Matthew Soteriou's The Mind's Construction (2013) is an ambitious, brilliant, and complex book, absolutely packed with argument and interest. His general strategy is to explore the connections between fundamental metaphysical issues concerning the ontological categories of various mental phenomena and a number of specific issues that arise in philosophy of mind but are normally discussed independently of such metaphysical concerns. He applies this strategy in great depth and detail to issues concerning both sensory consciousness and conscious cognition: the sensuous nature of perceptual experience; bodily sensation; representational versus relational accounts of perception; the experience of time and change; the characteristics of individual sensory modalities; the structure of spatial and temporal perception; perceptual imagination and recollection; introspection; mental action; conscious thinking and judgement; critical reflection and autonomy; practical self-knowledge; and belief.

The central idea is to provide an ontology that fits the phenomenology of consciousness and cognition in all of these areas in such a way that substantive progress is possible towards the solution of persistent philosophical problems concerning specific acts of consciousness and cognition. A unifying key to Soteriou's positive picture is the importance in all of these areas of the ontological category of ‘occurrent mental state’: a state of mind that is constitutively dependent upon and sustained by conscious mental occurrences (events and processes of various kinds). A state of the relevant kind cannot obtain without the occurrence of such conscious phenomenal events and processes. In many of the problem cases that he is concerned with, these sustaining occurrences are also in turn constitutively dependent upon the occurrent states in question. Events and processes of those kinds cannot occur without the obtaining of the relevant kinds of states.

Soteriou calls this the Interdependence Thesis: ‘The occurrence and state in question have an interdependent status, in so far as the nature of the occurrence is to be specified, at least in part, in terms of the kind of state that obtains when it occurs, and the nature of the state is to be specified, at least in part, in terms of the kind of occurrence that its obtaining is constitutively dependent on’ (249).

As well as providing an ontology that fits the phenomenology in the relevant area, a central part of the motivation and defence of Soteriou's picture concerns its capacity to make sense of competing and otherwise apparently irresolvable theoretical intuitions that hold in place the specific philosophical problems addressed.

The majority of my discussion to follow concerns the application of his central idea to conscious perception, and, in particular, to advancing a sophisticated and original argument for a relational view of perception in opposition to orthodox purely representationalist views. But in order to give some indication of the extraordinary richness and depth of the book as a whole, I shall briefly indicate some of the other applications of this central idea in the course of the book.

1. 1. Chapter 3 develops and defends an original account of bodily sensation according to which the experience of a located bodily sensation of pain, for example, is an experience that has as its object a mind-dependent bodily occurrence with phenomenal properties, in the sense that there is something it is like for the subject when it occurs. The nature of this bodily occurrence is to be specified in terms of the phenomenally conscious mental state of awareness that obtains when it occurs and that has that very occurrence as its object of awareness. This mind-dependent occurrence is at the same time bodily, in that it seems to constitute not only a modification of one's mental life, but also, and necessarily, a modification of a specific part of one's body. So, the account as a whole involves a dual application of the idea of an occurrent state (one mental and one bodily, as it were), both satisfying the Interdependence Thesis.

2. 2. Chapter 8 offers a compelling account of perceptual knowledge as an occurrent state of the subject whose obtaining over an interval of time depends constitutively upon the conscious perceptual experiences that s/he undergoes during that time. As Soteriou puts it, (187), ‘perceptual noticing marks the onset of this distinctive form of cognitive contact with a perceived item, and … when one looks at and watches a perceived item one engages in the activity of maintaining this distinctive form of cognitive contact with that perceived item’. Of course, not all knowledge gained through perception is of this form. Non-occurrent knowledge may be acquired on noticing a perceived item that outlasts the perceptual encounter in question. Furthermore, it is not clear to me whether the states of knowledge and their sustaining experiential occurrences in this case are supposed to meet the Interdependence Thesis. That is to say, could the experiential occurrences in question occur without the relevant states of knowledge obtaining? This seems to depend on what exactly the occurrences are: watching the object, or noticing its colour, for example. Plausibly the former could, but that latter could not, occur without the cognitive state of knowing that the object is red obtaining, for example if the lighting is poor or the subject's attention is taken up entirely with determining its shape as it moves past quite quickly. This, in turn, raises the general question of what the grounds are supposed to be of the Interdependence Thesis when it does obtain. On the assumption that experiential occurrences are on occasion particular events that may be picked out in many different ways, then one possibility is that any appearance that a given such occurrence necessarily sustains an occurrent cognitive state of knowing that p, say, is simply a product of the event in question being picked out in terms of that very state, effectively as what sustains it. But then that very event might have occurred without the cognitive state obtaining, in which case it would just not have been describable in terms of sustaining knowledge that p. Similar questions may also arise in connection with the converse constitutive dependence of occurrent states on phenomenal occurrences, although the picture of states as particulars with many and varied descriptions on which such dependence may be thought to turn is arguably less plausible. It is a delicate question whether any of this might be a threat to anything that really matters to Soteriou's overall project. But it would have been helpful to see more explicit discussion of the underlying metaphysics of constitutive dependence. I certainly had the impression that this is supposed to be something more than simply a consequence of choosing to describe various phenomena in terms of their contingent causes and effects.

3. 3. Chapter 10 presents an account of conscious judgement, which reconciles an important point that Soteriou derives from Geach with a realistic account of the mental life sustaining judgement. The Geach point is that the act of judging that p does not, as such, fill an interval of time that can be divided into subintervals, each containing a temporal part of that act corresponding to a part of the propositional content, p, that is judged. In this sense, ‘There is no succession within any one thought’ (30). On the other hand, the mental life of a subject consciously judging that p surely contains mental events sustaining that act that do have temporal extension, and hence temporal parts. The reconciliation is not strictly an instance of the Interdependence Thesis so far as I can see. For it does not appeal to occurrent mental states. But it is very similar in form and likewise ontological in nature. The idea is that such conscious mental events with temporal extension and parts constitute the vehicles of the conscious mental act of judging, which does not itself share their temporal profile. In particular, there is no need to accept that the temporal parts of such event vehicles in any way signify temporal parts of the content judged, or that the subject continued to judge that p for precisely the interval for which the vehicle events occur.

These examples are simply illustrative, and there is so much more in the book that I haven't even had chance to mention. Particular highlights for me that go well beyond the central idea of providing an ontology that fits the phenomenology of various problematic cases of consciousness and cognition include the following.

1. 1. Soteriou distinguishes a stronger from a weaker reading of the Moorean thesis of the diaphanousness of sensory experience. The stronger reading claims that introspection of one's perceptual experience reveals only the worldly phenomena that one is apparently aware of in having that experience. The weaker reading claims that one can only discern the conscious character of one's perceptual experience by attending to the worldly phenomena that one is apparently aware of, but leaves it open that one may thereby discern features of one's experience other than what it is experience of. Chapter 5 beautifully demonstrates the merits of the weaker reading by characterizing such structural features of experience in such a way as to illuminate, compare, and contrast the important ideas of the spatial and temporal sensory fields.

2. 2. Chapter 13 offers an ingenious and plausible account of our knowledge of what we are doing intentionally. Soteriou's key proposal is that an intention to ϕ induces a kind of self-binding that consists in accepting the proposition that one will ϕ as part of further practical deliberation. Then ‘When one's ϕ-ing manifests one's recognition of one's commitment to ϕ-ing, one accepts the proposition that one is ϕ-ing (or attempting to ϕ) and one's acceptance of this proposition can embody a form of practical knowledge of what one is doing’ (314).

3. 3. Chapter 14 considers the place of mental action in the metaphysics of mind in the light of its crucial role in the discussion throughout Part II of the nature of conscious thinking. The principal thesis is that an assumption of much current theorizing in the area is untenable. This is the assumption that ‘we can conceive of mental actions as reducible to mental processes (perhaps transitions between mental states) that enable and explain overt bodily action’ (333). Soteriou argues that this assumption has to be rejected in order to maintain the essential role of mental action in the epistemology of mind in relation to the conscious thinking of self-conscious subjects.

I now turn to a more detailed consideration of the fascinating extended discussion of the nature of conscious perception in Part I. A principal goal of this discussion is to establish a relational view of sensory experience, according to which ‘We need to appeal to the obtaining of a distinctive kind of psychological relation when characterizing the manifest image of conscious sensory experience … [where] the relevant psychological relation is one that subjects bear to material objects and their features, and not to the sorts of entities posited by sense-datum theorists’ (84–5). Clearly this view stands in opposition to the sense-datum theory of perception. But it is also opposed to the widely adopted representationalist approach, according to which a full account of the manifest image of conscious sensory experience may be given by appeal only to distinctive properties of mental states and events – in particular, their representational properties understood in terms of their intentional contents with veridicality conditions, along perhaps with additional non-representational properties/qualia of such mental states and events, too.

Soteriou's argument is extremely rich and quite complex. I focus in what follows upon what seem to me to be its two most important phases. The first concerns the processive nature of conscious sensory experience and introduces a role for occurrent mental states in our understanding of such experience. The second argues from our (apparent) perception of mind-independent events that are extended in time to the conclusion that the conscious mental occurrences upon which such states constitutively depend must involve psychological relations with those very worldly events, and therefore establishes the relational view, at least in connection with our perception of such events. I take these two phases of argument in turn.

My overall understanding of the first line of argument runs as follows.

1. (1) Conscious sensory experience is ‘stream-like’:= processive to the core in a way that cannot be analysed simply in terms of a sequence of mental states of the subject and the changes in those states.

2. (2) Pure representationalism, which characterizes conscious sensory experience solely in terms of representations individuated by their contents, entails that such experience is not stream-like.

3. (3) Impure representationalists may appeal to non-representational properties/qualia to capture the stream-likeness of experience. But they face difficult questions concerning the relations between the representational and non-representational features of perception.

4. (4) A promising alternative with significant diagnostic explanatory advantages characterizes conscious sensory experience in terms of occurrent mental states constitutively dependent upon conscious mental occurrences that jointly satisfy the Interdependence Thesis.

I begin with elaboration of (2) and (4) before turning to a critical consideration of (3) and (1).

Soteriou's motivation for (2) comes from Geach's discussion of the discontinuous character of thought in ‘What Do We Think With?’ (God and the Soul (London: Routledge, 1969)). The key claim is that acts of thinking that are individuated by their propositional contents cannot be broken down temporally into parts corresponding to the parts of the content in question: they are in this sense a ‘non-successive unity’. Similarly, according to Soteriou, experiential states individuated by their contents are likewise non-successive unities. Pure representationalists may appeal to mental occurrences, rather than states, to capture the more thoroughly processive nature of experience. But the postulation of a mental event, for example, bearing a representational content only makes a difference to what conscious experience is like for the subject if it sustains a mental state of the subject that has that content. Otherwise, it is merely subvenient subpersonal processing with no obvious personal level conscious significance. So, the pure representationalist is committed to an analysis of conscious sensory experience as a succession of non-successive mental states: such experience is not stream-like.

According to Soteriou's alternative proposal (4), it is right that, in order to understand what it is like for a subject of sensory experience we need to talk in terms of conscious mental states of the subject. But it is wrong to think that such states may be individuated simply in terms of their representational contents. For this implies that they may obtain in the absence of a stream of conscious sensory experience. Instead, they must be individuated by reference to the kinds of conscious mental events/processes in virtue of whose occurrence they obtain and upon which they are therefore constitutively dependent. Thus, the conscious events/processes cannot occur without the obtaining of the conscious states, and the conscious states cannot occur without the occurrence of the conscious events/processes. We have occurrent mental states constitutively dependent upon the conscious occurrences that jointly satisfy the Interdependence Thesis.

This proposal provides explanations of three intuitions that shape debates about the conscious character of perceptual experience. First, it explains the intuition that a purely representational account of perception misses a crucial ingredient. For states may plausibly share any of the representational properties of perceptual experience without themselves being phenomenally conscious. According to Soteriou's proposal, this is a reflection of the fact that typing perceptual states purely in terms of their representational contents fails to acknowledge their constitutive dependence on the conscious mental occurrences in virtue of which they obtain, and so fails to individuate phenomenally conscious states ‘as such’ (49). Second, it likewise explains the intuition that an adequate account of the phenomenal properties of perceptual experience must mention the directedness of perception on the world perceived. For, in order to specify the nature of the phenomenally conscious occurrences that sustain perceptual states, it is necessary to do so in terms of the kinds of perceptual states that obtain when they occur, and these states have intentional properties directed on the world. Third, since the relevant phenomenally conscious occurrences and states have to be specified in terms of one another, the proposal explains the intuition that knowing any purely physical conditions whose obtaining is sufficient for both the occurrence of a given event/process and the obtaining of its corresponding state is insufficient for knowing the phenomenal properties of the relevant perceptual experience. For such physical knowledge is insufficient for knowing the nature of the interdependent occurrence and state that supervenes on those physical conditions.

I have two comments about this overall line of argument.

First, the strength of the motivation that it provides for Soteriou's own occurrent state proposal depends on the force of the objections raised to the impure representationalist's alternative in (3). I am certainly not myself tempted by this approach. But it is fair to say that a number of influential representationalists are persuaded on independent grounds of the need to introduce non-representational properties/qualia into their account of the nature of perceptual experience, and there is a lively debate amongst them concerning the relations between such qualia and the representational features of perception. So there is certainly unfinished business here.

Second, a great deal hangs on the grounds given for the stream-likeness of conscious sensory experience understood as in (1). I suspect that this will be the point of attack of many opponents and it also raises larger issues about Soteriou's general strategy. Suppose that one grants the generalized Geach point that states individuate by representational contents obtain over intervals of time that cannot be broken down into subintervals corresponding to the parts of those contents. Suppose that one also accepts, therefore, that pure representationalists are committed to the idea that the course of our conscious sensory experience is to be analysed as a sequence of discrete such states and the transitions between them. Why must one accept that our conscious sensory experience is stream-like in a way that is inconsistent with this analysis? Why not insist instead that such states may on occasion be sufficiently short-lived and the successive transitions between them sufficiently frequent that we arrive at the illusory impression of a genuinely continuous flux? At this point in his discussion, Soteriou does not offer any arguments against this possibility, but rather endorses O'Shaughnessy's (Consciousness and the World (Oxford: Oxford University Press, 2000) eloquent insistence that it is incapable of capturing the genuine stream-likeness of experience. So there is more unfinished business here.

This line of objection raises larger issues about Soteriou's overall strategy of providing an ontology that fits the phenomenology of conscious sensory experience. We may grant that our impression of the phenomenology of experience is that it is, as O'Shaughnessy puts it, ‘occurrent to the core’ (2000, 49); and we are currently comparing two views of its ontology: Soteriou's on which this impression is veridical, and the representationalist's succession of states on which it is not, although the illusion is entirely explicable given the rapidity of the transitions between the mental states involved. The larger issue is whether we have on this basis alone any reason to prefer one ontology to the other.

I now turn to the argument of Chapter 4 for the relational view. The goal here is to establish the need for some appeal to the obtaining of a non-representational psychological relation between the subject and elements of the perceived world in an account of the conscious character of successful perception.

The argument is driven by two phenomenological claims: first, that introspection of experience reveals worldly events, unfolding over time, as well as objects and their properties; second, that in perceiving such things, one's experience seems to share their temporal location and duration, and indeed that this temporal transparency apparently applies equally to the temporal parts of one's experience and its unfolding event objects. I accept these data. Soteriou's argument proceeds from here by reductio.

Suppose that the relational view is false. In that case, the account of what it is like for a subject, S, of an experience, E, of a worldly occurrence, O, that unfolds over a time interval, t₁-t_n, may appeal only to the properties of E and the psychological properties that S has in virtue of being its subject. But, as Soteriou puts it, ‘Such properties, of E and S, do not themselves unfold over time. … They obtain over intervals of time … [and] continue to obtain throughout the interval of time over which they obtain’ (94). So, the theory faces a dilemma in attempting to accommodate the two phenomenological data: S's apparent awareness of an occurrence, O, with temporal extension, t₁-t_n, and the apparent concurrence of each temporal part of O with her awareness of it. On the one hand, if we suppose that S is in the state of apparent awareness of O throughout t₁-t_n, then we fail to capture the fact that what seems to her to be the case during distinct subintervals of t₁-t_n is different. On the other hand, if we suppose that S is in a succession of distinct states of apparent awareness of successive temporal parts of O during each of t₁-t₂, t₂-t₃, …, t_n-1-t_n, then we lose the fact that she is apparently aware of the whole occurrence, O, with temporal extension t₁-t_n. Hence, the relational view is correct in claiming that we must appeal in addition to the obtaining of a non-representational psychological relation between S and O in an adequate account of the conscious character of her perception.

How exactly does the relational view help? It proposes that S's perceptual state of awareness of O during t₁-t_n evidently depends upon the obtaining throughout that interval of a relation of acquaintance with O as it unfolds over that interval. So, as Soteriou puts it, ‘The phenomenology is therefore that of being aware of something that fills an interval of time over that interval of time; and being aware of something that fills a sub-interval of that interval of time over that sub-interval of time’ (101).

The crux of this solution is the rejection of what Soteriou calls the homogeneity of perceptual states down to instants: the idea that if a person is in a perceptual state over an interval t₁-t_n, then this is because she is in that very state at every instant in that period. Now, I myself endorse the relational view, although on quite different grounds (Brewer, B. Perception and Its Objects (Oxford: Oxford University Press, 2011)); but I can envisage its opponents questioning Soteriou's argument as follows. Is it not possible to reject homogeneity without appeal to any non-representational psychological relation with worldly occurrences? An alternative solution along these lines might proceed as follows.

Throughout t₁-t_n, S's perceptual state supervenes upon the relevant neural events and processes that concurrently register O's unfolding over that interval. The idea of her conscious experiential state at an instant in that interval does not make sense. We can, though, assign a perceptual state to a complete interval of time. This may be extremely small and may also expand beyond t₁-t_n to an empirically determined maximum ‘specious present’. Thus, for the interval t₁-t_n, S is apparently aware of the whole of O unfolding, for t₁-t_n/2, of the first half of O unfolding, for t_2-ε-t_2+ε, of O suddenly accelerating, and so on. There are many reasons why we might be interested in her condition for different such intervals; and we can only ever get a determinate answer concerning her condition for some such interval. So, we seem to be able to accommodate both Soteriou's data. It is true to say that, for t₁-t_n, S is apparently aware of the unfolding extended occurrence O. It is also true to say that, for any subinterval of t₁-t_n, she is apparently aware only of the temporal part of O corresponding to that subinterval. Although there would obviously need to be a great deal more discussion of the proposal, I can certainly envisage theorists opposed to the relational view welcoming the idea that the neural supervenience base of any conscious sensory experiential state is necessarily temporally extended, as a realization, perhaps, of its ‘occurrence to the core’, and also welcoming the richness and variety of experiential characterization that is introduced by the idea that determinate perceptual states are necessarily relative to (possibly nested and/or overlapping) intervals of time. It would be helpful to hear more from Soteriou about whether he thinks that anything along these lines might be made to work.

In any case, I hope that I have managed to convey what a wonderful book this is. I only wish that I had twice the space, or more, to engage in more detail with its powerful and original arguments and views.