A naturalistic answer to the grain problem may be hidden in assumption Al. In its first version, this problem consists in the fact that not only is a certain phenomenal property - namely the visual 'pink' quale - subjectively instantiated in the manifest experience of a pink ice-cube, but also that this property is homogeneously distributed over the object experienced. If the subjectively experienced object were to be identical with a neural activation pattern in a Leibnizian sense (i.e., if it shared all its non-intensional and non-modal properties with this pattern), then every region of this pattern of activity would also have to instantiate the property of 'pink', no matter how small. According to this analysis, 'homogeneity' would be a property of a functionally active representational whole: not a structural property, but a content properly.[28] We are now for the first time able to identify this property with a physical property, suggesting: The homogeneity of subjectively experienced qualities is the temporal homogeneity of the correlated states of the system. As a matter of fact, there is a complex physical property which can be found in all spatio-temporal regions in question, namely the synchronicity of neural activity. That is, qualia are not 'infinitely' homogeneous, homogeneity does not go all the way down, it rather is a representational phenomenon emerging at a certain level of complexity. Synchronicity can be experienced as homogeneity in exactly those cases in which an integration mechanism possesses a lower temporary resolution than the states of the system which it unifies. The system then represents temporal coherence as smoothness. On a higher level of representation, the synchronous activity of feature detectors of a certain type must therefore by nomological necessity appear as structureless and dense. A central aspect of the philosophical qualia problem consists in the fact that the qualitative character seems to render the sensory states in question impenetrable; it makes it impossible to recognize the causal role of these states. We are now in a position to say: The causal role is not realized through what we call 'qualitative character', but through the homogeneity of this qualitative character. In reality, I pick out redness introspectively in virtue of its homogeneity and precisely this is the realization of the causal role. On the level of simple phenomenal properties, homogeneity is the subjective correlate of the synchronicity of the activity of the feature detectors concerned and it is precisely through this synchronization that this activity can become functionally active and subjectively available to introspective experience. Thus, on the level of phenomenal content, the causal role is represented through its homogeneity and not through what we commonly call this content 'itself'. For this reason, I would like to suggest a 'double-aspect' - or 'two-component' - theory of qualia: Manifest qualities of conscious experience such as International Klein Blue possess an H-component and a Q-component. The H-component is the homogeneity of the state and the Q-component correlates with the phenomenal property which the state instantiates. What we pick out introspectively is not International Klein Blue, but always homogenous International Klein Blue. The interesting point is that, from an empirical point of view, it now becomes more and more plausible that it is actually the H-component which makes the Q-component causally active and at the same time available for experience. In virtue of the H-component, the respective state of the system can play a functional role and enter the space of conscious experience. The speculation S1 amounts to the claim that the H-component could be naturalized by means of the temporal coding hypothesis.

This would not solve the qualia-problem: We would still be unable to say what International Klein Blue actually is, because we still lack a convincing theory for the Q-component. But we would have an empirically plausible answer to the higher-order problem of the homogeneity of qualitative content: Elementary subjective qualities introspectively appear as atomic and irreducible because we are systems which bind the activity of their elementary feature detectors through temporal coding into a higher-order state of the system which possesses a holistic time-structure. Two important points must not be overlooked in this context: First, phenomenal properties never appear in an isolated fashion, but always as properties of higher-order wholes. A conscious pain is always localized within the body image, and even the coloured patches which we can sometimes see before falling asleep have a spatial extension, often even shape and a direction of movement. (Perhaps one could say: Colouredness is a dimension which we can structure through an internally generated context, but there are no isolated colours outside of this context. [29]) Secondly, isolated phenomenal individuals do not exist. In conscious experience pure individuals never appear, but only complexions of qualities, those which I earlier termed 'phenomenal Holon'. The elegance of the correlation theory lies in its potential to explain, in an ontologically parsimonious way, the transition from functional to phenomenal states. There are neither isolated phenomenal properties, nor is there a phenomenal individual to which these properties are 'attached': What exist in reality are sets of micro-functional events, which are bound to a coherent whole by synchronization. On the level of elementary properties, this coherence then shows itself introspectively as that which we philosophers like to call 'homogeneity'. The homogeneity of International Klein Blue could thus be the way in which the system represents the synchronicity of active feature detectors to itself - the way in which it experiences coherence. If these speculations point in the right direction, neuroscientific successor predicates will become available which are able to semantically incorporate the ultimate homogeneity in Sellars' sense.[30]

In order also to reduce qualitative content itself we would additionally have to be able to tell what it is that is bound into a higher-order state of the system by synchronization. But once what I have called the 'intrinsicality intuition' has been dissolved, a sophisticated relational analysis will be successful. We would need abstract descriptions of the corresponding new properties (the Q-component), which answer two questions: Why do elementary properties (for instance colour qualia) form phenomenal families? Why does subjective experience itself fail to tell us what this family resemblance, so concretely experienced by us, consists in? A complete reduction would therefore require mathematical models of those activation patterns, which are homogenized with regard to their temporal structure by temporal coding. These mathematical models would have to allow us to extract interesting objective similarity classes and project them on to the 'topology of subjective space'. If we one day succeed in connecting such formal descriptions of the respective concrete states of the system with the categorizations of folk-psychology, we will have moved another step closer to what Nagel dubbed an 'objective phenomenology'.

If the epistemic goal of our enterprise consists in illustrative descriptions which bridge the gap between scientific theory and our folk psychological understanding, one could for example say at this point: International Klein Blue is not a quality at all, but the high-dimensional 'form' of a neural activity state in my visual cortex. If we regard it as an internal representation, the concrete neural activity state in my brain which leads to my concrete experience of International Klein Blue possesses various abstract properties. Perhaps one could call these properties of the internal representation its 'format'. [31] Regarded as an event within the system's visual information processing, it is an activity state in a local neural net, i.e. in a certain subsystem of the brain. This local state of the network can be described abstractly in different ways: In principle, there are many mathematical models for this very special state as well as for its temporal dynamic. One possibility of describing it would be as point in, or a region within, a high-dimensional vector space, since one can describe the activity state of a neural net by means of activation vectors. If we then introduce a geometric model for this abstract algebraic description, we get illustrative and intuitively appealing pictures for very abstract objects of thought - spaces, sections or surfaces. Using such a geometric description one could then say: Every individual neural state has a form, possibly a very highdimensional form. Could that be what I see when I see International Klein Blue? Does the extremely high-dimensional form of a bodily state, which I do not recognize as such, lie concealed behind Yves Klein's 'dimensionless depth' of the subjective sensory quality International Klein Blue?

Let us not be deceived. All of this would still not amount to a complete reductive solution of the qualia problem: We would still not know why qualia are conscious experiences and why they are tied to a subjective perspective. Unless one already implicitly assumes a functionalist theory of subjective experience, it has not yet been shown that representational holism and homogeneity have anything to do with experienced holism and homogeneity.

S2

The principle of object formation is also interesting from an epistemological point of view. If a good theory of object formation within a single sensory modality as well as across different sensory modalities and for more abstract types of content becomes available, we are no longer forced to follow the classical philosophical topoi of the intentionality relation in postulating an integration or feature binding from above. There is no need to assume that the cognizing subject directs itself towards the world in a mysterious Husserlian manner and thereby grasps certain intentional objects. The emergence of active representational wholes can now be thought of as the self-organization of cognitive structures from the bottom up. Since the formation of such wholes should also be possible independently of external input, i.e. in the context of mental simulations, there are many promising possibilities not only for naturalizing central concepts of classical theories of mind, as for instance the concepts of an 'intentional act' and an 'intentional object', but perhaps also of spelling them out in a more precise way that preserves the original insights. This may also apply to many concepts of the classical phenomenology of perception, such as James' 'sensible totals', von Ehrenfels' Gestalt qualities, Meinong's higher-order objects, Husserl's figurale Momente, etc. Another consequence of this model may be of great interest for current debates in the philosophy of mind: The conceptual difference between 'construction' and 'representation' is defused, while its important semantic elements are simultaneously preserved in a new and empirically plausible idea. We can now understand ourselves as systems which under the pressure of input construct in themselves coherent states by means of feature binding, which then in turn function as representations in virtue of their informational content. The principle of object formation may also help in understanding the transition from a stream of massively parallel microfunctional events to symbol-type, serial processing. And it illustrates the interesting fact that 'higher-order representation' may be more than just abstraction or further mapping, that it may be a constructive type of representation by subsymbolic integration. All this may lead to new ideas about how naturalistic conceptions of intentional content can be developed. For the future it is therefore important to spell these observations out in more detail.

However, here we are concerned with phenomenal content. The central question goes like this: What is the relationship between active representational wholes and phenomenal wholes, i.e. those subregions of subjective space which I have characterized as 'phenomenal Holons' in the beginning? Representational as well as phenomenal wholes possess a striking Gestalt quality, which gives them their holistic character. And both lack a part of their internal structure. The system always deletes a part of the inner history of the different properties bound into a representational object through the process of synchronization and thereby renders it 'unknowable' to itself. The temporal structure of a neural impulse carries information about it's causal history, and by smoothing out that structure this information gets partly deleted. This might provide one of the reasons for the fact that we are systems which experience the world in a naive realistic mode: We are no longer able to recognize the phenomenal models which emerge in ourselves as models. On a representational level of description one can, in principle, understand how a number of active wholes can be differentiated from each other by means of desynchronization and how they can coexist without giving rise to a superposition catastrophe. If attention works by the same type of mechanism, it it easy to understand how they could be episodically tied into unified higher-order structures. In virtue of these three characteristics - holism, self-referential opacity and the capacity for higher-order embedding of representational entities - the correlation theory becomes interesting for a general theory of phenomenal consciousness. Still, the relation we are looking for cannot be the relation of identity. This is shown by a simple fact. Many of the crucial empirical investigations about object-formation and scene-segmentation have been successfully carried out on anaesthetized cats. This means that representational wholes can be generated without consciousness and without the first-person perspective. The formation of a perceptive object in the visual cortex of an anaesthetized cat amounts to artificially generating an isolated whole. This whole, however, is not functionally active because it is not bound into a highest-order Holon [32] together with other states of the same kind. Therefore, it is also unable to assist the cat in regulating its interaction with its environment. A phenomenal Holon on the other hand, is characterized by at least two further higher-order properties: A phenomenal Holon is always conscious and it always appears in connection with a perspectivally organized experiential space. Without a general theory of this space and the phenomenal self which appears in it and structures it, all reductionistic efforts will be doomed to failure. Phenomenal wholes possess a richer relational structure than isolated and functionally inactive representational wholes.

If the third of the assumptions bundled under A2 should prove to be justified, then we could understand how bound objects can in turn be represented as possessing external temporal relations. After the system has deleted the physical time interval on a fundamental level of representational content it is then in turn able to define higher-order relationships of temporal succession. [33] For instance, we are able to experience how a pink ice cube moves through the air and drops into a glass. In this way we can understand how seriality emerges from parallelity via a special form of object formation: An internal time, which is formed out of single moments that are variable in their duration, but connected with each other in a non-fragmented way.

If it could be demonstrated that the sensory-motor integration of the information flow can be explained with the help of the correlation theory, [34] this would provide us with a naturalistic strategy for the solution of all those forms of the mind-body problem which conceptually describe the perception of the world as an upward process and acting within the world as a downward one. Active representational wholes are - in virtue of their temporal structure - discrete states of the system, which may in many cases form functional modules. For this reason, they could serve as missing links in the precise specification of the causal chains which run through the system. Perhaps these considerations would even allow us to develop a holistic concept of action with the help of this general strategy. Moreover, one could ask at which point of such an analysis would the personal level of description enter? However, as it is very likely that many of these processes can take place independently of phenomenal consciousness and an active subjective perspective, this possibility is less interesting in our context.

Generally speaking, the possibility provided by the correlation theory of a 'liquid' embedding of active representational wholes in higher-order structures while preserving the holistic overall character reflects the third central feature of a phenomenal Holon, namely the simple phenomenological fact that singular experiential wholes can be bound seamlessly into larger wholes of the same type. Of course this thought becomes especially interesting when combined with the idea of phenomenal presence in the context of a subjective 'window of presence'. Probably we can nevertheless always imagine that all this happens without any kind of perspectival consciousness.

S3

Possibly some important explanatory ingredients for a naturalistic dissolution of the concept known in analytical philosophy as 'first-person perspective' can be found in the assumption A3. The phenomenal self, in my opinion, is the most interesting phenomenal Holon of all - amongst other things because it endows our conscious space with a very interesting structural feature: Centredness and perspectivalness. I have elaborated this point at great length elsewhere [35] and will therefore restrict myself to some very short remarks here.

The model of the self differs from every other mental model in an essential point. It possesses a part which is exclusively based on internal input: the part of the body image activated by proprioceptive input. Recent research concerning the pain experienced in phantom limbs seems to point to the existence of a genetically determined neuromatrix whose activation patterns could be the basis of the body image and body-feeling. The part of this neural activation pattern which is independent of external input produces a continuous representational basis for the body model of the self [36] and in this way anchors it in the brain. In almost all cases where there is phenomenal consciousness at all, there also exists this unspecific, internal source of input. It is the most 'certain' and stable region within the model of the self. In this way our consciousness becomes a centred consciousness. This is what makes us experientially embodied beings.

However, in order for the functional/representational property of centred-ness to become the phenomenal property of perspectivalness, the model of the system must become a phenomenal self. The pivotal question is: How does that which we commonly call the phenomenal first-person perspective emerge in a centred representational space? A first-person perspective - I would suggest - emerges in exactly all of those cases in which the system no longer recognizes the model of the self which it itself activated as a model. If it did, representational and functional centredness would remain, but the global phenomenal property of perspectivalness would disappear. In short: the system would have a self-model, but no phenomenal self.

On the level of inner experience, the phenomenal self is the paradigmatic case of a holistic structure. Its activation is accompanied by the instantiation of an interesting higher-order phenomenal property. This property creates a pre-reflexive form of self-presence and self-givenness within every single psychological moment. Let us provisionally call this property 'selfhood' or 'mineness'. It bundles the content of the phenomenal self into a holistic structure which appears as indivisible. (I hope that it has become obvious in the meantime why this structure must necessarily appear as indivisible to the system.) Moreover, the phenomenal self is embedded in a higher-order structure in a non-fragmented way, namely in a concrete phenomenal reality: We are also subjectively situated beings. The representational correlate - the self-model - is a functional module, episodically activated by the system in order to regulate its interaction with the environment. As we know from cybernetics, every good regulator of a complex system automatically has to be a model of the system. [37] If one assumes a PDP-inspired teleofunctionalism, then this model of the system appears as a kind of organ which emerges through the binding of a certain set of micro-functional properties and enables the system to represent itself in its environment to itself. So the self-model is a transient computational module, possessing a long biological history: It is a weapon, which was developed in the course of a 'cognitive arms-race'. [38] A real phenomenal self however, only emerges if the system confuses itself with the internal model of itself which it has generated. [39] Since the 'processuality' of the objective process of self-modelling is not represented on the level of content (at least in wide areas of the respective regions of representational space), the representational model of the system also possesses the aspect of presence in every individual psychological moment and the typical form of holism which cannot be transcended on the level of experience. If one thus assumes that the integrational mechanisms of temporal coding, which I have sketched above, play an important role in the activation of a unified and complex self-representatum, then we can begin to understand what the respective phenomenal Holon in the shape of a self-referentially opaque self-model could look like: While activating a representational object, the system gets caught in a naive-realistic self-misunderstanding and in this way generates a phenomenal subject.

But even this subject is not a phenomenal Holon in the strict sense. This is so because, of course, we can once again mobilize the modal intuition, which Peter Bieri has so aptly described at the beginning of this volume, and which in Germany has become known as the 'Tibetan prayer wheel question': Many of us can still imagine that all this happens entirely without any form of consciousness. For this reason we must now finally turn to the most important point. Can the speculative principle of GPFRW be of any help in attempting to gain a new perspective on the problem of consciousness?

S4

The classical philosophical model for consciousness - conscientia as higher-order knowledge about a subset of ones own internal states - is quite old. Many philosophers have continued to develop this fundamental thought using folk-psychological concepts like 'thought' or 'perception' in the construction of theories of higher-order thoughts or inner forms of perception. In his contribution to this volume, Guven Guzeldere has supplied a very lucid analysis of the difficulties associated with some current strategies of this kind. In my opinion, the time has now come to transform the classical concept conscientia into a technical term. From now on I shall call this new and speculative concept Highest-Order Binding, in short: HOB. This new concept (just like the concept of a 'phenomenal Holon') will serve to delineate in a first and provisional manner an explanandum which will, of necessity, be part of any theory of phenomenal consciousness. If one looks back into the history of philosophy and investigates the theoretical precursors of the modern concept of consciousness, one finds, besides accompanying higher-order knowledge, a second major semantical element in many of these early concepts (e.g. syneidesis or synesthesia): Consciousness is something that unifies or synthesizes experience. Interestingly, HOB has the advantage of preserving these two most important semantic elements, which kept on reappearing throughout many centuries of philosophical thinking about consciousness, by uniting them in an empirically plausible, naturalistic conception. However, it is far from clear how this could be done in an analytically convincing manner. The most intricate question is: How can one 'implement' the hierarchical logical structure of the concept of a global meta-representational function in an ultra-complex information-processing system, operating with holistic formats of representation? So 'HOB' is not only a new and speculative concept, but also the sketch of a possible research programme.

The most interesting theoretical problem here is that of understanding how the hierarchical logical structure of the concept of 'metarepresentation' could be projected onto a neurobiologically realistic model of parallel distributed information processing. If we integrate the basic assumptions of connectionism into our considerations, then the question arises of how global meta-representational knowledge can be understood, first, as sub-symbolic knowledge and secondly, as distributed knowledge. First, we have to comprehend how those inner processes which turn a part of the other inner processes into conscious inner processes can rest on non-linguistic forms of representation without a rigid constituent structure. Secondly, we need to clarify how the highest-order phenomenal property of 'consciousness' can be analysed as supervenient on or even as identical with a distributed network property of the system in question. Thirdly, we must do justice to the dynamic and holistic character of our space of consciousness.

The crucial speculation here is that the principle of GPFRW can help us to take the step from a functionalistic Global Workspace [40] to a highest-order phenomenal Holon. According to the assumption, synchronicity is the 'glue' responsible for the holistic character in question. As I have already said, representational levels or functions should be described through discrete classes of neural algorithms: Therefore, a mathematical model for the holistic character of our phenomenal space must exist. Now one may speculate that it is precisely the time-constants of a global metarepresentational function, physically realized by the process of HOB, which form the conceptual essence of the phenomenon which we call our subjective experience of the wholeness of our reality. HOB would have to be a continuous and rather unspecific mechanism which binds active representational content on the highest level by producing a dynamic and coherent global state. Accordingly, HOB is neither a form of concept formation nor inner perception - rather it is a constructive form of global metarepresentation: the self-organization of a coherent global state.

So if one continues to hold on to the highly speculative assumption that mechanisms of temporal coding (the process which I have first metaphorically called the opening of time-windows and then conceptually analysed as the generalized principle of the formation of active representational wholes) are operating right up to the 'highest' level of mental content, then the generation of a conscious experiential space in the brain of a biological organism can again be understood as a generalized case of feature binding and sub-symbolic object formation. A holistic representational space can be achieved if and only if a part of the information which is active in the system can be temporally integrated into a singular macro-representatum. If, on the highest representational level, a most general form of Gestalt formation or obiect formation is achieved through feature-binding (HOB), then a global macro-representatum is generated in this way. This macro-representatum, the system's actual conscious reality model, would therefore be a currently active informational structure which possesses very specific abstract properties. However, such a highest-order representational object could not be transcended by the system itself, because there would not be any larger internal data-structure with richer informational content with which it could be compared. Secondly, it is represented in the mode of 'immediate given-ness' because the utilized data-structures are semantically transparent. Thirdly, since the content of this data-structure appears within a window of presence generated by the system, one can understand how an experienced reality arises for the system - the presence of a global whole. In this global whole there is the qualitative homogeneity of sensations such as International Klein Blue and an ego-illusion which cannot be transcended: a self-model that is not recognized as a model.

Let us return to our metaphor of the time-window for the last time. Realized time-windows are holistic patterns, episodically active informational structures, which have been bound by an integrational mechanism. Returning to a poetical way of speaking once more, they are like flowers in our mind. These flowers open themselves so fast that we are no longer able to realize that they grow in ourselves. There may be many structures of this type, which influence behaviour as transient functional modules, but which nevertheless are unconscious because they are not bound into the currently active phenomenal model of the world. However, all those representational wholes, which become real phenomenal Holons because they are bound into the global space centred by a self-model now function as individuals in the flexible phenomenal ontology of the brain. As long as they are realized, i.e. as long as the synchronization process in question takes place, they are as a matter of fact indivisible for the system itself. In this sense (only in this sense), the phenomenal model of the world is itself such an individual. If, again, the glue of feature binding and of higher-order integration should be 'synchronization', one can even say the following: since there really is a concrete object characterized by a specific temporal Gestalt of its own, i.e. a largest (possibly 'slowest') time-window and since the mechanisms by which it is opened onto a multitude of time-windows which are already superimposed, do not themselves become objects of inner representation, it must of necessity become untranscendable for the system which generates this object in itself. In other words: the naive realism characterizing our form of phenomenal consciousness is a necessary result of the functional architecture of our brains. And in this manner it also becomes apparent how a model can become a reality. The content of the highest-order representational Holon possesses the structural feature of perspectivalness, it is represented in the mode of direct givenness and furthermore it is presented within a 'window of presence' generated by the system itself. These characteristics make it a highest-order phenomenal Holon.

In this contribution, I have investigated two higher-order properties of our experiential space, its holistic character and the homogeneity of qualitative content instantiated in it. I have tried to approach these properties from the inward perspective, from the outward perspective and by conceptual speculation. The purpose of the speculation was to test the usefulness of a conceptual model with regard to a theory of higher-order phenomenal properties. The assumptions on which it is based are extremely strong and, of course, I was not interested in the speculation for its own sake. I have nevertheless pursued this line of thought because, in my opinion, speculations of this type are nowadays for the first time rational speculations and because I believe that - even if they should prove to be imprecise or even false in many details - they may possess great heuristic value.

If I am correct in this point, they can, however, only develop this heuristic potential if they are successively replaced with empirically informative theories. This is the first project for future research into the phenomenon of conscious experience: We need to know the neural and functional correlates of phenomenal holism on all levels of instantiation. On a conceptual plane however, - and this is part of the second, philosophical, project - one has to investigate the modality of the relation between such higher-order functional properties of the brain and conscious experience.[41] Can we really imagine a system which belongs to the class delineated by GPFRW and a maximum realization of the assumptions Al to A4, which nevertheless lacks phenomenal states? And if we can in fact imagine this: What exactly does this mean?

However, the aim of this paper was much more modest. My intention was in part to point out that excellent empirical research is already being done with regard to the problem of the integration of mental content. In the philosophical debate of the last two or three decades this issue may have been underestimated, partly because of it's focus on qualia. If I have succeeded in showing the relevance of a generalizable solution of the binding problem and the problem of superposition for a philosophical theory of phenomenal consciousness, then I have achieved my purpose.