One of the main problems in the philosophy of causation is to identify the physical bases of causal relations. As Bertrand Russell (“On the Notion of Cause,” Proceedings of the Aristotelian Society 13 [1913]: 1–26) first pointed out, fundamental physics seems to leave no space for causal relations. Paradigmatic cases of causation involve local events, but physical laws relate global time slices of the world only. Moreover, causation is asymmetric, whereas physical laws are time symmetric. We thus face the pressing issue of reconciling the need for causal notions in the special sciences and in everyday life with the fundamentally a-causal picture of the world that physics gives us. The goal of Douglas Kutach in his excellent book is to offer a systematic explanation of the physical underpinnings of causation. Arguably the two most important contributions of the book are a new methodology for the metaphysics of causation and a new physical account of the causal asymmetry. This is a very rich and ambitious book, and I can only provide a rough sketch of its content.

The first chapter of Kutach’s book offers a new methodology for the metaphysics of causation. Borrowing a phrase from Phil Dowe’s Physical Causation ([Cambridge: Cambridge University Press, 2000], chap. 1), Kutach labels it ‘empirical analysis’. Like Dowe, Kutach is concerned with giving a physical explanation of what causation is in the actual world, but he offers a different conception of what this amounts to. Kutachian empirical analysis involves two steps. The first step is to identify the empirical phenomena that explain the utility of our folk concept of causation. The second step is to craft a scientifically improved concept of causation in order to give a physical explanation of those phenomena. Kutach endorses the view—standard since Nancy Cartwright’s “Causal Laws and Effective Strategies” (Noûs 13 [1979]: 419–37)—that our folk concept of causation is motivated in large part by our need to identify effective strategies to achieve goals. Hence, the goal of his empirical analysis is to articulate a concept of causation that can help us explain facts about effective strategies (e.g., our ability to advance future but not past goals) in terms of fundamental physics. By contrast to orthodox conceptual analysis, empirical analysis need not make all or even most folk platitudes about causation true, nor does it need to be responsive to intuitions about far-fetched cases (a nice feature, since there are good reasons to regard conceptual analysis of causation as a hopeless endeavor). It still deserves to be called an account of ‘causation’ because it is aimed at explaining the phenomena that make our folk concept of causation useful. As the rest of the book illustrates, empirical analysis is a very fruitful methodological framework for scientific metaphysics, and I look forward to seeing its application to other topics. (In other work Kutach has already applied his method to reductive identities and the time asymmetry of knowledge.) I would like to hear more about the empirical phenomena that are supposed to explain our repertoire of causal concepts. Presumably the need to identify effective strategies isn’t the only motivation for our use of causal concepts, since we also recognize causal relations involving unmanipulable causes such as the moon’s influence on tides. Indeed, Kutach insists that empirical analysis must make sense of causal regularities that do not involve human agency. I wonder what empirical phenomena motivate this aspect of our folk concept in Kutach’s view. It cannot simply be the fact that tides are associated with the position of the moon, since our (symmetric) concept of correlation is enough to capture that.

After introducing his methodology, Kutach begins his inquiry into causation with an examination of the ontology of our main fundamental physical theories (classical mechanics, electromagnetism, relativity, and quantum mechanics). His goal is to identify the components of fundamental reality that ultimately vindicate causal talk. In all fundamental theories one finds relations of ‘terminance’. A physical state of the world c termines another e if c nomologically determines e, or fixes a probability for e via the laws. Terminance relations are the building blocks of causal relations, but they lack the stereotypical features of causal relations. In particular, local events are almost never terminants, and in most of our fundamental theories terminance relations have no temporal direction: they stretch out in both directions of time. So the challenge is to explain how a-causal terminance relations ground local, asymmetric causation. This is the task of chapters 3–7, which form the core of the book.

Like many others, Kutach solves the locality problem by taking causation to be a matter of counterfactual dependence rather than nomological determination. He calls the relevant kind of counterfactual dependence ‘prob-influence’. Whereas counterfactual theorists of causation tend to focus on token causation, prob-influence is intended to represent the physical content of type-causal claims such as ‘lightning causes fire’. To test whether lightning prob-influences fire, one proceeds as follows. One starts with two sets of possible physical states of the world, A and Ā. Each member of those sets stretches sufficiently far out in space to serve as input to the global physical laws. The members of A and of Ā agree on everything, except that the members of A instantiate an occurrence of lightning in a space-time region R, whereas the members of Ā don’t. Each set is endowed with a probability distribution over its members. (Kutach argues that any reasonably smooth distribution will do.) When one evolves this probability distribution through the laws, one finds that the probability of a fire occurring near R is higher given A than given Ā. This is what it means for lightning to prob-influence fire. Kutach explores the consequences of construing causation as prob-influence for issues such as causal transitivity, Simpson’s paradox, and other problems in the philosophy of causation.

Kutach offers two explanations of the asymmetry of prob-influence. The first explains why prob-influence cannot ‘zigzag’ in time, that is, go in one direction of time first and then back in the opposite direction of time. Here I concentrate on the second explanation, whose goal is to account for the fact that we cannot directly influence the past. According to Kutach, the content of the claim that we can advance a certain goal E is revealed in the following experiment, which he calls the ‘asymmetry experiment’. Take a population of agents, and let a random device assign to some of them the task of securing E’s occurrence and to others the task of preventing E’s occurrence. Agents are suitably rewarded if they achieve their task. E is a goal that can be advanced just in case E occurs in higher proportion in the class of agents who are assigned the task of achieving E than in the class of agents assigned the task of preventing E. (Readers familiar with the work of Pearl and Woodward might think of the random device as playing the role of an intervention, i.e., eliminating possible spurious correlations between what agents do and the occurrence of E.) Kutach’s explanation for why we cannot advance goals in our past goes as follows. If we could advance a goal E in our past, there would have to be a bizarre correlation between the outcomes of the random device and the occurrence of E. Suppose, for instance, that one could influence now whether one had a slice of pizza for lunch yesterday. Then we would find that agents who are now randomly assigned the task of securing this outcome had pizza for lunch yesterday in other proportions than other agents. This sort of bizarre correlation never occurs in our world, for the same sort of statistical-mechanical reasons that explain the second law of thermodynamics (or so Kutach argues). By contrast, statistical mechanics doesn’t prevent the existence of correlations between being randomly assigned the task of having pizza for lunch tomorrow and the occurrence of this outcome a day after.

This is a sophisticated and ingenious explanation of the asymmetry of effective strategies. But I would like to hear more about the conception of effective strategies on which it relies. In Kutach’s view, an act-goal correlation is exploitable when the correlation survives in situations where the goal is randomly assigned to an agent. I do not deny that this account of exploitability is extensionally adequate, in the sense that it agrees with our opinions as to which acts are effective strategies for what. But it leaves an important question unanswered: why should we care about effective strategies so defined—why should the fact that certain act-goal correlations survive randomization be of such interest to us? After all, the sorts of goals we are routinely trying to achieve are not randomly assigned to us. So what is the point of having causal concepts that track the statistical patterns revealed in Kutach’s asymmetry experiment? Kutach doesn’t say much about this. But it seems to me that for his theory to achieve his goal (namely, articulating a physically grounded concept of causation that is of practical relevance to us), he owes us an answer.

These remarks are not intended to diminish the value of Kutach’s work. Causation and Its Basis in Fundamental Physics is an excellent book that pieces together many ideas to offer a coherent and original solution to the problem of causation’s fit with physics and various other issues as well. (In particular, the third part of the book, which I haven’t had space to discuss, contains an interesting discussion of the distinction between preempting and preempted causes.) I highly recommend it to anyone interested in causation, philosophy of physics, and the tension between the scientific and the manifest image.