Traditionally, we find a distinction in the literature between two categories: on the one hand conscious experiences, states, and processes to which subjects have access, and on the other hand unconscious processes to which they do not have access (Cohen & Dennett 2011). According to this general picture, access to these states and processes then includes in many cases accurate reportability, which is the reason why reportability, or accessibility to introspection, is central to any judgment about conscious states. But access can also be understood more broadly: not all access is conscious itself, and not all access results in behavioral or verbal reportability.
In general, if we have a conscious state and a corresponding unconscious state, there are two possibilities for how the two can differ.[5] The first option is that the representational content of a state determines the experience, at least in part, so that both states differ in content. My conscious belief that my partner is cheating on me has a different representational content than the corresponding unconscious belief. These accounts are first-order accounts. The second option is that the states have identical representational content, but there is a difference in kind in the way in which they are embedded in the system—in philosophical jargon, the functional role that each state plays differs. According to this position, my conscious and unconscious suspicious beliefs that my partner is cheating on me are two states with the same content—expressed in the that-clause—but the conscious belief causes different internal states and different behavior to my unconscious belief. For example, in the conscious case, I will have the conscious thought that he is not treating me respectfully, and I might verbally confront him right away; in the second, unconscious case, neither of these activities will happen.
The first option is consistent with the standard view of what determines a difference in experience. However, it has a disadvantage: we cannot explain why the two states “correspond” unless there is some significant semantic overlap between them. The functional role view has the advantage that it can explain the similarity between the two states, but the disadvantage that we need an explanation of what exactly it is that makes a state conscious, and we have to show why this difference results in a difference in experience.
Schooler seems to opt for the content or representational view. Picking up Dennett’s idea[6] that people can be inaccurate about their own mental going-ons and internal states, Schooler concludes that, at least in some situations, external observers can have better insight into a subject’s experience than the subject themselves (p. 8). However, as we saw in the quotes above, Schooler seems to interpret the internal states in question as conscious internal states.
This is consistent with the idea that the access to internal states changes the content of the state, i.e., the content view: accessing a state changes the content of the state. Since the content determines the experience, the experience of a non-accessed and an accessed state differ. Understood this way, Schooler’s criteria give us opportunities to know better than the subject himself what he consciously experiences. Access and the reports of subjects about their experience, and the experience itself can come apart. If this is right, it would be unexpected and not what the commonsense understanding of conscious states predicts. As for the first aspect, Schooler believes that mind-wandering gives us an empirical case, where accessing (in the sense of attending to) a process or state changes that very state.
I will start with a discussion of the motivation for the distinction (see p. here), and some general problems we seem to invite if we accept this distinction. Schooler, and with him others, presuppose that conscious experience and accessibility can come apart; moreover, there is an experience before it is accessed. In other words, we postulate a third category, besides conscious and unconscious states: there are now “conscious but not accessed” states. These thoughts seem to be in line with other considerations in this debate, which propose a new category of phenomenal consciousness with no access (Block 2011; Lamme 2003).
Schooler distinguishes between simply “having experiences”, which he calls that experiential consciousness, and explicitly “taking stock” or re-representing this experience, which he calls meta-awareness or meta-consciousness (Schooler 2002, p. 339). Meta-consciousness then, is “defined as the intermittent explicit re-representation of the contents of consciousness” (2002, p. 339), while a later he says it is “knowing that one is having that experience” (2002, p. 339). So meta-awareness is about a certain kind of access.
Because we can clearly distinguish both, mind-wandering seems an excellent empirical candidate for the study of consciousness. At one point we notice our mind-wandering; but what we notice, the mind-wandering itself, occurs earlier. In the meta-aware case, we re-represent the former state; in order to do this, we access it by re-representing it, and we “take stock”. Then the subject becomes meta-aware of the state, and we know that we are in this state, but this very process changes the content. Our experience of mind-wandering is different once we become meta-aware that we are mind-wandering.
But this seems conceptually puzzling. Access and (verbal) reportability are clearly not the same, such that missing (verbal) reportability cannot not be equated with general lack of access, especially at the subpersonal level. With knowledge, reflection, re-representation and meta-awareness, as well as meta-consciousness, we get additional and differing concepts. First, often “knowledge” is used as something that is itself conscious. Is the idea that we are aware only of the mind-wandering, or also of our knowledge that we are mind-wandering? The author alternates between both phrases. But both claims differ. I can be aware of an experience without being aware of my knowledge that I have this experience. The latter includes a meta-level of a different kind. While the first contains a meta-process regarding the experience, the second is a meta-process referring to a propositional state, knowledge, of the experience. As a result I am aware of being in the state and not just of the experience. Moreover, reflection is a vague term. How exactly do we reflect on a state, process, or content of a state? What exactly does this entail? So the question is: what is meta-awareness and what distinguishes it from simple awareness? Finally, re-representation is mentioned, yet another concept used to characterize meta-awareness. Without further explanation, re-reflection seems a very broad and vague concept that would include all kinds of re-represented contents. Do most of these occur unconsciously, as certain kinds of functional accounts, higher-order accounts, predict (Jackendoff 1987; Rosenthal 2005)? How is something re-represented? How exactly do the represention and the re-representation relate to one another?
The question of which types of neural processes might be sufficient for awareness is highly controversial in current debate, as is whether there can be any awareness of a state without access (see the exchange between Fahrenfort & Lamme 2012 and Cohen & Dennett 2011, 2012). Relatedly, the status of local recurrences is debated. Block and Lamme argue that there are perceptual cases in which subjects do not attend to a stimulus (in change blindness, inattentional blindness, and attentional blink) and as a result are not able to report the presence of the stimulus. They might nonetheless be phenomenally conscious of the stimulus because it induces local recurrence in perceptual brain regions. As a result, a subject’s reports are not to be trusted in all cases: subjects could be conscious of stimuli even when they themselves deny it. This sounds very close in spirit to Schooler’s idea. However, Schooler doesn’t tell us how his account, and pure mind-wandering versus meta-awareness of mind-wandering, relates to this debate.
Despite these unclear aspects, the underlying intuitive idea is clear: Schooler wants to distinguish phenomenally-conscious experience from a meta-level of consciousness, in the literature also referred to as meta-awareness, and sometimes as reflective awareness, reflexivity, or reflexive consciousness. But what exactly characterizes this meta-level remains unclear. We are simply not told, the used concepts seem vague, and, without further explanation, underspecified. But, of course, this does not imply that the main idea is not helpful, or that it is not possible to specify them.
However, Schooler seems to sympathize with Cohen and Dennett, so I take it that he thinks (like them), that awareness differs from behavioral reportability. However, Cohen and Dennett explicitly state that they do not see many reasons to think such conscious information exists before it is accessed (Cohen & Dennett 2012, p. 140). So they reject the very option, the third category, that Schooler wants to postulate. There seems to be a sharp tension between Schooler’s distinction and his agreement with Cohen and Dennett’s general approach: Whereas Cohen and Dennett argue that theories postulating inaccessible conscious states are intrinsically off-limits to investigation, Schooler not only defends an account along those lines, but also argues that his account gives us a solution strategy to overcome the tension between the first- and third-person. Obviously, there is a need for conceptual clarification of this highly original idea.
However, I think we can learn a few interesting things from this. First, we can rule out a very general understanding of reflection or meta-cognitive processes. Most theorists agree that part of what it is to be in a conscious state is to have a unified perspective on the world. So the possibility of distinguishing between me and the world, or a self, or some kind of self-consciousness is required as an indispensable part of conscious experiences of many kinds. One way of describing this is to say that experience includes some kind of categorization. In other words, it is a kind of meta-cognition on this highest and most general level. At least, we as humans keep track of this interdependence of action and perception/experience at the personal level. To mention a classical example, it seems very hard to experience pain if one doesn’t classify something as painful, or without seeing it as painful for me. Indeed, some kind of evaluation, conscious or not, seems to be required for something to classify as pain; just as, in order to see something visually as a cow, we have to classify or categorize it as a cow (Dretske 1993).
At first glance, an account like Schooler’s cannot allow for this because the standard view requires meta-cognition for conscious experience. Experience is cognitively penetrable, such that knowledge about categories influences how we experience an object. In contrast, Schooler distinguishes both, and wants to allow for experience before (any?) meta-level involved. At least he talks sometimes as if meta-cognition in general is the issue when it comes to meta-awareness of mind-wandering. When he talks about theory of mind and the areas involved in meta-cognition (Schooler this collection, p. 17), he suspects that because certain meta-cognitive processes and mind-wandering occupying both engage the same systems, specially the dorsal ACC and the anterior PFC, this might explain why it is so hard to catch oneself mind-wandering, i.e., to gain meta-awareness of mind-wandering. However, he notices that identity of brain regions does not imply a causal relationship, and that further research is necessary.
However, it would be hasty to conclude that Schooler cannot concede that meta-cognition can be involved in experience on his account. Though he talks frequently as if the issue were meta-cogniton in general, he is not committed to excluding any kind of meta-cognitive process. But what is needed is a differentiation between different kinds of meta-reflection or re-representation. Schooler needs to address the question of whether we see the same kind of meta-cognitive processes in different kinds of experiences, and how exactly this changes the experience. Interpreted this way, only a certain kind of meta-reflection or meta-cognition might establish meta-awareness. As I will show, this move avoids a number of other problems.
We know that experience depends on background knowledge, and that our knowledge and our classification processes change our experience in many cases. This seems to be the case not just in mind-wandering, Schooler’s favorite example, but also in many other cases. What matters is not just how I classify a state or process; many other internal states and contextual factors influence experience. Let’s assume that I am a big fan of Baroque music, but cannot stand twelve-tone music. I happen to blunder into a concert with music by Penderecki, and of course do not like what I hear. Simply by gazing at the program and learning that I am listening to Penderecki’s Saint Luke Passion, which uses references to motives by Johann Sebastian Bach and is in a sense a homage to a well-known Bach piece, how I experience this piece of music might change. Chances are that I am still not able to hear the references to Bach and the coded references to passages in Lucas in the middle of all the dense tone clusters. But my belief that it is a homage to my beloved Bach will change my experience in general. Other states, beliefs, and emotions influence my auditory experience and make it, in this case, somehow more enjoyable.[7] It is also well known that crossmodal influences change experience: ones taste experience changes with conflicting visual experience. So a pure strawberry juice tastes less like strawberry to us if it is colored blue, even if the juice itself is not altered.[8] How we experience a certain wine depends on knowledge about price, how famous the winery is, and many situational aspects. In these examples, the real question seems to be how exactly our experience changes, and how do particular internal and external factors contribute to the change. What changes in how we re-represent, and how fundamental is this change? And what is meant by these terms?[9]
So Schooler’s meta-awareness can come in many forms. “Meta-cognition” includes a broad range of phenomena. What they have in common is that subjects have some insight into their own cognitive functioning. It is not clear to me that it is an all-or-nothing affair between pure experience and meta-awareness or reflection. So a specification of what exactly is meant by meta-awareness, re-reflection, and access seems necessary. We also need to answer the question of how the two categorically differing states differ in content, and which exact kinds of meta-processes are relevant. “Reflection” and “re-representation” are notoriously vague terms. Some kind of reflection at least seem indispensable for a state to be conscious. But that doesn’t mean the distinction above is not justifiable. We just need to determine and specify the kind of reflection and/or re-representation. I will make some suggestions later in this paper (see p. here).
To be fair, while Schooler does not distinguish between different kinds of reflections, he indirectly assumes that there are differences. But in his view the phenomenon dictates what the criterion for introspective awareness is. He distinguishes classification under the concept of “taking stock”: “there are some mental states (e.g., mind-wandering) for which the crucial bottleneck in people’s introspective awareness stems not from their capacity to classify the experience, but rather from the fact that people only intermittently take stock of what is going in in their own minds” (Schooler this collection, p. 8).
This obviously implies that for other phenomena the crucial difference does stem from their capacity to classify an experience. As a result, we in effect have different criteria for introspective awareness and for mind-wandering and visual perception. I believe a more promising route is to allow for dimensions of reflection and complexity of experience along multiple dimensions, but to try to find as uniform criteria as possible. The experience and phenomenology in cases of thought and sensory states (broadly construed) might be different.[10] But some properties or property clusters have to bind instances of introspective or meta-awareness together.[11] Otherwise, what would justify classifying them as the same, if both the phenomenon and the properties associated with the phenomenon differ? We would just be talking about different things. I have already ruled out two kind of meta-cognitive processes the author cannot use for a more detailed characterization of the difference between conscious states and meta-aware states: categorization under concepts is one kind of meta-cognitive reflection that itself is unconscious, but necessary for conscious experience. Distinguishing between self and world is another dimension of reflection, at the highest level, that seems necessary. Meta-cognition always requires representational use (of some kind), because within it we find monitoring of cognitive affordances. But there are several ways in which this monitoring can take place. As I argue below, meta-cognition, the ability to monitor and control one’s owns cognition, and the ability to attribute mental states to oneself and others can occur in different ways; and both the self-other distinction, and self-awareness can occur in a number of ways.
A core idea in the target paper is the claim that there is a difference between an experience and an experience one is aware of having. Both states are experienced, but the idea seems to be that reflection could potentially change an experience in a certain way, because it focuses on the content of the intentional formerly un-reflected state. Interpreted this way, Schooler seems to defend the content view, though I do not think he is committed to it. He doesn't explicitly subscribe to it, but it seems implicit in what he says when he talks about the content of states and frequently switches back and forth between content talk and talk of experience. He seems to think that these are related. And he doesn’t say much about the functional role that the states in question play in other states, or how cognitive processes use them—something one would expect if he held the functional view. So it is tempting to interpret him as having the view that content determines experience (Block 2005). For example, in writing that there are “some situations in which observers might have better knowledge about a person’s mental state than does the person in question” (Schooler (this collection, p. 8), what he must mean is that observers have better insight into the content of people’s states. A little later, he claims, regarding misrepresentations, “while in the process of re-representing, one omits, distorts or otherwise misrepresents one’s mental state to oneself and/or others” (Schooler this collection, p. 10). Again, what we misrepresent is obviously the content of the state.[12] If he has a content view, than his view is that (at least in some cases) I have an experience first, and then, when I reflect on it, that very process changes the content of the initial intentional state. That then is the reason why the experience differs between mind-wandering as “purely experienced”, and mind-wandering experienced with awareness. The phenomenon of mind-wandering indeed introspectively changes after we reflect upon it, and become aware that we are mind-wandering.
But I think there is a larger issue here. Interpreted this way, it is tempting to judge that accounts claiming that what makes a state a conscious state is its functional role are inconsistent with Schooler’s account. Again, I think this would be too hasty. Let me explain. Assuming a representational theory of phenomenal consciousness,[13] there are accounts that provided in purely first-order terms and accounts that implicate higher-order cognition of one sort or another (see below) with conscious experience. If we accept Schooler’s distinction, a state is conscious before we are aware of it, or know that we are in this state, and, when we become aware of it, this changes the state, or its content, to be more precise, as Schooler seems to suggest. Thus, Schooler seems to defend a first-order account, namely an account in which it is claimed that the consciousness of a state is partly (or entirely) determined by its representational content, or sometimes the format of its representational content, not primarily at first the function it plays (Byrne 2001; Dretske 1993; Kriegel 2009).
In the class of functional[14] accounts we find a great range of different accounts, including second-order accounts, accessibility accounts (Prinz 2012), and global workspace accounts (Baars 1988). Many of these are close in spirit to Dennett’s. Though they differ, they have one thing in common: it is a certain functional relationship the states in question have to other states or within the system, which makes these states conscious states.
Second-order accounts, for example, would claim that what makes a state a conscious state is that the state is (or is disposed to be, in some versions) the object of a higher-order representation of a certain sort. This state is a meta-level state, a mental state directed at another mental state. Higher-order accounts differ on how exactly this higher-order representation is characterized and what the exact relationship between both states is. In some versions the higher-order representation is a higher-order thought (Rosenthal 1986, 2005), in others a higher order-order perceptual or experiential state (Lycan 1996), yet other versions see the higher-order state as dispositional (Carruthers 2000). There are also differences concerning the question of whether the higher-order state should be understood as entirely distinct from its target state (Rosenthal), or whether the higher-order thought is better viewed as intrinsic to the target state, which would imply that we have a complex conscious state with parts. There exist different versions of the intrinsic view, which all have in common the idea that instead of a separate higher-order state there is a global meta-representation within a complex brain state (Gennaro 1996; Van Gulick 2000; Metzinger 1995). For the purposes of this commentary, I will focus on Rosenthal’s higher-order thought theory, but my considerations generalize to many of the higher-order accounts. The existence of the higher-order state and the right connection between both (one is the object of the other) makes the lower level one a conscious state. The higher-level state, however, is itself unconscious, unless there exists a third-level state—the existence of which would result in awareness of being in a conscious state. In effect, the existence of a certain kind of meta-cognition is what makes the lower level state a conscious state, or even a state that we are aware of being in. In this framework, Schooler’s meta-awareness would require a third-order state.
Accessibility accounts, for example that of Jesse Prinz’ (2012), would claim that attention is both necessary and sufficient for states to be conscious. In global availability accounts[15] it is claimed that the functional role is the global availability, or the workspace. The idea is that there is competition among neural coalitions; the winning coalitions are the conscious ones. There are a lot of similarities between higher-order theories and the neuronal global-workspace theory, but we should not see them as theories of the same type. According to the neuronal global-workspace theory, a state is conscious due to the global availability of its content, whereas higher-order theories see a state’s being conscious as “consisting of one’s being aware of oneself as being in that state” (Rosenthal 2012, p. 1433). If one interprets Rosenthal’s reference to “oneself” as Metzinger’s phenomenal self-model (2003), then a higher-order theory requires the integration of an individual state in a coherent representation or inner model of oneself, in contrast to a global-workspace theory, in which all that is required is availability of the content. Both aspects, the kind of meta-representation (the number of higher-order steps) and a certain identification of the original state as my state are dissociable, and they are examples of what I mean by different dimensions of reflection.
I think Schooler’s account stands in natural alliance with both kinds of accounts, in contrast to what one might initially think. It is the vagueness of the term “meta-awareness” that is causing this unjustified reluctance. For example, higher-order thought accounts seem a natural way to specify what Schooler might have in mind when he talks about meta-aware states. According to Rosenthal, there can be unconscious pain states, if these are accompanied by the thought that I am in pain, I am experiencing pain, but the thought itself is unconscious. Only if there is a third-order state, the thought that I have the thought of being in pain, am I aware that I think that I am in pain. To me, this sounds close to Schooler’s meta-awareness of taking “stock of our ongoing experience and re-represent[ing] it to ourselves” (this collection, p. 8). However, there is an important difference: for Rosenthal there are only conscious and unconscious states; the presence of the third-order state gives us what Schooler might call meta-awareness. However, Rosenthal denies the very possibility Schooler claims exists, that one can be in a conscious state but not aware of it. “No mental state is conscious if the individual that is in that state is in no way aware of it” (Rosenthal 2012, p. 1425). Due to the existence of a third-order state with the right content, we get introspective awareness of a conscious state: a third-order awareness that makes one aware of the second-order awareness. Rosenthal expects such cases, in which we “are aware of focusing attentively on that state” (2012, p. 1427), to be rare. It seems to me that there is a natural fit between Schooler’s meta-aware states, in which we know that we are having a certain experience and Rosenthal’s introspective awareness of a conscious state. In Rosenthal’s framework, meta-awareness necessarily requires a third-order representation.
In addition, Schooler’s suspicion that “meta-awareness appears to be associated with rhythms of attentional flux” (this collection, p. 17) relates nicely to accessibility accounts.[16] But as I will claim in the next section, global availability accounts stand in another obvious alliance with Schooler. Again, it seems that it all depends upon our understanding and further specification of “reflection” or the “meta” in Schooler’s meta-awareness. Is reflection itself necessarily a conscious process? Is it a thought, or just any kind of representation for the purposes of monitoring one’s owns cognition or an explicit higher-order classification? Unfortunately, Schooler does not describe his meta-awareness in more detail.
It seems to me that we should concede that some kind of “reflection” might be required for something to be an experience. This leaves still plenty of room to specify different kinds of reflections, some of which might constitute more than awareness, namely meta-awareness. This becomes the real question. Is this reflection itself unconscious or even necessarily conscious? Is it a re-representation of some kind? If that is the case, what kind of re-representation is required? Schooler’s meta-awareness might require a rather demanding kind of reflection, and the relationship Rosenthal describes seems a good candidate. But perhaps what we have instead of a simple dichotomy between pure experience and meta-awareness is a full spectrum of dimensions of meta-representation. Then the question is, what are the dimensions of reflection required for Schooler’s “pure experience” and those for meta-awareness, and which other reflections are there? This search for a proper taxonomy of “reflection” seems the most pressing need. It will hence be my main focus, and I will suggest some building blocks for such a taxonomy (p. here). Rosenthal’s introspective awareness of a conscious state as an possibility for characterizing Schooler’s meta-awareness will be one element of this.
This brings us to another and more problematic issue. I find the general line of thought behind a rigid distinction between pure experience and meta-awareness of this experience problematic. First, it presupposes that we accept the distinction between access-consciousness and phenomenal consciousness—a distinction not everybody (to say the least) is happy to accept.[17] Second, and more fundamentally, such a new category would have to be motivated. How do we distinguish “conscious processes, which are not accessed” from unconscious activity? Are they de facto not explicitly re-represented, or is it impossible to re-represent them? What does it then mean to say that something is “conscious”? On might suspect that this new concept of “conscious” is not compatible with our common-sense intuitive understanding of the term. Moreover, the stronger reading of Schooler’s position might invite further problems. If we claim that access to a state would necessarily change the status of its content (or the content itself), it would be impossible to address whether it was of a phenomenal or unconscious nature prior to this conscious access. If such an “observer-effect” exists, it could potentially render the whole issue completely immune to scientific investigation (Kouider et al. 2012).
Another open question is how Schooler’s account relates to others that seem close in spirit. Dehaene et al. (2006) have presented a more modern and updated version of Freud’s concept of preconscious activity. They introduce a proposal with a carefully defended taxonomy of three categories: subliminal, preconscious, and conscious activity. According to Dehaene and Changeux’s workspace model developed a little later, dominant neural coalitions involving the workspace are accessed. In contrast, existing other weaker activations in the workspace, such as a connection that could be activated, for example by a shift of attention, are only accessible. Processes that are potentially accessible, but are not accessed at the moment because of sufficient top-down attentional amplification, are “preconscious” phenomenal conscious processes in Dehaene et al.’s terminology (2006, pp. 206-207). I am not sure whether what Schooler is proposing is another version of Dehaene et al.’s “preconscious” phenomenal consciousness. This is consistent with what he writes. In debates on the third category of phenomenally conscious but not accessed states, their distinction between cognitive access and cognitive accessibility is often used to defend the possibility of the aforementioned third category (see for example Block 2011). My own suggestion is related, although I will suggest more closely specifying different kinds of access (see p. here) and multiple levels of representation, instead of just distinguishing between accessibility and access.
I think Schooler’s account would profit from directly relating his terminology to other concepts already in use in the debate. However, there are problems looming: Dehaene at al. defend a version of a functional account, which Schooler seems to explicitly reject when he seemingly advocates a first-order account. But if Dehaene’s taxonomy is not what the author has in mind, what is the difference between the Schooler’s phenomenally conscious but unaccessed activities and Dehaene’s preconscious activities?
Let us take stock. I have argued so far for three closely related points. The basic distinction between being experientially conscious of a state and being meta-conscious of being in a state needs further conceptual clarification. Moreover, the combination of a first-order account of consciousness (the content view) and this very distinction might not be the optimal strategy. In fact, a functional or hybrid account seems to provide a more natural strategic alliance for Schooler’s main project. Finally, it seems there is no strict dichotomy between experiential consciousness and meta-awareness; we rather face a difference in many dimensions. From my perspective, both higher-order accounts as well as global workspace accounts might be helpful regarding this issue. They connect nicely with Schooler’s main project, and would help to clarify his basic distinction. But we might very well end up with a more complex understanding of different meta-cognitive dimensions and differentiations instead of a simple conceptual dichotomy. This is what I will provide later in this paper. In order to do this we need to take a closer look at Schooler’s second step; his argument that his conceptual distinction is something we find in cognitive capacities.