The Atlas of Reality

by Robert C. Koons,Timothy Pickavance

  Doesn't this solution still involve a problematic infinite regress of dependency? We can find an infinite series of events, each earlier than its predecessor in the series. Consequently, each event in the series would depend in some sense on its successor, ad infinitum. Isn't this an objectionable kind of infinite regress? How does the metaphysical primacy of the whole process help?

  What's needed here is a distinction between a grounded and an ungrounded infinite regress. We think that it's reasonable to believe in an infinite regress, in which event E1 depends on event E2, event E2 on event E3, and so on, so long as all of these dependency relations are themselves grounded in a common source, one that is independent of all of the members and that does not itself give rise to a further regress. The problem with Bradley's Regress (Section 7.2.1.3), for example, was that there was no way of introducing such an ultimate ground of the instantiation relation without simply falling into another regress.

  More specifically, when each new event emerges as the process unfolds, the new event E1 is immediately dependent on the whole process up to that point and, therefore, on each prior event within the process. This dependency is modal or counterfactual: if the whole process had not unfolded as it did up to the occurrence of E1, E1 could not have occurred. Thus, if E1, E2, and E3 all belong to the same process, with E3 the earliest and E1 the latest, then E1 depends immediately on both E2 and E3, through its dependence on the whole process, even if it is also true (for similar reasons) that E2 depends on E3. We can find a fundamental ground for each dependency relation in the process itself, and no Bradley-like regress threatens.

  With this clarification in mind, let's return to the question of whether all causation is continuous. If all causation were continuous, then all causally connected events would have to be parts of one and the same process. The whole world would consist of a single continuous causal process, without interruptions. Thus, the non-existence of discrete causation leads to a kind of Monism (11.2A). The world would consist of a single truthmaker.

  However, as we have seen, there is some connection between the possibility of our knowledge of modality and the existence of multiple truthmakers. A Monistic world would be a world that provided us with no knowledge about the causal powers of things, which would make it impossible for us to know anything about alternative, contrary-to-fact possibilities.

  Monist accounts of our knowledge of modality would seem to amount to a distinction between nomic and accidental regularities, probably in line with the Mill/Ramsey/Lewis (MRL) Theory of laws (see Section 5.2), according to which laws are the axioms of the best theory of the world, where being best consists in simultaneously maximizing simplicity, closeness of fit, and breadth of coverage. Sentences stating the boundary conditions of the universe are too messy and various to count as laws. Hence, such sentences count as being only contingently true, giving rise to alternative possibilities.

  However, we've seen a number of powerful objections to the Ramsey/Lewis Theory of laws, especially the difficulties that small worlds pose for the theory. In addition, it is difficult to see what connection there could be between non-lawfulness as defined by the Ramsey/Lewis Theory) and contingency. Why assume that the laws are necessary? Why assume that non-laws are contingent?

  On the alternative, Powerist account, we can have reliable knowledge of local necessities and contingencies, by way of gaining knowledge of the causal powers of particular things. Aristotelian Modality (15.2T.7), which fits nicely with Powerism, requires multiple, disjoint truthmakers, belonging to an evolving network of causes and effects. To gain knowledge of these truthmakers, we must encounter cases of discrete interaction between two or more processes. Thus, knowledge of modality seems to be connected to the occurrence of discrete causation. If we have no knowledge of alternative possibilities, we have no understanding of the actual world or of the contents of our thoughts or concepts.

  28.4 The Nature of Continuous Processes

  We've seen that continuous causation depends on cases in which two events are parts of the same process. When does a set of events constitute a single process? In addition to real processes (carriers of continuous causation), the world is filled with pseudo-processes (Reichenbach 1958). Moving spotlights and shadows are classic examples of such pseudo-processes. Suppose that a domed stadium has a rotating lamp at its center, projecting a moving spot of light that appears to revolve continuously around the stadium's inner wall. Each stage of the history of the moving spot is connected continuously with earlier and later stages. The intensity and position of the spot might change continuously. However, the various spot-events in this history are not really causally connected with one other. Each spot-event is dependent on the process by which light travels from the central lamp to the wall, but it isn't in any way dependent on the apparent process of the movement of the spot of light. You can't affect the future position or shape of the spot by trying to do something to the present condition of the spot.

  What distinguishes real processes from pseudo-processes? There are two kinds of answers that we could give: reductive and non-reductive.

  28.4.1 Reductive accounts of processes

  1. Epistemic analyses. Bertrand Russell (1948) recognized this problem. Russell called real processes “causal lines”, and he defined causal lines in terms of what we can correctly infer from facts about the parts of such lines:

  A ‘causal line,’ as I wish to define the term, is a temporal series of events so related that, given some of them, something can be inferred about the others whatever may be happening elsewhere. ]source[(Russell 1948: 459)

  Wesley Salmon (1984) objected that we need an ontic and not an epistemic account. Surely facts about which lines are real processes and which are pseudo-processes do not depend on us or on our epistemic practices of inference. There were moving shadows and spots of light and real processes of movement, heating, and growth long before there were any humans around to infer anything. Moreover, Russell's analysis puts the epistemic cart before the metaphysical horse. The reason that it is correct to make inferences in some cases and not others is to be explained in terms of the metaphysical difference between processes and pseudo-processes, not the other way around.

  2. Spatiotemporal continuity and quantitative constancy. Another simple but deeply flawed account defines a real process as any set of events that is spatiotemporally continuous and that is characterized throughout by a fixed amount of some fundamental quantity whose conservation is guaranteed by a law of nature, like mass-energy.

  Phil Dowe's (1995, 2000) account is an example of such a theory. He proposes that continuous causation consists simply in the transmission of some quantity whose conservation is guaranteed by physical law. Dowe thinks that whether a quantity has been transferred along a certain continuum of events is wholly determined by the facts about how much quantity there is at each point of spacetime, together with the relevant laws of nature.

  The central problem with this account is that there are pseudo-processes that satisfy this definition. Suppose that the spot of light moves in a continuous fashion across the wall, so that between any two times, the spot is located in a position spatially between the positions it held before and after, and the spot is located at the limit of the locations it occupies at each of the members of any infinite but bounded series of times. Suppose, in addition, that there is a fixed amount of mass and energy associated with the spot at each moment. Such a moving spot will satisfy this definition without being a real process.

  In addition to the problem of failing to exclude certain pseudo-processes, Dowe's account also fails to resolve possible ambiguities in nature, like cases where it is unclear on empirical grounds which of two sets constitutes a process and which a mere pseudo-process. Consider a case of ambiguous exchange involving a pair of interacting particles. There are four events: C, C′, E, and E′. There are laws of nature linking the four pairs (C-E, C′-E, C- E′, C′- E′) in the following way: each of C and C′ have a 50% probabil
ity of transferring their quantity of stuff (energy, momentum, or whatever) to each of E and E′. In fact, the sum of the conserved quantity in E and E′ is exactly identical to the sum of it in C and C′. Where did the stuff in C go? The laws of nature give no determinate answer. There are two equally good alternatives: a transfer from C to E (and a simultaneous transfer from C′ to E′), and a transfer from C to E′ (and from C′ to E). Even though we cannot tell empirically which process is real and which pseudo, there should always be a fact of the matter as to which is which, a fact that Dowe's theory must deny.

  Such ambiguity might occur even in deterministic cases. Suppose that two point-particles, each carrying an equal quantity of energy, are on intersecting paths. Two identical particles converge on the intersection point, and two identical particles emerge from it. Was there a collision or did the particles pass through one another without effect? The conservation laws can't answer that question, since, in either case conservation was preserved.

  Finally, such constant-quantity theories wrongly exclude real processes that are mereologically inconstant, gaining or losing matter, energy, information, and other quantities. Many biological processes are like this. Photosynthesis, for example, takes in energy from sunlight at a variable rate. Even locomotion at the macroscopic level loses energy through friction, dissipating its energy into the surrounding environment.

  3. Counterfactual dependency. One critical problem with the second theory was that it could not distinguish mere accidental continuity and quantitative constancy from the kind of continuity and quantitative constancy that are regularly and nomically associated with a real process. We might try to use counterfactual conditionals to distinguish between the two cases. Wesley Salmon's (1998: Chapter 16) Mark Transmission Theory is an example of such an approach. Salmon requires that real processes be capable of transmitting a mark, which comes to something like the following condition:

  28.1T Mark Transmission Theory. A set of events P constitutes a real process if and only if (i) the spatiotemporal locations of the events in P form a continuum C, (ii) some conserved quantity Q is associated with the events in P at a constant value throughout C, and (iii) there is some action M (the ‘marking’ action) which did not in fact occur during C, but which is such that, if M had occurred at some spacetime location within C, then every subsequent event in P (but none of the earlier events) would have been replaced by a different event associated with a different value of Q.

  The marking event M is something that would have altered the mass, energy, information or other conserved quantity associated with the process. It is clause (iii) of the theory that introduces the counterfactual conditional: if M had occurred at time t, then the subsequent events of P would have been replaced by a different set of events, a set of events appropriately marked by the action M.

  Not surprisingly, this counterfactual conditional theory of continuous causation is subject to some of the same objections as is the Counterfactual Conditional Theory of Causal Explanation (27.1A.2). First, there are problems concerning linkage, or the unity of causal processes.

  Second, the counterfactual conditionals can be finked (just as the counterfactual account of dispositions was: see Section 4.4). To return to the rotating spotlight example, let's suppose that the operator of the central lamp is disposed to move the spot of light in response to movements of a spectator located near the wall. If the spectator “bats” the spot as it moves past, the operator is disposed to alter the path of the spot, simulating a ricochet effect. In this case, the spectator could mark the spot with a new momentum in such way that if the spectator were to do so, the subsequent events in the pseudo-process would be affected. Thus, a pseudo-process can satisfy the Mark Transmission Theory.

  Finally, there is the problem of accounting for the asymmetry of continuous causation. If we run the tape backward, the conserved quantity will apparently be transferred from the effect to the cause. As we have seen, counterfactual conditionals do not seem to be capable of grounding the asymmetry of causal direction.

  4. Ehring's trope transmission theory. Douglas Ehring (1997) proposes that tropes can be used to provide a reductive, non-circular account of the difference between real processes and pseudo-processes. A set of events constitutes a real process just in case each event consists in the process of some trope at some spatiotemporal location, and the very same trope is present in every event in the set. The unity of a real process is grounded in the diachronic identity of its associated trope. One version of Ehring's trope theory would focus on tropes of a special kind, such as tropes of conserved quantities like energy.

  Ehring's account does a good job of accounting for the unity of real processes, but it founders in its attempt to ground causal asymmetry. As Ehring recognizes, trope persistence is an entirely symmetrical relation. Hence, it cannot provide a basis for causal asymmetry.

  In addition, Ehring's account doesn't seem to add anything to a simple, non-reductive account of processes (which we will discuss in the next section). Ehring assumes that tropes do not have fundamental temporal parts. Each trope is wholly present at each time during the lifespan of the process. In other words, tropes are simples extended in time. Why not simply take the processes themselves to be simple, extended, four-dimensional tropes?

  28.4.2 A non-reductive account

  We can draft a simple, non-reductive account of processes. Processes are mereologically simple, having no actual parts. A process is only potentially divisible into sub-processes. On this view, instantaneous events are dependent boundaries of processes.

  What sort of things are processes? They could be properties (i.e., universals or tropes) of enduring substances, properties that are extended in four dimensions (time as well as space). In fact processes could be instances of temporally extended structural universals (see Section 10.3).

  Let's take a simple example, namely, locomotion. Suppose that a baseball is struck by the batter and caught by the right fielder. The process of motion is a property of the ball that extends over time, from the time of the batting to the time of the catching, and across space, from the bat to the fielder's glove. The process has many potential parts, sub-processes that connect its starting point to various intermediate positions of the ball. Each sub-process is a distinct potentiality, namely, the ball's potentiality to be stopped at each of these intermediate positions.

  On this view, the distinction between real processes and pseudo-processes is a simple one: processes exist and pseudo-processes do not. In the case of the moving spot of light, there is nothing that possesses any property corresponding to the pseudo-process. There is no spot of light, nor is the spot constituted by a plurality of real things. In contrast, a moving baseball consists of a large number of small physical objects, each of which has a real property corresponding to the locomotion.

  This way of thinking about causal process fits well with the Intervalist and Proceduralist views discussed in Chapter 19:

  19.2A Procedural Intervalism. Some extended processes are not composed of metaphysically fundamental time-slices.

  If temporally extended processes are metaphysically fundamental entities, and not mere heaps of events, then the causal dependency of one event on another can be understood in terms of their common inclusion in a single, metaphysically fundamental process. The asymmetric causal dependency relations would then be relations between processes. A process P′ depends on process P just in case P is a proper part of P′, P includes every part of P′ at or before time t, and P′ extends beyond time t into the future. Although there will be infinitely many sub-processes that are causally intermediate between P and P′, the dependency of P′ on P does not itself depend any of those intermediaries.

  28.5 Processes and the Direction of Continuous Causation

  If some causation is discrete, and the exercise of causal powers provides a direction to discrete causation, then the causal direction of processes can be derived from the causal direction of discrete interactions.
For example, there will be some joint exercise of causal powers at the beginning of each process, responsible for the existence of the process. In contrast, if a process is ended by the exercise of some causal powers, those powers will be responsible only for shortening of the process, not for its very existence.

  In addition, it seems plausible to suppose that each causal power must pre-exist its first exercise. It doesn't seem possible for a thing to gain a new causal power and to exercise that power in the very same instant. We could justify this claim by appealing to the following principle:

  Temporal Separation of Power Acquisition and Exercise. If event A is the acquisition of some power by a thing, and B is an exercise of that very power by that thing, then A and B cannot occur simultaneously.

  The acquisition of a power and the exercise of that power are two distinct events, with the first a precondition of the second. If so, the necessary priority of the acquisition of powers to their exercises would provide a basis for distinguishing an intrinsic arrow of time. Processes may have natures that determine their evolution in a time-specific way. These seem to be true of many macroscopic phenomena, whether chemical, biological, social, and astronomical. Finally, we can appeal to asymmetric token necessitation. The later parts of a process token-necessitate the earlier parts, and not vice versa, since any token process could be ended by interruption at any time after its initiation.