Causation and Counterfactuals
Edited by J. Collins, N. Hall & L.A. Paul
Representation and Mind Series
Cambridge (Massachusetts): MIT Press, 2004
Reviewed by Geneviève Girard-Gillet
Université Paris III – Sorbonne nouvelle
The philosophical essays in this book, written to the memory of David Lewis, discuss the connection between counterfactuals and causation, as this approach to causality is a rather new idea. It replaces the well-known analysis by Hume of causation as constant conjunction, and was suggested by David Lewis in his groundbreaking paper "Causation", that appeared in 1973.(1) This does not mean that all philosophers agree and the papers by sixteen contributors which are collected in this book approve of, or dispute the thesis.
The book contains the complete version of Lewis's Whitehead lectures, "Causation as Influence", together with a more recent essay, "Void and Object", thus enabling the reader to follow the arguments that led to a change of perspective, as well as the counter-arguments developed by some. The issues of the transitivity of causation and the nature of the causal relata are part and parcel of the questions addressed.
The counterfactual approach posits that there is a link between event c and event e, accounted for as follows: event e and event c (the cause and the effect) both occur, but if c had not occurred, e would not have occurred. In other words, causal facts are to be explained in terms of counterfactual dependence.
The volume is well organised as it intends to capture the interest of readers not familiar with the subtle interplay of arguments. J. Collins, N. Hall, and L.A. Paul, in the first chapter, “Counterfactuals, and Causation: History, Problems, and Prospects”, provide indeed a long overview of what the counterfactual approaches propose. This background is actually necessary to appreciate the discussions of the eighteen chapters making up the volume. The "seasoned veterans" are advised to go straight to the contributions though.
The basic semantic features of counterfactuals are summed up by the authors. First, it is necessary to provide noncircular truth conditions for the general counterfactual conditional with antecedent A and consequent C: "if it were the case that A, then it would be the case that C." A way out was inspired by the development of possible worlds semantics in modal logic. Another problem is the temporal considerations, as causes precede their effects. Whereas the future depends counterfactually on the past and the present, the past is not counterfactually dependent on the present and future. So back tracking conditionals should be ruled out. Not so quickly. Discussions of Lewis's "divergence miracle"(2) and Elga's entropy-diminishing process suggest that backward causation has to be accommodated, even if it might end up as a special case.
The counterfactual approaches are "reductionist" in that they consider that causal facts can be reduced to categorial plus nomological facts. Most of the essays in this volume adopt this perspective. In "physical connection" accounts, the causation consists, for instance, in the transfer of energy. It relates events in the microphysical world, but says nothing about the macroscopic realm. The "nomological entailment", on the contrary, agrees that the fundamental laws play an essential role, but they do not play this role in so specialised a manner.
It seems that philosophers are convinced that some sort of nomological entailment account is correct, and that the counterfactual analysis is a promising one, even if some problems have arisen lately. Lewis's analysis, which makes it clear what role the laws of nature play in fixing the causal facts—but not too much of the laws—is questioned by the Davidsonian account, which requires that for each causal relation, a law will directly "cover" it, at least when the relata are described appropriately. But the difficulty faced by Davidson is that it places severe constraints on an account of laws that can meet its need.
Here are a few problems suggested by Collins, Hall and Paul: Bill and Suzy converse between noon and 12:30 one day. Let c be their conversation. What would have happened if c had not occurred? Lots of questions actually arise: would c have occurred if they had conversed away from where they in fact conversed? Could c have involved different people? And so on. To answer these questions, we must know which features of their conversation are essential, and which features are accidental. It is also necessary to have some sort of account of what events are. Several other examples are observed, which call for a philosophical theory of events.
After thirty pages devoted to explaining what counterfactual approaches imply, the methodology of the research is described. Intuitions about cases are the driving force, but the problem is how to account for new cases when they question the accepted theory. The strategies depend on how/to what extent the counter examples weaken the theory. Eight possibilities are briefly described and some of them are developed in the following chapters (Chapter 9, chapter 11).
The first chapter reads easily as it refers to some of the most frequently used scenarii in the papers by the contributors, and paves the way for a good understanding of their lines of reasoning: the balls moving on the billiard board, the shattering of a bottle by a stone thrown at it, the tossing of a coin, the presence of a bomb in front of someone's door, and so on. These examples can serve as illustrations of whether causation is transitive or not. Intuitions suggest it is, but compelling examples imply that it is not. They also raise the question of what theory of events is necessary for a complete theory of causation. When someone' s walking and talking occupy the same region of spacetime, can we speak of distinct events or not? Lewis, Menzies, McDermott, Ramachandran hold the view that any difference in time, place, or manner of occurrence makes for a numerically different event. But there are also other causal relata, that require considering. Causation is thought to obtain between states of affairs, where states of affairs consist in particulars having properties (L.A. Paul). Hence events are no longer primary causal relata, and the need to accommodate causation involving omissions could thus be simplified. Nevertheless, the question of the causation by omission remains difficult to grasp, since Collins, Hall and Paul say quite clearly that they do not understand why a relation of dependence cannot qualify as a cause, in the scenario where Billy's failure to show up for their lunch date causes Suzy to become disappointed. It may also be that different kinds of causation require different explanations (Hall, Armstrong in the volume). The eighteen chapters that follow develop one aspect of the question.
-Jonathan Schaffer analyses the process called "trumping premption" when two processes, which are sufficient to bring about an effect, go to completion. It is the example of the Major and the Sergeant both shouting the same order. The soldiers will obey the Major, as he is hierarchically higher. This "trumping" event is the only cause of the soldiers' response, and counterfactual account of this causation is not valid.
-David Lewis (chapter 3) gave the reasons why a counterfactual analysis of causation is to be adopted as rival approaches are less satisfactory. He takes up the scenario of the soldiers, among others, discusses trumping, commonplace preemption, late preemption, causation by absence, but also transitivity, with several examples. He imagines, for instance, that there is a conflict between Black and Red. Black makes a move that, if not countered, would have made him the winner. Red responds with an effective countermove, which causes his victory. So we can say that Black's move causes Red's countermove, Red's countermove causes Red's victory. But can we say that Black's move caused Red's victory?
-John Collins revisits "preemptive prevention" with the example of the ball heading for a window, but prevented from breaking it by being caught in its course by individual (a) happening to be there. If individual (b) was ready to catch the ball, in case individual (a) failed to stop the ball, individual (a)'s catch is an act of preemptive prevention. But can we say that, if a brick wall had stood between individual (a) and the window, individual (a)'s catch caused the window to remain unbroken?
-Stephen Yablo devotes his paper to arguing that causation is counterfactual dependence, and begins his argumentation by referring to Plato's distinction between the "cause and that without which the cause would not be a cause" (Phaedo). The example given is that of the bomb: Billy puts a bomb under Suzy's chair; later, Suzy notices the bomb and flees the room; later still, Suzy has a medical checkup (it was already arranged) and receives from her doctor a glowing report. This is a counterexample to transitivity: the bomb is not the cause of the glowing report.
-Peter Menzies thinks that Lewis (1973) is right, but his concept of cause as difference-maker is unsatisfactory, because of his conception of causation as an absolute relation, independent of any contextual parameters. If pragmatic factors are overlooked, the theory will generate various causes, as in the example of the lung cancer, where the presence of lungs in an individual is as well the cause of his cancer as his chain-smoking habit. He then analyses context-sensitive counterfactuals to explore the conception of a cause in relation to some background.
-Ned Hall (chapter 7) reviews some counter examples to the transitivity of causation, which he considers fall into two types, and wonders whether if it is transitivity or the theory of counter dependence that is defective. His conclusion is that philosophers were over optimistic when adopting the Dependence theory and that new tools are necessary to better understand the central concept of causation.
-L.A. Paul (chapter 8) is also interested with the question of transitivity and analyses some problems raised by the reductive analysis, among others. Then she suggests that taking into account aspects, property instances, might avoid some negative consequences, and conform more to common sense, as well as to current scientific pratice.
-Ned Hall (chapter 9) distinguishes between two concepts of causation: dependence (counter dependence between distinct events), and production (the bringing about of another event). As there are various versions of a counterfactual analysis of causation, he concentrates his critique on the simplest one, counterfactual dependence between wholly distinct events. He takes up well-known examples (the shattering of the bottle, the shooting down of the fighter) and studies them from the standpoint he is taking.
-David Lewis (chapter 10) comes back to the notion of causation by omission. He thinks that the possibility of this kind of causation requires an analysis that treats cases of causation without causal relation. He discusses Menzies's viewpoint on his theory. In conclusion, he gives two different, but equally right answers to the question: What is causation? dependent on whether causation is a matter of analytic necessity, or a matter of contingent fact.
-Helen Beebee aknowledges that people tend to consider that absences can be causes, when someone's failure to do something causes such and such an event to happen, but she objects to this interpretation. She does not believe that there is any such thing as causation by absence, and considers that the misjudgment comes from a confusion between causation and causal explanation (see Davidson's distinction and Mellor's view)
-D.H. Mellor starts off by saying that the notion of "causation" covers too vague a semantic domain and requires a stricter definition. He argues that causation is not only a relation, but that most causes and effects are not particulars but facts. Facts are the most basic kind of causal relata, and they can be expressed by: "E because C", where E and C are sentences and "because" a sentence connective.
-David Coady considers himself a defender of the naïve counterfactual analysis, and he argues against preemption, as it is defined by Lewis in his example of the shattering of the bottle (the asymmetry between Suzy's throw, which hit the bottle first, and Bill's throw). He suggests instead that we have cases of symmetrical overdetermination and not cases of redundant causation.
-Cei Maslen agrees with philosophers who think that causal statements are dependent on context. For her the causal structure of the world is an objective three-place relation between causes, contrasts and effects. She gives illustrations of what a context-dependent causation is, and she explains how the context helps understand what the relevant contrast situation is. Transitivity is defined only for binary relations, but in her approach, causation is not a binary relation, which avoids counterexamples to transitivity (example of the bomb outside the door).
-Igal Kvart suggests that philosophers might have abandoned the probabilistic analysis of causation too soon, and consequently presents a probabilistic analysis of token causation. He first argues that the major problem of the approach, namely the problem of causes that lower the probability of their effects can be overcome. An account of causal relevance must indeed play a crucial role there. He also tackles the question of late preemption, among others. His conclusion is that the counterfactual analysis of causation might be circular.
-Mural Ramachandran discusses Lewis's "causation as influence" theory (in this volume). He develops his discussion around three points, which are debatable: the transitivity of causation, the possibility of backward causation, the existence of cases of overdetermination. To support his argumentation, he proposes three special notions: the idea of background chance, of accumulative chance-raising and of potential chance-raisers.
-Christopher Hitchcock is in favour of a probabilistic theory, stemming from a desire to have a theory that encompasses causal relationships that are indeterministic or "chancy". The approach requires that causes raise the probabilities of their effects. He analyses seven scenarii to show a number of advantages to this line of reasoning. In each case there is a marker that distinguishes the genuine cause from the spurious probability-raiser. The fact that they are heterogeneous in nature requires further study on what they have in common.
-Tim Maudlin thinks it is time to evaluate the prospects of Lewis's program. He argues that causation cannot be analysed in terms of counterfactual dependency, no matter what improvements are suggested. What really matters is the laws that underwrite causal claims. He then distinguishes laws of inertia—how things behave if left alone—from laws of deviation—how things will deviate from their inertial behaviour.
-D.M. Armstrong criticises Lewis's analysis, and puts forward a singularist theory. Causation is conceptually primitive for him. It seems that each of us can have introspective access to some of our own mental goings-on. And ordinary language uses verbs that involve reference of the perception of causality in the singular case. This might suggest that we have observational access to singular causal facts.
This book is a very good read for linguists working on causation and the structures that are labelled as causative. Hume explained, in the words of David Lewis, that "we never perceive causation, but only repeated succession". We can then wonder how we can see that one thing causes the other. Do we infer it from what we know about the fundamental laws that govern what happens? David Lewis answers that he does not know and does not even know how to find out. This suggests that discussions about whether one event is the direct cause or indirect cause of another one in order to account for a given linguistic construction might be slightly vain. Lewis distinguishes, for instance, between the movements of billiard balls, and the causal mechanism whereby a dinner too low in carbohydrate causes low blood sugar, which is less easily accounted for. Things are getting even more difficult in case of causation by absences. One has to be very careful with counterfactuality, and use the right kind of counterfactual conditionals. It does not make sense to say that if the barometer had not fallen, there wouldn't have been a storm.
It is also worth reading for linguists working on events and their succession. Causation between event e and event c requires they be distinct. It also necessarily needs nonidentity and nonoverlap. Collins, Hall and Paul give an interesting example of the necessity to know what is meant by an event: Suppose that Suzy shuts the door, and in fact slams it. We may have two events—a shutting and a slamming—distinguished though because the first could have happened without the second. But, if she had not shut the door, she would not have slammed it. Can we then say that the shutting is the cause of the slamming? Or isn't it more accurate to say that the two events are not distinct, and that they do not stand in a causal relation? A similar question arises with the following scenario: You walk to class. As you walk, you talk to your companion. Is your walking one event and your talking a numerically distinct event? Or are they the same event?
It is, at the same time, interesting for analyses of conditionals. Is there a difference between "if this glass had been struck, then it would have shattered", in which the implication is that the glass was actually not struck, and "if this glass was struck, then it would shatter", where the truth value of the antecedent is an open question? All the essays in this volume enable the reader to capture the difficulty of understanding what causation really is, and suggest that our common sense alone is not the best tool to grasp the complexity of a notion which remains controversial.
(1) “Causation”, Journal of Philosophy 70 (1973): 556–567. Reprinted in Philosophical Papers, Volume II. Oxford: University Press, 1986.
(2) “Counterfactual Dependence and Time's Arrow”, Nous 13 (1979): 455–476. Reprinted ibid.
Cercles © 2011