Robert Van Gulick
Philosophy 541HL

Syracuse University

Syracuse, NY 13244-1170


Inward and Upward - Reflection, Introspection and Self-Awareness

We are conscious, self-aware and introspective. But whether those three ways we are depend upon each other, and if so how and why remains unclear. They form no doubt a triad of related aspects of our nature, but are they separate and merely similar, or are they linked in some more intimate way? They typically overlap in us, but could they come apart in special cases or in other creatures? Do ties of logical or natural entailment run in one-way or two-way links between some members of the trio, or are they three distinct and merely co-occurring and causally interactive aspects of our particular mental makeup. These are large questions to which I wish I could give complete and convincing answers, but for the present the best I can do is articulate the issues a bit and offer some tentative and partial hypotheses about what the links might be. I will do so in part by situating the problem within the context of the higher-order model of consciousness, i.e. the within the family of views that attempt to explain consciousness in terms of meta-mentalityand the self-reflexive turn by which the mind directs its intentional aim upon itself and its own states and operations.


There are many different higher-order theories of consciousness, but all treat the distinction between conscious and unconscious mental states as a relational matter about the presence or absence of a relevant meta-mental state (Armstrong 1980, Rosenthal 1986, Lycan 1987, Gennaro 1996). What makes one of a person P’smental states M a conscious state is not some intrinsic fact about M itself, but rather the fact that P has a simultaneous higher-order (i.e. meta-mental) state H which represents or asserts to P that P is in state M. A conscious mental state is a state of which one is conscious;e.g.,one consciously desires a cup of coffee if one both desires it and is aware of doing so. Most higher-order theories aim to explain consciousness (or at least the conscious/ nonconscious distinction) in terms of self-awareness, which they explicate in turn by appeal to higher-order states and meta-intentionality. Thus they seem to regard the notion of self-awareness as more basic than that of consciousness or of being a conscious mental state. Insofar as it is the addition of the relevant meta-intentional self-awareness that transforms a nonconscious mental state into a conscious one, such theories imply a dependence of consciousness on self-awareness. 

However, higher-order theories disagree strongly about the status of introspection. In part this reflects the major division among current higher-order theories about the psychological modality of the relevant conscious-making meta-mental states;some theorists take them to be higher-order thoughts, but others regard them as quasi-perceptual higher-order states produced by some system(s) of internal monitoring - the mind’s inward turning eye. Those in the former group, such as David Rosenthal (1986, 1992) are said to hold a HOT (higher-order thought) view of consciousness, while the latter, including David Armstrong (1980) and William Lycan (1987, 1996) are classed as offering a HOP (higher-order perception) model. On both models the higher-order states have a similar reflexive meta-intentional content, something like “I now have a desire for a caffé latt锓I now have (or feel) a sharp and cramping pain in my left foot”, or “I am now thinking about the election”. The content is more or less the same on the two models;what separates HOT theories from HOP theories is the psychological mode of the meta-mental bearer of that content:thought-like on the former and quasi-perceptual on the latter.

The difference has immediate implications for our present concerns, especially regarding the status of introspection relative to the other two members of our triad. On the HOP model, e.g. as in Lycan (1987, 1996), introspection is the process that generates the quasi-perceptual meta-mental states that make unconscious states into conscious ones. Thus introspection turns out to be the most basic feature of the trio on the HOP model. It produces the states of self-awareness that function as the meta-mental components of the relational complexes needed for conscious mental states. Were there no introspection, there would be no perception-like states of self-awareness, and without them no mental states could count as conscious. Thus on the HOP model, there is a very real sense in which introspection, understood as an inward-turned system or process of mental monitoring, is more basic than our two other features and might even be said to be their cause. We can show the priority relations in figure 1.

Introspection -->HOP --> Conscious Mental States

figure 1


The HOT model does not accord introspection a similarly foundational status. Rosenthal (1992) denies that the required higher-order states are perceptual or perception-like. He denies that there any organs of inner perception or that conscious-making meta-psychological states have any sensual or qualitative character except in the sense that they represent some other lower-order states as having such properties. My higher-order thought of my pain may make me aware of its hurtfulness but the thought itself has no qualia, painful or otherwise. The HOT model requires no inner-directed monitors through which introspection generates self-awareness on the HOP view. The main constraint that Rosenthal imposes is that the conscious-making meta-states be must noninferential;if I come to think that I am in pain or angry by observing my behavior and inferring that my mental state from the outward evidence then the resulting HOT would not make my sensation or emotion a conscious mental state. The HOT must arise noninferentially, but beyond that there is no requirement that it be produced by an inner monitoring system or even that it be caused by its lower-order mental object.

How then, if at all, does introspection fit in on a Rosenthal-style model? Rosenthal reserves the term “introspection” for those cases in which the HOT is itself a conscious thought .(1986)Though HOTs make their mental objects conscious, they are themselves not generally conscious;indeed if they were generally such, an infinite regress would be generated by the relational model of consciousness. A given HOT H* is itself conscious only if it isaccompanied by a yet higher-order thought H** whose content is “I am in state H*”. Rosenthal believes that we do not ordinarily have such third order thoughts. When I consciously feel a pain by being aware of it in thought, I do not typically also have the thought that I am aware in thought of my pain. I just have the second order thought, and by doing so I am aware of my pain but not of my thinking of myself as being in pain. However, even if they are not the norm of everyday experience, I can and sometimes do have such third-order thoughts. I am at times aware not only of my pain of but of myself being aware of my pain. It is just such cases that Rosenthal classifies as introspective. Thus rather than being the foundational operation of self-awareness and consciousness as it is on the HOP model, introspection on the HOT model is a special and derivative case of meta-self-awareness that is generated by the iterative operation of the higher-order thought looping back on itself. Diagrammatically we might represent the priority relations on the HOT model by figure 2.

HOT about(1st order M-state) --> conscious M-state/self-awareness

HOT about(HOT) --> conscious HOT/meta-self-awareness/introspection

figure 2


Unsurprisingly the foundational notion on the HOT view is that of higher-

order thought; it provides the basis not only for self-awareness and conscious states in its initial operation, but also produces introspective awareness through its iterative application.

Let us sum up the relations that the two higher-order models imply among our trio of mental features. HOP and HOT models agree that it is self-awareness that produces conscious mental states. However, they disagree about the primacy of introspection. HOP models accord it a foundational role as the source of the perception-like self-awareness that makes states into conscious mental states. Thus on the HOP model, conscious mental states depend upon self-awareness and self-awareness in turn depends upon introspection. The HOT theory accepts the former dependence but rejects the latter. It implies that conscious mental states depend upon self-awareness but of a thought-like type, and that introspection rather than serving as the ground floor basis of self-awareness comes in only at the level of third-order states. For the HOT theorist, introspection is not the source of consciousness but merely the product of the iterative operation of a basically thought-like process of self-awareness.

Thus if is one is sympathetic to the higher-order view - as I confess I am (Van Gulick 1988) - then it seems one could not clarify the status of introspection without deciding between the HOP and HOT versions of the theory. However, that assumes that those two variants exhaust the options open to the higher-order (HO) theorist, and as we will see below in Section III there are plausible alternative ways of developing the HO view that fit neither the HOP nor the HOT model and that offer quite a different account of the role and status of introspection. Nonetheless, the HOP and HOT models are the dominant variants discussed in the literature, and thus we should first review the evidence favoring one of them over the other before turning farther afield to consider other less familiar formulations.


The basic issue that divides the HOP and HOT approaches is whether the conscious-making meta-mental states are perception-like or thought-like in nature. The issue is somewhat ill defined in so far as it is unclear how the notion of “perception-like” is to be interpreted. No one supposes that we have a literal “mind’s eye’ with which we view our inner mental workings, but it remains unclear just what features should lead us to count a meta-state as perception-like or quasi-perceptual in the sense supposed by the HOP theory. Trying to define the notion of “perception” is notoriously difficult, and I do not intend to do so here. Hopefully, it will suffice to list some of the main features of paradigm cases of perception and consider which of those might be shared by our meta-mental states and the inner process that produces them.

Given the prominence of vision in humans, philosophers typically take some everyday example of seeing an ordinary object under normal conditions as their paradigm of perception, e.g., seeing a mug of coffee sitting on the desk beside my computer monitor. The focus on vision may not be theoretically innocent, but for present purposes let’s stick with the standard practice. We can list at least twelve features involved in such cases that seem relevant to their being perceptual.

1. The mental state M generally provides an accurate/veridical representation of the object O.

2. Error and illusion are possible, even if (necessarily) atypical.

3. M is informationally linked by some reliable channel to O.

4. There is a causal link between O and M, including more specific causal links between O’s having a given property F and those of M’s features that represent O as being F.

5. The content of M represents the roughly simultaneous nature of O, i.e. perceptual states in the first instance represent to the perceiver how the world is now. 

6.The process that produces M is noninferential (or at least involves no personal-level inferences of which the perceiver is aware.)

7. The process that produces M is (to a high degree) informationally encapsulated or modular in the sense that it is not penetratable by personal-level beliefs or other information outside the visual system. (Knowing the lines are the same length in the Müller-Lyer illusion does not make one see them as equal).

8. The process is on the whole nonvoluntary - we can control where we look but we can not control what we see when we look.

9. There are organs of sensation/perception involved in the link (e.g. eyes).

10. The object and its features are represented in a given sensory modality (e.g. represented visually as square vs tactilely as square.)

11. M’s content is presentational; M presents O to the agent rather than merely representing O. There is sense in which from the 1st person perspective the perceiver seems directly aware of O.

12. M’s mode of representation involves (or at least seems to involve) a sensuous medium of presentation with associated qualitative properties (or sensuous “feels”) such as the phenomenal greenness of my percept of the mug on my desk. 

There are thus two questions to answer:Which of these twelve features are shared by the relevant meta-mental states? And do the analogies that hold suffice to count the meta-states as perception-like or quasi-perceptual in the sense claimed by the HOP model? HOP and HOT theorists obviously give different answers to the second questions, but that is in part because they disagree about some cases in answering the first. Though there are surely some features they both regard as shared (or both regard as not shared) there are others about which there is no consensus. Moreover in many cases, the comparison is a matter of degree;the meta-mental state shares at most an analogous feature with the perceptual case, and there is room to disagree about both the closeness and significance of the analogy. For example, neither HOP nor HOT theorists believe there are inner sense organs, but HOP theorists do appeal to inner monitors or inner monitoring systems, and one might regard such monitors as functionally equivalent in important respects to organs of outer sense.

With that in mind, let us run through the twelve features and consider to what degree each is or is not paralleled in the meta-mental case. I can not do justice here to the many large issues that such a survey raises, but even a first approximation may help us get a better grip on the underlying dispute.

1 & 2. Accuracy and Illusion. Our meta-mental states are generally veridical at least those that concern occurrent present states, and contrary to past beliefs about the infallibility of self-awareness, most contemporary theorists including both HOP and HOT advocates accept the possibility of error and illusion in at least some such cases. Nonetheless self-awareness seems to have an intimacy, immediacy and epistemically privileged status that is not readily explained by either the HOP or the HOT model.

3. & 4. Causality and Channel Conditions. HOP theorists with all their talk of inner monitors and internal scanning clearly regard the meta-mental case as satisfying both the informational channel (3)and causalconditions, though Lycan has been at pains to acknowledge that there are likely many scanning systems that operate in a diversity of ways. By contrast Rosenthal, the most prominent HOT advocate, is noncommittal on the causal condition. He does not deny that lower-order mental states may typically be among the causes of higher-order thoughts about them, but he declines to build such a causal requirement into his account of when a HOT makes a lower-order mental state conscious. It is the intentional relation not any causal condition that matters. If for example, both the lower-order state and the HOT directed were related not as cause and effect but rather as joint effects of some common cause that need not on Rosenthal’s view exclude the HOT from making its lower-order object into a conscious state. Indeed once such possibilities are raised it seems a HOP theorist as well might accept some relaxation of the causal condition, as long as the inner monitors were still linked in some informationally reliable way with the states within their domain.

Moreover, it seems Rosenthal must concur in accepting the reliable informational channel condition, at least as an overwhelmingly likely empirical hypothesis about the relevant meta-mental states, even if he chooses not to build it into his analysis. For without such a channel condition, the general veridicality of the HOTs would seem like magic or wildly implausible coincidence. So it seems likely that regardless of whether or not he explicitly includes it in his analysis de jure, he is committed to regarding any states that in fact satisfy his analysis as also de facto satisfying the reliable channel condition. 

5. Simultaneity. HOP and HOT theorists agree in regarding rough simultaneity both as a condition on the co-occurrence of the higher-order state with its lower order object as well as an aspect of the intentional content of the higher order state. One can have higher-order awareness of states that one is not now in (e.g. I might now recognize or remember that I wanted to contact a friend last week), but such nonsimultaneous HO states do not make their lower-order objects conscious on either the HOP or the HOT model.

Although the HOP and HOT theorist agree in this regard, the HOP theorist may be better positioned to explain why it is so. Leaving aside cases in which the causal path from object to perceptual system is of long duration ( e.g. the extreme case of light traveling from distant stars or galaxies) we can normally perceive only what is now the case since our perceptual systems are input-driven and constantly tracking the current state. Our thoughts by contrast are not so bound to the present moment. Thus the HOT theorist has a further explanatory obligation to discharge where the HOP theorist has a ready explanation that falls out automatically from the very nature of perception. The HOT theorist needs to provide some special account of why the meta-thoughts involved in making states conscious can only concern one’s roughly simultaneous mental states. Since we can have thoughts about nonsimultaneous mental states, why do such HOTs not make their lower-order objects conscious? 

6. Noninferentiality and Informational Encapsulation. Higher order theorists typically include an noninferential condition in their analysis to exclude cases in which one comes to know of one’s lower order states indirectly by inference or reasoning, for example from evidence about one’s own behavior. (Rosenthal 1992, Lycan 1996)As the result of discussions that lead him to reflect on his patterns of behavior, a patient in psycho-therapy may come to recognize that he is often moved by a powerful desire to avoid criticism. Though such self-insight may be of great practical value, it does not involve the sort of HOTs or HOPs that make their lower-objects into conscious mental states. As result of indirectly and inferentially coming to be aware of one’s desire, one might go on to become directly aware of it as well in a way that did indeed make it a conscious desire, but the initial indirectly derived higher-order states would not by themselves do so on either the HOT or HOP models. 

The addition of a noninferential condition in the analysis may indeed be required to exclude counter examples, but it seems in need of further justification or explanation. Why should it matter that a given higher-order state was (or was not) arrived at by a process involving inference? If it is the intentional content of the higher-order state that makes its lower-order object conscious, why should it fail to do so when inference plays a role in producing the higher-order state? Even if the division between those HO states that make their objects conscious and those that do not should turn out to coincide with the division between those that are arrived at noninferentially and those that are not, that would seem at best an extensional coincidence. That is, based on facts about how we humans specifically operate, we might de facto be able to delimit the set of conscious-making HO-states by adding a noninferential condition, but we would not have explained why the two divisions coincide. It would seem that the real explanation of why some HO-states make their objects conscious and others do not must be found at a deeper level in some other feature(s) that covaries with the inferential/noninferential distinction rather than in that distinction itself. For example, the sorts of meta-representationsgenerated by the inferential route might be different in some important ways from those produce noninferentially. But immediate questions arise: What might that difference be? Why is it required for making a state conscious? And why does it covary with being produced noninferentially?

As with the simultaneity condition, the HOP proponent might claim that his version of the theory gains a small advantage. If as many believe, perception is a basically noninferential process but thought in general is not, then the HOP theorist could give a simple nonad hoc explanation of why the relevant conscious-making HO states are noninferential:they are so because they are a form of inner perception and perception is by nature noninferential. However, the assumption on which the explanation relies may not be justified by the facts;perception may more inferential than many suppose.

We admittedly draw a common sense distinction between what one literally saw or heard and what one inferred on the basis of what was perceived. In response to cross examination of his testimony about when the defendant left his house, a witness might admit that he heard only the car pulling out of the driveway and that he had inferred that the defendant had left home. Or a doctor might acknowledge that he saw only the pallor of the patient and inferred that she was in shock. Despite the ease with which we make such everyday divisions between what was perceived and what was inferred, the distinction blurs when we look closely at our current empirical models of the perceptual process.

Most perception involves the extraction of information about the environment from the features of the sensory signal through the computational derivation of a succession of representations of both the proximal and distal stimulus. Many of these derivational operations are at least quasi-inferential in that they involve content-sensitive processes whose conclusions or end points are more or less logically entailed by the earlier representations in the process. They are sometimes described as “ratiomorphic” to indicate the degree to which they resemble more ordinary cases of reasoning. For example, the human visual system uses stereopsis to compute the distance of an object in the scene on the basis of binocular disparity, i.e. from the slight differences in the images that the object projects onto the two retinae. The details are fascinating but not important here. What matters is merely the fact that the process moves in a content-appropriate way from representations of the retinal images and the disparities between them to representations of the 3D locations that object would have to occupy to produce such slightly offset pairs of images. (Marr 1982)Is such a process inferential? The answer would seem to be, “Yes and no.”It involves some, but probably not all, of the relevant features of paradigm inferences. The process produces representations whose content is rationally implied by the content of the representations from which it derives them. If an engineer carried out similar calculations we would have little reluctance in labeling it as inferential. On the other hand, the process is automatic, unconscious and relatively encapsulated and cognitively insulated. Other sources of information are not typically able to influence its outcome. Nor can the relevant representations enter into other content-sensitive rational processes;their activity is restricted to the internal operations of the steropsis module. To use a colorful term from Stephen Stich (1978), the representations that occur within the process of stereopsis are not very inferentially promiscuous ;the inferences into which they can enter are strictly limited to those that occur within the normal computation of distance. Although my steropsis module uses representations of binocular disparity to compute distance, I have no personal level access to them, and I can not integrate them logically with my general stock of personal level beliefs. Nonetheless, the process is at least “quasi-inferential” in so far as it moves through sequence of representations to conclusions that are implied by the joint content of those that occur earlier in the process.

Consider another visual example, that of the size and distance illusions induced in a viewer by a so called “Ames room”. The famous experimental set up, designed by the psychologist A. Ames (Ittelson and Kilpatrick 1951, Rock 1983), involves a trapezoidal room whose rear wall recedes at an angle, so that its far left corner is twice as far from a front central viewing point as its right corner. However, everything in the room is scaled to compensate for the increase in distance. The black and white checkerboard floor tiles and the window in the rear wall are all trapezoidal rather than rectangular and increase in size from right to left just enough to project the same retinal image to the viewing point as would a normally rectangular items in a standard rectangular room. Thus when the subject looks into the room from the fixed frontal viewing point she sees what appears to be an ordinary room. However, if two people of equal height are placed in the respective back corners of the room, they look to the viewer to be of enormously different sizes. The person in the right corners appears to be a giant, and the one in the left corner is seen as much shorter than an average person. If the two inhabitants of the room switch locations so too does their apparent size, though they both briefly “normalize” as they pass at the midpoint of the back wall.

Two levels of inferential-like processing are involved. First the visual system interprets the retinal image as indicating rectangular objects at a constant distance along the back wall. Given its learned familiarity with the rectangular construction of most rooms, its assumptions are well justified (though false), as is the conclusion it reaches about the likely nature of the distal stimulus. Interestingly the illusion does not appear to work (or at least not nearly so strongly) when it is tried on subjects who do not live in typically rectangular environments. For example when it is run with a subject population of southern African people who live in mostly circular huts, the illusion is not generated. Thus the relevant processes are at least to some extent cognitively penetratable by individually acquired information or beliefs. The second stage of the illusion is produced by inferring the size of the two inhabitants from the conjunction of their retinal image size and their erroneously represented distance. Once again as with the steropsis case, the process does not share all the features found inparadigmatic cases of conscious language based reasoning, but it shares enough to qualify in important ways as inference-like or quasi-rational.

As these examples show,once we begin to investigate the details of perceptual processing, the common sense distinction between what is actually perceived and what is merely inferred gets quite blurry and hard to draw. Thus the HOP theorist may not gain much of an advantage, if any,over the HOT theorist in motivating or explaining their common commitment to the noninferential nature of the higher-order processes needed to produce conscious mentality. The more inference-like perception turns out to be in general, the less help it provides to the HOP theorist in trying to explain why the specific forms of inner perception required for consciousness need be noninferential. Thus both HOP and HOT theorists are left in need of some explanations of why the relevant processes need to be noninferential. As noted above, it seems likely that the real work is being done not the noninferential condition per se, but by some more directly relevant feature that correlates with it. 

8. Involuntariness. What we perceive is not generally under our voluntary control. I can will where to direct my gaze but what I see is more or less determined by the structure of my surroundings. Though expectation, memory and acquired knowledge no doubt exercise some top-down influence, perception remains a largely input-driven process. Prior knowledge makes some non zero contribution to my present perception of a keyboard and computer monitor on the desk in front of me as I type, but nonetheless it is the physical stimulus of the light reflected from them that plays the overwhelming role. As long as I keep my eyes open and looking straight ahead, I can not choose to not see the monitor or to produce a visual experience in myself of some other sort of object. Given their evolutionarily based function of providing the organism with accurate current information about the immediate environment, perceptual processes operate automatically and beyond the reach of interference by volition. Our voluntary control is restricted to matters of orientation and attention. We can decide where to look, what to focus on and what to attend to, but we can not in general chose or even voluntarily affect what we perceive as when we direct our gaze. From the design perspective of natural selection that is probably a good thing.

The matter is quite different with thought. We far more able to direct and control the stream of our thoughts. I can voluntarily shift my thoughts from the paper I am writing to what I am planning to prepare for dinner or to the movie that I saw yesterday. Even if I keep my focus on my paper, I can move back and forth between various issues and direct my thinking down particular lines of reasoning in search of a more compelling or satisfying argument or explanation. If my search succeeds, what I will have “found”will be the mental product or result of an intentionally directed process of construction that operates largely under my voluntary control. Thinking far more than perception responds to volition, and the evolutionary rationale for such a difference is obvious. Nonetheless, ourcontrol of our thoughts has limits. There are things one probably can not will oneself to think, and sometimes we have thoughts that obsessionally resist our best attempts to banish them from our stream of awareness. Yet on the whole we are masters of our thoughts to a far greater degree than we are of what we perceive.

To what degree is introspection under our voluntary control. We obviously have some degree of control over our awareness of our mental states. Through an act of will, the patient in therapy becomes aware of previously unconscious desires, and in more ordinary contexts we can choose to observe desires, beliefs or even sensations that had been beyond our prior notice. However, this might be solely a matter of selective attention, directing our focus of inner awareness upon one or another area of our minds much as we may move our visual gaze around the scene before us. Perhaps in the inner case as well, we can determine where we “look” but not what we see there. If so, our higher order awareness may be no more under our voluntary control than is our vision. If I have a pain in my leg can I turn my attention upon it and yet choose not to be aware of it? If I yearn to win the prize, can I reflect upon my wants but voluntarily exclude that desire from my higher order 

self- representation? In some cases, the answers would seem to be negative. I can not by an act of will make my being in pain appear to me as my merely having a tickle. Perhaps I can distract myself and pull my attention elsewhere, much as I might avert my eyes from a disturbing scene. But in so far as my inner attention falls upon the sensations in my leg, I seem to lack the power to control how they appear to my higher order awareness. However, the situation is less clear with other sorts of mental states, especially those that are more dispositional in nature. Most of us seem able on occasion to hide our motives from ourselves, and self-deception though problematic seems at least possible in many such cases. The nature, indeed even the reality of self-deception is controversial, butin so far as it does occur, we seem to have voluntary control not only over where we direct our inner attention, but also some limited control over what we observe when we do so. However it’s not clear how much this counts as a disanology with the perception. In the external case as well, we accuse people of willful blindness, we say of such a person,“He sees (hears) only what he wants to.”Again there are limits;one may choose to not see a foul committed by one’s team or the misbehavior of a favored child, but can one easily chose not to see the lamp on the table before one’s open eyes.

To the extent that inner awareness seems beyond voluntary control, it may seem more like perception than like thought and thus to favor the HOP theory over its HOT rival. However, the HOT theorist has a ready reply. The HOT theory requires an assertoric higher order thought to make a state conscious. (Rosenthal 1992)One must not merely entertain or consider the thought, but think it in a way that involves treating it as a true, i.e. thinking it in a belief-like way. Once the assertoric requirement comes to the fore, our degree of voluntary control seems to shrink if not altogether disappear. To what degree, if any, we can voluntarily control our beliefs remains controversial and unresolved -Descartes (1641), William James (1897) and Bernard Williams (1973) not withstanding. Thus the HOT theorist may well argue that his view need not conflict in any way with the limited degree to which our higher order awareness is under our voluntary control. Assertoric HOTs may be just as involuntary as HOPs.

Thus if we construe the voluntariness of higher order awareness as primarily an attentional matter of where we direct our inner focus,it would seem to favor neither the HOP nor the HOT view. However,as we will see below, the distinction between where we focus and what we are are aware of it when we do so may not carry over so well from outer to inner awareness, especially if the shift of our inner attention often alters, transforms or even creates the objects that it brings into focus.

9. Organs of Sensation. As noted above, no one supposes that there is a literal mind’s eye that turns its gaze inward on the mental realm;there are no organs of inner sense that transduce physically encoded information into neural patterns for further processing as do our eyes and ear. Nonetheless, the HOP theorists are committed to inner monitors or inner scanners that in some way extract information about our mental states and actions from the mental flux of our mind/brain’s activity. The connection may be more direct and unmediated by any physical carrier of information that plays the role that patterned light and sound do in external sense perception. Nor need there be any anatomically distinct brain unit that has a separate and distinct monitoring function as is the case with external sense organs. Internal monitoring may be realized solely by the pattern of interconnections between brain regions rather than by any distinct organs of inner sense. Nonetheless, the HOP theorists might argue that the systems of inner monitoring no matter how diffusely realized still share enough similarities to sensory systems to count as perception-like or quasi-perceptual in the sense implied by the HOP model. The HOT theorist by contrast will emphasize the differences and deny the analogy suffices;he might well invoke the absence of any organs of inner sense as a basis for rejecting the HOP view.

10. Sensory modality. In ordinary cases of external perception, one is not merely aware of objects and their properties but aware of them in a way distinctive of one or another sensory modality. We see them, hear them or feel them. One may see a book’s rectangular shape or feel it. Although the two perceptual states share a common content, there are obvious differences in how they represent that content. Being visually aware of the shape is quite a different experience from being tactilely aware of it, even if both concern the same external state of affairs. Of course, we do not see or hear our mental states, but if the our self-awareness is perception-like one might expect there to be some distinctive modality associated with inner sense. Yet it is not obvious that there is any such aspect and HOT theorist may see that an mark against the HOP theory.

HOP theorists can respond in at least three ways. Firstly, they might concede that inner sense has no modality akin to those associated with outer sense, but argue that having some such aspect is not essential to perception but merely a contingent feature of our human external senses.

Alternatively they could deny that even our outer perceptions have any aspects of which we are experientially aware other than their representational contents. A defender of the strong representational view might agree that seeing the book’s rectangular shape is different from feeling it, but argue that the differences are all differences in content, i.e. differences in what properties the world is represented as having. (Tye1995)My visual perception of the book only partially shares its content with my tactile experience of it. The former’s total content concerns many other features of the book such as its color, its illumination, and the angle from which I am viewing it. Similarly the total content of the tactile state represents the book as having a certain hardness, texture and as resisting my fingers that push against it with a given pressure. According to the thorough going representationalist, such differences in total content exhaust the experiential differences between the two states. If so, weneed not invoke any modal differences distinct from content differences to explain why the two perceptual experiences seem so different. If there are no special modal aspects associated with external perception, then the HOP theorist should not be embarrassed by their absence in case of inner sense.

As a third option, the HOP theorist might argue that our self-awareness does involve a modality specific aspect, not of course one shared with any of our outer senses but rather a distinctively introspective aspect. Such a claim might be supported by an appeal to first person phenomenological evidence, and I must admit that based on my own self-observation the claim has at least some plausibility.

However, HOT theorists seem not to share such introspectively based intuitions. When they examine their own self-awareness they seem to find only bare propositional representations unclothed in a modally distinctive form. If they be correct, the HOT view might better fit the facts. If our higher-order states represent their lower-order objects in only an abstractly propositional way, then higher-order thoughts might seem more likely as the bearers of such content. Thus whether one counts the modal nature of perceptual states as favoring the HOP view or the HOT view turns in part on one’s intuitions about the phenomenology of self-awareness.


11. Immediacy and Presentness . Our ordinary perceptual experience of objects has an immediacy that is not typically paralleled in thought. In G.E. Moore’s (1922) famous phrase, perceptual experience is “diaphanous”;we “look through” our experience and are aware of objects right before our eyes. (See also Van Gulick 1988 and Strawson1994.)As a matter of phenomenology, the monitor, keyboard and lamp on my desk are present to me rather than representedto me. The objects themselves appear to me directly, or so at least that is how I experience them. Though my perceptual experience may depend upon representational processes, I am not aware of it as such. The underlying basis of my experience may be representational, but as a matter of first person phenomenology I simply see the objects on my desk, and it is they that appear present to my awareness.

The objects of thought do not seem similarly immediate or present to awareness. If I think now of the Eiffel Tower, of my breakfast a few hours back, or even of the lamp on my desk which I believe continues to exist when I close my eyes or turn my gaze away, the intentional objects of my thought do not seem present with the immediacy or directness I find in perception. Though my primary focus in on the objects I am thinking about, their status as objects of thought insinuates itself to some degree into the phenomenal content of my experience. I am thinking first and foremost of the Eiffel Tower;it is there that my attention is focused. But I am nonetheless also aware that it is not the Tower itself, butmy idea or mental act of thinking of it that is the immediate object of experience. Returning to Moore’s metaphor, thinking of an object seems less transparent than perceiving;the means or medium of representation is never fully lost from the experiential view. It is difficult if not impossible to “look through” one’s thoughts as fully as one typically does in perception. The distinction between mental act and object seems phenomenologically more salient in the case of thought and never completely invisible. 

As we will see below in Part III, our initial intuitions may mislead us here. As a phenomenological matter, perception may turn out to be more self-referential and at least implicitly less diaphanous than is commonly supposed. But for now, let us accept the common phenomen- ological assumption that perception presents its objects to awareness in an apparently more immediate way than does thought. If that were so, what consequences if any would that have for the HOP vs HOT debate?

At least initially the HOP view may seem to gain a slight advantage. Our awareness of our own mental states and process does seem to have the sort of presentational immediacy that commonly occurs in perception. When I am aware of my having a pain in my toe, of my being thirsty, or even of myself as now thinking of Paris, the mental object of my awareness does itself seem present to me in much the way my keyboard and lampappear as present to when I see them on my desk. Of course, the HOP theorist’s claim that such states of inner awareness are perception-like does not in itself explain their apparent immediacy;indeed that feature stands in need of further explanation with regard to external perception as well. Nonetheless, the HOP theorist might seem better situated to provide a satisfactory account of the immediacy of self-awareness in so far as that feature seems analogous to that which we find in other cases of perception. By contrast, the HOT theorist shoulders an extra burden or explaining why the objects of inner directed thoughts are immediate in a way that outer directed thoughts are not. Or so at least it seems.

The HOT theorists again has multiple avenues of reply:

•He might deny the alleged difference of immediacy in the ordinary casebetween first-order thoughts and first-order perceptions.

•He might challenge the alleged immediacy of inner awareness and deny that we have any relevant sort of direct acquaintance with our mental states and processes.

•He might acknowledge that our inner awareness of our mental states has an immediacy not typically present in our first order outer directed thoughts, but provide some satisfactory explanation of why HOTs should immediately present their objects in ways that first-order thoughts do not.

Of the three strategies, the third looks the most promising since the first two seem to require overturning powerful first person intuitions about the phenomenal immediacy or nonimmediacy of various forms of inner and outer awareness. Nonetheless, even well entrenched intuitions can mislead us and may have to be revised given enough evidence. So neither of the first two options should be counted out, even if the third offers better odds of success. To pursue the latter, the HOT theorist might appeal to the existence of a direct causal link between the sorts of HOTs that make a state conscious and their lower order mental objects. Given the noninferential nature of the link, one might suppose that as a matter of phenomenology the object of such a thought would be experienced as present in an unmediated way. I am not sure that would work, but perhaps something along those line might enable the HOT theorist to explain why HOTs present their objects with an immediacy that first-order thoughts do not. In sum, the HOP theorist seems at least initially better able to accommodate the apparent immediacy of self-awareness, but the HOT theorist has options for dealing with it as well. 

12. Qualia. We come at last to the most controversial feature on our list. Our perceptual encounters with external objects typically involve a rich variety of sensory properties such as colors, tastes, smells, sounds and feels. When I lift the coffee mug to my lips and drink:I see its celadon green color, I feel the smooth texture of its handle, I smell the aroma of the just brewed coffee, I taste the pleasantly bitter just-burnt edge of the French roast beans, and I feel the warmth of the liquid against my lips. Such experiences are often said to involve qualia or phenomenal properties. However, the status and nature of such properties remains a matter of ongoing philosophical dispute. There is no consensus on even the most basic matters:Are qualia identical with or distinct from the properties of the external objects? (Lycan 1987, 1996, Jackson1980, Shoemaker 1990)Are qualia features of inner objects or are they merely properties that inner states represent external objects as having? (Lycan 1996,Block1996) If there are any qualia, are they ineffable, private, nonrelational or intrinsic? (Dennett 1988)These are not questions we can settle or even begin to discuss here. Nonetheless, the fact that perception involves sensory qualities of the sort that lead many to believe in qualia is an issue that we must consider in assessing the state of play between the HOT and HOP theorist. So, let us set the controversies to one side and speak of qualia in relatively neutral way. That is, let us assume that external perception typically involves qualia, but where we take that to mean only that it involves the perception of sensory properties (e.g colors, tastes, smells, feels) of the sort that lead many to believe in philosophical qualia.

What implications might that have for the HOP vs HOT debate? The HOP theorist might seem to gain an advantage similar to that which he appears to have on the issue of immediacy. The situation with qualia looks to parallel that with immediacy;first-order externally directed thoughts seem to have qualia, but first-order thoughts appear to lack them. If, as many suppose, self-awareness involves qualitative character, then the HOP theorist might seem better set to accommodate them by analogy with cases of external perception, on a par with his advantage regarding immediacy. Again the mere fact that the relevant feature, qualia (or immediacy) is present in both the externally directed first-order case and in the internally directed higher-order case, does not by itself suffice to explain its presence in the latter, but at least it provides some hope of finding an explanation that relies on theperceptual nature of both sorts of states. If there is no similar parallel between externally directed thoughts and internally directed ones, then the HOT theorist would again be left needing an additional explanation of that disparity.

However, the phenomenological facts are far from clear. As long as we do not pack too much controversial philosophical baggage into the notion of “qualitative character” we can probably safely assume two things. First that external perception involves qualitative character, and second that self-awareness involves awareness ofqualitative character. 

When I am aware of myself as having a pain in my toe, I am aware ofits hurtfulness.

However,some of the other claims needed to make the alleged disparity argument work are more controversial. In particular, it is not obvious that first-order thoughts lack qualitative character, nor that our states of self-awareness - as opposed to some of their first-order mental objects - themselves have any qualitative character. If either of these two claims were not merely unobvious but in fact false, then the disparity to which the HOP theorist appeals would collapse. If first-order thoughts do have qualitative character, then the HOT theorist would be at no disadvantage in addressing the qualitative nature of self-awareness. Alternatively, if the only qualia involved in self-awareness are those present in some of its first-order objects, then there are no higher-order features that the HOT need be at a disadvantage in explaining. How might the HOT theorist try to motivate either of these two moves. Regarding the first, he might argue that in so far as there is something that it’ `s like to think, then there are qualitative features associated with such thoughts even if they are not the same qualia associated with perceptual experience. When I think of drinking of the coffee from my mug, my mental state may not involve any taste qualia or warmth qualia as it would were I really to sip it, but there is nonetheless a phenomenological aspect, a “what it’s like-ness” to be having such a thought. If so, the HOT theorist could argue that first-order thoughts have qualitative character and thus that he is no worse off in addressing the supposed qualitative character of self-awareness. The alternative move would be to challenge the belief that higher-order states themselves have qualitative character. Though widely held, the HOT theorist might attack that belief as the result of confusedly treating the character of the states of which we are aware as some character of the states of awareness. Rosenthal (1991) argues that when we have a conscious pain, we are aware of hurtfulness and when we have a conscious visual experience of a ripe tomato we are aware of phenomenal redness, but neither the hurtfulness nor the redness is a qualitative feature of the relevant higher order state. That state, which on his view is a HOT, has no qualitative character of its own. What has qualitative character is the first-order perceptual or sensory state that is the object of the HOT. By having a relevantly directed HOT, we make the first-order state into a conscious perception or conscious sensation and thus become aware of its qualitative character. Since on Rosenthal’s account the higher-order state has no qualitative character of its own, the HOT theorist need not be embarrassed or disadvantaged if thoughts lack qualitative character. Indeed Rosenthal tries to turn the tables and gain an point against the HOP theory. He argues that if the conscious-making higher-order states were perception-like, then they should have distinctive qualia associated with them just as externally directed perceptual states do. But, as he goes on to argue, we encounter no such distinctive higher-order qualia;the only qualia we met are those of the first-order states themselves. There are distinct qualia associated with a first-order state of seeing red, and we become of aware of them if our visual perception becomes conscious. However, there are no further qualia associated with the second-order state of being aware of that perception, nor do we become aware of any such second-order qualia if and when we form a third-order awareness of that second-order state. According to Rosenthal’s account of our phenomenology, the only qualia that ever show up are those of first-order states. Thus he believes the appeal to qualia supports the HOT view rather than the HOP. Once again, there appears to be no decisive resolution of the debate.

Having surveyed each of the twelve features of paradigmaticperception, let us recap their implications for the HOP vs HOT debate.

Conditions 1-5 appear generally compatible with both views and thus seem to favor neither;both seem able to accommodate a norm of veridicality, the possibility of error, the causal condition, the information channel condition, and the simultaneity condition. Neither view seems to have an adequate non ad hoc explanation of their shared noninferential requirement. Though the HOP theory may seem at first to garner some support from the alleged noninferential nature of perception, that benefit vanishes upon a closer examination of the nature of perceptual processing, yielding a draw on noninferential (6) and informational encapsulation conditions (7). On some of the other conditions, one side or the other seems to gain at least an initial advantage, but the opposing side always has resources with which to fashion a plausible reply. In that sense the HOP view seems initially favored by the nonvoluntary (8) and presentational (11) conditions, and the HOT side gets prima facie support from the organs of sensation (9) and sensory modality (10). But in each case the initially disadvantaged side has options for reply. Finally, the qualia condition (12) can tilt either way depending on what one takes the facts to be about the phenomenology of inner awareness. Thus we are left with no clear judgment in favor of either side, though I hope that working through the specific conditions has clarified the parameters of the dispute.


We come back thus to our original question about the relations among our triad of concepts:introspection, consciousness and self-awareness. The HOP and HOT views give different answers about which of the three is fundamental. Introspection in particular turns out to be foundational on the HOP view, but derivative on the HOT view. In so far as we can not decide between its two main variants, the higher-order theory per semay seem to provide us with no real answer regarding the status of introspection. However, we may gain some insight by considering an alternative form of the theory that differs from both the HOP and HOT models.

Moreover, such a move might be independently motivated by some common problems that confront both of the mainstream versions, in particular a challenge raised by Fred Dretske (1993, 1995) among others (Byrne 1997) that we can call the generality worry. The problem in brief is as follows. Why and how do higher-order states turn their lower-order mental objects into conscious states, given that in general perceiving some x or thinking of x does not make x conscious. If I perceive the pencil on my desk or think of the the snow that fell yesterday, I do not make either the pencil or the snow conscious. They become objects of my awareness, but it would be bizarre to hold that I had thereby created conscious writing instruments or conscious precipitation. But how then can the relational fact of my having a higher-order state about some first order belief, desire or sensation convert what had been an unconscious state into a conscious one? We might mean by “conscious state” merely one of whichI am aware, but in that sense the pencil as well is conscious when I perceive it. It too is an “x of which I am aware.”But the distinction between conscious and unconscious desires, beliefs, or sensations seems to mark much more than that. The higher-order theorist - whether HOP, HOT or something else - needs to explain what is special about the meta-mental case in virtue of which merely perceiving or thinking of something nonconscious can make it conscious. Nor will it do to claim it is just a definitional matter;i.e. that we apply the word “conscious” to mental states of which we are aware, but do not apply it to nonmental items of which are aware in either perceptual or thought-like ways. A deeper explanation is needed of why we use the word that way and of the real distinction that we believe we capture in doing so. To use Thomas Nagel’s famous phrase, a conscious state is one “that it’s like something to be in.”(1974)There is something that it’s like to have a conscious desire, a conscious sensation or a conscious thought. Conscious states have a subjective or phenomenal aspect, and the generality argument turns on the apparent inadequacy of the higher-order theory to explain how a state without any such aspect could be transformed into one with it by the merely relational addition of a meta-state having the first state as it intentional object. It seems that converting the first state into one that there is “something that it to be in” would require some change in the state itself rather than just making the object of a higher-order thought or perception. 

Higher-order theorists have responded to the generality objection in several ways, but all seem less than adequate. Rosenthal (1991) has defended the HOT theory by arguing that nonconscious sensations and perceptions have qualia, but since the states themselves are not conscious we are not aware of their properties. Thus there is nothing that it is like to be a person having such a nonconscious state. It is only when we come to have an appropriate HOT about the sensation that we are aware of its qualitative properties and thus that there is something that it is like to be in such a state. On this view, the first order state had its qualia all along. When the person forms a HOT about the state,she does not change that state. However, she does change her overall state of mind to one that there is something that its like to be in because now, and only now, is she aware of the qualia in virtue of her HOT directed at them.

AcceptingRosenthal’s account would indeed entail that there is something that’s like to have a conscious state, but it seems oddly off target. It implies that there is something that it’s like for the agentto be having a HOT about a first order perception, but that subjective aspect seems stranded between the two states. It is not an aspect of the first order state itself since there need be nothing that it’s like to be in that state. Nor is it really an aspect of the higher order state since that state on Rosenthal’s account has no qualia. The subjective “what it’s likeness” thus exists in a sort of twilight zone between the two;it’s not really a feature of either state but yet there is something that it’s like for agent. In that respect Rosenthal’s reply to the generality argument is not likely to allay worries about how the addition of a meta-state, itself without qualitative character, somehow moves the agent from a state that there is nothing that “it’s like to be in” into one that there is something that it’s like to be in. Merely adding a non-qualia thought about qualia seems incapable of producing such a transformation.

Lycan (1997) has defended his HOP version of the theory against the generality objection in part by invoking a “divide and conquer” strategy. He denies that the HOP model is supposed to explain the “what it’s like-ness” of conscious states, a notion about which he is deeply skeptical and which we regards as engendering a wealth of confusions. According to Lycan, the HOP model aims merely to mark the distinction between conscious and nonconscious states. It does so by defining the former as states ofwhich we are aware in the relevant inner perceptual mode. The HOP model in not intended to explain qualia or a mental state’s having a qualitative aspect. Lycan deals with qualia in another part of his overall theory of consciousness, which offers a purely representational theory of qualia as properties that actual objects (or possible objects) are represented as having. Given the divided nature of his explanatory project, to complain that the HOT model does not adequately account for the qualitative aspect of conscious mentality is to fault it for failing to produce a result it never aimed to achieve. Fair enough, but one might still fault the HOP model for doing less than one would like or expect of the higher-order theory . One can distinguish conscious mental states in many ways. Lycan’s reading of it as “state of which we are aware” is one plausible reading, but we also often use it to distinguish states with a qualitative aspect from those without. In so far as the generality argument appeals to that latter notion, one may reasonably criticize the HOP model for explaining less than we may want, even if it is modestly eschews any aim to explain such matters. We would like to have a convincing account of what the difference is betweena conscious desire with a felt experiential aspect and unconscious unfelt desire. Lycan may perhaps be able to satisfy our explanatory demand with other parts of his theory, but the generality argument seems to show that the HOP model by itself it does not seem to do so. Thus it would be nice if we could find an alternative version of the higher order theory that did.

Thus we need to consider what alternative forms the higher-order theory might take that would differ from both the standard HOP and HOT models. Both agree in regarding the relevant higher-order state as distinct and separate from its lower object. This is perhaps clearest and most explicit on Lycan’s HOP model, since he accepts a generalized language of thought view of representation and describes the higher-order perceptions as being realized by token representations that occur within the operations of the monitors that produce higher-order inwardly directed perceptions. (1996)However, Rosenthal’s HOT model is equally committed to the token distinctness of the higher and lower-order states. As we just noted, he accepts that many first-order states have qualia, but he denies that higher-order states have any qualia of their own. Thus by a simple application of Leibniz’s law they can not be very same states. Although both HOP and HOT theorists assume distinctness or nonidentity, it is not entailed by the higher-order theory per se, and one HOT theorist (Gennaro 1997) has disavowed it. Thus one could try to develop the higher-order view in a way that rejected or at least weakened that assumption, and for the remainder of the paper I will explore that option.  Although the idea may seem initially odd and little more than a mere logical possibility, closer examination will reveal it to be far more plausible than one might at first suppose. Indeed it may offer a line of reply to the generality argument that is stronger than any of those produced by more mainstream HOP or HOT theorists.

Three strands from the current literature offer important clues about how one might go about constructing a nonstandard version of thetheory with a lessened commitment to the nonidentity of lower-order and higher-order states:Daniel Dennett’s theory of consciousness as cerebral celebrity, Chris Hill’s view of introspection as a matter of volume control, and the widely accepted hypothesis that the neural correlate of consciousness is a globally distributed brain state. Let me say a bit about each.

•Consciousness as cerebral celebrity. According to Dennett’s multiple drafts theory 1991), the distinction between conscious and nonconscious mental states is blurry, admits of degrees and turns on two principal dimensions. The first concerns the degree to which a mental state (or content fixation) influences the subsequent development of the system’s states and its outputs. This is what is meant by “cerebral celebrity”;to put it crudely, the more effect a given content fixation has on what other content fixations occur, the more “famous” it is. Conscious states take a more powerful and broader range of content-relative effects throughout the agent’s mind;a conscious perception (thought or desire) and its content will be accessible to other processing areas, more able to affect other states (thoughts, desires, memories) and have more impact on those states driving the system’s output, especially on the system’s reports about its state of mind since conscious state are normally ones that we can report ourselves as being in. All these aspects of influence admit of degree, and in general the greater the impact of any given state the greater its level of cerebral celebrity. Thus in so far as being a conscious state is a matter of such “intra-mental fame”, whether or not a state is conscious need not have a strict yes or no answer.

The other dimension of consciousness on the multiple drafts model is the degree to which a given content gets integrated into what Dennett describes as the ongoing serial narrative the system constructs from the “stream of consciousness”. This is not a separate meta-narrative that is produced independently or over and above the system’s lower-order content fixations. Rather it is an assemblage of activated lower-order contentful states that cohere together in such a way that they form a more or less integrated set from the perspective of a unified self. Dennett denies that there is any separate self that constructs or views the sequence;rather it is the other way round. It is the coherent serial narrative that is fundamental and the self is merely a virtual entity that exists as the perspectival point which is implicit in the narrative and from which the narrative hangs together as unified.

Dennett’s multiple drafts theory is thus a higher-order theory of a sort, though it differs greatly from more mainstream HOP and HOT models. A state with a high degree of cerebral celebrity will typically be one that the agent can report being in, and such a report would express the relevant higher-order thought. Indeed Dennett, like Rosenthal,relies heavily on a tight link between a state’s being reportable and its being conscious. The second aspect of his theory also has a decidedly higher-order slant, since incorporation into the serial narrative carries with it the status of being represented as a state in the stream of the (virtual) self, which at least implicitly involves higher-order representation. 

•Introspection as volume control and activation. Chris Hill (1991) has faulted the “inner eye” model of introspection as overly passive. He has argued that introspection is active in the sense that it often alters its lower-order mental object.

In a case of paradigmatic external perception, as when I see the lamp on my desk, my awareness of the object does not change it. The lamp is unaffected by being seen. However, inner awareness does often seem to alter its objects. When I turn my inner attention to the lingering taste of the olive that I ate a few minutes ago or to the ache in my lower right molar, directing my awareness upon those sensations can change many of their features. The sensation often gains in intensity and vividness;various sensory properties may become more specific, shift from one specific character to another or even emerge where no detailed character was previously present. Of course, as noted above in section 3, a redirection of attention typically leads to changes in external perception as well, but there it is usually only the perceptual state that changes not its object. When I visually scrutinize my desk lamp, I become aware of many details that were previously unnoticed but the properties of the lamp itself remain unchanged. Admittedly in some external cases, the act of observation does change its object. That is apparently so at the quantum mechanical level and obviously so in many social situations. Indeed designing non-obtrusive measures is a perennial problem in the social sciences. However, in the interpersonal case, it is not the act of observation per sethat produces the change but rather the subject’s awareness at some of level of being observed that does so.

Hill thus contrasts the “inner eye” model of introspection with alternatives that he refers to as “volume control” and “activation” to emphasize the respects in which the intensity, character, or even the existence of a sensation (or other lower-order state) can be affected by the occurrence of a higher-order awareness directed at it. He seems to regard this as a problem for the perceptual view of introspection and thus for the HOP model of consciousness. Lycan (1996), however, denies any such negative consequence follows for the HOP view. He accepts the active nature of inner awareness and the many ways in which it may alter its lower-order object, but denies that the HOP view is committed to a passive model of inner perception as the “inner eye” analogy might suggest. Thus Lycan accepts the data Hill presents but claims they are fully consistent with the HOP theory. For present purposes, we need not settle that latter dispute over consistency;it is the active nature of introspection that matters, and about that they agree. 

• Globally distributed neural correlate of consciousness. Current scientific evidence on the neural correlates of consciousness indicates that there is no special local brain area(s) that are the unique (or special) basis of conscious experience. Rather any given conscious state appears to be realized by a globally distributed pattern involving many different cortical and subcortical regions that are simultaneously active and bound together in some way, perhaps by regular oscillations that entrain neural firing patterns in disparate areas of the brain. An important consequence of this result is that the very same regions that are involved in the processing and realization of nonconscious mental states are also among the correlates or realization bases of conscious mental states. For example the areas of visual or auditory cortex that are active when one nonconsciously perceives a stimulus are also components of one’s conscious perception of such a stimulus. The difference between the neural correlates of the conscious and nonconscious states is not that the information gets passed on and re-registered or re-represented elsewhere but rather that those same areas get integrated into a larger unified pattern of global brain activity in the conscious case.

Given these three suggestions from Dennett, Hill and the neuro-imagers, we are in a position to construct an alternative higher-order model that more intimately links the meta-mental states with their lower-order objects and thereby offers a possible solution to the generality problem. We can tentatively label it the higher-order global state (HOGS) model. The basic idea is that a mental state becomes a consciousness state by being recruited into a globally integrated pattern of brain activity that is the momentary neural realization of an episode in the experiencing subject’s stream of consciousness. Contrary to the HOP and HOT models, the transition from unconscious to conscious status on the HOGS model does not result from the production of an independent meta-representation, but rather from the original lower-order representation’s taking on a new systemic role through its integration into the larger pattern associated with the transient and shifting dominant focus of neural and mental activity.

Nonetheless the HOGS model remains a type of higher-order theory in so far as the change that occurs in a lower-order state’s function as it is integrated into the momentary global correlate of self-awareness transforms its content in ways that involve a heightened element of experiential self-reference. Let me spell this out a bit more fully. Though some might disagree, let us assume that an intentional state’s content is (at least in large part) a function of its functional role with the system that contains it (however that might be defined, whether narrowly, widely or otherwise). Thus when there is a significant change in a state’s function, there will typically also be a corresponding change in its content. Thus if a lower-order state changes its function by being integrated into a larger global pattern, its content may well change accordingly. What sort of content change might occur? The global patterns associated with the HOGS model are the neural correlates or realizations of the sequence of states of a self-consciously experiencing subject. Thus when a previously nonconscious state is recruited into such a global pattern of activity, one would expect it to take on a role and content relative to the mental reality of that larger unified organization. A nonconscious perception of a small red book in front to the right might be transformed so as to contribute an analogous aspect to the conscious subject’s experiential stream, which at least implicitly involves a self-referential element. The content of the relevant conscious experience might aptly be specified as “I am now aware (visually) of a small red book in front and to the right of me”. To some degree this self-referential aspect may be present in a limited and wholly implicitly way even in states that we typically regard as nonconscious. Even nonconscious perceptions or a nonconscious desires may incorporate some implicit elements of self-reference in their satisfaction conditions:a complex such as “me-seeing: a tree/ here/now”or “me-desiring: a drink of water/here/now”may do a better job of capturing the intentionality of such nonconscious states than would the merely objectual “tree/here/now”. In so far as this is so, the change in content that accompanies the move from unconscious to conscious status is not the totally de novo addition of a self-reflexive aspect but rather the transformation from a limited and implicit self-referential aspect to a richer and more explicit one.

Thus, although the HOGS model does not deny that there are content differences between first and second-order states, it recognizes more continuity than either the HOP or HOT models. In particular it allows for the existence of at least some forms of limited implicit meta-mentality in nonconscious states. The transition from nonconscious to conscious involves a significant increase in the quantity, quality and explicitness of that meta-mental aspect rather than its initial onset. 

Developing the HOGS model is detail is not something I can do here, nor can I be sure that I would not meet insurmountable obstacles were Ito try to do so elsewhere. But at least prima facie, it seems to offer a promising alternative to both the standard HOP and HOT models. Moreover, it may provide a natural and satisfying way out of the generality problem. On the HOGS model, lower-order states are retained in a somewhat transformed way as constituents of the integrated complexes that realize the transition to conscious status. For example, the representations in visual cortex that constitute my nonconscious perception of the red book are retained as elements in my conscious perception of it. Those cortical representations may not be exactly the same, since they may be altered in the sorts of ways suggested by the volume control and activation models described by Hill. Indeed integration into the temporary focus of global activity will typically involve an intensification of local activity in response to reciprocal resonant amplification through out the global pattern, as well as perhaps some strengthening from attentional mechanisms such as those mediated by the specific thalamo-cortical pathways. Even should the local pattern remain largely unchanged in its intrinsic features, it would still likely change both its overall functional profile and its associated intentional content in virtue of the new patterns of global activity into which it enters. Nonetheless, either the original lower-order state (or one of its near analogues that results from these various changes) gets incorporated into the larger complex that makes it conscious. Thus if the original nonconscious state had sensory qualia, they will be included in a transformed way in the associated conscious complex. Moreover, it is only when such a state is integrated into the larger global pattern of activity that it functions as an aspect of the experiential life of the conscious subject, and thus only then that there is something that it’s like to have such a qualitative perception.

The possible solution this offers to the generality problem is similar in some respects to that offered by Rosenthal in defense of the HOT theory, but also importantly different in ways that may make it more satisfying. Both the HOGS model and Rosenthal’s defense say there is nothing that it’s like to be in an unconscious perceptual state, even if the state has nonconscious qualia. Both also say that there is something that it’s like to be in a conscious qualitative state. However, for Rosenthal the relevant qualia are properties only of the lower-order state, and as we noted above it is puzzling why there should be something that it’s like to have a HOT about a qualia state given that the HOT itself on Rosenthal’s account has no qualia.

On the HOGS model, the higher-order state has qualia, and indeed it alone has qualia in the form needed to produce an experiential what it’s likeness. The qualia are those that were present in the lower-order state since that state is preserved in a somewhat transformed way as a constituent of the global complex that realizes the conscious state with its added higher-order aspects. It is only when those qualia are incorporated via the larger state into the self-conscious experiential stream of the agent that there is anything that it’s like to be in the relevant state. The transforming higher-order process is not one of generating a separate non-qualia thought or representation about the original qualitative perceptual state, but a matter of activating that state and integrating it with its associated qualia into the momentary global state as an aspect of the experiential agent’s phenomenal world. This then provides the avenue of response to the generality argument. Having a thought or a perception about a non-mental item such as the lamp on my desk does not make the lamp conscious, since the lamp itself can not be recruited into a global pattern of activation realizing the conscious experience of an agent. Because it can not become a constituent of any such global state it can not contribute any qualia to it, nor does it have any qualia to contribute. However, nonconscious mental states can generally be recruited into such large scale complexes, and if such a nonconscious recruit has qualitative features then they may contribute to the experiential “what it’s like-ness” of that stretch of the agent’s consciousness.

The problem for Rosenthal was the puzzle of why or how the addition of a nonqualitative meta-representation of a qualitative state should create a “what it’s like-ness” where none previously existed. The HOGS model appeals to the part/whole nature of the link between the nonconscious state and the conscious complex into which it is recruited to answer that worry. Although recruitment into the HOGS state need not involve the creation of any new higher-order qualia, the HOGS state incorporates the qualia of the recruited state and transforms them into aspects of the ongoing phenomenal world of the experiential agent. It is in just those circumstances that we would expect to find consciousness in the “something that it’s like to be” sense. 

Let me back up a just a bit, since I have been speculating at a pretty rapid rate. I have not done any more than offer the HOGS model as a promising variant of the higher-order theory. I have not worked it out in rigorous detail, and certainly not shown that it is the best form of the higher order view. Nonetheless, I hope to have made it at least prima facie plausible and given good reasons for seriously pursuing it as an alternative to the more standard HOT and HOP versions of the theory. I believe on balance it has many advantages, but I do not claim to have shown that here. Doing so will have to wait for another occasion.

Let me close by considering how the HOGS answers our initial questions about the relations and dependencies among our triad of notions:introspection, consciousness and self-awareness. As we noted above, the HOP and HOT models both view conscious mental states as the result of a type of self-awareness, involving distinct higher-order states directed onto lower-order objects, which are made conscious by the very fact of being meta-represented (whether in a thought-like or perception-like way). However, the two views differ about the status of introspection. It is primary on the HOP model as the process that generates the required HOPs, but it is derivative on the HOT model as merely a special case in which a HOT itself becomes the object of a yet higher-order thought. What are the comparable relations on the HOGS view? 

The HOGS model agrees with both the HOT and HOP views on the first issue in so far as the transformation of a nonconscious state into a conscious involves an element of self-awareness. However, the way in which self-awareness comes into the process is quite different on the HOGS model. It is not a matter of producing a distinct separate meta-state. Self-awareness figures instead as an essential but largely implicit aspect of the overall global state into which the lower-order state is recruited. By being integrated into that larger pattern of activity the previously nonconscious state becomes part of the phenomenal experiential stream of the conscious agent. Self-awareness in at least implicit form as a basic feature of the structure of phenomenal experience. The empirical reality of “world over against self” or of “self confronting world” is,as Kant (1781) showed us two centuries ago, built into the very structure and organization of experience. It is in that sense that the transformation of a nonconscious state into a conscious one essentially involves or depends upon the addition or enhancement of self-awareness. We might take a cue here from Dennett’s hypothesis about the virtual structure of the self as implicit in the intentional perspectival structure of the serial narrative. What makes a system of representational states into a self (or perhaps it would be better to say more self-like) is largely a matter of the intentional coherence and causal unity that binds the elements of the system together. (Dennett1988b, 1991;Van Gulick 1988)A HOGS theorist might go one step farther and argue that recruiting a nonconscious state into the global pattern is a way of integrating it into the coherent structure of the conscious self. Given the implicitly self-reflexive nature of phenomenal representation, the very act of integrating the recruited state into the global unity necessarily embeds it within a more self-aware structure. There need be no further discreet explicit higher-order representation to produce self-awareness as there is on the HOP and HOT models. The very act of integration itself suffices. 

Introspection on the HOGS model is neither as fundamental as on the HOP model nor as marginal as it seems to be on the HOT model. On the HOGS model, the transition from nonconscious to conscious states is not produced by introspection, at least not in so far as introspection is thought of as a form of inner perception as it seems to be on the HOP model. The model of introspection as a kind of “inner vision”does not fit all that well with the sorts of processes that are involved in recruiting a state into the dominant pattern of global activity. Thinking of introspection in terms of Hill’s alternative models of volume control or activation may better link it to the process of recruitment, but it is not clear that introspection in any (but the most expansive form) need be involved in every case of recruitment. We can and do direct our inner attention to various aspects of our minds, and by doing so we can often bring a previously unconscious state or process to consciousness. But the moment to moment flow of consciousness is not normally controlled by any such active “looking” or attending. Other factors - perhaps the intensity of new local regions of activity produced by powerful or salient external inputs or perhaps the internal dynamics within the unified experiential global pattern itself( e.g. when I work through a conscious bit of inferential reasoning) are at least as likely to be responsible for the shifting character of the stream of consciousness. Each shift in the stream involves the recruitment and integration of new mental elements that are thereby transformed into conscious mental states. But in so far as many (most) such shifts do not rely upon introspection, it can hardly be regarded as foundational for consciousness as it seems to be on the standard HOP model. On the other hand, introspection does seem to play some role at least some of the time in determining which states get integrated into the larger global pattern, especially if we interpret introspection as including Hill’s notions of volume control and activation along with the more common inner vision variety. In such cases introspection plays an active role in producing the first level of consciousness. It is not confined to the derivative third-order role involving only HOTs about HOTs to which Rosenthal seems to restrict it. Thus if the HOGS model fulfills its initial promise as a viable alternative to the HOP and HOT models, it would yield importantly different conclusions about what relations our trio of notions bear to to each other and to the higher-order theory of consciousness. Further investigation is clearly called for.



Armstrong, David M. (1980) “What is consciousness?” In D. M. Armstrong The Nature of Mind and Other Essays. Ithaca, NY: Cornell University Press.

Block, Ned (1996) “Mental paint and mental latex.” In E. Villanueva (ed.) Philosophical Issues: Perceptual Content. Atascadero, CA: Ridgeview Publishing 19-50..

Byrne, Alex (1997) “Some like it hot: consciousness and higher-order thoughts.” Philosophical Studies, 86: 103-29.

Dennett, Daniel C. (1988) “Quining Qualia.” In E. Bisiach and A. Marcel (ed.) Consciousness in Contemporary Science, Oxford: Oxford University Press 42-77.

Dennett, Daniel C. (1988b) “Why everyone is a novelist.” Times Literary Supplement,1016-22.

Dennett, Daniel C. (1992) "The interpretation of texts, people and other artefacts.”Philosophy and Phenomenological Research 50: 177-94.

Dennett, Daniel C. (1991) Consciousness Explained. Boston: Little Brown and Co.

Descartes, Rene (1641) Meditations of First Philosophy. Paris: Michel Soly.

Dretske, Fred (1993) “Conscious experience.” Mind 102: 263-83.

Dretske, Fred (1995) Naturalizing the Mind. Cambridge, MA: MIT Press.

Gennaro, Rocco (1997) Consciousness and Self-Consciousness. Amsterdam: John Benjamins Publishing Co.

Hill, Christopher (1991) Sensations. Cambridge: Cambridge University Press.

Ittelson, W. and Kilpatrick, F.(1951) “Experiments in Perception.” Scientific American 185: 50-55. 

Jackson, Frank (1982) “Epiphenomenal qualia.” American Philosophical Quarterly, 127-36.

James, William (1897) The Will to Believe and Other Essays in Popular Philosophy. New York: Longmans, Green and Co.

Kant, Immanuel (1781/1965) Critique of Pure Reason. Translated by Norman Kemp Smith. New York: St. Martins Press.

Lycan, William (1987) Consciousness. Cambridge, MA: MIT Press.

Lycan, William (1990) “What is the subjectivity of the mental?” In J. Tomberlin (ed.) Philosophical Perspectives, 4: Action Theory and Philosophy of Mind. Atascadero, CA: Ridgeview Publishing, 109-30.

William Lycan (1995) “A limited defense of phenomenal information.” In T. Metzinger (ed.) Conscious Experience. Imprint Academic.

Lycan, William (1996) Conscious Experience.Cambridge, MA: MIT Press.

Marr, David (1982) Vision. New York: W.H. Freeman. 

Moore, George E. (1922) Philosophical Studies. New York: Harcourt Brace. 

Nagel, Thomas (1974) "What is it like to be a bat?” Philosophical Review 74: 339-56.

Rock, Irvin (1983) The Logic of Perception. Cambridge, MA: MIT Press.

Rosenthal, David (1986) “Two concepts of consciousness.” Philosophical Studies 97: 329-54.

Rosenthal, David (1991) “The independence of consciousness and sensory quality.” In E. Villanueva (ed.) Philosophical Issues: Consciousness. Atascadero, CA: Ridgeview Publishing.

Rosenthal, David (1992) “Thinking that one thinks.” In M. Davies and G. Humphreys (ed.) Consciousness. Oxford: Basil Blackwell.

Shoemaker, Sydney (1990) “Qualities and qualia: what’s in the head?” Philosophy and Phenomenological Research 50: 109-31.

Stich, Stephen (1978) “Beliefs and subdoxastic states.” Philosophy of Science, 45: 499-518.

Strawson, Galen (1994) Mental Reality. Cambridge, MA: MIT Press.

Tye, Michael (1995) Ten Problems of Consciousness. Cambridge, MA: MIT Press.

Van Gulick, Robert (1988) “A functionalist plea for self-consciousness.” Philosophical Review 97: 149-88.

Williams, Bernard (1973) “Deciding to believe.” In B. Williams Problems of the Self. Cambridge: Cambridge University Press 136-51.