Perhaps the most noticeable effect that OOP has had upon contemporary metaphysical debates has been to bring renewed attention to the status of relations between entities.1 According to Harman, the battle lines dividing up this conceptual terrain pit a contemporary relationist orthodoxy—which dissolves objects into their relations with one another—against his own substantialist heresy—which defends the independence of objects from all such relations. For Harman, correlationism is implicitly responsible for this orthodoxy to some extent by invariably making objects dependent upon some vertical relations to subjects. He also identifies more explicitly metaphysical champions of relationism—split roughly between those he draws influence from (Whitehead, Latour, and DeLanda) and those he merely opposes (Deleuze, Badiou, and Ladyman and Ross)—who make objects dependent upon some horizontal relations to one another. However, as we have already seen, the genesis of Harman’s substantialism occurs in his opposition to Heidegger, both as a reaction against Heidegger’s identification of substance (ousia) with presence (Anwesenheit), and as a corrective to the putatively relational ontology of readiness-to-hand (Zuhandenheit) that Harman finds in his work.2 Indeed, the two halves of Harman’s notion of withdrawn substance can both be traced back to his appropriation of Heidegger’s work on relations: the epistemic excess of objects over their presence is obtained by means of a general model drawn from the account of intentional relations; and this excess is converted into their causal independence from one another by means of specific cases (synchronic dependence and diachronic affection) drawn from the account of functional relations. Although we have already examined this genesis and its flaws in detail, further consideration will reveal deeper problems with the manner in which Harman frames the overarching question regarding the metaphysics of relations.
To begin with, it is important to understand that Heidegger does not present us with anything like a general theory of relations. He does provide an account of intentionality as a system of vertical relations between subjects (or Dasein) and objects, and an account of functionality as a system of horizontal relations between objects, locations, and functional roles; but neither of these is intended as an account of relations qua relations—they are species of relation, rather than the genus—nor is one meant to be analysed in terms of the other—functional relations are not a subspecies of intentional relation (or vice versa). Of course, Harman is aware of this, and explicitly acknowledges that Heidegger would not countenance his universalisation of intentionality. Nevertheless, this does raise the question of why one might think intentionality is an adequate model not just for functional relations, but for relations per se. Moving beyond Harman’s appropriation of Heidegger, the core motivation for this idea can be found in the way he positions himself in relation to Kant:
We might summarize the philosophical position of Kant by saying that he makes two basic claims: 1) Human knowledge is finite, since the things-in-themselves can be thought but never known; 2) The human-world relation (mediated by space, time, and the categories) is philosophically privileged over every other sort of relation; philosophy is primarily about human access to the world, or at least must take this access as its starting point.
Object-oriented philosophy agrees with the first Kantian point and disagrees with the second, while for speculative materialism it is precisely the reverse. For object-oriented philosophy, the things-in-themselves remain forever beyond our grasp, but not because of a specifically human failure to reach them. Instead, relations in general fail to gasp their relata, and in this sense the ghostly things-in-themselves haunt inanimate causal relations no less than the human-world relation, which no longer stands at the center of philosophy.3
What this shows is that Harman’s universalisation of intentionality is motivated by a radical form of ontological humility. He takes it that to grant any sort of special metaphysical status to the human would be hubristic, and that to restrict the provenance of intentional relations of knowing to humans is to do just this, regardless of whether they can be successful or not. It is a correspondingly radical form of epistemological humility that then denies the possibility of genuine knowledge of things, and thus, in collusion with its ontological counterpart, implies that all relations must be modelled as failed attempts to understand an object.
Now, we might wonder precisely why the restriction of knowledge relations constitutes an illegitimate metaphysical privilege when the restriction of other types of relation does not (e.g., not all objects are able to ‘consume’, ‘magnetize’, or even perhaps ‘marry’ something else). However, pursuing this question would mean probing the methodological relationship between epistemology and metaphysics in far greater depth than we have so far, and perhaps even asking why we should treat epistemic relations as having any metaphysical status at all. This is a bigger can of worms than I care to open here. Nevertheless, there is more to be said about why Harman thinks knowledge (or intentionality more generally) is capable of providing an adequate model for other relations.
According to modern logic, relations are a type of predicate. Predicates are loosely defined as mathematical functions that take objects as arguments (e.g., Fx) and return truth or falsity as values (e.g., Fa returns true iff a is F and false otherwise),4 or as open sentences that require the addition of singular terms or quantified noun phrases in order to form complete sentences (e.g., ‘…is red’, ‘…is out of tune’, ‘…knows a good place to park his car’, etc.). The examples just given are monadic predicates, which only have one argument place. These are usually taken to denote properties or qualities. However, predicates can easily have a different addicity, or number of argument places (e.g., Fxy, ‘…is larger than…’). It is these n-ary predicates that correspond to relations. This means we can have three-place relations such as ‘…is between … and…’, which is true of something if it is located between two other things, and even eleven-place relations such as ‘…compose a soccer team’, which is true of some group of eleven people if they are in fact members of the same soccer team. Moreover, two-place relations are usually characterised by their algebraic properties: as potentially reflexive (e.g., if x is near to y, then x is near to x), transitive (e.g., if x is larger than y, and y is larger than z, then x is larger than z), and symmetric (e.g., if x is married to y then y is married to x). Given that knowing is a two-place relation that is neither reflexive, transitive, nor symmetric, how is it supposed to provide a model for relations with different addicities or algebraic properties?
The answer to this question is that Harman never really intended knowledge (or intentionality) to provide a model for these cases, because he uses the term ‘relation’ in an idiomatic fashion that doesn’t include them. The scope of Harman’s usage is made clear in the following passage:
[E]lements are the basis of all relations, not just sentient ones. For not only is sentient perception object-oriented, bonded to fugitive objects in the night—but also interaction in general is saddled with this fate, and elements are the vehicle through which this destiny is enacted.5
The crucial point to take away from this passage is that Harman treats ‘all relations’ and ‘interaction in general’ as synonyms. The scope of this idiomatic usage becomes more explicit later in the same book:
We have said repeatedly that every relation immediately forms a new object, since every relation has a full inner life not exhaustible by any outer perception of it. But this might be cause for confusion, since the word “relation” is generally used to describe the relation between two things that do not fuse together into a new object. The numerous keys and toothpicks lying before me can obviously be said to relate to each other in a certain sense, but we have been employing the term “relation” for a closer kind of fusion between parts that give birth to a new thing […] Instead of saying that the various side-by-side elements of perception are related, we will say instead that they are contiguous or adjacent.6
That contiguity, which eventually becomes a full-blown category (SO–SO) in Harman’s ontography, is not a ‘relation’ in his sense of the word should make it clear that anything like locative relations (e.g., x is to the left of y, x is earlier than y, etc.) and perhaps even comparative relations (e.g., x is larger than y, x is less important than y, etc.) are not ‘relations’ either, because they do not describe interactions between objects that can themselves be viewed as further objects.7 As already noted, the only relations that Harman deals with (as ‘relations’) are the specific cases he develops out of Heidegger’s account of functionality: synchronic dependence (execution) and diachronic affection (causation), or the interactions through which entities persist and change, respectively. The reason these are at least prima facie susceptible to analysis in intentional terms is that they share (or can be made out to share) the asymmetrical two-place structure of intentional relations.8 One might be forgiven for thinking that this is nothing but an issue of terminology. However, not only does Harman’s peculiar sense of relation distort his presentation of the metaphysical debate between relationism and substantialism; it also generates blind spots in his thinking that warp his treatment of certain fundamental topics.
The distortions engendered by Harman’s terminology are often masked by his choice of interlocutors. Latour’s metaphysics of networks is principally concerned with precisely the sorts of synchronic dependence relations that Harman begins with, in virtue of its origin as an extension of his sociological methodology (Actor Network Theory [ANT]) for describing how certain relatively stable social systems function; but it does not distinguish between these and other types of relations (e.g., spatio-temporal configuration), precisely because Latour’s avowed relationalism converts them into ersatz dependence relations (e.g., by making a node’s identity dependent upon its relative location within the network). This enables Harman to castigate Latour for rendering impossible the diachronic causal interactions through which networks are rearranged, while simultaneously leveraging his model of ersatz dependence to portray all other relations as if they were synchronic causal dependencies, and thus amenable to analysis in intentional terms. By contrast, Whitehead’s metaphysics of prehension already combines causal interaction and perception in a manner resembling Harman’s intentional model, but it does not present this as an account of relationality per se.9 However, those who take up Whitehead’s banner when Harman assimilates his relationalism to Latour’s are not always so precise in making this distinction, and thereby acquiesce to his framing of the issue.10 Though Harman’s debate with Latour and Whitehead is not substantially affected by the conflation of interaction and relation that it tends to obfuscate, it nevertheless serves to distort his debate with Deleuze, DeLanda, and Ladyman and Ross by presenting them as more similar to Latour and Whitehead than they in fact are.
To demonstrate this point it is first necessary to rehearse the core complaint that Harman levels against relationism, namely, that it makes change impossible. As we hinted earlier, this complaint is prefigured in his engagement with Heidegger’s supposed relationism in Tool-Being, which has two basic components: the claim that defining objects in terms of their relations inevitably collapses into some form of holism, and the claim that this holism precludes change.11 Harman’s argument is essentially that the indefinite ramification of functional dependence relations implied by their referential structure inevitably leads to a sort of functional saturation (a ‘world-machine’) that leaves no room for functional disruption, and therefore reconfiguration or change. However, though the complaint against Latour and Whitehead is the same, the argument cannot be, because Latour’s relationism isn’t holist and Whitehead’s holism isn’t functional. Harman thus needs a more robust argument to establish the conclusion that relationism implies stasis more generally. The argument he settles on approaches the same point—the inseparability of an object’s identity from its effects upon other objects—from two different angles.12 On the one hand, he argues that this prevents objects from instigating change, because they cannot be exhaustively constituted by their current effects and yet capable of novel future effects. On the other, he argues that it prevents objects from undergoing change, because if they were to produce novel effects then they would no longer be the same objects. The problem with this argument is that the characterisation of relationism on which it rests does not apply either to Deleuze and DeLanda, or to Ladyman and Ross, albeit for different reasons.
In the case of Deleuze and DeLanda’s process metaphysics, the problem with Harman’s argument is that it characterises relationism as reducing individuals to their relations with one another. The reason this is problematic is that both thinkers distinguish between actual individuals and the virtual multiplicities that provide the conditions of their individuation, and it is the differential relations constitutive of the latter which they take to have something like an ontological priority over the individuals constitutive of the former.13 This is Deleuze and DeLanda’s way of articulating Gilbert Simondon’s idea that individuals must be understood in terms of the ‘pre-individual’ factors underlying their genesis. Harman does try to engage with these ideas to some extent:
These positions try to enjoy the best of both worlds, defining a unified realm beneath experience that is not completely unified. Instead of a total lump-world, it is one animated in advance by different ‘pre-individual’ zones that prevent the world from being purely homogeneous. This position has the following supposed benefits: it prevents things from being overdetermined by their current actuality (an admirable object-oriented gesture), while also slyly bridging the gap between things without doing the required work (a merely ‘radical’ move in the sense that must be rejected). For instance, DeLanda wishes to establish the possibility of a ‘continuous, yet heterogeneous space’. The same is true of Gilbert Simondon, that posthumous rising star. As Alberto Toscano describes Simondon’s position, ‘whilst [preindividual being] is yet to be individuated, [it] can already be regarded as affected by relationality. This preindividual relationality, which takes place between heterogeneous dimensions, forces or energetic tendencies, is nevertheless also a sort of non-relation […]. Being is thus said to be more-than-one to the extent that all of its potentials cannot be actualized at once’. Simondon like DeLanda wants the world to be both heterogeneous and not yet parcelled out into individuals. In this way, specific realities lead a sort of halfhearted existence somewhere between one and many.14
However, he does not actually consider the nature of ‘preindividual relationality’ and how the differential relations Deleuze and DeLanda take to be constitutive of the virtual might differ from the ‘relations’, or rather, causal interactions, that he deals with exclusively.
It is hard to do justice to this difference without providing a thorough survey of Deleuze’s metaphysics (or its DeLandian variant), but it suffices to point out that differential relations (e.g., dx/dy) do not hold between individuals, but variable quantities (i.e., x and y). These variables may correspond to quantities of individuals (e.g., populations of foxes and rabbits, as in DeLanda’s example of a predator-prey system),15 but they can equally correspond to velocities, spatial orientations, temperatures, and the whole range of degrees of freedom characteristic of dynamic systems. Moreover, even in cases where they do correspond to quantities of individuals, they do not correspond to specific individuals or groups thereof (i.e., the differential relation between number of foxes and number of rabbits captures a tendency governing the relative rates of change in population that applies across successive generations of foxes and rabbits, without care for which particular foxes and rabbits these are). There is a legitimate metaphysical debate to be had about whether it makes sense to give this sort of relation ontological priority over individuals, but it is crucial to understand that this is precisely not the debate that Harman engages in:
An obvious question to raise is why the relations between real attractors that build up a multiplicity are any less problematic than the relation between the real and the actual, or between two actual things. If no actual trajectory ever does justice to its underlying attractors, it should also be the case that no real multiplicity ever does justice to its own real components. In both cases it is a matter of relations, and relations are simply unable to exhaust their terms.16
This passage is little more than an attempt to sidestep the broader debate regarding the metaphysics of relations by converting it back into his preferred terms, or by portraying his own concern with ‘relations’ as more fundamental. It does this by treating the attractors (or singularities) that populate the vector field (or multiplicity) which is generated by the relevant differential relations as individuals that can be unproblematically separated from these relations, as if they were discrete components out of which the vector field was constructed, rather than topologically invariant features of a continuous curve.17 This doesn’t just presume what it means to establish—that description in terms of individuals is more fundamental than descriptions in terms of relations—it does so by completely misrepresenting the logic of the mathematical structures in question.
Although Harman successfully avoids engaging Deleuze and DeLanda on their own terrain by obfuscating the extent of the difference between their relationism and that of Latour and Whitehead, the argument he deployed against the latter still lacks purchase upon them, not least because their avowed endorsement of the idea that ‘relations are external to their terms’ clashes with the characterisation of relationism upon which his argument turns.18 He compensates for this by repurposing the strategy he deployed against Heidegger—he emphasizes the opposition between the continuum of virtual multiplicities (what Deleuze calls ‘the plane of immanence/consistency’) and the discreteness of actual individuals, and substitutes it for that between the real absorption of the world-machine and the apparent distinctness of its parts:
We find then that DeLanda’s actual world is made up of sterile nodules unable to affect one another or to relate in any way, while the non-actual zone of reality has no difficulty forming relations at all. There everything bleeds together in a continuum […] and the fact that it was woven together from initial heterogeneity does not prevent it from being a single continuum.19
It is possible to take issue with Harman’s characterisation of DeLanda’s account of individuality here. DeLanda does hold that there is a sort of quasi-causality operating at the virtual level that is responsible for the genesis of actual individuals and their behaviour, but he does not for all that deny that this behaviour is causal. It is perhaps better to look at this as a case of Harman putting the cart before the horse—his repurposed strategy is to show that the impossibility of change is an unintended consequence of Deleuze and DeLanda’s holism. The idea thus seems to be that, if causal interactions between actual individuals emerge from their mutual envelopment within a virtual continuum, then it is sufficient to attack this envelopment to undermine causality and thereby the possibility of change.
However, it is not so easy to reconstruct precisely how Harman’s attack on the virtual is supposed to work—although it is clearly directed at the claim that the virtual is both continuous and heterogeneous, it is less clear why this is a problem. The most in-depth presentation of this objection is to be found embedded in the narrative of Harman’s Circus Philosophicus, during a fictional exchange between Harman and his erstwhile Deleuzian paramour, in which he recounts a dream wherein various famous metaphysical holists are punished for their intellectual errors by being submerged in a lake of molten lead. At the end of this exchange, he engages his interlocutor’s Deleuzian holism explicitly:
[Y]ou do not claim that the world is simply a united whole, as the full-blown thinkers of apeiron do. Rather, you contend that the world is both one and many at the same time. Any given object is already interwoven with all others in a sort of continuum. Whatever happens in the world does not result from contact between one individual entity and another, but happens at the level of a united apeiron, though you hedge your bets by calling it both heterogeneous and continuous. Since I am not fully myself, and the shark and tree also not fully themselves, we are all laced with difference. No causal relation exists at the level of individual things; such individuals are not really cut off from each other in the first place. But the pernicious consequence is that the same thing will be simultaneously ‘a battleship, a wall, and a human being,’ in Aristotle’s memorable phrase […] [Y]ou always respond that the various individual things are not just ‘potentially’ distinct, but ‘virtually’ so. Yet here is the problem. Either the various beings dissolved in the lake of lead remain distinct, or they do not. If not then we have monism, and there is no reason that different entities would ever emerge from it. But if they do remain distinct, then there is the rather different problem of knowing why they are more than merely one. For how is the virtual shark different from the virtual tree? You want them to be a continuum, but this is a step that Aristotle knew could not be taken.20
The argument put forward in this monologue is essentially that individuals cannot be enveloped in a virtual continuum without ceasing to be distinct individuals, and that the specific relations of mutual envelopment (e.g., between populations of foxes and rabbits in a particular ecosystem) that are supposed to enable specific causal interactions between them (e.g., a particular fox eating a particular rabbit) are impossible without distinct relata.
The problem with this argument is perhaps best encapsulated by Harman’s complaint that the virtual ‘merely plays the double game of saying that true reality in the universe is both connected and separate, both continuous and heterogeneous’.21 What this shows is that Harman equates heterogeneity with separation (or discreteness) and thereby takes the very idea of ‘continuous heterogeneity’ to be internally inconsistent (i.e., as tantamount to ‘continuous discreteness’). This supposed inconsistency is presumably why the argument against the heterogeneity of the virtual is so rarely articulated. Nevertheless, this supposition plays a crucial role in the above argument, because it not only blurs the lines between discreteness, heterogeneity, and difference, but also confuses continuity, homogeneity, and identity. It is this latter confusion that enables Harman to argue that the mutual envelopment of distinct individuals within the same continuum inevitably results in them becoming identical. The irony looming behind this assumption is that Deleuze’s Difference and Repetition—the key text for his metaphysics and DeLanda’s appropriation of it—is best described as an attempt to articulate an account of difference that is not subordinated to the opposition between identity and distinctness, and thus to provide a consistent account of heterogeneous continuity. Moreover, the heart of this project is his interpretation of the significance of differential calculus, and the differential relations it describes. There is much to be said about this, but it is sufficient to observe that mathematical curves present the exemplary case of continuous heterogeneity, or difference that is not understood in terms of some prior discreteness. This is because they are not principally composed of distinct lines or infinitesimal points that differ from one another in their direction or position, but are first and foremost continuous lines that exhibit an equally continuous change in gradient.22 Although it can be differentiated to determine the gradient of a tangent to any arbitrary point along its length (x/y), this heterogeneous continuity can be exhaustively captured by a single differential relation (dx/dy). This is what it means to say that the difference (as captured by the differential relation) is internal to the curve, rather than a relation it bears to another line. It is important to emphasize that this does not necessarily establish the metaphysical applicability of the concept of continuous heterogeneity, let alone the consistency of Deleuze’s account of the virtual as a whole. It simply shows that the concept is prima facie consistent, and cannot be simply dismissed without engaging in precisely the sorts of metaphysical debates that Harman sidesteps.
Of course, this does not constitute a complete response to Harman’s worries about the relation between the virtual and the actual individuals that are enveloped in it. Such a response is not really possible without delving much deeper into Deleuze and DeLanda’s projects than would be appropriate in the current context. It is no easy matter to outline how every variable characteristic of every physical system in the universe could in principle be incorporated as dimensions of a single continuum which would thereby informationally encode the complete actual state of those systems along with their virtual tendencies, let alone how this continuum can still be divided into discrete chunks corresponding to individual systems and their specific tendencies.23 Nevertheless, it is worth making a final point about the relation between this form of holism and the possibility of change, given the general thrust of Harman’s attack on relationism. It is important to see that whether we consider individual physical systems, or the totality of all such systems, the differential relations which constitute their virtual multiplicities essentially encode information about the way they change over time. They cannot be considered in isolation from change, even if they can be considered in isolation from the causal interactions between specific individuals that are involved in these changes. Harman might respond that these systems are executing a pattern through which they persist as the systems that they are, but that this does not thereby account for the possibility that this pattern could become otherwise. However, although there is something to be said about complicating the typology of change, one can only keep separate the two types of change this objection is predicated upon by framing them in functional terms (i.e., as either acting towards or against the system’s end); the advantage of the dynamic systems approach that Deleuze and DeLanda adopt is precisely that it is capable of capturing complex behaviours without appealing to these terms (i.e., by means of the mathematics of attractors in phase spaces).24
In the case of Ladyman and Ross’s ontic structural realism, the problem with Harman’s argument is twofold. On the one hand, it characterises relationism as being concerned with specifically causal relations (albeit synchronic ones), so as to undermine the notion of causation more generally (by precluding diachronic ones); on the other, it characterises relationism as treating individuals as dependent upon all such relations. It is quite easy to see the error underlying the first problem: the sorts of mathematical structure that Ladyman and Ross take to be realised in the physical structure to which they give ontological priority involve the whole range of types of relations alluded to earlier. For instance, the phase space descriptions of physical systems at the heart of Deleuze and DeLanda’s metaphysics discussed in the last subsection are merely one example of such structures, and as we’ve seen, the differential relations between variables these involve simply aren’t causal relations in Harman’s sense. Furthermore, Ladyman and Ross examine the mathematical structures of systems described in the frameworks of both quantum mechanics and general relativity in some depth, in order to argue that they need not presuppose (and indeed, tell against) pre-individuated relata.25 We need not even assess the validity of these arguments in order to point out that the mathematical structures in question are composed of relations whose logical features are completely different from those of Harman’s ‘relations’. This is aptly demonstrated by the following passage, in which they distinguish relative and weak forms of discernibility from the absolute form demanded by thoroughgoing substantialism:
Two objects are ‘relatively discernible’ just in case there is a formula in two free variables which applies to them in one order only. Moments in time are relatively discernible since any two always satisfy the ‘earlier than’ relation in one order only. An example of mathematical objects which are not absolutely indiscernible but are relatively discernible include points of a one-dimensional space with an ordering relation, since, for any such pair of points x and y, if they are not the same point then either x>y or x<y but not both. Finally, two objects are ‘weakly discernible’ just in case there is two-place irreflexive relation that they satisfy […] Clearly, fermions in entangled states like the singlet state violate both absolute and relative discernibility, but they satisfy weak discernibility since there is an irreflexive two-place relation which applies to them, namely, the relation ‘is opposite spin to’.26
This should be enough to establish that any genuine debate with ontic structural realism over the ontological priority of relations must explicitly concern itself with the sorts of locative and comparative relations (e.g., ‘earlier than’ and ‘is opposite spin to’, respectively) that Harman’s framing of the issue ignores.
When it comes to the idea that relationism reduces an individual to the totality of its relations, the problem is somewhat more complex. On the one hand, it is easy to show that this is not Ladyman and Ross’s position: at worst, they reduce the individuals posited by the mathematical descriptions of physical structures to the specific relations that are involved in those descriptions—which is to say, to those that are considered scientifically relevant to capturing the phenomena in question. This means that any argument against relationism that proceeds by exploiting the apparent absurdity of an object’s identity depending upon seemingly irrelevant relations to other things (e.g., the gravitational interaction between Mars and my shoe) cannot get off the ground here. On the other hand, the manner in which Harman exploits this misunderstanding in his criticism of ontic structural realism exemplifies his persistent conflation of epistemological and metaphysical issues under the heading of ‘relations’. I do not aim to assess all of Harman’s criticisms of Ladyman and Ross’s position, some of which I am highly sympathetic to.27 I will instead restrict myself to commenting on the objection contained in the following passage:
A pattern, for these authors, is a bundle of relations no less than a bundle of qualities. The reasonable objection that there can be no relations without relata is quickly dismissed by the authors as an old-fashioned gimmick, in the eye-rolling spirit of ‘here we go again’. And yet they must tacitly concede that our knowledge of specific subject matter is never exhaustive at any given moment; science changes and advances. For this difference between representational and extrarepresentational real patterns is the key to their whole position, since it is all that allows them to maintain realism against an idealism that would hold that whatever science thinks at any given moment is always true. Our knowledge of the planet Neptune is surely incomplete, and hence our current mathematization of that planet is at best a translation of the real pattern Neptune itself, even if it were granted that certain mathematical aspects of our current translation will survive into any future understanding of it. In short, the real pattern Neptune is something more than our or anyone else’s relation to it. This means that they already accept a distinction between relation and relata at one level, at least. But as soon as representation is taken out of the picture and we move to the realm beyond representation, we supposedly find that Neptune belongs to a giant relational structure rather than being a discrete individual. In other words, although Neptune cannot be dissolved into observers’ current relations with it, Neptune itself is supposedly dissolved into the relational structure of the world, having no status as an individual except when viewed by an observer from a specific scale.28
In order to see what is going on in this paragraph, it is necessary to understand that although ontic structural realism is a metaphysical position, the motivation for it is epistemological. It is meant to solve the problem of how successive scientific theories can be understood as presenting incrementally better descriptions of the same real phenomena even while the individual entities they purport to refer to differ (e.g., the phlogiston and oxygen theories of combustion). It does this by holding that there are strictly no individual entities to be referred to, only physical structures to be mathematically modelled, and that successive theories can preserve mathematical structure that successfully represents this physical structure through isomorphism despite permutations of reference.29 What Harman does in the above passage is to posit a tension between Ladyman and Ross’s epistemological realism and the metaphysical relationism it motivates, by suggesting that whereas the former demands that the thing modelled is independent of the model qua mathematical structure, the latter makes the thing dependent upon the model by incorporating the representational relation within the thing qua physical structure.
It is important to understand why this objection doesn’t survive the restriction of physical structure to relevant relations. For instance, the example of quantum mechanical systems might incline us to think that our observational relations to systems are relevant to understanding their behaviour, and so must be incorporated in our models of them, and thus, presumably, in the physical structures modelled thereby. However, this inclination misunderstands these relations in at least two ways. Firstly, not only does it erroneously infer that the need for a general model of observational effects upon a system (e.g., the system involved in the double-slit experiment) implies that the system’s identity is dependent upon particular observation relations (i.e., that it is a different system when observed by me than when observed by you, or not observed at all); it illegitimately extends this dependence beyond those systems for which observational effects are relevant. Secondly, and perhaps more damagingly, it ignores the difference between observation and modelling: it treats the representational relation between mathematical model and physical structure as equivalent to the perceptual relation between an observer and what they observe. This is important, because even though quantum mechanics must account for the general effects of observing a system within the models it bases on these observations, there is no additional need for it to account for the effects of its models upon the systems they model.30
Ultimately, what this reveals is that the tension between epistemology and metaphysics that Harman locates in ontic structural realism is based upon the same logic underlying the conflation of intentional and causal relations we addressed as the argument from independence.31 However, we are now in a position to examine this logic in more depth, and to show how Harman’s way of discussing ‘relations’ disguises it. As we saw in the last chapter, the account of representation at the heart of Harman’s theory of intentional relations takes perception as its model, and this undermines its ability to account for the differences between observation and modelling in empirical investigation. Moreover, we saw that the central feature of this model is Harman’s refusal to properly distinguish between perception and inference, or between what is directly given in observation and what is indirectly inferred from it. We are now in a position to recognise that this refusal to distinguish between perception and inference is crucial to maintaining the conflation between causal and intentional relations, because perceiving an object involves being both causally and intentionally related to that object (i.e., sensing and representing it), whereas inferring something about it does not (i.e., representing it without sensing). This exclusion of inference facilitates the non sequitur underlying the argument from independence, by allowing Harman to elide the difference between representational and causal success, but it equally underwrites the above objection against Ladyman and Ross, by giving the impression that mathematical models must be somehow causally enmeshed with the physical structures that they model.
Once more, it is Harman’s choice of Latour as an interlocutor (and as an authority on the scientific enterprise) that works to disguise the assumptions about representation that motivate his arguments.32 His use of Latour’s notion of ‘translation’ in setting up the Neptune example is illustrative, because it enacts a sort of anthropological bracketing that dissolves the differences between epistemic activities such as perceptual observation and inferential modelling. Through this lens, the whole process of interacting with phenomena through observation and experiment, analysing the resultant data, constructing theoretical models, and elaborating their consequences, is collapsed into a single operation of translating those phenomena into theories.33 However, Harman is not just exploiting the epistemological homogeneity of translation here, but also Latour’s account of the ‘networks’ these translations compose. For example, Latour famously sees neutrons as essentially bound up in a network of relations that includes all the various elements of the historical process through which Joliot made them available as an object of technological manipulation and political action.34 This process of translating between neutrons and politics constitutes a particular relation between neutrons and politics in which Joliot is the mediator—that is to say, a relation in which ‘neutrons’ and ‘politics’ are viewed as individuals in the same sense as is Joliot. This is in distinct contrast to the general relations that physics concerns itself with modelling, such as the relation between neutrons and protons viewed as general kinds. Through this lens, the metaphysical difference between individuals and general kinds is collapsed and subsumed into the function of network-node—a node in a network of slapdash references deployed in anthropological descriptions of scientific practice, as opposed to those deployed in scientific explanation itself.35 This metaphysical homogeneity of networks combines with the epistemological homogeneity of translation to obscure the difference between the representational relations involved in observing individual systems and modelling their general behaviour. Once we see past these Latourian distortions, the supposed tension between the epistemological and metaphysical sides of Ladyman and Ross’s relationism disappears.
Now, we have already managed to locate a number of metaphysical blind spots in Harman’s theory by tracing his peculiar usage of the term ‘relation’ through a number of different debates with other thinkers, but there is one particular blind spot that deserves special attention, both because it is so extreme and because it proves so central to his system as a whole. I did not mince my words when I described Harman’s theory of qualities as a ‘conceptual disaster’, but I must be blunter still when it comes to his theory of space and time: it is in my opinion the most catastrophically inept aspect of his metaphysical system. This unapologetically harsh judgement is motivated by Harman’s persistent inability to engage with the topic of spatio-temporal relations, on the one hand, and his consequent failure to thematise the essentially temporal basis of his theory of objects on the other. The remainder of this chapter is dedicated to examining these problems and their consequences.
It is worth reminding ourselves of what Harman thinks about space and time before exploring why it is problematic. In The Quadruple Object, Harman introduces the topic by querying why space and time are simultaneously bound together and separated from other metaphysical structures in most philosophical speculation (e.g., Kant’s division between space and time as pure forms of intuition and the categories as pure concepts of the understanding).36 He goes on to argue that, in fact, space and time should be counted alongside essence and eidos as two of the four fundamental tensions implied by the fourfold split between the sensual, the real, objects, and qualities, and against the three radiations and three junctions that make up the remainder of the ten ontographic categories it demands.37 What distinguishes these four tensions is that they are object-quality relations, as opposed to the quality-quality relations of the radiations and the object-object relations of the junctions. What distinguishes space and time from essence and eidos is that they involve sensual qualities: space consists in the tension between real objects and their sensual qualities (RO–SQ), and time consists in the tension between sensual objects and their sensual qualities (SO–SQ). What this means is that they delimit the bounds of experience (or vicarious relation): time enables relations of confrontation (time-fission) between real objects and the sensual facades that populate their experience, and space enables relations of allure (space-fusion) between them and their real counterparts hiding behind those facades. By contrast, the other tensions individuate these objects and their facades (the vicarious relata): eidos (SO–RQ) constitutes the identity of a facade across variations within experience, and essence (RO–RQ) constitutes the identity of a hidden object between appearances in other objects’ experience; these enable the remaining intra-experiential relation of theory (eidos-fission) wherein the real object moves from confronting an individuated facade to tracing the eidetic anchor that ties its individuation to an already individuated hidden object, and the sole extra-experiential relation of causation (essence-fusion) wherein a hidden object moves from alluding to its own depths through its facade’s features to allowing its own features to modify the encountering object.
There are two points that we should take away from the above rehearsal of Harman’s categorial schema. Firstly, with an eye to Harman’s use of the term ‘relation’, it is now clear that although the tensions are the conditions under which certain complicated ranges of interaction occur (time: confrontation-theory and space: allure-causation), they are not themselves interactions, and thus not ‘relations’ in Harman’s sense of the word. This is hardly surprising given his insistence that contiguity (SO–SO) and withdrawal (RO–RO) are not ‘relations’ either, despite their categorial status being discussed in the same putatively relational terms. Secondly, and more importantly, it is now equally clear how important the difference is between the three intra-experiential relations (confrontation, theory, and allure) and the extra-experiential one (causation). Although space remains a condition of causation insofar as causation proceeds via allure, it is not clear that the same holds for time. Even if allure must itself proceed via confrontation, and thus causation via confrontation, this would still grant space a priority with regard to causation that should be cause for concern.
If we are to address the first of these points, it is important to explain what precisely it means to say that there is a ‘tension’ between an object and its sensual qualities. This is stated most concisely during the introduction of the concept of time in The Quadruple Object:
When we speak of time in the everyday sense, what we are referring to is a remarkable interplay of stability and change. In time, the objects of sense do not seem motionless and fixed, but are displayed as encrusted with shifting features. Nonetheless, experience does not decay in each instant into an untethered kaleidoscope of discontinuous sensations; instead, there seem to be sensual objects of greater or lesser durability. Time is the name for this tension between sensual objects and their sensual qualities.38
It is clear from this passage that the tension proper to time consists in the fact that an intentional relation to a sensual object persists across variations in its sensual qualities. Harman has addressed these sorts of phenomenal shifts in his discussion of Husserl’s notion of perceptual adumbration and the corresponding method of eidetic variation, wherein we encounter the same object from differing perspectives, be it really or imaginatively. The question is then whether the tension proper to space consists in a similar sort of variation. This is addressed further on in the same passage:
When we speak instead of space, everyone will recall the old quarrel between Leibniz and Clarke over whether space is an absolute container or simply a matter of relations between things. But in fact it is neither: for space is not just the site of relation, but rather of relation and non-relation. Sitting at the moment in Cairo, I am not entirely without relation to the Japanese city of Osaka, since in principle I could travel there on any given day. But this relation can never be total, since I do not currently touch the city, and even when I travel to stand in the exact center of Osaka I will not exhaust its reality […] This interplay of relation and non-relation is precisely what we mean when we speak of space […].39
This ‘interplay of relation and non-relation’ consists in the fact that a real object persists across variations in which other objects are intentionally related to it by way of its sensual facades. The role of sensual qualities consists in the fact that, although a sensual facade can persist across minor variation in its qualities, there are major variations that it cannot persist across. Given that the intentional relation between two real objects cannot persist without the sensual object that mediates between them, these major variations are capable of altering the overall distribution of intentional relations between objects. As such, for Harman, space and time are not ‘relations’ because they are the conditions under which the connections between objects are created, maintained, and dissolved by eddies in the sensual ether that supports them.
The problem with this account is that it says nothing about specifically spatio-temporal relations (e.g., x is to the left of y, x has the same orientation as y, x happened before y, x had the same duration as y, etc.), because these are precisely the sorts of relation that Harman’s idiom excludes. However, the situation is much worse than this. The above passage displays the point at which Harman’s idiom moves from simply excluding types of relations to actively conflating them. This is his invocation of the debate between Leibniz and Clarke, and subsequent framing of the spatio-temporal relations Leibniz is talking about as ‘relations’ or interactions. This mistake is repeated elsewhere with more explicitly absurd consequences:
It would be mistaken to follow Leibniz literally and say that space is simply generated by the relations between things. For it is just as true that space is the site of non-relation between things. If space were simply made up of relations, we would have a systematic gridwork with each object utterly defined by its relations with all the others, and the universe would become a single lump interrelated to the point of homogeneity. Such a lump provides no room for anything like space, which by definition would contain only one position: that of the lump as a whole.40
This caricature of Leibniz’s position is perhaps a result of confusing the idea that space and time are epiphenomenal—spatio-temporal relations can be derived from the non-spatio-temporal properties of monads—and the idea that each monad mirrors every other—the non-spatio-temporal properties of every monad can be derived from any monad. The result is the same crude accusation of radical holism Harman levels at Heidegger, Whitehead, and Deleuze, among others.
The core of the debate between Leibniz and Clarke is whether we need to metaphysically distinguish between locations in space (i.e., the regions and points of space viewed as a coordinate system) and their occupants (i.e., the events, processes, and seemingly persistent objects that can be located within this coordinate system). Following Newton, Clarke held that this distinction is necessary to make sense of the mechanics of motion, and that these independently subsisting locations are part of an absolute space. By contrast, Leibniz held that this distinction is unnecessary, because the relations between locations out of which fixed coordinate systems are composed (e.g., in space [x, y, z], point <1, 2, 3> is located at -3 along the x axis from point <4, 2, 3>) are abstracted from spatio-temporal relations between objects (e.g., my table is to the left of my chair, which is to the right of my bed, and so on) and thus needn’t presuppose an absolute space. The debate is very complex, and becomes even more so when it is extended to the advances in physics beyond Newtonian mechanics, and the contentious question of how to unite the differing conceptions of space presupposed by quantum mechanics and general relativity.41
However, we needn’t resolve this debate to see that Harman’s discussion of ‘space’ and ‘relations’ is at best entirely irrelevant to it, and at worst actively misrepresents it. That Harman is not in Osaka but can potentially travel to Osaka is not a ‘non-relation’ is Leibniz’s sense, but a spatio-temporal relation that could be described in a variety of different coordinate systems abstracted from their spatio-temporal relations to their peers (e.g., by means of a three-dimensional planetary map that can account for distance through the earth, by means of latitude and longitude over the earth’s surface, or by means of a network graph that restricts the available paths between the two nodes even further, etc.). Moreover, even if we decide that these various coordinate systems are pale reflections of an absolute coordinate system within which everything can be located, we still need not concede that the ability to specify unique spatio-temporal relations between everything thereby makes them indiscernible, because it does precisely the opposite—it makes them absolutely discernible.
One might nevertheless want to champion Harman’s account of space against the idea of absolute space on the basis that, as the condition under which relations of interaction can be formed and broken, it gives us a way of thinking about how these various coordinate systems are expressions of ‘space’ without reducing them to ‘a’ space. Although the general shape of this idea is unmistakably attractive, the problem with it is that there is no real account of why systems of spatial relations are expressions of ‘space’, or how it enables us to unify them. What needs to be thought through is the relationship between spatial relations and causal interactions: how proximity relations function as a condition of causal interaction (e.g., how ‘touching’ Osaka is dependent upon ‘being in’ Osaka), on the one hand, and how possible interactions function as a condition of coordinate systems (e.g., how different ranges of movement motivate different choices of map), on the other. Harman’s account of space provides us with no resources to formulate, let alone address these questions, and so shouldn’t be taken seriously.
If this is bad, the situation is much worse when it comes to Harman’s account of ‘time’. However, to see this, it’s worth examining some of Harman’s comments about the supposed difference between space and time, which requires quoting the relevant sections of Guerrilla Metaphysics extensively. To begin with, he makes a seemingly bold break with traditional thinking about space and time:
Everyone puzzles over “time’s arrow” and why and whether it only flows in one direction, but no one has ever asked about “space’s arrow,” since reversibility seems to belong to the very essence of spatial movement. But we can now see a way in which the opposite is true. Namely, time is always reversible, because on the interior of a relation it makes no difference when the Taj Mahal cycles from pink to blue to yellow to orange to black and finally to pink again. The sequence can go in any order and reverse itself any number of times without shifting the regime of objects. But this is not so with space. When I move from Chicago to Davenport and back to Chicago, it is space that has changed, since objects are to some extent no longer what they were: houses have been torn down or rehabilitated, brain cells have developed or died, friendships have formed or decayed, old wounds have healed slightly […] Time can always be reversed on the interior of an object, because the shifting gales of black noise within have no direct consequence for the regime of objects. But space can never be reversed, and we can never return to the same airport twice—the regime of objects will have shifted.42
When reading the above passage, one might, and indeed should, get the sense that something has gone awry in his discussion of space and time, such that what he calls ‘space’ and ‘time’ have become completely unmoored from their everyday reference. It is important to show that this is not some brilliant metaphysical rebellion against our common ways of thinking about space and time, but amounts, at best, to radically changing the subject of debate without changing its terms. The best way to make this clear is to quote Harman’s elaboration of this passage, and to highlight the offending phrases:
If someone now rephrases the traditional question and asks whether “space’s arrow” can flow in both directions, the answer must be in the negative. The reason for this is interesting. When objects enter into a relation, the relation cannot necessarily decompose again into the same objects: two chemicals might necessarily mix to form a third, but this does not imply that the new fluid is able to break down into its original parts. There is an asymmetry of cause and effect, and this is why space is irreversible. There are lasting consequences to space, but none to time, that transient fulguration along the surface of things—or rather, in the molten cores of things. Time itself creates nothing, while spatial changes create lasting monuments.43
It should now be easier to see precisely what is wrong: Harman distinguishes space from time by means of temporal terms. Even if one disputes the idea that talk of cause and effect is inherently temporal, one cannot deny that the distinction between ‘lasting’ and ‘transient’ consequences is a temporal one; but it is the phrase ‘spatial changes’ that is most indicative, by condensing the conceptual confusion into a bitesize oxymoron—thinking about change presupposes time, which means one cannot oppose spatial changes to temporal ones. Moreover, this cannot be viewed as an isolated logical slippage—consider the reformulation of the question ‘Is time finite or infinite?’ in the conclusion to his essay ‘Space, Time, and Essence: An Object-Oriented Approach’:
“Is time finite or infinite?” Under the object-oriented model time unfolds only on the interior of an object. As long as objects exist, time must exist. The question can thus be rephrased as follows: “must objects always exist?”44
This subordination of time’s extent to objects’ persistence simply presupposes a deeper time in which they might or might not persist—a time whose logic is indistinguishable from the everyday time out of which this metaphysical question arises. Harman’s ‘time’ is thus not only not what we usually mean by the word, but is actually parasitic upon it—‘time’ makes sense only insofar as we implicitly understand time but refrain from explicitly thematising it.
Once one sees the deep time lurking behind the surface ‘time’ in Harman’s metaphysics, it cannot be unseen. Worse, one cannot help but see it everywhere. For instance, the contrast between types of variation through which we articulated the tensions constitutive of ‘space’ and ‘time’ is essentially about change, and thus presupposes deep time. However, if we pull on this conceptual thread it unravels Harman’s system all the way back to the beginning. The contrasts between execution and causation, persistence and change, and synchronic dependence and diachronic affection, through which we have managed to make sense of Harman’s system, are all fundamentally temporal. It is not simply that one term in each distinction is temporal (causation, change, diachronic affection), but that both terms denote types of occurrence (executing, persisting, depending, etc.). The twisted temporal logic underlying these distinctions was evident in the temporal tension in the argument from execution, but it is most prominent in Harman’s repeated appeals to the reality of change in his arguments against relationism and holism in Heidegger, Whitehead, and Latour.45 The whole trajectory of Tool-Being out of which the account of vicarious causation emerges turns around the recognition that things must change within the subterranean realm of withdrawn objects, even if we only experience the ripples these changes produce in the glimmering surface world. These real changes and the time they presuppose are explicitly not accounted for by what passes for ‘time’ at the level of sensual change, and are treated as mysterious enough to demand a whole new theory of causation. In essence, the rift between changes that belong to the seamless functioning of an existing apparatus (surface ‘time’) and the changes that belong to the erupting malfunctions that disrupt and reconfigure these regimes (deep time) is not a theoretical consequence of Harman’s metaphysical system, but the pre-theoretic foundation upon which it is built.
Having examined the problems with Harman’s theory of space and time in detail, we are now left to extrapolate their consequences. This brings us at long last to the central claim of my interpretation of Harman’s philosophy: namely, that it is not a critique, but rather a consolidation of correlationism. Harman’s theory of space and time provides the most illustrative case of convergence between correlationism and object-oriented philosophy. To see this, we simply need to return to the initial problem out of which Meillassoux’s discussion of correlationism grows in After Finitude—what he calls ancestrality, or the confrontation between empirical science and the arche-fossil:
The question that interests us here is then the following: what is it exactly that astrophysicists, geologists, or paleontologists are talking about when they discuss the age of the universe, the date of the accretion of the earth, the date of the appearance of pre-human species, or the date of the emergence of humanity itself? How are we to grasp the meaning of scientific statements bearing explicitly upon a manifestation of the world that is posited as anterior to the emergence of thought and even of life—posited, that is, as anterior to every form of human relation to the world? Or, to put it more precisely: how are we to think the meaning of a discourse which construes the relation to the world that of thinking and/or living—as a fact inscribed among others, inscribed in an order of succession in which it is merely a stage, rather than an origin? How is science able to think such statements, and in what sense can we eventually ascribe truth to them?46
As others have noted, this is not meant to provide an argument against correlationism (which comes later in the book), but to confront it with an unpalatable consequence of its restriction of knowing to the human-world correlate. Harman’s presentation of this unpalatable consequence in his book on Meillassoux is admirably clear and concise:
The problem for correlationism is that it cannot give a literal interpretation of scientific statements […] The literal claim that the earth dates to 4.56 billion years ago must give way to a second, more sophisticated interpretation of this statement. We have seen that, instead of saying the earth is 4.56 billion years older than humans, the correlationist says that the earth is 4.56 billion years older than humans—for humans. No matter how skilled the sciences become at dating pre-human entities, the correlationist always has the trump card of turning all ancestral dates into dates for us […] Correlationists do indeed claim that they are not merely trapped in a human interior, when they ‘readily [insist] upon the fact that consciousness, like language, enjoys an originary connection to a radical exteriority (exemplified by phenomenological consciousness […] transcending toward the world)’. But Meillassoux rightly calls this supposed exteriority ‘a transparent cage’, and notes that in this way ‘contemporary philosophers have lost the great outdoors, the absolute outside of pre-critical thinkers […] that outside which was not relative to us […] existing in itself whether we are thinking of it or not’.47
The problem for Harman is that this is an equally clear and concise description of the consequences of his own theory of time. For him, the surface ‘time’ of all phenomenal experience (including empirical investigation) is not the deep time in which the real reconfigurations of objects occur (including the emergence of humanity), and this means that whatever systems of temporal relations empirical science employs to date such events are for us in precisely the sense he outlines, along with the systems of spatial relations in terms of which it locates them. To all appearances, this leaves Harman trapped in the same transparent cage as Kant and his successors.
Thus the question is what, if anything, differentiates Harman’s supposed realism from the appeals to an ‘originary connection with radical exteriority’ that he and Meillassoux so rightly scorn. This question can only be answered by properly delimiting the convergence between object-oriented philosophy and correlationism. The obvious way to begin this delimitation is by returning to the origin of correlationism in Kant’s distinction between phenomena and noumena, and determining its proximity to Harman’s distinction between the sensual and the real. There are obviously numerous parallels between these distinctions, but they are essentially united by their epistemological implications: we cannot know noumena/real objects, because we only ever encounter phenomena/sensual objects. The crucial point to take away from the above discussion is that Harman’s theory of ‘space’ and ‘time’ implies that the particular spatio-temporal systems in terms of which we locate and date entities and events operate at the phenomenal/sensual level, in essentially the same way as forms of intuition in Kant’s theory of space and time. Harman might wish to claim that he has regained the great outdoors, because his real objects exist in themselves ‘whether we are thinking of [them] or not,’ but he cannot articulate this independent existence in terms of any anteriority circumscribed by empirical science. He might insist that there is nevertheless some sort of deep anteriority that can be distinguished from surface anteriority, or a sort of primitive past that is demanded by the present but which can never be dated in relation to it.48 However, not only would this involve precisely the sort of thematisation of deep time that his work eschews, but it is not clear how it could possibly be justified within his framework.
This is made more problematic by the fact that merely insisting that real objects exist is insufficient to distinguish them from Kantian noumena, insofar as this claim is precisely what distinguishes Kant’s weak correlationism from the strong correlationism that follows it:
We claimed above that Kantian transcendentalism could be identified with a ‘weak’ correlationism. Why? The reason is that the Critical philosophy does not prohibit all relation between thought and the absolute. It proscribes any knowledge of the thing-in-itself (any application of the categories to the supersensible), but maintains the thinkability of the in-itself. According to Kant, we know a priori that the thing-in-itself is non-contradictory and that it actually exists. By way of contrast, the strong model of correlationism maintains not only that it is illegitimate to claim that we can know the in-itself, but also that it is illegitimate to claim that we can at least think it.49
If OOP is to avoid becoming indiscernible from this weak correlationism, we must be able to articulate the precise point at which it radicalises the latter, just as Meillassoux’s speculative materialism radicalises strong correlationism. Meillassoux’s description of correlationism delimits this point of radical divergence as well as it does the preceding convergence:
[C]orrelationism is not a metaphysics: it does not hypostatize the correlation; rather it invokes the correlation to curb every hypostatization, every substantialization of an object of knowledge which would turn the latter into a being existing in and of itself. To say that we cannot extricate ourselves from the horizon of correlation is not to say that the correlation could exist by itself, independently of its incarnation in individuals. We do not know of any correlation that would be given elsewhere than in the human beings, and we cannot get out of our own skins to discover whether it might be possible for such a disincarnation of the correlation to be true.50
Harman’s radicalisation of Kant turns upon this question of disincarnation: he insists, pace Kant, that we do in fact know that the correlation is ‘given elsewhere than in human beings’. This is not a metaphysical response to weak correlationism, but the conversion of weak correlationism into a metaphysics. The difference between Kant’s noumenon and Harman’s real object is thus that, whereas the former is fundamentally an epistemological notion that circumscribes the limits of empirical knowledge, the latter is a metaphysical notion that grounds these limits. For Kant, the thinkability of the in-itself does not provide us with any positive knowledge of it; our knowledge of its existence and non-contradictoriness is merely a critical delimitation of what it means for there to be something-we-know-not-what. For Harman, the in-itself is not merely thinkable but genuinely knowable, even if all that can be known about it are the conditions that make it otherwise unknowable. Although this (metaphysical) knowability of (empirical) unknowability is entirely consistent, we are left to wonder what unique characteristic of metaphysical knowledge makes it possible when all else is impossible.
This wonder gets to the heart of the relation between what Harman thinks and why he thinks it, and thereby reignites the methodological worries that have plagued our attempt to make sense of his work from the beginning. We have seen several crucial points at which the development of Harman’s metaphysics has been guided by appeals to certain fundamental facts about the real: the existence of real discreteness, the existence of real change/causation, and most recently, the existence of real time/anteriority. The important thing to understand about these facts is that, even though the value of Harman’s fourfold schema depends upon its ability to make sense of them, it does not, for all that, imply them: it is consistent with there being only phenomenal discreteness (Apeiron),51 it is consistent with there being only phenomenal change (world-machine), and it is consistent with there being nothing but a phenomenal past (ancestrality). There is a distinct lesson to be learned from each of these appeals to metaphysical fact.
In the first case, we saw that Harman attempts to use the ‘glaring experiential fact’ of discreteness to underwrite its reality. This seems somewhat paradoxical given the incommensurability between the real and the sensual, as his account of individuation prevents us from inferring the number of real objects from the number of sensual objects. This paradox is indicative of the peculiar intersection of phenomenology and metaphysics that defines Harman’s work, as it suggests that the reality of multiplicity is made manifest by the appearance of multiplicity. The second case is similar to the first, insofar as Harman seems to use the appearance of change and causal interaction to underwrite their reality. This is complicated somewhat by his tendency to associate these with the reality of discreteness in his arguments against radical holism. However, there is more to this appeal, because Harman is able to introspectively secure the reality of causation by using the way that we are emotionally affected by experience as the paradigm of real causal interaction and change. In both of these cases there is an appeal to a peculiar sort of phenomenological intuition capable of providing oblique access to the real. The final case is somewhat different from the others, because it is less an explicit appeal to metaphysical fact than an implicit dependence upon a metaphysical frame through which the other facts are understood. The reality of a deep time in which objects can come into being and cease to be provides the phenomenological background against which the intuitions of discreteness and causation emerge. Without this unthematised conception of time, his picture of vacuum-sealed objects that are nevertheless capable of violent interactions makes no sense.
So, what does this tell us about the form of metaphysical knowledge that Harman takes to be possible? It indicates that the oblique access to the structure of the real provided by his phenomenological method is of the same kind as the oblique access to real objects provided by allure: sensual multiplicity alludes to real multiplicity; causation is not merely alluded to, allusion is the very paradigm of causation; and time is the implicit and properly inexplicable background that is only ever alluded to. Of course, this is consistent with Harman’s own thoughts about both the general importance of allure and the specific character of philosophical method as opposed to that of the sciences.52 However, the idea that the only possible knowledge of the real and its structure depends upon allure as a mode of access, and thereby upon allusive language (i.e., metaphor) as a mode of expression, should give us cause for concern. The problem that ancestrality poses for correlationism is precisely that it is unable to take the claims of science literally. It makes their apparent truth compatible with the impossibility of knowing the real by redescribing them as true claims about appearance (i.e., as true-for-us). This is the semantic component of its epistemic scepticism. In defending the possibility of metaphysical knowledge against this semantic strategy, one might think that Harman had finally enabled us to take some claims about the real literally, but his interpretation of this knowledge as founded upon allure does precisely the opposite. We are left in a position where what we took to be literal claims about the world (e.g., ‘the earth is 4.56 billion years older than humans’) are covertly figurative, and what appear to be figurative claims about the world (e.g., ‘the world is composed of vacuum sealed objects with molten cores’) are prototypically literal. Seen aright, this is not a counterintuitive insight that is to be welcomed, but an intractable contradiction symptomatic of the deep and abiding influence of correlationism upon Harman’s thought. If there is one overriding irony we have exposed in this chapter, it is that the supposedly metaphysical connection that Harman draws between correlationism and relationism has been very effective at concealing the far more significant epistemological convergence between correlationism and OOP.
1. Cf. S. Shaviro ‘The Actual Volcano’, in Bryant, Srnicek, and Harman (eds), The Speculative Turn, and Harman’s ‘Response to Shaviro’ in the same volume.
2. Chapter 2.1, subsection II.
3. ‘The Road to Objects’, 171.
4. In some interpretations (e.g., possible world semantics) predicates do not directly return truth values, but return propositions which are themselves understood as functions from contexts (e.g., possible worlds) to truth values. This is to account for the fact that non-mathematical propositions such as that expressed by ‘the earth is the third planet from the sun’ are true when evaluated in ordinary contexts (e.g., in talking about the actual world), but might be false if evaluated in a different context (e.g., in a hypothetical discussion of a possible world in which the solar system formed differently).
5. Guerrilla Metaphysics, 169, my emphasis.
6. Guerrilla Metaphysics, 195.
7. Harman once suggested, during the Real Objects or Material Subjects conference at Dundee (2010), that comparative relations might be viewed as exclusively constituting sensual objects, but this was an off the cuff remark.
8. However, as pointed out in our discussion of the argument from independence, that causal relations and intentional relations can both be portrayed as asymmetric does not mean that they can be portrayed as sharing the same direction (chapter 2.3, subsection II).
9. This is hardly surprising given Whitehead’s contributions to modern logic.
10. Cf. Shaviro, ‘The Actual Volcano’, and ‘The Universe of Things’, <http://www.shaviro.com/Othertexts/Things.pdf>.
11. Chapter 2.1, subsection I.
12. Prince of Networks, 186–7.
13. Cf. G. Deleuze, Difference and Repetition, tr. P. Patton (New York: Columbia University Press, 1994), chapters 4–5; M. DeLanda, Intensive Science and Virtual Philosophy (London: Continuum, 2002), chapters 1–3. It is worth noting that the precise strength of this ontological priority is disputed by interpreters. For example, following Badiou (Deleuze: The Clamor of Being, tr. L. Burchill [Minneapolis: University of Minnesota Press, 2000]), Peter Hallward criticises Deleuze for granting the virtual (qua unitary whole) a strong ontological priority over the actual individuals, or creatures, that emerge from it (P. Hallward, Out of this World: Deleuze and the Philosophy of Creation [London: Verso, 2006], chapter 3). By contrast, DeLanda’s appropriation and development of Deleuze’s metaphysics seems to treat actual individuals (and the intensive processes that produce them) as distinct from but not inferior to the virtual multiplicities they incarnate (cf. his remarks on ‘flat ontology’ in Intensive Science, 153–4). Regardless, it is clear that both Deleuze and DeLanda wish to reject the total assimilation of pre-individual differential relations to relations between actual individuals that Harman advocates.
14. Prince of Networks, 188.
15. DeLanda, Intensive Science, 166–8.
16. Towards Speculative Realism, 178.
17. In dynamic systems theory, a system is represented by a phase space whose dimensions correspond to its degrees of freedom (the system’s variables). If a system has three dimensions [x, y, z], then its vector field is a manifold whose curvature is determined by the differential relations between their variables [dx/dy, dx/dz, dy/dz], and its attractors are the topologically invariant features of this curvature (e.g., cycles, saddles, points, etc.). For a more thorough overview of these ideas, consult DeLanda, Intensive Science (chapter 1), or Ian Stewart, Does God Play Dice: The Mathematics of Chaos (London: Penguin, 1989).
18. Cf. G. Deleuze, Empiricism and Subjecitivity: An Essay on Hume's Theory of Human Nature, tr. C. V. Boundas (New York: Columbia University Press, 2001); M. DeLanda, A New Philosophy of Society: Assemblage Theory and Social Complexity (London and New York: Continuum, 2011).
19. Towards Speculative Realism, 177–8.
20. Circus Philosophicus (Winchester: Zero Books, 2010), 23–4.
21. Prince of Networks, 187.
22. It is worth remembering that a line is just a one-dimensional surface, and that the vector field/virtual multiplicity is an n-dimensional surface whose curvature is determined by differential relations between many variables.
23. See my paper ‘Ariadne’s Thread: Temporality, Modality, and Individuation in Deleuze’s Metaphysics’ (<http://deontologistics.files.wordpress.com/2011/03/deleuze-mmu.pdf>) for an attempt at an outline.
24. It is worth noting that catastrophe theory was developed to provide mathematical tools to study precisely the sorts of changes that this objection takes to be outside of the scope of the dynamic systems approach.
25. J. Ladyman and D. Ross, Every Thing Must Go: Metaphysics Naturalized (Oxford: Oxford University Press, 2007), §§3.1–3.2.
26. Ibid., 137.
27. In particular, I think he is entirely right to criticise their refusal to account for the difference between mathematical and physical structure (Every Thing Must Go, §3.6). I am even willing to admit that there is something untoward about the manner in which they aim to derive metaphysical conclusions from epistemological premises, but this is of a piece with Harman’s own methodological problems, rather than something his criticisms identify.
28. ‘I am also of the opinion that materialism must be destroyed’, 786–7.
29. Ladyman and Ross, Every Thing Must Go, chapter 2.
30. There is a case to be made that this sort of reflexivity occurs in certain areas of social science, such as economics, insofar as the availability of models of our own behaviour can lead to both self-fulfilling and self-refuting prophesy. However, these cases are hardly representative.
31. Chapter 2.3, subsection II.
32. Cf. R. Brassier, ‘Concepts and Objects’, in Bryant, Srnicek and Harman (eds), The Speculative Turn.
33. Cf. Prince of Networks, chapter 1, §A.
34. Ibid., 73–5.
35. Harman’s failure to notice the elision of generality on Latour’s part is hardly surprising given his own track record on the matter (see chapter 3.2, subsection II).
36. The Quadruple Object, 99.
38. The Quadruple Object, 100.
39. Ibid.
40. Towards Speculative Realism, 161–2.
41. Cf. Ladyman and Ross, Every Thing Must Go, §3.2.
42. Guerrilla Metaphysics, 251.
43. Ibid., 252, my emphasis.
44. Towards Speculative Realism, 166, my emphasis.
45. See chapter 2.2, subsection 1, and chapter 3.3.
46. Meillassoux, After Finitude, 9–10.
47. Philosophy in the Making, 11–12.
48. This is not dissimilar to Iain Hamilton Grant’s account of anteriority in Schelling (I.H. Grant, ‘Does Nature Say What-It-Is?’, in Bryant, Harman and Srnicek [eds], The Speculative Turn, 66–83).
49. Meillassoux, After Finitude, 35.
50. Ibid., 11.
51. See chapter 2.2, subsection III.
52. Cf. Weird Realism, 11–23.