at best. And you know that because…?

    To lay more of my weasel cards on the table, here’s what I think. If one definition (feeling is the mark of the mental) leads to agonizing philosophical conniptions and another does not (mental states are states playing certain causal roles (yes, I know, weasel words, as you see ‘em)), then, unless there is compelling reason to do otherwise, go with the non-conniption-causing definition.

    Another possibility. Perhaps over time, as we learn more and more about non-feeling brain-internal processes, we’ll revise our definition of ‘mental’ so that it (non-loosely) covers some of them. But that would be to change the meaning, surely! Well, maybe (see the last paragraph), but maybe it’s just a revision and not a change of meaning. How are we to tell the difference anyway (see Quine, 1951, 1960, e.g.)?

    You also say:

    “And the existence of unfelt states certainly does not help explain the existence of felt ones!”

    Why? Because only felt states can explain the existence of felt states? Well, then it’s pretty straight forward that there can’t be a noncircular explanation here. But what is the justification for this claim? Is this supposed to be obvious? Perhaps as a weasel I don’t even have feelings, so I don’t see the obvious here. Or perhaps you’ve smuggled in the idea of inexplicability to the word ‘feeling.’ But that would be weaselly in the other direction. My guess is that you’re convinced by various philosophical thought experiments about these definitions of ‘feeling’ etc. No weasel-word worries there!

  76. Matthew: But the only place where a difference is found between the two conditions was the dlPFC and in the long SOA condition there is a HOT that is missing in the short SOA condition, right? At the light of the evidence we have the dlPFC is the most plausible candidate to realize this HOT.

    Michal: I hope you are feeling better, no need to apologize, quite the opposite I am grateful that despite of being sick you are participating in the discussion.
    I fail to see the forest, sorry. I think that for there being an argument (I know that you are just sketching the idea of an argument) you need to argue that this distinctive awareness is constitutive of consiousness and for the connection between this kind of awareness and memory. Even for HOT theories, there are many experience that are accompanied by the right kind of HOT and that we tend to forget.
    “That is something anyone besides him thinks is true?”
    The scriptwriter of “Inception”? 😉

    Josh: I know that this is outside the discussion but it would help me: is there any difference between non-conscious qualitative states and non-qualitative states? Are there non-qualitative states that represent?

  77. APOLOGY (Reply to Josh Weisberg)

    I am very sorry if my reference to “weasel words” offended you. I did not at all mean it personally.

    I have been increasingly criticizing the use of synonyms for consciousness because I think they give the illusion of substance and progress where there is none. I think “weasel word” is in fact the right expression to describe this, but I certainly don’t mean it in an ad hominem way, since the practice is so very widespread.

    Please accept my apologies,

    Stevan Harnad

  78. Great discussion!!
    Let me close it by saying that I am enormously grateful to Matthew Ivanowich and Josh Weisberg for their really interesting and thoughtful commentaries. I also want to express my HUGE gratitude to all the participant in the discussion: Jake Berger, Richard Brown, Stevan Harnad, Michal Klincewicz, Hakwan Lau, Myrto Mylopoulos and David Rosenthal.

    I have learned a lot from your comments and enjoyed a really productive discussion; it should be obvious that despite of the disagreement in some issues and the heat of the discussion, I am, as anyone working in this field, profoundly indebted to the work of the proponents of the theories that I was arguing against.

    Finally I want to thank again Richard for organizing such an amazing and successful conference. Long live to the consciousness online conference!!

  79. Great job, everyone. That was a blast!

    Thanks especially to Miguel for the stimulating paper and the spirited discussion. For a while there, it was you against the HOT gang! And thanks to Richard, the crazy weatherman of consciousness! Awesome.

Comments are closed.