Relating to Responsibility: Essays in Honour of Tony Honoré on his 80th Birthday 9781472562388, 9781841132105

Citation preview

PREFACE The essays in this collection are based—some more closely than others—on papers presented at two workshops, one in Canberra in November 1999, and the other in New York City in March 2000. Although organised independently of each other and attended by different people, the two workshops shared a format and a focus. Four papers were discussed at each event,1 and all of them were reactions to Tony Honoré’s remarkable post-retirement writings on responsibility, as represented in his 1999 collection Responsibility and Fault. Tony himself was present on both occasions and had the opportunity to respond to the papers. That dialogic feature is carried over into this collection, in the form of a ninth essay by Tony himself replying to criticisms and, in characteristically generous spirit, welcoming new ideas. A Festschrift for Tony was published in 1986, marking his sixty-fifth birthday. This one marks his eightieth birthday in 2001. The two books are quite different in conception. The Legal Mind embraced the daunting breadth of Tony’s scholarship, its contributors testifying to his status as a giant of comparative law, legal history and legal philosophy alike. This volume shifts the emphasis from breadth to depth. Rather than attempting to cover the many fronts on which Tony works—all of which have remained as well-defended as ever in his eightieth year—this book is a celebration of his profound and enduring insights in one area. Tony’s interest in that area, namely the area of moral and legal responsibility, goes back at least to his collaboration with H.L.A. Hart on the work that was to become Causation in the Law (1959). Hart too emerged from that collaboration to write many important and influential solo essays on responsibility, especially as it figures in the criminal law. Yet Hart deliberately steered clear of certain aspects of the subject, particularly those bearing on the philosophical foundations of the Causation project. It was left to Tony, in principal charge of preparing the long and rightly venerated preface to the second edition of Causation in the Law (1985), to bring these issues to the surface. There Tony instigated the deep and difficult fifteen-year line of inquiry that adds up to Responsibility and Fault. That line of inquiry is strictly speaking three tightly interwoven lines of inquiry. The first, and the one that remains closest to the relatively familiar issues traversed by Hart and many other modern writers on responsibility, concerns human capacities. How, if at all, do people’s capacities and incapacities 1. Cane, Pettit, Smith and Stapleton in Canberra; Gardner, Lucy, Perry and Ripstein in New York.

v

 bear on their moral and legal responsibility? Tony has a novel answer, first sketched in 1964, which involves focusing on certain capacities and testing them at a certain level of generality. In this volume the idea is picked up by Michael Smith and Philip Pettit who both revisit some classic puzzles about responsibility with Tony’s clarifications and distinctions in mind. Although the same clarifications and distinctions remain in view in the four essays that follow, these essays also introduce the second strand of Tony’s explorations. In respect of this strand Tony is not only the architect of a distinctive answer but also the engineer of a distinctive question. Why are people responsible, indeed responsible above all, for the harm they actually bring about? Why (in Tony’s expression) are responsible people, above all, outcome responsible? Among the contributors to this volume we have two (Arthur Ripstein and Peter Cane) who have serious doubts about whether responsible people really are outcome responsible above all in the relevant sense, as well as two (Stephen Perry and John Gardner) who set out to corroborate the assumptions of Tony’s question. When they come to consider the answer, however, the four essays form a variety of cross-cutting alliances, and each reinforces some parts of Tony’s edifice while attempting to shake others. Of these four essays on outcome responsibility, only Cane’s delves into the third aspect of Tony’s inquiry, the one which provided the original theme of Causation in the Law. Granted that we are (in some sense) outcome responsible, what counts as an outcome in the relevant sense? Which relations are causal ones, how do causal relations differ among themselves, and which causal relations are responsibility-grounding? These questions are the ones to which Jane Stapleton gives over her essay. She tests the theses at the heart of the Causation book that are brought back to the boil in, or sometimes merely simmer under the surface of, Responsibility and Fault. Her essay is followed by William Lucy’s broader summative attempt to locate Tony’s whole enterprise —from the detailed study of case-law in Causation in the Law right through to the most speculative of the material gathered in Responsibility and Fault— on the map of legal theory. Is it lawyers’ philosophy or philosophers’ law? How do the philosophical sweeps interplay with the legal niceties? Like the contributors to The Legal Mind, the contributors to Relating to Responsibility relate to Tony and his work in various different ways. Some have worked closely with him and others know him mainly through his published ideas. Some have traced the steady development of his views over the years while one or two have come to Responsibility and Fault as a more or less finished product. All, however, have been touched in their thinking and feeling by the rare combination of humanity, rigour, inspiration, and learning that Tony brings to this most testing of subjects, as to so many others. Nor should vi

 it be forgotten that the published work discussed here represents only one dimension of Tony’s contribution to the theoretical study of legal doctrine. This year, 2001, marks not only Tony’s eightieth birthday, but also the fiftieth anniversary of the first series of seminars in which Tony (then collaborating with A.D. Woozley) introduced Oxford law students to the deep philosophical puzzles nesting within apparently prosaic rules of legal liability. Although conducted with various different collaborators over the years, those seminars have continued virtually uninterrupted until the present day. As we write this, indeed, Oxford graduate students are preparing, as they have about fifty times before and as they will for years to come, for tomorrow’s crucial seminar on fault and strict liability. Only now, thankfully, they have Tony’s illusionbusting and perspective-restoring writings to help them tell the philosophical wood from the trees. Peter Cane and John Gardner St Valentine’s Day 2001

vii

ACKNOWLEDGEMENTS Thanks are owed to the Law Program in the Research School of Social Sciences at the Australian National University and to the Center for Law and Philosophy at Columbia Law School for the support they provided, financial and otherwise, to the workshops from which this book emerged. Thanks also to Richard Hart for his usual encouraging ways.

xi

ABOUT THE CONTRIBUTORS Peter Cane is Professor of Law in the Research School of Social Sciences at the Australian National University. His main research interests are in public law (especially administrative law) and the law of obligations (especially tort law). Recent work includes The Anatomy of Tort Law (1997) and Atiyah’s Accidents: Compensation and the Law (6th edn, 1999). He is currently writing a book entitled Responsibility in Law and Morality. John Gardner is Professor of Jurisprudence in the University of Oxford. He has held appointments at Columbia University in New York City and at King’s College London, as well as various other Oxford posts. His main interest is in the philosophical analysis and critique of legal doctrines (such as those of the criminal law and the law of torts). He has convened graduate seminars with Tony Honoré for the past twelve years. Tony Honoré, who has a South African background, works in three main areas: South African trust law; Roman lawyers and government in the second to sixth centuries ; and legal theory, especially problems of causation and responsibility. Since 1946 he has been in Oxford where he was Regius Professor of Civil Law from 1971 to 1988. His book Responsibility and Fault, the main focus of this collection, was published in 1999. William Lucy is Professor of Law at Keele University, having moved there from the University of Hull in 1999. His main interests are in jurisprudence and private law. The author of Understanding and Explaining Adjudication (1999), he is currently working on a book introducing the philosophy of private law. He is also taking a fresh look at the democratic critique of adjudication. Stephen Perry is John J O’Brien Professor of Law and Professor of Philosophy at the University of Pennsylvania. He previously taught at McGill University and has held visiting positions at the University of San Diego and New York University. His current research is in the areas of tort theory and general jurisprudence. Philip Pettit is Professor of Social and Political Theory at the Research School of Social Sciences, Australian National University, and a regular Visiting Professor of Philosophy at Columbia University. He is author of a number of books, including Republicanism: A Theory of Freedom and Government (1997), xiii

   Three Methods of Ethics (1997, with M. Baron and M. Slote) and The Common Mind: An Essay on Psychology, Society and Politics (1993). He also has a new book in press A Theory of Freedom: From Psychology to Politics (2001). Arthur Ripstein is Professor of Law and Philosophy at the University of Toronto. He was Laurance Rockefeller Visiting Fellow at Princeton in 199596, and was a Connaught Fellow in 2000. He is the author of Equality, Responsibility and the Law (1999) and co-editor of Law and Morality (1996, 2nd edn, 2001), and Practical Reason and Preference (2001). He has also published numerous articles in legal and political philosophy. He is an associate editor of the journal Ethics. Michael Smith is Professor of Philosophy at the Research School of Social Sciences, Australian National University. Author of The Moral Problem (1994) and editor of Meta-Ethics (1995). His main research interests include moral and political philosophy, philosophy of law, moral psychology, and philosophy of mind and action. Jane Stapleton is Professor within the Law Program of the Research School of Social Sciences, Australian National University, before which she was Fellow in Law at Balliol College, Oxford. Currently she is Adviser to the Restatement (Third): General Principles project and is writing about causation and comparative product liability.

xiv

Tony Honoré Photograph: Bernd Noelle, Schönbergerstr 12, D-81679 München

i


1 RESPONSIBILITY AND SELF-CONTROL Michael Smith*

In “Being Responsible and Being a Victim of Circumstance”, Tony Honoré tells us: “Before imposing sanctions or attaching blame, law and morality requires something more than that the person concerned is responsible for what they have done. One further requirement, common to both . . . is that in the circumstances the agent had the capacity to reach a rational decision about what to do. When this capacity is present, blame for bad behaviour is appropriate and criminal liability may, depending on the state of the law, be imposed. But, though capacity has often to be treated as an all-or-nothing matter, since an offender must be found guilty or not guilty, in real life our ability to decide rationally is a matter of degree. So different degrees of blame, punishment and censure correspond to the extent to which the agent’s capacity is impaired.” (Honoré, 1998: 138)

Though Honoré is doubtless right about this, the fact that we impose sanctions and attach blame in this way raises several difficult questions. My aim in this paper is to raise some of these difficult questions, and then hopefully to suggest some answers. The general idea, let’s agree, is that the only people we see fit to sanction and blame are those who have rational control over their conduct. We no more sanction and blame those who have no rational control over their conduct—those who are, in Honoré’s phrase, “victims of circumstance”—than we sanction or blame floods or earthquakes or lightning bolts for the harm for which they are responsible. The main question on which to focus, then, is what precisely it means for someone to have rational control over their * An earlier version of this paper was read at Responsibility in Law and Ethics, a symposium held in honour of Tony Honoré at the Research School of Social Sciences, Australian National University, November 1999. I would like to thank all those who participated in the discussion at this symposium, but especially Tony Honoré. The paper builds upon ideas that first appeared in “Frog and Toad Lose Control” (Kennett and Smith, 1996). Ss. 2, 3 and 4 include a translation of material that first appeared in “Quelques énigmes concernant le contrôle de soi” (Smith, forthcoming). I am grateful to the editor of Philosophiques for his permission to include that material here. Finally, I would like to thank Lloyd Humberstone for a crucial suggestion, and Peter Cane and John Gardner for their very helpful comments on the final draft.

  conduct, and, in particular, how it can be that this ability, and the consequent sanctions and blame, comes in degrees. So as not to muddy the waters with too much unnecessary controversy, my discussion will focus on a particular case in which issues of rational control arise, but which raises no issues of legal or moral significance. My aim is, if you like, to examine and describe the psychological structures required for an agent to possess and exercise the capacity for rational control in a noncontroversial case, and then simply to assume that those same psychological structures will be in place, playing much the same roles, in the more controversial cases. With that in mind, here is the case on which I will focus. Each day on his way to work, John stops at the local supermarket to buy his day’s supply of chocolate bars. Despite his love of chocolate, he knows that he shouldn’t eat as much chocolate as he does. Given that he is over forty and does no exercise, he realises that the amount of chocolate he consumes simply adds to an already significant weight problem. But his belief that eating so much chocolate will make him fat and induce heart disease simply does not move him. Try as he might, John cannot control himself— or so he says. He buys and eats chocolate notwithstanding his beliefs about what he should do. With this case before us, let’s turn to consider the various questions that arise.

1. H               -       

Note a crucial feature of John’s case, as described. John says that when he eats chocolates he is out of control. But it would surely be quite incredible to suppose that he does not intend to eat chocolates. Indeed, as described, it seems that John quite evidently has a standing intention to eat chocolate each day. This provides us with a first puzzle. The puzzle arises because the concept of intention would seem to bring a concept of rational control in its wake. The constitutive role of an intention is, after all, to ensure that agents act over time in accordance with a plan (Bratman, 1987). Thus, for example, the constitutive role of John’s intention to eat chocolate each day is to ensure that John’s behaviour, over time, fits in with an overall plan of action that sees him eating chocolate each day. His intention to eat chocolate is thus what ensures that he leaves for work in time to stop at the supermarket before he begins his day’s work; it is what ensures that he takes the route to work that goes via the supermarket that sells chocolate; it is what ensures that he has some money with him when he walks into the supermarket; it is what ensures that he takes 2

  - the chocolate he buys with him to his office, and doesn’t leave it in the car; and so on and so forth. The mere fact that John has an intention to eat chocolate each day therefore guarantees that he exercises quite a lot of control over his behaviour by placing quite severe constraints on the way in which he conducts himself over time. Very roughly, it ensures that, whatever else he does over time, he does not act in ways inconsistent with his eating chocolate. Moreover, it ensures that he does the things that he needs to do in advance for his subsequent eating of chocolate. But now the puzzle should be evident. For how can we suppose that John lacks rational control when he plainly exercises so much rational control? The solution to this first puzzle lies in recognition of the fact that, though there is a sense in which John exercises control over his behaviour simply in virtue of having a standing intention to eat chocolate, there is also a sense in which he doesn’t. The reason is that intentions themselves are appropriately arrived at as a result of deliberation. When they deliberate, agents reflect on what it would be good to do, or what they should do, or what it would be rationally justifiable for them to do, and, on the basis of these reflections, they form their intentions. But what cases like John’s bring out is that an agent’s intentions do not invariably answer to the considerations that he takes into account when he deliberates. Sometimes, as in John’s case, an agent may intend to act in one way even though, when he reflects, he commits himself to the view that it would be bad to act in that way, or that he should not act in that way, or that his doing so would not be rationally justifiable. Here, then, we find a residual sense in which an agent may fail to control his behaviour despite the fact that he intends so to behave. For an agent to be in control of himself when he behaves, it is not enough that his behaviour conforms to his intentions. His intentions must in turn conform themselves to his deliberations, that is, to his beliefs about what it would be good to do, or what he should do, or what it would be rationally justifiable for him to do. John thus looks to be a likely candidate to lack control, in this sense—control of himself—because his behaviour, though controlled by his intentions, is not controlled by intentions that are in turn controlled by his deliberations. 2. H           

The solution to the first puzzle about rational self-control makes an important assumption. It assumes that agents can have beliefs about what it would be good to do, or what they should do, or what it would be rationally justifiable 3

  to do. But what exactly are these beliefs about? What would make them true or false? Moreover, how are such beliefs supposed to impact on our actions? Thomas Hobbes provides the orthodox answer to this question. “. . . whatsoever is the object of any mans Appetite or Desire; that is it, which he for his part calleth Good: And the object of his Hate, and Aversion, Evill; And of his Contempt, Vile, and Inconsiderable. For these words of Good, Evill, and Contemptible, are ever used with relation to the person that useth them: There being nothing simply and absolutely so; nor any common Rule of Good and Evill, to be taken from the nature of the objects themselves . . .” (Hobbes, 1651, part 1, ch. 6: 120)

Hobbes’s idea is that our beliefs about what is good and bad can impact on our desires and aversions—which, in turn, impact on our intentions— because these beliefs are beliefs about our desires and aversions. However, if Hobbes is right about the meaning of “good” and “bad”, then it turns out that the distinction made in solving the first puzzle about self-control is a distinction without a difference. John says that he shouldn’t eat so much chocolate. But, translating what John says into Hobbes’s terms, what he says is, apparently, that he is averse to eating so much chocolate, or desires not to. But, as the story has been told, not only is this manifestly false, John knows it to be manifestly false. John is not at all averse to eating so much chocolate. Rather, he has very strong desire to eat chocolate, a desire so strong that he has a standing intention to do so, and in recognition of which he reaffirms his view that he shouldn’t be eating so much. The Hobbesian translation of John’s claim that he shouldn’t eat so much chocolate thus casts doubt on the possibility of distinguishing claims about what an agent desires to do or intends to do from claims about what it would be good for him to do, or what he should do, or what it would be rationally justifiable for him to do. Here, then, is the second puzzle about self-control. We have to give an account of what the “should” means when, in cases in which we lack control, we say that we know we desire to act in ways that we shouldn’t. But the orthodox account, Hobbes’s account, will not do. It suggests, falsely, that we contradict ourselves in saying what we say. But nor, importantly, will the obvious variations on Hobbes’s account do either. For example, it will not do to suggest that when John says that he shouldn’t eat so much chocolate, what he is saying is that he has a desire that his first-order desire to eat chocolate not be effective: that he would prefer that a first-order desire to refrain from eating chocolate be effective in action instead (Frankfurt, 1971). For, as Gary Watson has pointed out, an agent’s 4

  - second-order desires are simply further desires he possesses (Watson, 1975). They have no special status that would allow them to give content to claims about what we should do. Thus, though it is certainly true that if John has a first-order desire to eat chocolate and a second-order desire that a first-order desire to refrain from eating chocolate be effective in action instead, then the desires that he has will not be cosatisfiable, there would seem to be no reason why it is the first-order desire John has to eat chocolate that should be changed to match his secondorder desire, rather than his second-order desire that should be changed to match his first-order desire. In other words, there seems to be no reason why John’s lack of control should reside in his possession of a conflicting first-order desire, as opposed to residing in his possession of a conflicting second-order desire. The solution to the second puzzle about self-control thus requires a more radical departure from Hobbes’s view. Let’s assume, for the moment, that our desires and aversions can sometimes be the product of irrationality. On that assumption it is simply implausible to suppose, as Hobbes does, that to say that something is good or bad, or that it should or shouldn’t be done, is to say that we desire or are averse to it. It is implausible because we would all readily agree that at least some of our desires and aversions are desires and aversions that we should not have in a totally uncontroversial sense, that is, in the sense that we would not have them if we were fully rational (Smith, 1994; 1995). What would still seem plausible, however, is a modified form of Hobbes’s view, a modification in the spirit of the Enlightenment idea, due to Kant, that our desires and aversions must themselves be formed via rational processes. On this view, what each of us, for our own part, calls “good” and “bad” is still a matter of a relation that we stand in to the things that we call “good” and “bad”, but the relation is not that of being something that we actually desire, or to which we are actually averse, but is rather that of being something that we would desire, or to which we would be averse, if we were in a more fully rational state: that is, if we had desires and aversions that eluded all forms of rational criticism. The question we must ask is thus whether the assumption that our desires and aversions can be liable to rational criticism is correct. And the answer must surely be that it is. Imagine, for example, that John has both a desire for pleasure and a stronger desire for health, and that he believes both that eating chocolate will lead to pleasure and that refraining will lead to health, but that only the weaker desire for pleasure transmits its force across the means–end relation. In that case John’s resultant desire to eat chocolate would be liable to rational criticism on the grounds that it is the product of 5

  instrumental irrationality. John might then be able to say quite truly that, notwithstanding his desire to eat chocolate, it would be more desirable for him to refrain from doing so: that is, that he would more strongly desire himself to refrain if he were fully rational. Or imagine instead that John has both a desire for pleasure and a weaker desire for health, and that he believes both that eating chocolate will lead to pleasure and that refraining will lead to health, but that the relative strengths of his desires would change if his desire set as a whole was more coherent and unified. In that case John’s resultant desire to eat chocolate would be liable to rational criticism on the grounds that it is the product of the incoherence or disunity of his desire set as a whole. John might then once again be able to say quite truly that, notwithstanding his desire to eat chocolate, it would be more desirable for him to refrain from doing so: that is, that this is what he would more strongly desire himself to refrain if he were fully rational. Or imagine instead that though John has only a desire for pleasure and a belief that eating chocolate will lead to pleasure, and though his desires are not a product of either means–end irrationality, or incoherence or disunity in his desire set as a whole, that he is ignorant of certain facts, and that, if he weren’t ignorant of those facts, he would have a much stronger desire to be healthy. Imagine further that this desire would be so strong that, given that he believes that refraining from eating chocolate will lead to health, he would then desire even more strongly not to eat chocolate. In that case John’s resultant desire to eat chocolate would be liable to rational criticism on the grounds that it is the product of ignorance. John might then, again, be able to say quite truly that, notwithstanding his desire to eat chocolate, it would be more desirable for him to refrain from doing so: that is, that this is what he would more strongly desire himself to do if he were fully rational. Here, then, we find a natural interpretation of what John is saying when he says that he should stop eating so much chocolate, notwithstanding the fact that his strongest desire is to do just that. He is expressing his belief that this is what he should do all things considered, and this, in turn, is a belief about what he would most want himself to do if he were fully rational: that is, if he had knowledge of all the relevant facts, was fully instrumentally rational, and if his desire set as a whole was maximally coherent and unified. This is a natural interpretation of what John says because we are all familiar with situations in which we have no desire at all to act in the way we believe we would act if we were fully rational. Failures of memory and imagination, ignorance, means–end irrationality, incoherence, and all manner of other non-cognitive personality disorders as well, can readily cause us to lose desires that we would have in a more fully rational state, or cause us to have desires that we would 6

  - not have in a more fully rational state. These sorts of failures of reason are thus what explain our need to exercise rational self-control. 3. H        -     

The second puzzle was to explain what the “should” means when we say, as John says, that though we may desire to act in certain ways in certain circumstances, when we think that we need to exercise self-control in those circumstances what we also believe is that we should not act in those ways. The third puzzle is to make some coherent sense of the idea that the needed exercise of self-control is so much as possible. The source of the problem here lies in a truism in the philosophy of action, a truism popularised by Donald Davidson (1970). According to the truism, if an agent desires most to act in a certain way, and he believes himself free to perform that action, then if he tries to perform any action at all, that is the action he will try to perform. Thus, according to the truism, if right here and now I desire most to continue writing, and I believe myself able to do so, then that is what I will try to do, if I try to do anything. Of course, I might not try to do anything. I might faint, or fall into a coma, or drop dead. But if I do try to do anything, then what I will try to do is to continue writing. Though the truism no doubt requires more careful formulation than I have given it here, I hope that it does sound truistic. It certainly should, given the ways in which the concepts of desire and belief and action are interdefined. An action is, after all, simply defined as a doing that is the causal upshot of a desire and belief pair, and in cases in which an agent has a variety of desires and beliefs, what she wants most to do is simply defined to be the action that is the object of the desire and belief pair that has the greatest causal power. Despite the fact that the truism requires more careful formulation, then, it seems safe to suppose that it, or something along similar lines, is indeed a truth. Given the truism, however, it is hard to see how anyone who needs to exercise self-control could ever succeed in doing so. For, as we have seen, an agent who needs to exercise self-control desires most to act in one way while believing that he should act in another. But if an agent desires most to act in one way while believing that he should act in another, the truism tells us that if he tries to do anything at all, then what he will try to do is to act in the way he most wants to act, not in the way he believes he should. Indeed, since the truism purports to be a conceptual truth, it seems to follow that it 7

  is logically impossible for someone who needs to exercise self-control to succeed in doing so. For that would require him both to desire most to act in one way, and yet to desire more strongly to act in the way required for selfcontrol instead, and that is an out and out contradiction in our description of the agent in question. Return to the case of John, just to drive the point home. John desires most to eat chocolate, but believes he shouldn’t. According to the truism, however, it follows that if John tries to do anything at all then he will try to eat chocolate. But if John is going to try to eat chocolate, if he tries to do anything at all, then he evidently isn’t going to try to exercise self-control. Though we have succeeded in explaining why John needs to exercise self-control, it thus seems that the needed exercise is a logical impossibility. This is the third puzzle about rational self-control. How are we to solve it? It seems to me that this puzzle arises because of assumptions that need to be made explicit and then questioned. Note, for example, that there are at least two quite distinct times at which we can exercise self-control. Suppose we envisage, at time t1, that we will be out of control at time t2, at least absent an exercise of self-control. The two distinct times at which we can exercise self-control reflect the fact that t1 and t2 might be the same time, or different times. The puzzle, on the other hand, arises only on the assumption that t1 and t2 are the same time. Imagine that t1 is an earlier time and that t2 is a later time. We are then in a position to ask ourselves, at t1, what we most want to do at t1, and the answer might well be that, believing as we do that at the later time t2 we have the potential to lose control, what we most want to do is to ensure that at t2 we do not lose control. And here, accordingly, is one completely straightforward way in which we can perform an action, and so exercise control over our own subsequent actions, at least provided we are not out of control at t1. We can exercise self-control diachronically, at the earlier time, by so arranging the circumstances of action that we will face at the later time so as to remove the possibility of our then losing control. Thus, for example, if at an earlier time when, say, he was instrumentally rational, John had foreseen that he would no longer be instrumentally rational when passing the supermarket on the way to work—perhaps the sight of the supermarket makes especially salient the possibility of buying chocolate, and this causes him to become instrumentally irrational and only then to intend to eat chocolate—then, if he had most wanted to do so, he could have made sure that he was unable to act in an instrumentally irrational way at that later time by, say, ensuring that the only option available to him then would be the one that he would have desired if he had been fully instrumentally 8

  - rational. He could have driven to work by a different route, say, so that he didn’t ever get to see the supermarket; or he could have so arranged things that he had no money in the car on the way to work; or he could have taken someone with him in the car who would talk him out of stopping to buy chocolate; and so we could go on. In this way he could have ensured at the earlier time that, despite his potential to act in an instrumentally irrational way at the later time when driving past the supermarket, that potential was never realised. In short, then, one solution to this third puzzle about self-control lies in the mundane observation that diachronic self-control is possible. When we exercise diachronic self-control our trying not to do what we most want to do may indeed require that our strongest desire is not to act on our strongest desire. But there is no contradiction involved once we see that these desires are had at different times. When we exercise diachronic self-control at t1 our strongest desire at t1 may be that we cause ourselves not to act on what will be our strongest desire at the later time t2. The trick lies in the fact that the exercise of diachronic self-control at t1 itself ensures that we are unable to act on our strongest desire at t2 by making it the case that the only acts available at that later time are acts that satisfy weaker desires. 4. H        -           

What if t1 and t2 are the same time? At the very moment that he desires most strongly to eat chocolate, is John able not to act on this, his strongest, desire? This is the fourth puzzle about self-control. Granting that diachronic selfcontrol is possible, is synchronic self-control possible too? Is it possible for an agent both to need to exercise self-control and to succeed in doing so at the very moment of vulnerability? Some people are sceptical about the possibility of synchronic self-control. One reason for their scepticism is that many examples of what appear to be cases of synchronic self-control turn out, on closer inspection, to be cases of diachronic self-control where the times in question are just very, very close together. Thus, for example, imagine a case in which if John catches a glimpse of the supermarket where he regularly buys all of his favourite chocolates on his way to work then that glimpse will cause an irrational shift in his desiderative profile. It will cause him to become instrumentally irrational, say, and so to desire most strongly to stop and buy some chocolate, whereas before he had the glimpse he had no such desire. If that is right then John might well do something to prevent himself from ever having that glimpse just a moment before he 9

  has it. He might look the other way, or shield his eyes with his hands, or distract himself, or whatever. But though, in such a case, it would be natural to describe John as having pulled himself together at the very moment of vulnerability, the fact is that we imagine John’s pulling himself together an instant before he would otherwise have lost control. It is therefore a case of diachronic, not synchronic self-control, and, as such provides us with no embarrassing contradiction in our description of what John does. We can simply imagine John wanting most to prevent himself from having a glimpse of the supermarket, and acting on this desire, an instant before his desires would otherwise have changed. What cannot be imagined, however, is a case in which an agent, at one and the same time, both wants most to act in one way—the way required for a loss of control—and yet wants even more to act in another way—the way required for an exercise of self-control. John, for example, cannot at one and the same time both want most to eat chocolate and want even more to prevent himself from eating chocolate, for that is an out and out contradiction in our description of John. Sceptics about synchronic self-control quite rightly emphasise this point. The question we must ask, however, is whether the exercise of properly synchronic self-control requires any such thing. An example might help. Suppose, as seems common enough, that John exercises self-control by having certain thoughts or engaging in certain imaginings. At the very moment at which he wants most to eat a chocolate, suppose he thinks of the chocolate that he is about to eat as a lump of fat, and that he imagines that lump of fat curdling in his stomach after he eats it. This imaginative exercise might well prevent John’s desire to eat a chocolate—which by stipulation is the strongest desire he has at that instant—from having its characteristic effect. The exercise of self-control counts as synchronic, because it happens at the very time at which John desires most to eat a chocolate. The question we need answered, however, is what the cause of those imaginings is supposed to be. The strongest desire John had at the very moment at which he engaged in those imaginings didn’t cause it, that’s for sure. For, by stipulation, what he wants most to do at that very moment is to eat a chocolate, and his imaginings simply undermine that. Nor did the strongest desire he had an instant earlier cause it either. For we can plainly imagine that what he most wanted to do an instant earlier was also just to eat a chocolate. But in that case, what did cause those imaginings? The answer seems to me to be implicit in what we said at the very beginning about what the “should” means when John says that he knows he shouldn’t eat so much chocolate. Remember, John needs to exercise self-control because, though he believes that he shouldn’t eat chocolate, he wants most to do so. As I have already 10

  - argued, this belief of John’s—his belief that he shouldn’t eat chocolate—is itself in turn simply the belief that he would most want himself not to eat chocolate, in his present circumstances, if he was fully rational: that is, if he was fully informed; if he had the set of desires he would have if he were fully instrumentally rational; if his desires as a whole formed a maximally coherent and unified set; and so on and so forth. John’s problem, then, is that the strongest desire he actually has fails to accord with the strongest desire he believes he would have in this more fully rational and coherent state of mind. But now consider the following two psychologies. One comprises both an agent’s belief that he would want himself, in his present circumstances, to act in a certain way if he had a maximally informed and coherent set of desires and, in addition, a desire of his to act in that way. The other psychology comprises his belief that he would want himself, in his present circumstances, to act in a certain way if he had a maximally informed and coherent set of desires but does not comprise, in addition, a desire of his to act in that way. Perhaps it comprises indifference, or aversion to acting in that way. What can we say about these two psychologies, from just what we have said about them so far? The answer seems plain enough. What we can say is that the first psychology exhibits more in the way of coherence than the second. The mere fact that agents fail to have desires, as regards what to do in their present circumstances, that they believe they would have if they had a maximally informed and coherent set of desires, itself constitutes a kind of incoherence, or disequilibrium, in their psychology. It constitutes a kind of incoherence or disequilibrium because these agents fail by their own lights. Note that this is not to say that the desire that they believe they would have if they had a maximally informed and coherent set of desires would indeed be an element in such a set. The point is rather that one source of incoherence in the psychology agents have lies in the mismatch between their desires about what they are to do in their present circumstances and their beliefs about what they would want themselves to do, in these circumstances, if they had a maximally informed and coherent set of desires. The fact that this is so is in turn very significant. For it is independently plausible to suppose that rational agents possess a quite general nondesiderative capacity to acquire and lose psychological states in accordance with norms of coherence (Pettit and Smith, 1996). It is rational agents’ possession of this capacity that explains why, for example, they tend to acquire beliefs that conform to the evidence available to them. Moreover, it also explains why, when they do not acquire such beliefs, they take themselves to be liable to censure and rebuke. Rational agents quite rightly feel shame when they fail to believe in accordance with the evidence available to them because, 11

  given that the evidence dictates that belief, norms of coherence entail that they should have acquired the belief, and because, in the light of the fact that they possess the capacity to acquire the belief, they could have acquired it. They therefore rightly feel shame because they failed to acquire a belief that they should and could have acquired. Similarly, rational agents’ possession of the quite general non-desiderative capacity to acquire and lose psychological states in accordance with norms of coherence explains why they tend to acquire desires for the believed means to their desired ends and why, when they do not acquire such desires, they likewise take themselves to be liable to censure and rebuke. Rational agents quite rightly feel shame when they fail to desire the believed means to their desired ends because, given that coherence augurs in favour of the acquisition of such desires, they should have acquired them, and because, in the light of the fact that they possess the capacity to acquire these desires, they could have acquired them. They therefore rightly feel shame because they failed to acquire desires that they should and could have acquired. If I am right, however, that agents who believe that they would desire themselves to act in a certain way, in their present circumstances, if they had a maximally coherent psychology, but then fail to have a corresponding desire, display a lack of coherence in their psychology as well, then it seems to follow that the capacity rational agents possess to acquire psychological states in accordance with norms of coherence has the potential to explain not just why their beliefs tend to evolve in conformity to evidence, and their desires in conformity to their desires for ends and beliefs about means, but also why their desires as regards what they are to do in their present circumstances tend to evolve in conformity to their beliefs about what they would want themselves to do, in their present circumstances, if they had a maximally informed and coherent set of desires. In agents who never fail to exercise this capacity we might well expect to find that their beliefs about what they would want themselves to do, in their present circumstances, if they had a maximally informed and coherent set of desires cause, or at any rate causally sustain, their having of corresponding desires quite generally. They never need to exercise self-control because they never find themselves believing that they would desire themselves to act in one way, in their present circumstances, if they had a maximally informed and coherent set of desires, while yet failing to desire to act in that way. But in agents who aren’t as superhumanly coherent as that, we might well expect to find that they possess back-up capacities, capacities that enable them to get back on track when their desires fail to match their beliefs about what they would want themselves to do, in their present circumstances, if they had a maximally informed and coherent desire set. We might expect them to be 12

  - disposed, for example, to engage in certain processes of thought or imagination which prevent their divergent desires from having their characteristic effect, and which cause them to have desires that lead them to do what they would have wanted themselves to do if they had had a maximally informed and coherent set of desires instead. Here, then, lies the explanation of how, despite the fact that he wants nothing more than to eat chocolate, John can none the less imagine chocolate to be a mere lump of fat curdling in his stomach. He can engage in that imaginative exercise because he possesses a quite general non-desiderative capacity to acquire and lose psychological states in accordance with norms of coherence, and that capacity, when exercised, amounts to no more or less than his thinking such thoughts and engaging in such imaginings as will restore his psychology to a more coherent state given its present state of incoherence. In short, John imagines chocolate to be a lump of fat curdling in his stomach because he is a rational creature and that is a rational thing for him to imagine at that time. It is a rational thing for him to imagine at that time because it causes him to lose his desire to eat chocolate—the desire whose presence makes for incoherence in his psychology—and instead causes him to desire most to do what he believes he would most want himself to do if he had a maximally informed and coherent and unified set of desires. In this way John restores coherence to his psychology. The fourth puzzle about self-control arises because of the apparent logical impossibility of properly synchronic exercises of self-control: that is, cases in which agents both need to exercise self-control and succeed in doing so at the very moment of vulnerability. These cases seemed to be impossible because their description seemed to involve a contradiction. It looked like agents had to both want most to act in the out-of-control way, and yet to want even more to act in the way required for the exercise self-control. The solution to this puzzle lies in the fact that exercises of properly synchronic self-control aren’t caused by desires and means–end beliefs. When agents exercise properly synchronic selfcontrol they engage in various thought processes and imaginings that are caused by a non-desiderative capacity they possess, in particular, by the capacity to acquire psychological states in accordance with norms of coherence. 5. W         -,   

I have suggested that despite the fact that John wants nothing more than to eat chocolate, he can still exercise self-control. He has the capacity to do so, 13

  even though he might fail to exercise that capacity. But what exactly makes it true that John can exercise self-control when he fails to exercise it? What is the difference between an unexercised capacity for self-control and no capacity for self-control at all? This is the fifth puzzle about self-control. In “Can and Can’t” Tony Honoré offers the makings of an answer. Honoré suggests that we need to distinguish between two “cans”: “can” (particular) and “can” (general). In these terms, to say of John that he can (particular) exercise self-control is, according to Honoré, to say of him that he does or will exercise self-control. “[S]uccess or failure, on the assumption that an effort has been or will be made, is the factor that governs the use of the notion. If the agent tried and failed, he could not do the action: if he tried and succeeded, he was able to do it. If he will fail however hard he tries, he cannot do it; if he will succeed provided he tries, he can”. (Honoré, 1964: 144)

What makes it true of John that he can (particular) exercise self-control is thus that he succeeds in doing so. His failure to exercise self-control suffices for the truth of the claim that he can’t (particular). Plainly, this can’t be what makes it true of John that he can exercise self-control when he fails to exercise that capacity. This is where Honoré thinks that we need to appeal to the idea of what an agent can (general) do. He suggests that “can” (general) is most commonly used in connection with types of performance in order to claim a general competence, or ability, or skill. “[A] condition sufficient for asserting ‘he can (general) do such-and-such a type of action’ is that, when the agent tries, he normally succeeds in doing an action of that type. . . . ‘Can’ (general), when used of particular actions, differs from ‘can’ (particular), when used of particular actions, in that its correct use does not depend on actual or prospective success or failure.” (Ibid.: 145–6)

Thus, according to Honoré, the sense of “can” in which we are interested— the sense in which John can exercise self-control, though he might fail to do so—is presumably the sense of “can” (general). What makes it true of John that he can (general) exercise self-control is thus that he usually does exercise self-control when he tries to do so. Though, as we will see, I have some sympathy with Honoré’s idea, it seems to me that we must reject his suggestion as it stands. True enough, our best evidence for the truth of the claim that John can exercise self-control, when he fails to do so, is very often that he usually succeeds when he tries. But it is hard to believe that the truth of the claim that John can exercise self-control, when he fails to do so, consists in the fact that he usually does when he tries. It is hard 14

  - to believe for the simple reason that John’s possession of the ability to exercise self-control doesn’t seem to require that he exists for longer than a moment. In order to see that this is so, consider the following thought experiment. Imagine that we have the technology to create an exact replica of John in a laboratory, a molecule for molecule duplicate. Suppose that the replication process takes place, but a few moments after the process is complete, through some mishap or other, the replica is destroyed. To my mind, so long as we are convinced that the creature we created really was a molecule for molecule replica of John, we wouldn’t hesitate for a moment in ascribing to him all of the psychological states and capacities that we would be willing to ascribe to John himself, including John’s capacity for self-control. Thus, suppose that during the moments the replica existed we offered both John and his replica a piece of chocolate and that they both ate it. Let’s suppose further that we are willing to say of John, at that time, that he could have exercised self-control by refraining from eating the chocolate and that, what’s more, this is true. It seems to me that we should then be willing to say exactly the same thing of John’s replica, and that what we say of John’s replica should in that case be true too. But since John’s replica doesn’t exist long enough for it to be true of him that he normally succeeds in exercising self-control on such occasions when he tries, it follows that what makes it true to say of him that he could have exercised self-control by refraining from eating chocolate cannot be that he normally succeeds in exercising self-control on such occasions when he tries. And, in that case, when we say of John that he could have exercised self-control by refraining from eating the chocolate on that occasion, it cannot be that that is what makes what we say of him true either. Honoré thus seems to me to be wrong that the sense of “can” in which we are interested—the sense in which John can exercise self-control, though he might fail to do so—is the sense of “can” (general). But if Honoré is wrong, then we might begin to wonder whether any coherent sense can be made of the idea that John can exercise self-control, and yet fail to do so, at all. Perhaps we should conclude instead that our commonsense assumption that there is a difference between an agent who has, but fails to exercise, a capacity for self-control, and an agent who has no capacity for self-control at all, is simply mistaken, and revise our allocations of moral and legal responsibility accordingly. However this would, I think, be premature. The sense of “can” can coherently be spelt out (Smith, 1997). We want to know what the difference is between, on the one hand, the possible agent who has the capacity to exercise self-control by performing an action, but fails to exercise that capacity, and, on the other hand, the possible agent who has no such capacity, and who therefore fails to perform the action 15

  because he could not have performed it. My suggestion is that the difference between these two agents consists in the differential similarity relations that obtain between the possible worlds in which they each fail and the possible worlds in which they succeed. Roughly speaking, the first of the two possible agents possesses an intrinsic feature which makes the possible world in which he succeeds in performing the action more similar, relatively speaking, to the possible world in which he fails, than the possible world in which the second of the two possible agents succeeds in performing the action is to the possible world in which he fails. This intrinsic feature possessed by the first agent is in turn, I suggest, what his capacity for self-control consists in. For this is what explains why it is possible (relatively speaking) for him to succeed in exercising self-control, whereas it is not possible (relatively speaking) for the second of two agents to succeed in exercising self-control (Smith, forthcoming). Note that this suggestion explains why the capacity to exercise self-control comes in degrees. For two agents may be alike in that they both possess the capacity to exercise self-control and yet differ in that the one may find it easier to exercise self-control than the other. Suppose we fix on a possible world in which three agents fail to perform some action. It might be true of two of them that they possess an intrinsic feature which makes the possible world in which they succeed in performing the action in question more similar, relatively speaking, to the possible world in which they fail, than is the possible world in which the third agent succeeds to the possible world in which he fails. Yet it might also be the case that the possible world in which one of the two succeeds is more similar to the world in which he fails, relatively speaking, than the possible world in which the other succeeds is to the possible world in which he fails. This is why sanction and blame come in degrees as well. It would plainly be unfair to sanction or blame, to the same extent, two people who differ in the crucial respect that one of them is more like someone who shouldn’t be sanctioned or blamed at all. The attraction of this suggestion should be plain. The difficulty is to explain what makes claims about our capacities true or false, and the solution is to suppose that what makes claims about our capacities true or false is exactly the same sort of thing that makes any other modal claim true or false: namely, facts about the similarities that obtain between possible worlds. The idea of an unexercised capacity for self-control, as opposed to a lack of self-control, therefore turns out to be no more mysterious than the idea of a possible world in which there is something that is really quite similar to the way that that very thing is in another possible world, as opposed to a possible world in which there is something that is not similar at all to the way that that very thing is in another possible world. 16

  - I said above that I had some sympathy with Honoré’s suggestion that to say of someone that he has, but fails to exercise, a capacity for self-control is to say of him that he usually succeeds in exercising self-control. The reason that I am sympathetic should now be clear. If the modal facts I have described are what makes claims about our abilities true or false, then it comes as no surprise that the best evidence that we have for the truth of some particular claim to the effect that someone or other can exercise self-control, when he fails to do so, will often be that that person usually does succeed in exercising self-control when he tries to do so. Regular patterns in actuality are, quite in general, what provide us with such evidence as we have for the similarities and differences that obtain between the actual world and other possible worlds. But we must not let this epistemological point obscure the metaphysics of capacities. What makes claims about our capacities true or false are the modal facts I have described, not the regular patterns in actuality that would provide us with evidence of those modal facts. Here, then, lies the solution to the fifth puzzle about self-control. The difference between an agent who has, but fails to exercise, a capacity for selfcontrol and another agent who has no capacity for self-control at all, lies in the relative nearness or remoteness of the possible worlds in which such agents succeed in exercising self-control from the possible worlds in which they fail to exercise self-control. To repeat, the best evidence for the nearness or remoteness of such possible worlds lies, much as Honoré suggests, in whether or not the agents in question usually succeed in exercising self-control when they try. But this claim about our evidence for the truth of ascriptions of the capacity for self-control must not be offered as a substitute for what makes such ascriptions true.

C

I said at the outset that my aim was to raise, and hopefully to answer, some of the difficult questions that arise given that we restrict sanctions and blame to those who have rational control over their conduct. The main conclusions can be summed up as follows. Though possession of an intention suffices for agents to be in control in one sense, it does not suffice for their being in control in another, and more important, sense. For agents are in control in this more important sense when their intentions are suitably responsive to their deliberations, that is, to their reflectively formed beliefs about what they would want themselves to do if they were fully rational. The capacity for self-control can thus be seen to embody 17

  this responsiveness. It is the capacity rational agents possess to have desires corresponding to those they believe they would have if they were fully rational, a capacity which, in turn, is an instance of a more general capacity they have to acquire and lose psychological states in accordance with norms of coherence. Armed with this definition of the capacity for rational self-control we can define the idea of someone who, though capable of acting intentionally, remains a victim of circumstance. An agent who, though capable of acting intentionally, remains a victim of circumstance is someone who, on the one hand, has the capacity to act on his desires and intentions, but is also someone who, on the other hand, has desires and intentions that are beyond the reach of the capacity he has for rational self-control. The limits of an agent’s capacity for self-control is in turn fixed by the nearness or remoteness of the possible worlds in which he succeeds in exercising self-control from actuality. This explains why the capacity for rational self-control comes in degrees. The distinction between synchronic and diachronic exercises of selfcontrol is, however, crucial at this point. For an agent to truly lack selfcontrol, and hence to be truly a victim of circumstance, more must be true of him than that he has a desire which is he is unable to conquer synchronically. In other words, more must be true than that no feat of the imagination or thought which was within his reach at the time could have stopped the desire from having its effect. The desire must also be one which the agent could not reasonably have foreseen that he would have at a time at which he was not out of control. An agent who could have foreseen that he would be out of control if he were to find himself in certain circumstances in the future, and who failed to take such steps as were available to him to ensure that those circumstances did not arise, though he may well have a desire that is beyond the reach of his capacity for synchronic self-control, does not have a desire that is beyond the reach of his capacity for diachronic self-control. Such an agent is thus not a victim of circumstance, notwithstanding his inability to exercise synchronic self-control. R Bratman, Michael (1987) Intentions, Plans and Practical Reason (Cambridge, Mass.: Harvard University Press). Davidson, Donald (1970) “How is Weakness of the Will Possible?” reprinted in his Essays on Actions and Events (Oxford: Oxford University Press, 1980), 21–42. Frankfurt, Harry (1971) “Freedom of the Will and the Concept of a Person” reprinted in Gary Watson (ed.), Free Will (Oxford: Oxford University Press, 1982), 81–95

18

  - Hobbes, Thomas (1651) Leviathan (Harmondsworth: Penguin, 1968). Honoré, Tony (1964) “Can and Can’t” reprinted as “Appendix: Can and Can’t” in his Responsibility and Fault (Oxford: Hart Publishing, 1999), 143–60. —— (1998) “Being Responsible and Being a Victim of Circumstance” reprinted in his Responsibility and Fault, 121–42. Kennett, Jeanette and Michael Smith 1996. “Frog and Toad Lose Control”, Analysis 56, 63–73. Pettit, Philip and Michael Smith (1996) “Freedom in Belief and Desire”, Journal of Philosophy 93, 429–49. Smith, Michael (1994) The Moral Problem (Oxford: Basil Blackwell). —— (1995) “Internal Reasons”, Philosophy and Phenomenological Research 55, 109–31. —— (1997) “A Theory of Freedom and Responsibility” in Garrett Cullity and Berys Gaut (eds), Ethics and Practical Reason (Oxford: Oxford University Press), 293–319. —— (2000) “Quelques énigmes concernant le contrôle de soi”, Philosophiques (27), 287–304. —— (forthcoming) “Rational Capacities”, in manuscript. Watson, Gary (1975) “Free Agency” reprinted in Gary Watson (ed.), Free Will , 96–110.

19


2 THE CAPACITY TO HAVE DONE OTHERWISE: AN AGENT-CENTRED VIEW Philip Pettit*

I

Whenever we hold someone responsible for a given action we assume that in some sense they could have done otherwise. Perhaps, unbeknownst to them, they would have been forced to perform that action had they not chosen to perform it voluntarily; perhaps they would have been induced by neuro-scientific or hypnotic or strong-arm tactics to act in that same way (Frankfurt, 1988). But even in such a case they could presumably have tried to do something else instead (Otsuka, 1998; Fischer, 1999). And were that not so then it is hard to see how we could continue to hold them responsible for what they did. Not only does the fact of holding someone responsible commit us to believing in this sense that they could have done otherwise. Believing that they could have done otherwise also involves us in believing that they were responsible, at least in some measure, for what they did; they were responsible in the sense that there is nothing wrong about praising them or blaming them for what happened. If the agent could have done otherwise, as we naturally understand that condition, then there is no ground on which they can avoid having the action put down to their credit or discredit.1 I shall say no more here to defend the assumption that responsibility for an action and the capacity to have done otherwise go together. Assuming the linkage alleged, the paper is concerned with the question of how we should understand the capacity to have done otherwise. The standard line is that it * My thanks to Victoria McGeer, François Schroeter and Michael Smith, and to Peter Cane and John Gardner, for useful comments on an earlier draft. 1. The linkage between being responsible for an action and having the capacity to have done otherwise is quite consistent, notice, with responsibility for other matters—say, responsibility for negligent inaction—not involving any parallel capacity. And it is quite consistent, of course, with something less than responsibility proper—say, strict liability under a regime of law—not involving such a capacity. Those are separate topics and require separate treatment.

  must consist in a special feature of the way the action was produced. I think that this act-centred line has proved unproductive and I want to explore instead an approach that would equate it with a more general, agent-centred capacity. The approach is in the spirit of Tony Honoré’s (1999: 11) claim that “in morals and law capacity means a person’s general capacity to perform successfully”; it draws on a point of view that I have tried to develop more fully in a recent book (Pettit, 2001; see, too, Pettit and Smith, 1996). This paper is in three main sections. In the first I look at the problems that bedevil approaches that give the capacity to have done otherwise an actcentred reading, and argue that they all fail at least one of two intuitive constraints, one naturalistic, the other normative. In the second section I introduce the alternative approach, according to which the capacity is agent-centred in character. And then in the third section I argue that this agent-centred approach promises to satisfy both of the constraints introduced earlier. The general line defended is summed up in a brief conclusion. 1. T - :  

The standard approach to explicating the capacity to have done otherwise assumes that the agent displays this capacity in virtue of the way the action arose. The idea is that whatever was true of the agent in other respects, the process leading to the action satisfied a certain distinctive sort of condition and that the satisfaction of this condition makes it the case that he or she could have done otherwise. All versions of the standard approach agree on that central assumption. They diverge from one another only when it comes to addressing the further question as to exactly what act-centred condition makes it the case that the agent could have done otherwise. The big divide that appears as soon as that question is asked is the division between approaches that are compatible and approaches that are incompatible with the truth of determinism (Watson, 1982). Incompatibilists argue that for it to be the case that the agent could have done otherwise, it must be that the laws which govern the natural world, together with the history of the world up to that point, left open the possibility that the action-generating process should have had a different output. There must be a gap in the causal order of the world such that nothing that happened before the action can have made it inevitable that the process led to precisely that upshot (Van Inwagen, 1983). Compatibilists hold, by contrast, that this is too strict and that all that is required is something weaker: viz., that the sorts of antecedents that we spontaneously treat as hostile to agency—antecedents like natural obstructions, 22

     :  -  psychological compulsions, idées fixes and the like—should not have ensured, all on their own, that the process led to a particular result. The action will have been fully determined by its antecedents taken as a whole—antecedents that at the psychological level will include the agent’s normal beliefs and desires—but it will not have been fully determined by hostile antecedents alone. The agent will not have been robbed of his or her decision-making capacity by the intrusion of such “constraining causes” (Ayer, 1982: 21). There are two constraints that any account of the capacity to have done otherwise should satisfy and all versions of the standard approach, compatibilist and incompatibilist, fail to meet one or the other of these. So, at any rate, I shall argue. The first constraint is that the account should be consistent with what we may describe as a naturalistic picture of the universe. A picture of the universe will count as naturalistic just so far as none of the entities or forces that it posits are add-ons to those entities and forces—whatever they are—that are recognised in the scientific image of the world.2 None of them is of a kind with the res cogitans postulated in Descartes’ theory of mind, for example, or with the vis vitalis posited in early theories of life. The constraint holds that an account of the capacity to have done otherwise should be naturalistic in this sense; it should not force us to reject the image of the world projected in natural science. The second, normative constraint is that the account should also make sense of why someone who could have done otherwise is thereby taken as fit to be held responsible for it in some measure, being subject to praise or blame for what was done. The capacity to have done otherwise is associated with the agent’s being fit to be held responsible, as we saw earlier, and the account must therefore explain why it is permissible to praise or blame someone who possesses the capacity for doing what they did. It will be permissible to praise or blame them in this sense so far as it is permissible, not just to make an assessment of what they did—this could be quite detached—but to react to them in the positive mode of gratitude or appreciation or in the negative mode of resentment or indignation (Strawson, 1982).3 2. How to demarcate the entities and forces that can be recognised in the scientific image? I equate them with the entities and forces that are guaranteed to figure in the world by virtue of the way things are in the microphysical realm postulated by physics (Pettit, 1993b). But other lines are also possible on that demarcation problem—see Jackson (1998)—and we do not need to judge between them here. 3. In the case where an agent acts on behalf of a principal, in particular on behalf of a group, it may be unclear who should be the target of such a reaction, the individual or the collective. My own view, defended in Pettit (2001), is that even if the collective is the entity that ought to be praised or blamed for what was done through its agent, still that agent will be properly subject to praise or blame—assuming that nothing like duress or coercion was involved—for having done what the group required.

23

  Let us consider incompatibilist accounts in the light of these constraints. One such account is naturalistic in character, arguing that natural laws may leave it to objective chance whether one or another event occurs at a given time; the laws may ensure that A does not occur, while leaving it possible that either of B or C occur. Such an indeterministic picture might make it true, for some actions, that the agent could have done otherwise: it might allow that though the agent performed an action, B, the laws of nature and the history of the world up to that point were equally consistent with the actiongenerating process leading to C instead. But while this incompatibilist account would satisfy the naturalistic constraint, it would clearly fail the normative one. For why would the fact that an action was underdetermined by the laws and history of the natural world make it permissible to praise or blame the agent for performing it? All that the story implies is that the action was the product of objective chance and that scarcely permits us to lay it at the door of the agent. What incompatibilists clearly need to do is not just to deny that the laws of nature uniquely determined what was done by a responsible agent but also to postulate that it was determined within the agent in a way that implies responsibility: the action was produced and owned by the agent in a distinctive manner. The problem for this version of indeterminism, however, is how to give a naturalistic account of such “agent-causation” (Chisholm, 1982). Under any naturalistic story, the only causes in the natural world will be causes that operate in virtue of the natural laws that are operative there, and the causation whereby an agent is held to close the gaps left by indeterministic, natural laws cannot itself belong in that naturalistic category. This variety of incompatibilism will violate the naturalistic constraint. Compatibilist accounts of the responsible agent’s ability to have done otherwise, where this is also understood as a special, act-centred capacity, run into parallel difficulties. Such accounts are well-fitted to satisfy the naturalistic constraint but none of them succeeds in meeting the second, normative constraint. They identify the capacity to have done otherwise with a naturalistically intelligible, act-centred condition but none of the conditions canvassed makes sense of why it is permissible to praise or blame the agent for an action performed when it obtains. The usual compatibilist account is hypothetical in character. It says that an agent could have done otherwise at a given time just in case he or she would have done otherwise if a certain condition had been fulfilled. One line takes the condition to be that had the agent chosen, then he or she would have done otherwise (Moore, 1911; Ayer, 1982). But this is no good. Choosing or willing or any such cognate is an action, so that there will always be a question as 24

     :  -  to whether it itself satisfies that condition, and this question will open up an indefinite regress. The now more standard line takes the condition to be that had the agent not desired to act in that way, then they would not have done so (Davidson, 1980). This proposal is not subject to the same difficulty as the one that invokes choice, since desiring is not an action and the question does not arise as to whether it itself is a free action. This standard proposal, however, is problematic in another respect. For all the condition about desire stipulates, it may be that in order for the agent’s desires to have gone the other way, the conditioning and drilling to which the agent was subject as a child would have had to be different from what it actually was (Chisholm, 1982). Perhaps the agent could have done otherwise, in the sense that he or she would have done otherwise in the event of having a different desire. But it may be that the agent could not have desired otherwise; it may be that the desire that actually produced the action was more or less hard-wired into the agent’s make-up. The problem involved is, of course, that under this compatibilist account of the capacity to have done otherwise, it is not intuitively permissible to praise or blame the agent for an action. If an agent does something under the influence of a desire that is more or less hard-wired into their make-up—and this, through no fault or virtue of their own—then they can hardly be praised or blamed for what is done. If a person is subject to such an unavoidable motive, then by ordinary criteria it will not be permissible to praise or blame them for acting on that motive. Can this problem be resolved by requiring that the desire from which the action issues be one that the agent desires to be moved by, as distinct from being a desire that operates willy-nilly (see Frankfurt, 1988)? I do not think so. For just as the first-order desire manifested in the action may have been hardwired into place, so the same may be true of the higher-order desire to be moved by that desire at the first-order level (Pettit, 2001). This short review of variations on the standard approach to our problem should be sufficient to explain why I am pessimistic about finding any that will answer to the naturalistic and normative constraints. No naturalistic account is likely to satisfy the normative constraint and no account that satisfies the normative constraint is likely to be naturalistic. Or so at least it seems. 2. I  - 

The alternative approach to explicating an agent’s capacity to have done otherwise starts from quite a different reading of the remark: “X could have 25

  done otherwise”. Rather than assuming that that remark is meant to direct us just to something about the way the action was generated within the person, it suggests that the intended interpretation bears on the sort of agent that X more generally is. A good way to introduce this alternative reading may be to consider the import of parallel remarks in other contexts. Think of a mechanic who reflects on how two cars did in a particular race and says that one performed up to its limits while the other was capable of achieving a better time. Or imagine a horse trainer who considers the performance of two animals in a dressage event and says that while the first gave of its best, the second could have done better. Or think of an engineer who says that while two missiles did equally well in homing in on a target, one of them—a smart missile, let us suppose— could have achieved a greater level of accuracy. The brunt of such remarks is to comment on the cars or horses or missiles in general terms, not to reflect on something that holds just in virtue of how the performance was generated. What we are told in each case is that one of the pair in question has a general capacity that is lacking in the other and that the similar performances of the members of each pair do not reflect this difference in capacity. To say that one car or horse or missile could have done better is to say that it is a better car or horse or missile and that if this was not reflected in actual performance, then that was due to an accident of circumstance. It is to suggest that we should not judge the entities in question—in particular, we should not judge their general capacities—on the basis of this performance. That performance was not typical; it was not one in which the car or horse or missile performed to type. The agent-centred approach to explicating the responsible agent’s capacity to have done otherwise suggests that something similar is true in this case. When we say of an agent that he or she could have done otherwise, so the idea goes, we are presupposing the relevance of certain background standards and, in the case where the agent fails to meet those standards, we are saying that this failure was not typical. The agent could have done otherwise, we remark, intending to convey the thought that he or she is capable of better. The actual choice they made may have fallen away from those standards but this should not be taken as indicative of the sort of agent they are; it should be put down to the influence of a more or less incidental factor. It need not be the case under this reading that there was any point in the process leading to the action where we can see the presence of something we might describe as volition. For all that is implied, the process leading to the action performed may have made it absolutely inevitable that the agent should have acted as they did. The remark that the agent could have done 26

     :  -  otherwise, interpreted on the proposed lines, means only that an incidental feature of the circumstances played an important role in the process leading to action and that the behaviour would have been different—in particular, it would have been more typical of the agent’s character—had that feature been absent. The incidental feature that played this role may be an event that distracted the agent’s attention, or the presence of passion or fatigue or boredom, or just a glitch in the way the agent’s memory worked. The possibilities are endless, though how we enumerate them—what we count as perturbers that interfere with the exercise of a capacity, without undermining its existence— will depend on our background view of agents in general and that agent in particular.4 This may explain what is meant by saying in the case of failure to meet certain standards that an agent could have done otherwise. But what of the case of success? Suppose that the agent does act to type and satisfies the relevant standards. What does it mean to say in that event that the agent could have done otherwise? We might say it means that they would or might have acted otherwise had an incidental feature of some kind thrown them off form. But while that would mirror the story told for the case where the agent fails, it would not give us an intuitive account of the content of the remark. When we say that someone who does well could have done otherwise we surely mean to convey something positive about the agent, not merely the negative message that they might have been put off their stroke by this or that perturbation. Is it possible to vindicate that intuition within the agent-centred approach? I think that this is possible. There are two interpretations of what it means for agents to be disposed to act so as to satisfy certain standards—to track certain standards—and we have only been allowing for one. Under a first, weaker interpretation it means that the agents are disposed to act in those ways that, as it happens, are in line with the standards. Under a second, stronger interpretation it means that they are disposed to act in line with the standards, whatever the standards should happen to require: they are actually disposed to act in the ways that, as it happens, are in line with the standards, but had the standards required different modes of action then they would have been disposed in that counterfactual event to act in those different ways. The agents in the first case are cued to the behaviours that happen to satisfy the 4. Some may even think that the required capacity can remain in place consistent with more or less continuing failure. They may think that the type to which we assign the agent is not to be determined by empirical performance, or not just by empirical performance, but by an independently sourced sense of counterfactual possibility.

27

  standards, the agents in the second case are cued to the standards themselves: they are aware of them as standards that they should meet; they are reliably disposed to enact whatever they take the standards to require in any context; and they are reliably disposed to interpret the demands of the standards correctly. If we stipulate that agents have the general capacity to meet certain standards in the second, stronger sense, then it turns out that we can get over the difficulty raised. Suppose that an agent does well on a particular occasion, manifesting the general capacity envisaged. If we think that he or she has that capacity in the strong sense, then the remark that they could have done otherwise need not mean that they might have been thrown off form by an incidental factor. It can mean that their satisfying the standards is no accident. They would have acted so as to satisfy the standards even if the standards had required a different form of behaviour. They are explicitly focused on the demands of the standards and they are reliably disposed to meet those perceived demands. The picture emerging, then, is this: if an agent does badly and we say that he or she could have done otherwise, then we mean to suggest that they have a general capacity to do better and that the failure should be treated as an accident, not as something typical. If an agent does well and we say that he or she could have done otherwise, then we suggest in similar vein that they have a general capacity to do well—they would have done well even if the standards had required a different response—and that the success should not be treated as an accident but as something typical of them. To say that an agent could have done otherwise is always to speak about the agent involved in the act, not just about the act itself, and it is always to make a positive comment about that agent. 3. E  - 

How is such an agent-centred account likely to fare with the naturalistic and the normative constraints? Does it postulate a capacity that we can imagine a regular, naturalistically unmysterious creature possessing? And does it postulate a capacity such that any action done in the presence of the capacity is one for which it is permissible to hold the agent responsible: to praise the agent if the capacity present is actually exercised, to blame them if it is present but unexercised? We turn to those questions in this final section. The naturalistic constraint does not raise any distinctive problems for the agent-centred approach. There is no particular difficulty in giving a naturalistic account of a capacity or disposition, since nature is rich in propensities 28

     :  -  of that kind. True, the approach postulates a strong capacity to satisfy standards, which in turn requires an explicit awareness of standards and a disposition to try to meet their perceived demands. And true, there is no agreed account of what makes such normative sensitivity possible. But the challenge thereby raised for naturalistic approaches is not specific to the area of our discussion and need not concern us here. What, however, are the prospects of the agent-centred account meeting the normative constraint? Suppose that an agent is disposed to track relevant standards and so has the capacity, in the agent-centred sense, always to have done otherwise than he or she did. Does this make it permissible to praise or blame the agent for what they do on any occasion? Does it make it permissible to react to them with gratitude or resentment, appreciation or indignation? Some may say that this will be permissible, so far as praising or blaming someone serves to shape their behaviour; and that it will serve in this way, so far as people care about being praised or blamed for meeting relevant standards: that is, pursue the good opinion expressed in praise and flee the bad opinion expressed in blame. The idea is that the susceptibility to praise and blame of the person who tracks certain standards—the person who has the capacity, in the agent-centred sense, always to have done otherwise—will make it useful and therefore permissible to subject them to the influence of praise and blame. This response makes the activity of holding someone responsible counterintuitively strategic and manipulative (Strawson, 1982). The posture of holding someone responsible for action will be of a kind with our disposition towards the dog—or perhaps the young child—when we expose it to rewards and penalties that are designed to shape its behaviour. If we praised or blamed an agent only because of hoping to reinforce or alter their behaviour in such a manner—and were this a matter of common knowledge, as it inevitably would be—then praising or blaming someone would be a highly disrespectful act and would be a reasonable ground for their resenting us. Praising or blaming a person is intuitively respectful in character, involving an acknowledgement of their agency and autonomy, and whatever makes the activity permissible, it cannot just be people’s susceptibility to the shaping effects of praise and blame. The basis of permissibility must be more subject-friendly than that. This observation provides the cue for what I think is the right answer to the question before us. For the uniquely subject-friendly basis on which it might be permissible to praise or blame an agent is that the agent gave his or her permission for this to happen. And it turns out that we can identify such a basis of permissibility under an intuitive development of the agent-centred account. 29

  Suppose that agents represent themselves to others as tracking certain standards, where this representation is overt: it is a matter of common knowledge among the parties involved that this is what is happening. If agents do this, then it is equally going to be a matter of common knowledge that others will expect them to meet those standards and may act out of reliance on their doing so; we would not give credence to the agents’ representation of themselves as tracking standards, after all, if they expressed surprise at others’ reacting in this way. But if agents overtly represent themselves to others as tracking certain standards, despite its being a matter of common knowledge that others will therefore form and act on corresponding expectations, then they presumably acquiesce—again, as a matter of common knowledge—in others holding them to those expectations; this must be a matter of presumption so long as they do not reject reciprocity and community with the others involved. They license others to feel aggrieved about any failure on their part to satisfy the standards, and they license them to feel gratified by any success. They give others permission to blame them for failure and to praise them for success; they invite those responses by the way in which they represent themselves. None of this should be surprising. In overtly representing themselves as tracking the standards involved, the agents will not just have reported their attachment to those standards, as if this were a matter of merely idle interest. They will have avowed and committed themselves to those standards as criteria by reference to which others are permitted to assess them and are permitted to react, as appropriate, with gratitude or complaint, praise or blame (Bilgrami, 1998). In putting themselves forward as committed to the standards in question, they will have accepted the right of others to judge them on the basis of those standards; to react negatively or positively, depending on the quality of their performance, and indeed to expect this reaction to have some effect. The agent-centred account of the capacity to have done otherwise equates it with an ability to track certain standards. That ability is consistent with the naturalistic constraint on a satisfactory account, as we saw earlier, and it now begins to seem that it is consistent also with the normative constraint. The ability of agents to track certain standards will make it permissible to praise or blame them for what they do in the presence of that ability, provided that they avow those standards that they track. If this proviso is fulfilled then it will be permissible to praise or blame agents for what they do in the presence of that ability because the agents will themselves have licensed or permitted that reaction. But will the avowal proviso have to be fulfilled in those cases where we think it is permissible to praise or blame an agent, in particular an agent who 30

     :  -  could have done otherwise in the agent-centred sense? Or will it be possible for the agent to escape responsibility just by cancelling that avowal? I argue that there are standards relevant in all these cases such that it will not be possible for the agent to cancel the avowal. We would not think it permissible to hold agents responsible unless we saw them as potential, conversable interlocutors: unless we saw them as persons with whom we could in principle reason about how things stand and about what should be done. Agents who lay beyond the reach of reason and discourse in this sense would be at best like mute animals and would not present themselves as the sorts of creatures it makes sense to hold responsible. It is striking in this connection, after all, that we do not praise or blame mute animals in any serious sense: we do not react to them, or at least not if we are being sensible, with attitudes of resentment or gratitude. But if we regard an agent as conversable in the required sense then we must believe that they are disposed to track certain standards and indeed to avow those standards. So at any rate I argue. The standards that agents must be disposed to track, on pain of not really counting as conversable subjects, are those standards that people must fulfil under intuitively favourable circumstances, if they are to count as having intentional states like perceptions, beliefs and desires and if they are to count as expressing and enacting those states in the manner of a creature with whom we can reason. Other things being equal, conversable subjects must be disposed to form the belief that p on perceiving that p; to form the belief that q on coming to believe that p, where they already believe that if p then q; to form and act on the intention to X on coming to believe that by X-ing they can bring it about that r, where they already desire that r; and so on. In particular they must be disposed to respond in such ways by virtue of recognising that those responses are supported by relevant reasons, since otherwise they would not count as creatures with whom we could reason (Pettit, 1993a: ch. 2; McGeer and Pettit, 2001). Not only are they disposed to form the belief that p on perceiving that p, as a mute animal might do; not only are they disposed to be triggered into believing that p by the perception that p. They are disposed to recognise that the perceptual evidence supports the belief that p and, other things being equal, to be led by that recognition to form the belief that p. And so on in the other cases (McDowell, 1996). So much by way of illustrating the standards that people must be disposed to track, on pain of not really counting as conversable subjects. But conversable subjects have to be disposed to do more than just track such standards. They also have to be disposed to avow them as standards to which they can be held. 31

  When someone engages us in conversation, they have to put themselves forward as worthy of being addressed and worthy of being heard; there is no point in talking or listening to the wall. And this means that they have to avow those standards that any worthy interlocutor—any conversable subject— must generally be expected to satisfy. The interlocutor who proved indifferent to criteria of inductive evidence, logical consistency, or argumentative coherence would soon lose any hold on us. And this being a matter that is knowable in common to all, any interlocutor who aspires to connect with us —to reach our minds—must avow such standards as criteria on the basis of which we are entitled to assess and respond to their performance (Pettit, 2001; Pettit and Smith, 1996). We will treat subjects as conversable only so far as we think that they have the capacity to connect with us, or at least to connect with some others, in this way. And so we must treat such subjects as being disposed, not just to track the standards in question, but also to avow them as standards they embrace. It follows then that with any subjects that we take to be conversable—as we must take anyone whom we praise or blame to be conversable—we have to think that they are disposed both to track and to avow the standards of reason illustrated earlier. We must assume, not just that they will generally adjust as the standards of reason require, but that in any discursive engagement with others they will avow those standards as guidelines by which they can permissibly be judged. When we assume in this way that conversable subjects are disposed to avow standards of reason, what we assume is quite substantive. It is not just that such subjects are disposed in discursive exchanges with others to avow those standards as guidelines by which they can be judged in the course of such exchanges. Rather it is that they are disposed to avow those standards as guidelines by which they can be judged in the course of any performance, whether in discursive exchange with others or in non-discursive contexts. An interlocutor would have little claim on being treated as someone worth talking to if we thought that while he or she could do quite well within the confines of an exchange, they were not generally a creature of reason. This being so, the interlocutor who lays claim to being taken seriously has got to avow standards of reason as standards by which they can generally be judged. And if we take a subject to be conversable, therefore, we must take them to be disposed to avow standards of reason in that general way. The upshot of this line of argument is that the avowal proviso that I mentioned above is going to be reliably fulfilled with agents that are fit to be praised or blamed. Such agents will have to be conversable. Conversable agents will have to be disposed not just to track but also to avow the 32

     :  -  standards of reason illustrated. And being disposed to avow those standards, they will be equally disposed to give us permission to react to their performance—their performance generally—with feelings of resentment and gratification, blame and praise. They cannot put aside the disposition to avow the standards and so they cannot withdraw the disposition to permit such reactions. There is a difference, of course, between a person’s avowing standards of reason and actually giving us a licence to react with resentment or gratification and their just being disposed to do those things; but the difference does not make for a problem with the argument. For suppose that we react on the basis of the agent’s disposition alone, manifesting resentment or gratification, blame or praise. The agent will not be in a position to deny our title to react in that way, since they will have to admit that they were disposed to license such reactions and that this was indeed something that we were able—and that they were able to see that we were able—to discern. We may have been a little presumptious in reacting as we did but we were not presumptious in a degree that the agent could seriously condemn.

C

That an agent could have done otherwise in any action means, under the agent-centred account on offer, that the agent acted in the presence of a capacity to track standards of reason, in particular standards of reason that he or she is disposed to avow. And that an agent could have done otherwise in this sense, so it transpires, explains why it is permissible to praise or blame them for what they did. In particular, it explains why this is permissible without forcing us to regard them as creatures with capacities that transcend the resources of the natural world. The agent-centred account meets both the naturalistic and the normative constraints introduced at the beginning. The position defended offers an ontologically distinctive account of what makes true the claim that the agent could have done otherwise in a given choice. The truth-maker is not a discrete and punctual feature of the process in which the choice was generated. It does not reduce to the sort of thing postulated in those incompatibilist and compatibilist theories discussed in the first section: not to a distinctive sort of agent-causation, for example, and not to a desire that might have been otherwise than it was. It consists rather in the nature of the agent at the time of action: in the fact that he or she was possessed at that moment of a capacity to track those demands of reason that are avowed in any discursive relationship. 33

  This feature of the approach has an interesting implication. Consider two agents who each act under the immediate influence of certain beliefs and desires, performing more or less identical actions in more or less identical situations. According to the agent-centred account defended here, it may still be true that one agent could have done otherwise and the other not. The difference will consist in a difference in their general make-up, not in a difference in the specific aetiologies of their actions. It will consist in the fact that the one agent had the capacity to track the demands of reason and the other not, though that capacity may have played no causal role in generating the first agent’s behaviour. One question, in conclusion. If an agent had the capacity to do otherwise in a given choice, as that capacity is interpreted here, does this mean that the agent enjoyed the fullest degree of freedom in making that choice? It certainly means that the agent had the basic capacity to track the avowed demands of reason; the agent was compos mentis and faced a choice in which the demands of reason were relevant. But freedom in the full sense imposes two other sorts of requirements that may or may not have been satisfied in the case on hand (Pettit, 2001). First, if an agent is to count as fully free in a certain choice then their capacity to track the avowed demands of reason must not have been reduced in any measure—that is, rendered difficult but not impossible of exercise—by psychological factors like obsessiveness, compulsion, fatigue and the like. And, second, if an agent is to count as fully free in the choice then the demands of reason that the agent had the capacity to track must not have been primed or rigged by others—at least not primed in the negative manner associated with coercion—so as to lead the agent in a particular direction; things must not have been fixed against the agent’s will, for example, so that he or she faces a threat of punishment in the event of taking a certain action. Freedom in a basic sense may be assured by the presence of a capacity to track the demands of reason, as that has been elucidated here, but freedom will exist in full measure only where the capacity is unreduced and the demands unrigged.

R Ayer, A.J. (1982) “Freedom and Necessity” in G. Watson (ed.), Free Will (Oxford: Oxford University Press). Bilgrami, A. (1998) “Self-knowledge and Resentment” in B.S.C. Wright and C. Macdonald (eds), Knowing Our Own Minds (Oxford: Oxford University Press), 207–41.

34

     :  -  Chisholm. R.M. (1982) “Human Freedom and the Self” in G. Watson (ed.), Free Will, 24–35. Davidson, D. (1980) Essays on Actions and Events (Oxford: Oxford University Press). Fischer, J.M. (1999) “Recent Work on Moral Responsibility”, Ethics 110, 93–139. Frankfurt, H.G. (1988) The Importance of What We Care About (Cambridge: Cambridge University Press). Honoré, T. (1999) Responsibility and Fault (Oxford: Hart Publishing). Jackson, F. (1998) From Metaphysics to Ethics: A Defence of Conceptual Analysis (Oxford: Oxford University Press). McDowell, J. (1996) Mind and World (Cambridge, Mass.: Harvard University Press). McGeer, V. and Pettit, P. (2001) “The Self-regulating Mind”, Language and Communication 21. Moore, G.E. (1911) Ethics (Oxford: Oxford University Press). Otsuka, M. (1998) “Incompatibilism and the Avoidability of Blame”, Ethics 108, 685–701. Pettit, P. (1993a) The Common Mind: An Essay on Psychology, Society and Politics (pb edn 1996, New York: Oxford University Press. —— (1993b) “A Definition of Physicalism’, Analysis 53, 213–23. —— (2001) A Theory of Freedom: From the Psychology to the Politics of Agency (Cambridge: Polity Press). —— and Smith, M. (1996) “Freedom in Belief and Desire”, Journal of Philosophy 93, 429–49. Strawson, P. (1982) “Freedom and Resentment” in G. Watson (ed.), Free Will. Van Inwagen, P. (1983) An Essay on Free Will (Oxford: Oxford University Press). Watson, G. (1982) “Free Agency” in G. Watson (ed.), Free Will.

35


3 PRIVATE LAW AND PRIVATE NARRATIVES Arthur Ripstein*

Responsibility and Fault 1 collects Tony Honoré’s recent essays about the familiar form of responsibility that he calls “outcome responsibility”, the kind of responsibility that is invoked when an agent is held responsible for some outcome to which he or she has contributed. Outcome responsibility is a familiar and pervasive feature of human life—persons think of themselves, and of each other, in terms of the impact that they have in the world. It is difficult to imagine what human life would be like without it. Our conception of interpersonal interaction is shaped by it, as is our autobiographical conception of ourselves as persons. I apologise to you if I jostle you in the corridor, because your dropping the pile of papers is the result of my doing. In the quiet of my study, I take credit for solving this week’s crossword puzzle, or feel foolish for having yet again spilled coffee into my keyboard. As Honoré puts it, “This responsibility is an essential constituent of our character and identity, without which we would lack both achievements and failures” (76). This type of responsibility is the inevitable concomitant of the fact that when humans act, they make their way in the world, changing it in countless ways. Once Honoré has drawn our attention to it, the idea of outcome responsibility is familiar, and everywhere. The book’s main focus is on the way in which this familiar idea animates the law, particularly the law of torts.2 One of the central puzzles of tort theory can be posed as a question about outcome responsibility, and the particular * A draft of this essay was presented at a conference celebrating Tony Honoré at Columbia University in March 2000. Thanks to all present, especially my commentator Jeremy Waldron. Benjamin Zipursky, Emer O’Hagan, Dennis Klimchuk, John Gardner, and Peter Cane for comments on earlier versions of this paper, and to Jules Coleman and Avishai Margalit for helpful discussion. I trust that my debt to Tony Honoré will be apparent in what follows. Many thanks to the general editor and the review editor of the Oxford Journal of Legal Studies for so readily agreeing in advance that this essay could be published here as well as in their pages. 1. Oxford: Hart Publishing, 1999. All parenthetical page references are to this volume. 2. The book also contains stimulating and insightful discussions of topics including no-fault social insurance schemes, the distinction between acts and omissions, and the ways in which the criminal law should take account of an offender’s circumstances.

  ways in which people can be related to consequences. There are two aspects to this question. The first of these centres around a knot of interpretive questions, about what counts as a consequence in the legally relevant sense, and about how those consequences matter. Honoré has done groundbreaking work on this first question, both on his own and together with H.L.A. Hart. Causation and the Law 3 offered a careful and detailed account of the role of causation in tort law, an account that remains as important as ever more than four decades later. That account is meant to defend two common-sense claims: first, causal attributions are factual rather than normative, and second, the fact that one person caused an injury to another provides a reason for holding the injurer liable. But the more recent book’s main focus is on an importantly different question, which Honoré’s earlier work touched on but did not develop: Why do consequences matter in tort law? What is the moral status of the causation requirement? To be sure, answering that question requires taking a stand on which consequences matter, and how they matter. As a result it cannot be separated entirely from a specific view of how the causation requirement operates in tort law. One essay in Responsibility and Fault returns to that set of questions, and offers further defence of the basic account that he and Hart developed. But the central question Honoré sets himself is ultimately one of justice: under what conditions is it morally acceptable to hold a person responsible for the causal consequences of his or her acts? Under what circumstances is it morally acceptable to hold that person liable in damages for the consequences of those acts? The questions get part of their force from a series of worries about the relation between causation, responsibility and luck. How things turn out is usually a matter of luck: two drivers may be equally careless, but one may get in an accident causing horrendous damage, while the other, through good fortune, does no damage at all. There is a perspective from which this result seems arbitrary, because whatever is wrong with the first driver’s driving (or character, or whatever) is also wrong with the other’s.4 Both should have exercised more control. Given that the two are alike in morally important respects, why should things turn out so differently for them? The thing that seems arbitrary in this sort of example is just the operation of causation. Why hold people liable for what they have caused, rather than for the things over which they could exercise some measure of control? 3. Oxford: Clarendon Press, 1st edn 1958; 2nd edn 1985. 4. For a forceful statement of this view, see Jeremy Waldron, “Moments of Carelessness and Massive Loss” in David Owen (ed.) Philosophical Foundations of Tort Law (Oxford: Clarendon Press, 1995).

38

     One approach to causation, increasingly influential in American torts scholarship, is to explain it away. This sort of approach says that people are not related to the consequences of their actions in any deep way. Instead, we hold them responsible because it is useful to suppose that they are, because holding people responsible will reduce accident costs, since the prospect of liability will provide them with an incentive to exercise greater care. Honoré’s answer, by contrast, takes both the causation requirement and the idea of responsibility at face value. It takes the causation requirement at face value because it accepts the pre-theoretical idea that a person is held liable for an outcome because she is responsible for it. Honoré’s task is to defend this intuitive idea by providing a theoretical underpinning for it. We do a person no wrong in holding him responsible because so holding him is part of a system of responsibility, one about which the person held responsible cannot rightly complain. The basic strategy is to show that a system of liability that holds people responsible in those ways is fair overall, so that its application in any particular case is not unfair. Thus the two equally careless drivers may be treated differently on some occasion, but across time the system as a whole will be fair to both of them, so that neither may complain of unfairness. Rather than showing that liability is fair in each particular case, then, Honoré shows that it is fair in general. Particular cases inherit their justification from the more general justification of the institution, simply by showing that they fall under its general principles, which are fair to all concerned. A system of tort liability is useful, because of its role in reducing unwanted conduct. But, as Honoré rightly points out (72), the difficult and interesting questions about the justification of tort liability do not concern its efficacy but its fairness. How does one go about showing that the system as a whole is fair? Honoré’s account has four strands. The first of these, developed almost 40 years ago, is the idea that whether a person can be held responsible for something depends on whether he has the requisite general capacities, not on whether those capacities were exercised on the particular occasion. The second is the idea that the things for which a person is responsible in tort depend on some idea of a fair distribution of risks. Tort law is sometimes said to be concerned exclusively with corrective justice, which applies apart from considerations of distributive justice. Honoré argues that, although there is a sense in which this is true—for example, an undeservedly wealthy person may recover substantial damages from someone who is much poorer—there is another sense in which it is not, because tort law rests on a distribution of risks within society. Both of these ideas are original, interesting and deeply illuminating of important features of tort law. 39

  In developing the ideas of capacity and risk distribution, Honoré also appeals to two other ideas, both of which have broad significance for moral life, but seem to me to apply to tort law only with a great deal of qualification. These are, first, the idea that our sense of who we are depends on being held responsible, so that tort principles express a significant part of our conception of ourselves as persons, and second, the idea that it is fair to make people shoulder burdens because they recover corresponding benefits on other occasions. Both of these ideas divert attention from Honoré’s core insights about the ways in which risks are distributed, for both are ideas that are subject-centred in a way that tort law and the idea of risk distribution are not. If I seem to be overemphasizing the difficulties with these ideas, it is because I think they stand in tension with Honoré’s most illuminating ones.5 While I have some sympathy with the underlying pluralism about values that his account suggests, it seems to me that some things—the law of negligence in particular—are best understood as centring around certain values, and supposing others to be irrelevant, regardless of their importance to other aspects of life.6

C

The earliest essay included in the collection, first published in 1964, is “Can and Can’t”. It is addressed to the traditional philosophical problem of freedom of the will. The problem, simply stated, is that the natural sciences conceive of the world as deterministic, while human beings think of themselves as responsible for their actions. Yet if all events are determined, those events we think of as human actions are also determined. Thus there seems to be no room left for the idea of human beings as free and responsible agents, since their actions, like all other events, are the outcome of antecedent conditions. Yet we think of ourselves, and of each other, as free and responsible, and in thinking of ourselves in this way we suppose what we could have acted differently than we did. 5. In the introduction to Responsibility and Fault, Honoré notes that legal philosophy must face scepticism from both philosophers and lawyers. In my case, scepticism about one part of his account grows out of admiration for another. 6. Stephen Perry has distinguished between what he calls the “social” and “agency” understandings of outcome responsibility, and gone on to develop an account of tort liability in terms of the latter. See Perry, “The Moral Foundations of Tort Law” Iowa L Rev 77 (1992), 449 at 490. Honoré claims (79) that the two accounts are compatible. I agree with Perry that there is a tension between them, but think that the “social” account incorporates an interesting account of agency. I have criticised Perry’s account in detail in Equality, Responsibility and the Law (Cambridge: Cambridge University Press, 1999), ch. 4.

40

     Honoré draws our attention to the way in which we talk about whether a person could have done something which they did not, in fact, succeed in doing, and reminds us that such talk is contrastive. The relevant contrast is between that person’s performance on the particular occasion, and the more general capacities that person has. When we say of a golfer who has just failed to hole a six-foot putt, that he could have sunk it, we do not thereby suggest that he had the capacity to sink that very putt. Instead, we implicitly contrast his failed performance with his more general putting skills. Honoré generalizes this distinction, between general and particular uses of “can”, to explain the place of responsibility in a deterministic world. In the general sense of “can” most persons not only can do things other than they did, but can have intentions other than those that they have. Thus when we say of a person that she could have done something other than what she did, we are not making a mysterious claim about an alternative history of the universe, or an exemption from causal laws. Instead we are appealing to the capacities that person has, capacities which make her the person who she is. The relevance of this analysis to legal contexts should be apparent: in holding someone liable in tort because he is at fault, our concern is with his conduct on the particular occasion, as measured against his more general capacities. The general capacities provide for the sense in which he could have avoided liability, even if he did not avoid it on this occasion. Moreover, Honoré contends that the system of liability is fair because people are held liable only for acts that they can (general) avoid. Thus, conduct that attracts liability will be an expression of an agent’s character, even if on a particular occasion he could (particular) not have avoided doing what he did. In the long run, a person’s general ability to avoid injuring others will enable him to be free of liability, unless he is a careless person. As Honoré puts it (38), “the fault system penalizes those whose conduct displays a bad disposition, but at the same time affords people fair opportunities of escaping sanctions for botches and bad outcomes most of the time”. If the idea of general capacity is relevant to tort law, its scope needs to be narrowed to explain the law’s use of it, for people are not liable for everything they could have avoided in the general sense. They are only liable for a small subset of those things. In the sense of “can” and “could” that Honoré shows to be relevant, the person who drives safely enough to avoid liability is not liable even if he could (general) have driven more safely, or stayed at home. Nor does the person who consistently falls below the law’s requirements escape liability. Those whose capacities are limited, “shortcomers” in Honoré’s phrase, are still answerable for the consequences of their deeds. And those who are more able than most are not answerable for every failure to live 41

  up to their own abilities. We may ask more of the experienced surgeon than the novice, and more of either than of the person on the street. But we do not hold the more experienced driver to a higher standard than the less experienced, or the experienced user of flammable solvents to a higher standard than the less experienced. Instead (outside of professional contexts), we hold them to a common standard, and ask only whether a reasonable person of ordinary prudence could have avoided the injury in question.7 The particular defendant’s actual capacities are not to the point. Instead, the law focuses on the capacities of the representative, reasonable person. The law’s use of the idea of general capacity is thus subtly different than the use which Honoré’s more general account suggests. Rather than looking at what are the particular person’s general capacities, the law looks to capacities that people more generally can be expected to have. Where Honoré’s can (general) applied to the golfer’s other, golf-specific abilities, in tort law can (general) applies to the defendant’s general abilities to conform his conduct to the law, the pair of abilities that common law judges have long referred to as foresight and prudence. Those capacities do not guarantee success on any particular occasion, but they are only imputed to those who are capable of success most of the time. Other, more specific abilities, such as the ability to see over a fence if six feet tall, are treated differently. Far from showing the idea of general capacity to be irrelevant to tort law, the use of the representative, reasonable person shows just how important the general strategy is. But that raises a further question: why deploy it in this way? That is, why are those particular capacities the relevant ones, given that the defendant may lack them?

“R-D J”

Why focus on the capacity to avoid accidental injury to others, rather than the capacity of this defendant to avoid this injury to this plaintiff? While this question might be answered in a variety of ways, Honoré’s approach is to focus on the idea that tort liability presupposes a prior distribution of risks in a society. We live, and act, in a world of uncertainty. As a result, everything we do can be thought of as taking certain risks and accepting others. The risks 7. I take it that the exception for professional contexts reflects the asymmetries involved in professional relationships. Ordinarily, the law seeks to protect people equally from each other; where asymmetrical relationships create special vulnerabilities, special duties may be imposed. In addition, people with special skills may be held to a higher standard for contributory negligence. But manufacturers of solvents do not owe a lesser duty of care to users with expertise. See for example, Lambert v. Lastoplex (1971) 25 DLR (3d) 121 (SCC).

42

     in question are risks that holdings will change in particular ways. I bear the risk that my house will be destroyed by a storm (unless I get others to bear it, either through contract, or the state holds such risks in common); you bear the risk that my house will be destroyed by your nearby blasting. Honoré suggests that this can be thought of as expressing an idea of a fair distribution of risk. The distribution is fair in two distinct senses: first, over time it tends to work to the advantage of all, even if it does not work to some particular person’s advantage in some particular case.8 Second, it is fair across persons. As Honoré puts it, “It places on every member of the community the burden of bearing the risk that his conduct may turn out to be harmful to others, in return for the benefit to himself that will accrue should his conduct turn out as he plans”. So risks are distributed fairly to each person when every burden comes with a matching benefit. Risks are distributed fairly across persons because the principle of risk distribution applies reciprocally: “Corrective justice is a genuine form of justice only because the just distribution of risks requires people of full capacity to bear the risk of being held responsible for harming others by their conduct even when they are not at fault for doing so” (80). That is, I get credit for my deeds, but am responsible for the burdens they place on you; in return, because you are responsible for the costs your deeds place on me, I am free of those costs (142). We thus get two things in return for being held responsible for the bad consequences of our acts: credit for the good consequences of those acts, and freedom from the bad consequences of the acts of others. In sum, “This gives people a degree of independence from one another and from the power of government” (72). The ideas of a distribution of risks and outcome responsibility thus express part of the ideal of the rule of law. The rule of law requires that people conform to public standards of conduct; so long as they conform, they are free to order their own lives as they see fit. A distribution of risks determines what people owe to one another by way of norms of conduct; because all owe each other 8. The first argument looks at a single person’s advantage across time, the second at fairness across persons. Hart and Honoré had offered a version of the first argument in passing in Causation and the Law, above n. 3, 1st edn at 243, 2nd edn at 268. The claim there is that it is not unfair to hold someone responsible for an unforeseeable extent of damages because a person can fail to exercise appropriate care on other occasions but not be held liable. There is much that is puzzling about this version of the argument. As Joel Feinberg long ago remarked, it appears to presuppose that some sanction would be acceptable, or even appropriate, for those who fail to exercise appropriate care. Given that such a sanction is waived in ordinary circumstances, a more severe sanction when harm ensues is less troubling. Such an account appears to presuppose a sanction instead of explaining it. Feinberg notes the parallel between Hart and Honoré’s argument and the Augustinian view that, since none deserve to be saved, the damned cannot complain if God saves others through grace alone. Some independent argument is needed to show that those who are careless deserve some sanction.

43

  the same thing, attributions of responsibility—the sense of can (general)—are keyed to the capacity to conform to those standards. The sense in which risks are distributed is open to misunderstanding. Honoré does not suppose that risks are, or should be, distributed along with other goods in a process designed to ensure that everyone gets a total bundle of goods determined by merit, or that is equal in value to the bundles others receive. Risks are distributed, but are so distributed independently of other sources of wealth and opportunity.9 The relevant notion of distributive justice is thus confined to the distribution of a specific sort of thing. As a general strategy, this approach has much to recommend it. For example, many people who suppose that health care should be allocated on medical grounds alone have no objection to considerable disparities in wealth and access to other goods.10 Instead, they suppose that the allocation of health care should be protected from the effects of these (otherwise acceptable) disparities. No doubt such a supposition is rooted in views about the importance of health in a human life, views which do not regard it as just one of many possible inputs to well-being. Moreover, in systems in which health care is distributed on the basis of need, any deviation from this standard of distribution is frowned upon precisely because someone gains something to which they are not entitled, and gains it at the expense of another. Others may appeal to the same principle at a different level, and so demand that the sickest patients in an emergency room receive immediate attention, even if others have been waiting for hours. (“First come, first served” is a competing distributive principle.) Honoré’s key insight is that the same kind of strategy is presupposed by the ways in which particular people bear particular risks, quite apart from whatever other holdings they might have. We assign risks to people in this way because we do not regard the risk of injury by others as just one of many possible inputs to some overall scale of well-being. Instead, we regard particular risks as concomitants of certain types of conduct, and suppose that persons of full capacity can (general) moderate their behaviour in the interest of the safety of others, and must accept responsibility when they fail to do so. 9. Risks might be thought to be different from other things that are distributed because nobody takes much of an interest in them unless they are realized, so that our primary interest is in who is injured, rather than in who bears which risks. In fact, people bear and trade unrealized risks all the time. They bear risks whenever they decide to take a chance on something, and they trade them both by purchasing insurance and by bargaining around them in more ordinary contract settings. 10. On equality in health care, see Bernard Williams “The Idea of Equality” in Peter Laslett and W.G. Runciman (eds) Philosophy, Politics, and Society, 2nd Series (Oxford: Basil Blackwell, 1962). Williams’s strategy is generalized in Michael Walzer Spheres of Justice (New York: Basic Books, 1983).

44

     Just as the principle that medical care should be distributed on the basis of health can be applied either to a particular emergency room or to an entire health care system, so too can the idea of risk-distributive justice be applied at different levels of generality. We place the same expectation of taking responsibility for risk on organisations, such as corporations, when we hold them responsible for the torts of their employees. Such vicarious liability is not puzzling, insofar as we suppose that risks characteristic of the activities from which they seek to profit rightly belong to them. Nor is strict liability puzzling on this view; here too, we can understand it in terms of an allocation of risks. The same account makes room for the idea that some risks may be removed entirely from a system of individual responsibility through social insurance, by supposing that the community as a whole is the unit to whom they are distributed.11 The question of the appropriate unit of accountability cannot be answered through the concept of risk distribution alone. Doing so requires substantive judgments about the justice of distributing particular risks in particular ways (91).12 The idea of a distribution of risk, or some idea like it, helps us to understand why the capacities relevant to negligence law would be those of the ordinary or reasonable person, rather than those of the defendant in question. Risks are distributed across persons; although I am entitled to the benefits and liable for the burdens of my conduct, to say that I am so entitled and so liable is implicitly contrastive: it is to say that other persons are not. To say I must bear the costs that my conduct imposes on you is to say that you are entitled to be free of them. To say that I am entitled to the benefits that flow from of my conduct is to say that you are not entitled to them.13 But to say such things, both of us must be brought under a common standard. Otherwise, there will be leftover, undistributed risks. But of course there cannot be undistributed risks, since any injury that results from them will still be somebody’s problem, and so its risk will turn out to have been distributed after all. The way in which considerations about the distribution of risk make particular capacities relevant to tort law illustrates a more general point about the 11. Honoré puts this point in terms of benefits and burdens. I discuss that formulation of the point below. 12. Honoré goes further, suggesting that a system of social insurance for injuries is not incompatible with corrective justice. This seems to me to overstate matters in two respects. First, if risks are distributed in this way, there is no maldistribution to correct, and so corrective justice does not have anything to which it might apply. The unit that bears the risks bears its cost. Second, the community as a whole is not outcome responsible for the injuries it must repair, in the ways in which both individuals and corporate persons can be. 13. Although you may well receive them as a side-effect of my conduct, you are not thereby entitled to them. I elaborate on this point below.

45

  relation between capacity and outcome responsibility. Attributions of outcome responsibility frequently use the idea of capacity, and are probably applied only to beings with general capacities for controlling their behaviour. But the idea of outcome responsibility by itself does not always require appeal to an understanding of the agent’s relevant capacities. The careful driver who chastises himself for injuring a child who runs out from between parked cars need not suppose that he could (general) have done otherwise than he did, except in the entirely open-ended sense in which he could have taken a different route, stayed home and so on. Yet he may well regard himself as outcome responsible, even though neither he nor anyone else had the capacities necessary to avoid it. The person who is thrown by a sudden and unforeseen impact, and has his fall broken by another person may also think of himself as responsible in this sense. It is only in contexts in which responsibility must be allocated to one person rather than another that questions of capacity are necessary to imputations of outcome responsibility.14 It is perhaps worth remarking that this idea of a fair distribution of risks is specific to tort law in a way that the general ideas of capacity and outcome responsibility are not. Neither that idea, nor anything very much like it, needs to be invoked in order to explicate the idea of outcome responsibility more broadly construed. We can make sense of outcome responsibility in cases in which there is no obvious sense in which particular risks are assigned to particular persons. If one person injures another non-tortiously, the injured person has no claim in corrective justice, because she bore the risks of the type and occasion of injury that materialised. But the injurer may nonetheless be outcome responsible for the injury. It may be appropriate for her to apologize for it. Indeed, she may have recurring thoughts about it, perhaps even nightmares, because she sees herself as responsible in a deep sense, even if social institutions do not treat her as responsible. If, as a result of the negligence of its owner, the cable on a crane snaps, dropping its load on people standing below, tort law will hold the negligent owner liable, but the crane operator will no doubt regard himself as outcome responsible for any injuries.15 14. The golfer who cannot (in general) sink a six-foot putt may feel elated when he does, because it is his accomplishment, even if it is not repeatable. We say of such a golfer that he has holed a six-foot putt, but not that he can do so. Another golfer who cannot sink a six-foot putt may feel he has let his teammates down when he fails to sink such a putt. His sense of who he is is deeply connected to what happens in the world as a result of his actions, together with his sense of who he is and the roles he occupies. 15. In Dooley v. Cammell Laird & Co Ltd [1951] 1 Lloyd’s Rep 271, the plaintiff crane operator successfully sued the owner when he suffered nervous shock in a similar incident. (His view was obstructed. As it turns out, nobody was hurt below.) Had he been a mere bystander, his case would have been weaker; because he was outcome responsible, his response to what appeared to be his outcome responsibility is understandable.

46

     When tort law holds persons responsible for outcomes, its task is primarily allocative, in the sense that assigning particular consequences to particular persons is a way of determining who will bear which costs. “Corrective Justice” is corrective insofar as it seeks to allocate costs to their proper bearers, so that a cost which one person has wrongfully placed on another is returned to the person to whom it properly belongs. Not all attributions of responsibility are allocative in this sense, not even all interpersonal allocations. Consider one of the most familiar manifestations of outcome responsibility, the case in which one person injures or inconveniences another, and apologizes. As Honoré notes, one may well have reasons to apologise for things for which one is outcome responsible even where there is no issue of either fault or legal liability. If I bump into you on the street, and you drop your package, I have reason to apologise even if, as it turns out, you were the one who appeared from between parked cars without any notice. Again, the careful driver who injures a pedestrian is outcome responsible and must acknowledge that responsibility, even where there is no question of fault. And if I am unavoidably delayed in meeting you, I must apologize for the delay even as I seek to excuse myself. If you and I, each lost in thought, collide as we walk down the corridor, we may each accept full outcome responsibility for the books that lie scattered on the floor, each apologizing to the other as we rush to pick up the other’s books before our own. Our interpretation of our own responsibility, and our responses to it, need not be compatible, because from the first-person perspective in which outcome responsibility is so often important, there can be more (or less) outcome responsibility than there are outcomes. This is because some first-person takings of responsibility are exclusively autobiographical. When I apologize as a way of acknowledging my role in your misfortune, my autobiography and your autobiography need not be compatible, any more than each of our own sense of whom we are needs to be compatible with that of the other. Outcome responsibility is an important ingredient in the private narrative of my life that I construct for my own interest. Its role is subordinated to the coherence of the larger narrative of which it is a part.16 Appropriating oneself as a person by making up one’s own story is a matter of shifting emphases that may or may not be systematically related to its impact on others. Consequences that are too remote to attract legal liability—or even the notice of those they affect—may be treated as central to a narrative. For a key feature of narratives (as Freud was fond of pointing out) is that events only acquire 16. Notice the tension between Nietzsche’s dictum that “the thing is the sum of its effects” and his larger claim that to live a life is to attempt to create an interesting story.

47

  their full significance in the retelling, not as they happen. As a result, the significance of my bumping into you for your life and for mine may be vastly different. In the story I tell myself, it may come to symbolise the point at which I realised that I needed to get more sleep, or give up driving. Yet the same part might have been played by a near miss. (Which plays the role depends upon how heavily the camel’s back was burdened as it waited for an appropriate straw.) Or it may be a minor member of a litany of examples of things that went wrong in a particular difficult day, or period of my life. Again, in the story you tell yourself it may be deeply significant or a trifling inconvenience. Which it is for each of us depends primarily on what else was going on in our separate lives at around the same time, that is, on what other things it needs to cohere with, how each of us supposes the rest of our life stories to hang together. It is only when we need to have a single answer to questions about who is responsible, or how responsible each of several persons are, that outcome responsibility needs to be disciplined by an idea such as the distribution of risks, for it is only in such circumstances that a just resolution requires that there be a single and determinate answer. That is, there are some cases in which outcome responsibility figures primarily in autobiographical imputations; there are others in which it figures in interpersonal allocative practices. Consider again the example of apologising. If I interrupt you, I owe you an apology not only to acknowledge the impact of the interruption on each of our lives. Its impact on either of us is transitory. In such contexts apologising is also an imperfect way of restoring you to your prior uninterrupted status in our conversation. As I apologise, I give you back your place, and undo the effects of my interruption. The law’s approach is even more austere, and even more heavily driven toward coherence across persons. The law’s concern is not with how each of our lives cohere separately, but with whether our interaction amounts to a violation of either of our rights. The law must thus look to the systematic considerations of the sort captured in Honoré’s idea of risk distribution in order to assess the significance of any interaction. From this perspective, the impact of my deed on you is measured in terms that are public and impersonal. If imputations of legal liability are allocative in a way that other practices involving responsibility are not, they are also more determinate than other allocative practices. Honoré has argued elsewhere that legal institutions are needed to make moral obligations more determinate, by allowing people to know what is required of them, where morality requires, but does not provide, clear directives.17 Rules governing property provide a clear example, as 17. “The Dependence of Morality on Law” Oxford Journal of Legal Studies 13 (1993), 1–17.

48

     do definitions of rights more generally. There is much to be said for this idea, but it is important to see that it does not apply to either the idea of a general capacity or that of outcome responsibility. Law serves to make morality sufficiently determinate to provide guidance in cases in which that determinacy must be interpersonal, that is, in those cases in which a plurality of interacting persons must act on a common understanding of what is required of each of them.18 Those who wish to interact with others on fair terms do not know what those terms require unless they also know what others, also wishing to interact on fair terms, suppose them to require. Positive law makes such obligations sufficiently determinate to enable people to treat each other justly.19 By contrast, legal institutions are not always needed to make outcome responsibility more determinate, because outcome responsibility does not always need to be interpersonal. Each of us needs to have a more or less determinate sense of ourselves, but we can make it determinate enough in a variety of ways, and in different ways at different times. Whether or not one wants to judge a life without much determinacy as successful, it is still recognisable as a particular person’s life. To sum up, both outcome responsibility and the idea of capacity are only relevant to tort liability when their scope is narrowed. Honoré brings both under the discipline of an idea of distributive justice, according to which corrective justice is responsive to a prior distribution of risks in a society. The idea of a distribution of risks across persons is able to play this role because it is essentially relational. The idea of outcome responsibility can be invoked in an autobiographical sense, so that the fundamental distinction it draws is between outcomes that are mine and those that are not. The idea of a fair distribution of risks, like all ideas of distribution, is inherently interpersonal: that which is not mine is always somebody else’s in particular. As a result, the idea of a fair distribution of risk makes outcome responsibility both allocative and determinate. It is allocative because particular outcomes are allocated to particular persons. It is determinate because (ideally, at least) each outcome is allocated to exactly one person, or, where they are not, each outcome is 18. Legal responsibility is frequently narrower than moral responsibility—in Bolton v. Stone [1951] AC 850 (HL) the cricket club took itself to be morally (outcome) responsible for Mrs Stone’s injury, and offered to pay damages, despite the fact that they denied legal liability. It is also sometimes wider. Some may well think that the defendant in Vaughan v. Menlove (1837) 132 ER 490 (CP) is morally faultless, while accepting that the structure of legal liability requires that he pay damages to the plaintiff. 19. To be sure, law also provides a coercive back-up to induce those who do not care about how they treat others to treat them justly. But the moral obligation to treat others justly often needs positive law to make it more determinate.

49

  divided up among persons so that the total amount of outcome responsibility (measured in liability) adds up to the total magnitude (measured in damages) of the outcome. With this idea in place, outcome responsibility becomes an essential part of corrective justice, because its role is to keep track of, and undo, interruptions of the appropriate distribution of risk.

M P

Focusing on the idea of risk distribution shows us that outcome responsibility turns up in two quite different contexts. It shows up when we are decide who a particular outcome belongs to by deciding whose risk it was, and it shows up when each of us is trying to make sense of his or her own life. I now want to suggest that the sort of outcome responsibility that shows up in tort law is not simply a refined version of the sort at stake in our deepest selfunderstandings. We might call the sort of outcome responsibility that is tied up with a person’s conception of who she is as she makes her way in the world “implication responsibility”—my agency was implicated in your injury, and so I must face up to my role. Thus I must apologise, or call for help, or perhaps help you up. What I must do may be largely symbolic—if I knock you over, I must help you up even if you would not have had any trouble getting up on your own. The norms governing implication responsibility are sometimes norms of repair, which may or may not supervene upon the violation of any norms of conduct. But they always depend on the fact that something was my doing. Implication responsibility is a special case of a much broader feature of human life. Just as people identify with, and act in light of, some of the consequences of their deeds, so too do they identify with, and act in light of, features of their situation that they did not bring about. I can be implicated in something in this sense without doing anything. That is not only because I am sometimes responsible for my omissions as well as my acts. I may think of myself as implicated in something that cannot meaningfully be described as an omission either. I may find myself implicated in the deeds of my country’s government, or something done by a family member. In such cases, there may have been something that I could have done, but didn’t, to disentangle myself, perhaps by taking a public stand against something. But in such cases, my entanglement cannot be traced to my failure to disentangle myself. For example, many Americans saw themselves as implicated in the first moon walk, and took a certain kind of credit for it, a kind of credit, moreover, which they reserved for themselves, as against the rest of the 50

     human species.20 Yet the basis of their readiness to bask in Neil Armstrong’s reflected glory is not anything they did—such as paying taxes—or declined to do—such as moving to Canada. Other times, people feel responsible for bad consequences in which they are implicated in virtue of their citizenship, for example, American slavery. Of course, nobody supposes that such responsibility is a matter of foresight or avoidability. It is rather that people are implicated in virtue of institutions and practices in which they participate, as well as in the upshots of their deeds. Nor is this sort of implication purely subjective. You may tell a rather different story about the pivotal events in and aspects of my life that I do. But the general constraints on any story you might tell are the same as the constraints on the story I tell— coherence across time in a single life. That is why the stories a third party tells about the impact of a single incident on different people need not line up in any systematic ways. Some may question the wisdom of a person thinking of him or herself as implicated in all of these ways. Wise or unwise, such implications are familiar, because the general idea of implication is tied up with who a person is. That sense depends, in large part, on how he or she makes a path through the world. But it also depends on the world in which that person finds him or herself – the world into which one is “thrown” to use Heidegger’s phrase. In our more rationalist moments, we may wonder whether a person’s sense of herself, or other people’s sense of her, should depend on such things. But those very same moments lead us to suppose that we are at most responsible for a proper subset of the consequences of our actions. It is difficult to deny that a person’s sense of him or herself is implicated more directly when the event in question is a direct consequence of his or her deeds than when it is not, and even more so if found liable. The passerby may wish he had taken a different route, and so not witnessed the traffic accident; he understands his responsibility for seeing the accident, while the careful driver, also wishing she had taken a different route, understands hers in terms of the outcome. Still, bringing about an outcome is but one of the many ways in which one can be implicated in a situation. The type of outcome responsibility that is central to tort law is different. Just to give it a name, we could call it “displacement responsibility”. While implication responsibility shows up both as one interacts with other persons and with the natural world, displacement responsibility only shows up in relation to a plurality of interacting persons. If I injure you through my 20. At the other extreme, Ludwig Feuerbach introduced the idea of a “species being” to suggest that, simply in virtue of being human, we can regard ourselves as participating in all of the achievements of other humans. See The Essence of Christianity (Prometheus Books, 1989).

51

  negligence, I do not simply have to acknowledge the role of my deed in each of our lives; I have to pay you damages, to “make you whole”. Rather than acknowledging my deed, I must undo its effects, or take them back. A similar idea comes up in non-legal contexts as well: if I interrupt you in a conversation, a prompt apology restores you to your proper place in the conversation. A rough and ready principle of distributive justice governs conversations—it says that speakers should wait their turn. Just as implication responsibility is a special case of the manifold ways in which a person’s sense of who he or she is can be implicated in a situation, so too is displacement responsibility a special case of the morality that governs interacting persons. The duties of repair enforced by tort law bear a close connection to proprietary duties to return another’s goods, or restitutionary duties to return a mistaken payment. If I receive a mistaken payment, I must repay it, even though my receipt of it is not my doing in any sense, and even though I do not identify with it in any sense. In each case, the holdings in society have changed in the wrong ways, and the imbalance must be corrected. Displacement responsibility is not distinctive only because it yields liability. Instead, responsibility is the reason for imposing liability. I am responsible for injuring you if my injuring was a wrongful interruption of the ordinary course of things; liability in damages is a way of making it as though that interruption had never taken place. Now if the interruption is to be understood in terms of the private stories that either of us tells ourselves, money damages would be a peculiar and ineffectual means of restoring normality. But if the interruption is understood in terms of the rightful positions that we occupied prior to the incident, damages, while still imperfect, seem much more promising. For they serve to return the victim of a wrong to the position that he was entitled to occupy. Having put that right in peril through his deeds, defendant has taken a risk with plaintiff ’s security. That risk is now his;21 should it ripen into an injury, it is his, rather than the plaintiff ’s to deal with. The court does not care whether injuring you has changed my life, or whether liability will change it; it conceives of each of us as juridical persons, and keeps track of the changes each of us makes in the world in those exclusive and narrow terms. It is a familiar feature of legal systems that a tortfeasor is liable for harm done, including ulterior harm, such as loss of income. Yet the same tortfeasor cannot reduce his liability by maintaining 21. I make what I take to be a similar point in terms of risk ownership in Equality, Responsibility, and the Law, above n. 6.

52

     that he has conferred an ulterior benefit on the plaintiff. The negligent driver who causes someone to miss a flight that subsequently crashed cannot maintain that he had benefited the plaintiff, despite the fact that, from the point of view of making one’s way in the world, the plaintiff surely did benefit.22 If the plaintiff has any sense, his narrative of his own life will make the event both pivotal and positive. The law does not take account of such matters in assessing damages. Tort law creates a special public identity for persons, on the basis of some understanding of a fair division of risks. It keeps track of human action in terms of those public personae. We are then held outcome responsible for our deeds if we violate that division of risks, because it is fair to us both as tortfeasors and as plaintiffs. To suppose that in so doing it is addressing itself to our deepest nature as reflective and social persons is to suppose that we must understand ourselves only in terms of what others may demand of us, or that the demands of others are unjust or arbitrary whenever they diverge from our deepest self-understandings. Each person’s sense of who he or she is is non-relational; although my relations to others are likely to be important to it, it does not need to co-ordinate systematically with other persons’ sense of who they are.23 The two ideas come apart, because they answer to such different criteria. One requires narrative coherence within a single life; the other requires a single standard applying across a plurality of persons. B  B

Gambling metaphors are always inviting in thinking about responsibility. Kant once suggested that the person who tells a lie “plays a game of chance” with another person’s agency, and so is answerable for the full consequences of the lie.24 Honoré offers a different sort of gambling metaphor when he writes: “Any principle that can justify responsibility for bad luck must be fair. If it is to be fair it must entail that when we bear the risk of bad luck we also benefit when our 22. This example is drawn from Ernest Weinrib “Right and Advantage in Private Law” Cardozo L Rev 10 (1989) 1283. 23. Bernard Williams remarks “No conception of public responsibility can match exactly an ideal of maturity because, among other reasons, to hold oneself responsible only when the public could rightfully hold one responsible is not a sign of maturity.” See “Voluntary Acts and Responsible Agents” in his Making Sense of Humanity (Cambridge: Cambridge University Press, 1995), at 32. 24. See “On a Supposed Right to Lie from Benevolent Motives” in Lewis White Beck (ed.), Kant’s Critique of Practical Reason and other Writings in Moral Philosophy (Chicago: University of Chicago Press 1949).

53

  luck is good. Allocation according to luck must cut both ways. So the system will be fair only if there are situations in which we implicitly bet on the outcomes of our actions . . . In calculating the odds for achieving the favourable outcome we have to discount the chance that we may not be able to do X or that the outcome of X, if we do it, will not be what we predict . . . Despite the uncertainties, over a span of time more outcomes are likely to redound to our credit than to our debit, so that we are not permanently saddled with a losing ticket. Risk and benefit go hand in hand”. (26)

The betting metaphor is inviting, but it is also ambiguous. Much depends on who I am betting against. Consider some examples: (1) at the racetrack, I bet on one horse rather than another. Here I bet against the bookmakers, and through them against other gamblers. If I win, my gains come out of the losses of others; if I lose, my losses become other people’s gains. (2) I buy a piece of farmland in the hope that coming urbanisation will see its value increase considerably. Here I do not bet against anyone in particular; I bet only “against” the market. If the market as a whole grows, my gain will not be tied to anyone else’s loss. If I lose, the persons who unloaded their worthless land on me may gain, but perhaps nobody will.25 (3) I bet I can cook and eat an elegant dinner and still arrive at the Opera on time. Here I bet only against myself; if I succeed, I gain at nobody else’s expense, and if I lose I lose only at my own expense. (4) I bet I can detect a gas leak using a lighted match. In each case, my gamble may work to my advantage or it may not. From one perspective—my own, first-person interest in my own success—the gambles are all the same, because I take a chance in the hope of gain and anticipation of loss. In each case, if I lose, those observing my plight can say to me “you made your bed, now lie in it”. Yet the reason that I must accept the consequences of failure seems to be different in each of the cases. In the first case, I must accept them because accepting them is a condition of making the bet in the first place: I must put money down in order to have a claim to money if I picked the right horse. My liability—if we are to call it that—is contractual. In the second, I bear the loss in the land’s value because it is my land. In the absence of contractual arrangements or government programmes, there is no rationale for picking anyone else to bear the loss. In the third, there is again nobody else who might bear the loss instead. The fourth example seems different. In the first example, the risk was mine because the bookmaker and I created that risk through our 25. If I do gain, it will not be my doing alone: among the sources of my success will be both the countless private transactions and a variety of acts by public authorities such as zoning commissions.

54

     agreement. Different horses may run around the track at different speeds; the risk of loss, and the chance of gain, come about only because we put something else at stake. In the second and third examples, I bear the risk of loss because nobody else is involved. I bet against myself, or perhaps against nature. In the fourth example, however, the risk of injury to others is my risk because others are entitled to be free of its effects. Not every case of imposing a risk on others, even where the risk-imposer stands to gain from it makes the risk-imposer answerable for the consequences of the risk. For example, if I bet that I can undersell my competitor, I do so in the expectation of gain, but I do not need to underwrite the losses that she suffers as a result. (A similar principle applies to suitors competing to woo the same person, though perhaps for different reasons). Indeed, other, more familiar situations can also be thought of as bets which I can make without thereby taking responsibility for all of the harm that might accrue. Not everything that makes injury more likely counts as a bet. If I engage in an ordinary activity with sufficient care, I am entitled to the benefits that accrue, but not answerable for the costs that activity may impose on others. As I drive carefully, or operate my gas grill carefully, I do not eliminate all of the risk to persons and property that those activities carry with them. If you are injured as a result, the injury counts as a bet you made with your own safety, rather than as one I made with it,26 even though I expect to gain from the convenience of driving or outdoor cooking. By contrast, if I fail to exercise appropriate care, or am engaged in an especially dangerous activity, such as transporting explosives, the law treats any characteristic injuries that result as costs of my activity. In Honoré’s terms, the law supposes that I have made a bet with the safety of others, and, having lost, must bear its burdens. What determines which risks I take, and which ones you take? Each of us can be described as making a series of bets; when we both act, and an injury to one of us results, the law needs to decide which bet to assign that injury to. We can call this the problem of the individuation of bets. In a world of uncertainty, every injury is the joint product of “bets” made by various people, including the victim. Some further principle is required in order to explain why, in some cases but not others, the plaintiff ’s injury is treated as the result of the defendant’s unsuccessful bet rather than the plaintiff ’s. 26. In this example, your bet was that you could live safely among neighbours who engage in ordinary activities that are not free of risk. “In the crowded conditions of modern life, even the most careful person cannot avoid creating some risks and accepting others” Bolton v. Stone [1951] AC 850 (HL) per Lord Reid.

55

  The benefit/burden idea is puzzling on other grounds as well. A successful bettor doesn’t get to keep all of the benefits he causes. If I own a hotel and your nearby restaurant prospers because of my clientele, I confer a benefit on you but have no legal claim to it. I might try to reap those benefits, threatening to close down unless I receive what I have bestowed on you. But even here, it may cost too much to make a credible threat; if you refuse to pay, I may find it too expensive to shut down. Again, if you benefit from the afternoon shade my building provides to your garden, I can threaten to tear down my house unless you pay me the value of the benefit you receive. But you would do well to call my bluff and refuse to pay.27 These familiar examples of the law’s distinction between nonfeasance and misfeasance remind us that although I have a right to refuse to confer the benefit on you, I have no entitlement to the benefits that I cause you to receive. I do have an entitlement to occupy my land as I see fit, casting or removing a shadow at will. And I can refuse to do what is convenient for you unless you pay me whatever price I name. But I do not have a right to the benefits that might accrue to you as such. So I cannot demand an accounting from you for the benefit you receive of my shade or the water flowing under my land in the way that I can demand an accounting if, through no fault of yours or mine, you receive a deposit that I was supposed to receive, or accrue benefits from wrongful use of my property.28 It may be that these examples are not to the point. Perhaps the benefit–burden principle should be understood in terms of the idea of risk distribution. The reason defendant should bear the loss is that it would be wrong to make plaintiff bear it. If I take an unacceptable29 risk with the safety of others, I am responsible for their losses, whether or not I gain, or expect to gain, over the long run from taking risks of that sort. The only gain I receive is the freedom from the wrongful imposition of risks by others. On this reading, “risk-distributive justice” solves the problem of the individuation of bets, because it says that particular risks properly belong to particular persons. 27. Such was the fate of the defendant in Bradford v. Pickles, [1895] AC 587 (HL). In concluding that diverting the flow of percolating water so as to make plaintiff pay for it, defendant committed no nuisance, the court upheld his right to do so, and so upheld his right to demand payment from plaintiff as a condition of restoring its flow. But the plaintiff town refused to negotiate, and the won the waiting game that ensued. The excavation and legal costs bankrupted Pickles, and he was forced to sell his land and move to Canada. 28. This is the converse of the rule that precludes liability for pure economic loss, that is, economic loss that is not occasioned by the invasion of a right. 29. A risk may be unacceptable even though the defendant is not prohibited from imposing it. Cases of strict liability for unusually dangerous activities fit this model. In such cases, the duty imposed by tort law prohibits injuring others in a certain way, instead of prohibiting risk imposition. The same point applies to negligence; carelessly exposing others to risk invites no legal response unless injury ensues.

56

     Honoré is right to suppose that defendant is liable because plaintiff ’s injury is in some important sense his doing. But we need some account of which things count as the defendant’s doings. The difference between implication responsibility and displacement responsibility remind us that the ways in which we keep track of who did what depend on the social context. Private law is a peculiar social context, because it regards persons, and their doings, in terms of their rights against each other, rather than in terms of other things that, from other perspectives, matter vastly more. “Risk-distributive justice” is an illuminating way of thinking about this point in the tort context, because it offers an explication of the sense in which the law keeps track of who does what by creating entitlements to impose certain risks and be free of others.

C: J T L

Unless we construe it relationally, the benefit/burden principle appears to answer the wrong question. It appears to provide an answer to the question “is it fair for the state to single out this person to bear the cost”? and answers with a principle that works well for many things such as tolls on bridges, that is, the principle that those who receive the benefit from some activity should also bear the costs associated with it. But the invocation of the principle in that sort of case rests on an implicit contrast between the user paying and the general public being asked to pay. Thus it might serve as an answer to a challenge issued by an advocate of some scheme of generalised social insurance: those who engage in activities dangerous to others should pay for the accidents they cause, rather than leaving the social insurance scheme to cover the costs. So understood, it seems to leave out the most fundamental and familiar feature of a tort action: an aggrieved plaintiff comes before the court seeking recourse against the defendant whom she alleges to have wronged her. The fact that she has been wronged is the crux of her argument, not the fact that the defendant can, over the long run, expect to benefit through the activity which produced her wrong. Of course, the argument made by the plaintiff, or the argument accepted by the court, do not exhaust the question of the justification of tort law. Any number of external criteria, such as fairness across time, efficiency, or some other good, might be brought forward to justify a system of tort liability. But the plaintiff ’s argument has a special relevance to questions of the normative significance of causation, on two quite distinct grounds. First, any attempt to justify a practice needs to be sure to justify that practice, not something else 57

  that differs from it in significant ways. If the plaintiff prevails, she receives what she has demanded: damages from the person who she alleges has injured her, for the injuries brought about through conduct that she claims a right to be free of. The key point here is not just that the court will pay no heed to ulterior arguments about the benefits defendant has received through his past deeds, or the benefits society might hope to receive if it holds him liable. It is rather that such damages as plaintiff will recover will not be the ones that any such arguments will suggest.30 Any justification for its doing so does well to focus on what it does, rather than beginning with the idea that one could legitimately be held liable for all of the harm one’s conduct causes, and then going on to provide independent rationales for the apparently systematic ways in which such liability is limited.31 Second, the shape of the plaintiff ’s argument has special relevance because it addresses both sides of the dispute: her claim is that it would be unfair to her to relieve the defendant of liability, because the costs associated with this particular risk properly belong with the defendant rather than the plaintiff. An account that looks only at what liability means to the defendant misses that crucial feature of tort liability. The idea of a fair distribution of risk captures that idea, because it allows us to understand plaintiff ’s claim that her injury is the defendant’s problem. Hart and Honoré’s account of causation is relevant here. On that account, the condition we identify as “the” cause of an event depends on an implicit appeal to some background understanding of a normal course of events that has been interrupted. The idea of a fair distribution of risks is a way of specifying the normal course of events which the tort in question interrupts. If I fail to exercise reasonable care, or engage in a hazardous activity, I interfere with the normal distribution of risks; if you are injured, my deed is the cause in the relevant sense, and your injuring is my doing in the relevant sense. I am responsible for the outcome that is the realisation of the risk that I took with

30. This seems to me to pose problems for Honoré’s claim (87), that tort liability is properly governed by a retributive principle of proportionality, and that such a principle explains the exclusion of damages that are too remote. Yet proximate damages may be enormous, and far out of proportion to the defendant’s fault. (See Waldron, n. 4 above). Risk-distributive justice provides an account that is both compelling and in keeping with tort law’s practice: remote damages are ones which were not the realisation of a risk that properly belonged to the defendant. As Honoré says, the line between proximate and remote damages is seldom clear. But riskdistributive justice explains the point of that line, if not the point where it will be drawn. 31. No doubt the issue here turns, at least in part, on the interpretive question of what counts as a consequence in the legally relevant sense. If a negligent tortfeasor is responsible for the harm she causes, the pattern of liability in need of justification looks different than liability which must be within the ambit of the risk that makes her conduct negligent.

58

     your safety.32 Corrective justice serves to restore the distribution of risks; in so doing it also restores the distribution of other goods. But it depends on standards governing the ways in which people are allowed to treat each other. Honoré sometimes seems to suggest that tort law is a way in which the state reinforces and encourages people to make sense of their own lives. Honoré’s idea of risk distribution draws our attention to its more significant role in upholding standards for how people treat each other.

32. But consequences that are part of the normal distribution of risks count as too remote. If a ship is damaged as a result of another ship’s negligence, the negligent shipowner is liable for the full costs of the accident, including the costs of lost business due to any delay occasioned by repairs. But should the ship subsequently be destroyed by a storm or another accident, the defendant is not liable, though it be the case that plaintiff ship would have been elsewhere at the time of the collision, were it not for defendant’s negligence. The law treats the situation as having returned to normal once the ship has been repaired. Salmond, The Law of Torts (6th edn, London: Sweet and Maxwell, 1924), 153–5.

59


4 HONORÉ ON RESPONSIBILITY FOR OUTCOMES Stephen R. Perry*

In addition to his many other contributions to moral and legal philosophy, Tony Honoré has introduced two notions that, in my opinion, are of crucial significance to a successful non-consequentialist account of tort law. These are, first, the idea of responsibility for outcomes, and second, the idea of general capacity. While my own work in tort theory has drawn heavily on both these ideas, I will devote the greater part of this essay to a discussion of the former notion, focusing in particular on the role that it plays in our theoretical understanding of tort law. Towards the end of the essay I will also add a few words about the significance for tort theory of the notion of general capacity.

I

Honoré has characterized the notion of responsibility for outcomes in a number of different and not always compatible ways. In his essay “Responsibility and Luck” he first introduces the notion, under the label of outcomeresponsibility, by means of an analogy with a wager.1 Choosing, he says, is inescapably like betting. In making a decision to do X rather than Y “we are choosing to put our money on X and its outcome rather than Y and its outcome” (25). In opting to attempt a U-turn rather than to go on to the next roundabout, for example, we implicitly bet that we will arrive at our destination more quickly. But we might lose the bet by getting involved in an accident, say, or by discovering that going via the roundabout would have been * I would like to thank the participants in the Responsibility and Luck Conference, held in honour of Tony Honoré at Columbia Law School in March, 2000, for their discussion of the version of the paper that I presented there. I am particularly grateful to my commentator, Jules Coleman, for his very perceptive remarks. I also received a helpful set of comments and queries from Peter Cane, one of the editors of this volume. 1. Tony Honoré, “Responsibility and Luck: The Moral Basis of Strict Liability” in Honoré, Responsibility and Fault (Oxford: Hart Publishing, 1999), 14–40. All page references contained parenthetically in the text are to Responsibility and Fault.

 .  the faster route after all. It is important to see that Honoré is not at this point saying anything about compensation. The payoffs of the bet are credits and discredits in a social ledger of some kind. Thus in the U-turn example we receive a credit if we are successful in getting to our destination quicker, but “if we botch it, have an accident, or mistake the route, that is chalked up against us” (25). Moreover, we receive the discredit even if the botch or miscalculation was not our fault. Choosing to act is said to be an implicit bet on outcomes, where neither the stakes nor the social payoffs are necessarily known in advance. This characterisation appears to envisage outcome-responsibility as a social system of a certain kind, which Honoré refers to as “outcome allocation”. The outcomes being allocated are social credits and discredits. In the case of a harmful outcome to another person, outcome-responsibility should not, Honoré says, “automatically involve a legal duty to compensate” (27). An “extra element” is needed to ground legal liability. Sometimes this extra element is fault, and sometimes it is the fact that the actor’s conduct gave rise to a special risk of harm that in fact materialized. In this latter case, strict liability is appropriately imposed: “In certain areas of life where there is a special risk that what we will do will have a harmful outcome, society insists on pressing our responsibility for outcomes to its limits. Our fellow citizens insist that we pay up in terms of compensation rather than just apologize or ring for the ambulance or comfort the injured”. (27)

As Honoré explains it, strict liability is justified because “a consequentialist argument for avoiding serious harm reinforces the non-consequentialist arguments for imposing outcome-responsibility” (27–8). If outcome-responsibility is a social system, then in some sense it has been socially adopted, and presumably it is capable of being changed. If we are to continue to retain the system we would therefore do well, as Honoré recognizes, to offer a justification for it. The justification Honoré advances looks to the idea of fairness. Any principle that can justify responsibility for bad luck must be fair, and “[i]f it is to be fair, it must entail that when we bear the risk of bad luck we also benefit if our luck is good” (24). As so far described, this principle seems simply to suggest that if we must bear the costs of our actions, then we should also be entitled to the benefits.2 But in “Responsibility and Luck” Honoré goes on to elaborate three conditions that must be met if the 2. As we will see below, Honoré turns this around in a subsequent article, suggesting that if we are entitled to the benefits of our actions, then fairness requires that we bear those actions’ costs as well. Both formulations are reminiscent of a common theme in libertarian thought.

62

     “system” of outcome allocation is to be defended as fair. These are that the system must in its operation be “impartial, reciprocal, and over a period, beneficial” (26). The first two of these conditions are aspects of most general moral concepts, and need not be considered in detail here. It is the third that is potentially problematic. With respect to that condition, Honoré says that the system “must work so as to entitle each person to potential benefits that are likely on the whole to outweigh the detriments to which it subjects him” (26). Honoré maintains that while this makes it unfair to apply the system to those who are incapable, “people of full capacity stand to win more than they lose from the system of outcome allocation” (28). At this point we touch on the second of Honoré’s two notions that I said are significant for tort theory, namely, the idea of capacity. I put this second notion aside for the time being, but will return to it in section II. Let me call Honoré’s characterisation of outcome-responsibility as a system for allocating social credits and discredits the “social” understanding. It is supplemented later in his article by a strikingly different characterisation, according to which “[w]e could not dispense with outcome-responsibility without ceasing to be persons” (15). Let me call this the “personhood” understanding of outcome-responsibility. The idea seems to be that personhood is inescapably bound up with agency, but we cannot be agents capable of choice without accepting, in some very fundamental sense, that we are responsible for the consequences of the choices we make. Pursuing what he apparently regards as the same line of thought, Honoré also says that “outcomeresponsibility is crucial to our identity as persons” (29), by which he apparently means that it is crucial to our identities as individuals and not just to our having the general status of personhood. The sense of this claim is perhaps clarified by his later statement that “it is outcomes that in the long run make us what we are” (31). Our sense of identity as individuals is at least partially determined by the outcomes of the choices we make in life. However, at least at first glance, the relationship between outcome-responsibility (in the personhood sense) and the concept of personal identity is not immediately clear. I will return to this point later. Although Honoré does not explicitly distinguish between the social and the personhood understandings of outcome-responsibility, there nonetheless seems to be, as I have suggested elsewhere,3 a tension between the two. The personhood understanding treats outcome-responsibility as a morally fundamental notion in the sense that “we can neither choose [responsibility and personal identity] nor choose to give them up” (30). But when he speaks of 3. Stephen Perry, “The Moral Foundations of Tort Law”, Iowa L Rev 77 (1992), 449 at 488–96.

63

 .  the system of outcome allocation, i.e., outcome-responsibility in the social sense, Honoré says that “[b]y allocating credit for the good outcomes of actions and discredit for bad ones, society imposes outcome-responsibility” (14). The idea that outcome-responsibility is socially imposed would seem to be very much at odds with the claim of the personhood understanding that we could not dispense with the concept without ceasing to be persons, since if this latter statement is true then outcome-responsibility is simply not the sort of thing that society could impose; it is too fundamental a notion to be the subject of a social decision.4 As Honoré himself puts the point, we could not make a bargain to assume responsibility in exchange for personal identity, because “the exchange would not even, properly speaking, be in our interest; for to be responsible is part of what it means to be a person and hence to have interests” (30). If outcome-responsibility is as fundamental a moral notion as this suggests, then not only is talk of adopting or retaining it as a social system out of place, but so is talk about whether such adoption or retention is justified. This is true whether the proposed justification is Honoré’s fairness idea, or whether it takes some other form. As Honoré says, outcomeresponsibility “bind[s] us willy nilly” (29). In a later article, entitled “The Morality of Tort Law”,5 Honoré attempts to clarify some of these matters. First, he brings in a principle of corrective justice, which had not made an appearance in “Responsibility and Luck”. This principle “requires those who have without justification harmed others by their conduct to put the matter right” (73). (Honoré calls this a “wide” characterization of corrective justice because reasons other than fault may support a duty to compensate.) As for outcome-responsibility—by which Honoré apparently means, in the passage about to be quoted, outcome-responsibility in the

4. Peter Cane has suggested to me that the two notions of outcome-responsibility can be reconciled if personhood is “socially constructed”. It is not entirely clear to me, however, what it means to say that personhood is socially constructed. Whatever it means, I do not believe that such a claim could serve to reconcile the two understandings of outcome-responsibility unless society “constructs” personhood in very much the same way that it “constructs” (on Honoré’s account, at any event) the social system of outcome-responsibility, namely, by adopting that system. Even if there was never at the outset an explicit social decision to adopt a social system of outcome-responsibility—perhaps, like certain conventions, it just developed—it is nonetheless clear that Honoré thinks there is a standing possibility of change, i.e., a standing possibility that society might make an explicit decision (presumably a political one) to alter the current system. (See further the discussion below on pp. 67–68.) This would be an extremely implausible claim to make about personhood, even if that notion turned out to be in some sense socially constructed. 5. Tony Honoré, “The Morality of Tort Law—Questions and Answers” in Honoré, Responsibility and Fault, above n. 1, 67–93. As noted earlier, references to this book are given parenthetically in the text.

64

     personhood sense—he suggests that it applies to all outcomes, both good and bad, that my actions brings on myself and others: “I win the one hundred metres and you lose. Outcome responsibility makes me responsible for your defeat as well as my victory. But the nature of the race justifies me in inflicting that setback on you. The same is true of other forms of competition, for example in trade, business, politics, literature, and love. If some succeed, others fail. When, however, there is no justification for inflicting a loss on another, outcome-responsibility supports the claims of corrective justice. Since I am responsible for a loss inflicted on you without justification, I have a duty to answer for what I have done, and to make whatever amends are appropriate to the situation”. (78)

Corrective justice thus involves a subset of instances of outcome-responsibility (in what I am calling the personhood sense). This subset consists of losses that, without justification, my actions caused others to suffer. Similarly, justified impositions of tort liability constitute a subset of instances of corrective justice. The state may properly impose tort liability and compel me to make good your loss, “if my conduct was undesirable and your loss an infringement of your rights, provided that to do so is not inconsistent with other values important to maintain” (78). Honoré goes on to say that although outcome-responsibility “supports” the wider view of corrective justice, “we must note that the justification for imposing outcome-responsibility on those who cause harm to others rests not on corrective justice but rather on distributive justice” (78). He continues: “Perry rightly points to the distinction between outcome-responsibility from the agent’s point of view—something that helps him foster a sense of his personal identity, character and history—and outcome-responsibility as a justification for holding people liable to others for the harmful outcome of their conduct. But I do not agree with him that these two aspects of outcome-responsibility are inconsistent. The argument for holding people responsible to others for harmful outcomes is that it is fair to make the person to whom the advantages will flow from an uncertain situation over which he has some control (or which he has chosen to enter into) bear the losses that may equally flow from that situation”. (78–9)

Honoré continues that this “principle of risk” rests on a form of distributive justice, distributing throughout society the risks of harm attributable to human conduct: “It places on every member of the community the burden of bearing the risk that his conduct may turn out to be harmful to others, in return for the benefit to himself that will accrue should his conduct turn out as he plans” (80). 65

 .  It is not entirely clear, however, what the relationship of this distributive risk principle is to the general concept of outcome-responsibility. As was noted in the preceding paragraph, Honoré seems to suggest that outcomeresponsibility in the personhood sense applies to all outcomes that causally flow from my conduct, including both my victory in the race and your defeat. To the extent that outcome-responsibility plays a role in constituting our identities as persons, this makes some sense. But the risk principle applies to harmful outcomes only—indeed, it apparently applies only to legal liability for harmful outcomes—and its stated justification rests on a notion of fairness. We seem to be back in the realm of the social understanding of outcomeresponsibility. In the article to which Honoré is referring,6 I did not mean to say that the two understandings of outcome-responsibility were inconsistent, but only that they seemed not to be the same thing. That point still holds. It is simply not clear what outcome-responsibility in the social sense—now stated more precisely as the risk principle—has to do with the personhood understanding. One could make Honoré’s argument about risk distribution without even referring to outcome-responsibility in the latter sense. I will come back to outcome-responsibility in the personhood sense in section II. First, I would like to discuss the risk principle in somewhat greater detail. Although, as already noted, this principle seems to be a more precise characterization of the social understanding of outcome-responsibility that was introduced in “Responsibility and Luck”, there are also a number of differences. First, the argument applies to “losses” and “benefits” rather than to some general notion of social credits and gains. This strikes me as an improvement, since while there is a definite sense in which I benefit from making the U-turn if I get to my destination more quickly without having an accident, there is no clear sense in which I receive a “social credit” for having done so. Similarly, if I botch the U-turn and cause an accident, it makes little sense to speak of a discredit unless I am legally required to pay compensation for the other person’s loss. Still less does it make sense to speak of a social discredit if I do not have an accident but simply mistake the route and take longer to get to my destination. I have certainly suffered a loss of some sort in this latter circumstance, but it is implausible to think that society cares one way or another about my plight, let alone that it frowns on me for getting into it in the first place.7 6. Perry, “The Moral Foundations of Tort Law”, above n. 3. 7. Peter Cane has pointed out to me that conduct that does not attract legal liability might attract other forms of social discredit. This is of course true; it is merely an attribute of the particular example of the U-turn that I receive no non-legal social discredit—in the form, say, of general disapprobation—for taking longer to arrive at my destination than I expected. The main point

66

     The second difference between the risk principle and the original argument in “Responsibility and Luck” for “imposing” outcome-responsibility concerns the notion of fairness. As we saw, fairness is taken to mean, in the earlier article, that the system of outcome allocation must be impartial, reciprocal, and over a period beneficial (26). It is the last of these conditions that is the crucial one, since it holds that “[a]ll those who possess a minimum capacity stand to profit from the system of outcome allocation most of the time and, if there is a minority of permanent losers, they teeter on the edge of incapacity” (27). Unless minimum capacity is simply defined as not standing to profit ex ante from the system of outcome allocation—and Honoré comes close to doing this in a footnote—the claim that the system is beneficial to all seems to be, empirically, simply false. We all know people, or know of people, who apparently possess whatever minimum capacity is required to get by in the world and be properly regarded as a person, who nonetheless seem to be (and to be destined from the outset to be) life’s perennial losers. At any rate Honoré completely abandons this view of fairness in the second article, in favour of a quite different conception. The argument now is that it is fair to make a person who stands to benefit from his actions also bear the losses that might equally flow from those actions. There is no requirement that all such persons must stand to gain more than they stand to lose from this arrangement. This second version of the fairness argument bears a strong resemblance to certain aspects of libertarian legal and political philosophy. I will return to that point later. For the moment the point I wish to emphasise is that if the issue is taken to be the fairness of adopting one social arrangement over another, it is not enough to argue that a particular system is fair tout court. It is necessary to argue that the proposed system is the fairest of various possible arrangements, or, perhaps more generally, that it is the best possible arrangement taking account of fairness and other values. This is as true of the beneficial-to-all claim as it is of the claim that it is fair to make people take the bad with the good. We cannot evaluate the fairness of possible social arrangements in isolation from one another, and indeed towards the end of “The Morality of Tort Law” Honoré does consider the fairness of alternative here is that the existence of social credits and discredits other than legal liability seems quite haphazard and unsystematic, and does not seem to be centrally related to the general problem of responsibility for outcomes. The reason we are concerned with responsibility for outcomes is that some outcomes are beneficial for oneself, while others are harmful either to oneself or to other people. It should be remembered in this regard that outcome-responsibility is, according to Honoré, one of the theoretical concepts that underpins tort law, and the central concern of tort law is the allocation not of social discredits but rather of harm-caused losses.

67

 .  arrangements. He suggests that there is an argument for transferring the whole or the part of the duty to compensate from the harm-doer to the taxpayer or the contributors to an insurance fund: “If it is fair for everyone to have to contribute through taxes to the defence of the country, since everyone in the country benefits from its being defended, so it is fair for everyone who owns or drives a vehicle, or who benefits from the existence of a transportation system, to contribute to the accident costs that such a system carries with it. . . . The level at which risks should be distributed in a particular area of community life seems pre-eminently a matter of political judgment”. (91)

Notice that these fairness arguments are, like the risk principle that apparently lies at the core of outcome-responsibility in the social sense, based on a conception of what constitutes a just distribution of risks. In other words, Honoré is acknowledging that there are a number of different riskdistributing arrangements that can plausibly be envisaged, and also that there is a social choice to be made among them. At least one of the grounds of choice, if not the sole ground, is the relative fairness of these competing arrangements, as viewed from the perspective of distributive justice. It is true that the system of outcome allocation—i.e., outcome-responsibility in the social sense, now understood as resting on (or perhaps coinciding with) the risk principle—assigns losses to particular individuals who causally contributed to those losses. But there is nothing morally fundamental about this. It is just one possible social arrangement among others, where the choice of which one to adopt is to be made at a political level. Outcome-responsibility in the personhood sense, not to mention the principle of corrective justice, seem to have entirely disappeared from view. I should emphasise that I am not offering a criticism of what Honoré has to say about outcome-responsibility in the social sense. I am just pointing out that (1) social outcome-responsibility seems to have little to do with, and certainly does not presuppose or flow from, outcome-responsibility in the personhood sense; and (2) it must be regarded as competing with other social arrangements that are true practical alternatives and not just theoretically imaginable constructs. There is a strong moral case to be made for a New Zealand-style compensation system or for a compulsory no-fault auto insurance plan, and of course there exist jurisdictions that have substituted one or another of these social arrangements for at least part of tort law. Honoré’s particular conception of social outcome-responsibility is based on the risk principle, i.e., on the idea that it is fair to require individuals to bear the costs of their actions in return for the right to keep any benefits. As was noted earlier, the risk principle, thus understood, is strongly reminiscent of 68

     the libertarian approach to tort law, particularly as embodied in the early work of Richard Epstein.8 Libertarians in fact envisage the right to compensation as morally fundamental rather than as the upshot of a fair, politicallydetermined distribution of risks. As I have argued elsewhere, the morally fundamental character of this libertarian right is best regarded as deriving from the idea of self-ownership—i.e., the idea that one owns one’s physical person and one’s powers of agency—together with the idea that one also morally “owns” the outcomes, both good and bad, that causally flow from the exercise of one’s agency.9 So far as beneficial outcomes are concerned, the resulting principle is that one is entitled to keep the fruits of one’s labour. So far as harmful outcomes are concerned, the resulting principle is that the actor has a moral obligation to compensate for all harms to others that were caused by her actions. The standard of liability, in other words, is, according to libertarianism, absolute, based on causation alone. But the libertarian view is problematic because the underlying notion of ownership of outcomes conceptually requires that outcomes be uniquely attributable to one responsible party. Because causation cannot plausibly be regarded as a basis for unique attributions of this kind, a problem of indeterminacy arises. I will elaborate more fully on this in section II. Because A’s right to be compensated for harm that B has caused her, together with B’s corresponding obligation to pay A compensation, are viewed by the libertarian as morally fundamental, libertarian legal theory is best regarded as adopting a conception of outcome-responsibility in the personhood sense. I will argue for this point more fully in section II. Honoré’s risk principle is a conception of outcome-responsibility in the social sense, but even so it seems to resemble the libertarian theory in its content: “[I]t is fair to make the person to whom the advantages will flow from an uncertain situation over which he has some control (or which he has chosen to enter into) bear the losses that may equally flow from that situation” (78–9). On the surface, at least, this seems to refer to all advantages and all losses, in which case Honoré’s social conception of outcome-responsibility is as subject to the indeterminacy problem as is the libertarian conception. But Honoré would no doubt wish to invoke the theory of causation that he developed with H.L.A. Hart in Causation in the Law,10 and it may well be that the risk principle, 8. Richard A. Epstein, “A Theory of Strict Liability”, J Leg Stud 2 (1973), 151. 9. Stephen R. Perry, “Libertarianism, Entitlement, and Responsibility”, Phil & Pub Aff 26 (1997), 351 at 358–73. In the same article I contrast a fairness-based argument for the libertarian position—an argument not dissimilar to Honoré’s—with the morally more fundamental argument based on self-ownership. Ibid. at 368–70. 10. Oxford: Clarendon Press, 1st edn 1959, 2nd edn 1985.

69

 .  when constrained by that theory, does not give rise to indeterminacy after all. Although I cannot defend this view here, I believe that the theory Hart and Honoré develop in Causation in the Law is best regarded, not as an account of pure causation, but rather as an account of responsibility. It is, in fact, a partial account of outcome-responsibility. It may well be that the risk principle can be combined with the theory of Causation in the Law to create a complete social conception of outcome-responsibility that is not subject to the indeterminacy problem. These are not matters that I can pursue in the present paper. I raise them simply to make the point that, despite the superficial resemblance of Honoré’s social understanding of outcome-responsibility to the libertarian conception, the former may not suffer from the deep flaws of the latter.

II

The main point that I wished to make in the preceding section is that Honoré has two different understandings of outcome-responsibility, and that these are independent of one another.11 More precisely, the discussion in section I demonstrated that the social understanding does not depend on the personhood understanding. In this section, it is the personhood understanding that will take centre stage. Although I will not argue for the following conclusion directly, one of the points that will emerge is that the personhood understanding is independent of the social understanding. The social understanding of outcome-responsibility is not, as we have seen, a morally fundamental notion. Rather it derives from a certain kind of social decision. In addition, the argument for the risk principle envisages outcomeresponsibility as legal liability (which is of course socially-determined), and nothing more. The personhood understanding, as Honoré characterises it, is 11. It is worth pointing out that an observation along somewhat similar lines applies to the theory of tort law developed in Jules Coleman and Arthur Ripstein, “Mischief and Misfortune”, McGill LJ 41 (1995), 91. Coleman and Ripstein argue that “both tort law and the institutions of distributive justice can be understood as responses to the question: who owns which of life’s misfortunes?” Ibid. at 94. In Stephen Perry, “The Distributive Turn: Mischief, Misfortune, and Tort Law”, in Brian Bix (ed.), Analyzing Law: New Essays in Legal Theory (Oxford: Clarendon Press, 1998), 141, I argue that Coleman and Ripstein’s notion of “ownership of misfortune” is ambiguous. Sometimes it refers to what I call “strong ownership”, which is a morally fundamental notion that is roughly analogous to Honoré’s personhood understanding of outcomeresponsibility. Sometimes Coleman and Ripstein use “ownership of misfortune” to refer to what I call the “allocation” of misfortune, which is a concept in political morality roughly analogous to Honoré’s social understanding of outcome-responsibility. Coleman replies in Jules I Coleman, “Second Thoughts and Other First Impressions”, in the same volume, 257 at 297–316.

70

     different on both counts. It is morally fundamental, at least in the sense that we could not dispense with it without ceasing to be persons. Moreover, it is a broad form of moral responsibility the existence of which is a necessary but not, usually, a sufficient condition for legal liability. As Honoré says, “responsibility for a harmful outcome should not automatically involve a duty to compensate” (27). In “The Morality of Tort Law” he characterizes the further principle which is required to ground a duty to compensate as corrective justice. There is thus a two-step process involved: first, a finding that the defendant is outcome-responsible (in the personhood sense) for the plaintiff ’s loss, and second, a determination that the principle of corrective justice applies so as to place the defendant under an agent-relative obligation to compensate the plaintiff for her loss. I will not have much to say about corrective justice in the present paper.12 Instead, I will focus on the question that both Honoré and I regard as normatively prior to questions about the nature of corrective justice, namely, what is outcome-responsibility in the personhood sense? Honoré has several things to say about this. For one thing, he says that “outcome allocation is crucial to our identity as persons” (29), by which he apparently means, as was noted in section I, that it is crucial to our identity as individuals: “If actions and outcomes were not ascribed to us on the basis of our bodily movements and their mental accompaniments, we could have no continuing history or character” (29). Bodies would exist and each would have a certain continuity, “[b]ut having decided nothing and done nothing these entities would hardly be people” (29). “Finally, it is outcomes that in the long run make us what we are” (31). Honoré is surely right that, in the long run, it is outcomes that make us what (and who) we are. It is the outcomes of actions that give us a history as persons, and our history as persons contributes in very large measure to our identities as individuals.13 However, it does not follow from the fact that outcomes make us what we are, that we are responsible in some normative, i.e., reason-changing, sense for those outcomes. Outcome-responsibility does not derive from identity; if anything, outcomes contribute to our identity precisely because we are responsible for them. So the point about identity does 12. See Stephen R. Perry, “On the Relationship between Corrective and Distributive Justice”, in Jeremy Horder (ed.), Oxford Essays in Jurisprudence: Fourth Series (Oxford: Clarendon Press, 2000), 237. 13. Arthur Ripstein, in his contribution to the present volume (“Private Law and Private Narratives”), emphasises the “private narrative” aspect of Honoré’s discussion of outcomeresponsibility. In my view, Ripstein fails to see that the fundamental concept underpinning Honoré’s understanding of outcome-responsibility is not the notion of first-person narrative but rather that of third-person attribution; outcomes are attributed to agents, who are then properly regarded as the authors of those outcomes.

71

 .  not help us answer the question, what is outcome-responsibility in the personhood sense? A clue emerges, I think, when Honoré says that human bodily movements and their mental accompaniments are ordinarily interpreted as actions and decisions, which means that they are “ascribed to authors, who accordingly count as persons” (29). In a similar vein he maintains that “people who possess adequate general capacities are the authors and originating cause of their conduct and its outcome” (39). In other words, outcome-responsibility is a moral notion (and perhaps, as Honoré at one point suggests (29–30), a pre-moral notion as well)14 for ascribing, or attributing, outcomes to human beings, who accordingly count as authors of both the outcome and the originating action. The crucial concept here is agency. We do not just move about in the world; we perform actions. Actions (almost) invariably produce change in the world—outcomes—even if the change is no more than a shift in the position of our bodies. Outcomeresponsibility is the concept that links outcomes to agents qua agents. If someone is outcome-responsible for a particular outcome, it means we can say that not only was there a causal connection between her action and the outcome, but that she did it; she is appropriately regarded as its author. The concept of outcome-responsibility is a necessary concomitant of the concept of agency, since not every outcome to which our actions causally contribute is properly attributable to us as agents; outcome-responsibility is the concept we use to draw the line between those outcomes that can be so attributed and those that cannot. Outcome-responsibility is crucial not so much to our identity as persons as it is to our status as persons, where the critical aspect of personhood is the fact that we are agents. Of course something more must be said about how outcome-responsibility goes about drawing the line between those outcomes that can be attributed to our agency, and those that cannot. I will return to this question shortly. First, however, I would like to try to place the concept of outcome-responsibility in 14. It is not entirely clear what Honoré means when he writes that “the normative principles [of personhood and responsibility] may be regarded in a pre-moral sense as well-founded, since they embody a balance between identity and responsibility” (30). I take it that the main idea is that outcome-responsibility may have normative roots that go deeper than morality. If that is so, it is worth noting that that fact does not call into question the status of outcome-responsibility as a morally fundamental principle. Outcome-responsibility is clearly a moral concept, and it is morally fundamental insofar as it does not directly derive from some other moral notion or notions. (This does not rule out the possibility that a number of different, and equally fundamental, moral concepts might be properly characterized by reference to one another.) The fact that outcome-responsibility does not derive directly from some other moral notion does not, of course, entail that it has no normative foundations at all.

72

     a more general moral framework.15 One familiar form of responsibility concerns responsibility for our actions as such. Let me refer to this general concept as action-responsibility. Theories of action-responsibility set out rival accounts of when and why an action should be regarded as blameworthy or deserving of punishment. Perhaps, it might be suggested, outcome-responsibility in the personhood sense is not a morally fundamental notion after all, but is really just derived from, or a form of, action-responsibility. After all, the concern of outcome-responsibility is not with outcomes considered just as states of the world, but rather with outcomes that have been brought about by human agency. In addition, we can only refer to a specific action by employing a description of it, and the descriptions of many actions make reference to an outcome of some kind. Thus an act of killing includes, under that description, the outcome of someone’s death. If, moreover, a particular act of killing was blameworthy, it seems intuitively clear that the killer would also be outcomeresponsible for the death. But despite the connections that undoubtedly exist between outcomeresponsibility and action-responsibility, it seems unlikely that the former will turn out to be normatively equivalent to, or just a special case of, the latter. For one thing, the same action can generally be described in different ways, some of which make no reference to a harmful outcome. Thus the action of killing John can also be described as the action of, say, firing a gun at John. The action is presumably still blameworthy under this latter description, even though no mention is made of the outcome of death. Conversely, an outcome for which it could well seem appropriate to hold a person responsible might simply be too remote from the originating action to count, under any plausible description, as part of the action. More importantly, it seems intuitively plausible to think that a person can at least sometimes be held outcomeresponsible for the consequences of a non-blameworthy action. A related point is that action-responsibility appears to find its most appropriate legal expression in criminal punishment, whereas the natural legal home of outcome-responsibility would seem rather to be the remedy of compensatory damages. In other words, outcome-responsibility is more properly associated with the law of torts than it is with the criminal law. Thomas Nagel has suggested that there are three broad sources of agentrelative, non-consequentialist reasons for action.16 These are deontology, autonomy, and obligation. Deontology concerns the moral constraints on 15. In this and the following paragraph, I draw on my article “Libertarianism, Entitlement, and Responsibility” above n. 9, at 354. 16. Thomas Nagel, The View from Nowhere (New York: Oxford University Press 1986), 164–88.

73

 .  what an agent may do to other people, or on how he may treat them. (Actionresponsibility, which centres on the notion of culpability, figures in the determination of when conduct violates deontological constraints or other norms, or at least in the determination of when norm-violating conduct is subject to blame or punishment.) Autonomy concerns the interests and desires an agent may have that generate reasons for action for him, but do not generate impersonal values; thus the personal value to me of climbing Mount Kilimanjaro may give me a reason to perform that particular action without giving anyone else a reason to help me succeed. By obligation, Nagel means the special obligations that we owe to, for example, members of our families. I wish to suggest that outcome-responsibility is a fourth source of agentrelative, non-consequentialist reasons for action.17 The basic idea is that whether or not the causally-precipitating action is subject to a deontological constraint or is otherwise morally prohibited or blameworthy, our status as moral agents requires us to acknowledge certain of the outcomes of our choices to act as ours; we are their authors, and they are properly attributed to us as agents. It is not just that we may choose to acknowledge such outcomes as ours; we are required to do so. This can affect our reasons for action in various ways. As Honoré says, sometimes we just have a reason to apologise or ring for the ambulance or comfort the injured (27). Sometimes, however, outcome-responsibility can serve as the basis of a moral obligation to compensate. There are different theoretical accounts of how outcomeresponsibility can give rise to such an obligation, and indeed there are different theoretical accounts of outcome-responsibility itself. I will have something to say about these matters shortly. One point that should be emphasized, however, is that because the general concept of outcomeresponsibility does not make responsibility depend on the prior violation of a norm, it allows for the possibility of a non-consequentialist defence of strict liability. We are of, course, not committed to saying that strict liability is ever an appropriate standard of liability until we have spelled out a substantive theory which entails that result. But, as Honoré rightly emphasises, the general notion of outcome-responsibility gives us the conceptual room to develop such a theory. Let me return to the question of how outcome-responsibility goes about drawing a line between those outcomes that are properly attributed to our agency, and those which are not. The most plausible answer, it seems to me, 17. See further Stephen Perry, “Responsibility for Outcomes, Risk, and the Law of Torts”, in Gerald Postema (ed.), Philosophy and the Law of Torts (New York: Cambridge University Press, forthcoming).

74

     has to do with the notion of control.18 If the agent had the appropriate degree of control over an outcome to which her action casually contributed, then she is properly designated its author. Outcome-responsibility is, as I indicated earlier, a general concept, and different conceptions of it are grounded in differing views of what it means to have control over an outcome. There is a range of possibilities here.19 At one extreme, a person is regarded as having had the morally appropriate kind of control if she acted with the intention of bringing the outcome about, either as a means or as an end, and succeeded in doing so. Let me call this the “achievement conception” of outcomeresponsibility. At the other extreme, a person is regarded as having had the appropriate kind of control over an outcome if she acted and, in acting, caused that outcome. This is the libertarian conception. The achievement conception of outcome-responsibility, which is applicable to both good and bad outcomes, is an important moral notion that is relevant not just to our practices of blaming and demanding compensation, but also to our practices of praising and giving credit. It can, however, co-exist with a separate form of outcome-responsibility, which focuses exclusively on harmful outcomes. That is Honoré’s chief concern, and mine as well. The libertarian approach is one conception of this form of outcome-responsibility, but there are others. The basic idea underlying the libertarian approach is that I have a choice about whether or not to become active in the world. If I chose to act on a particular occasion, then I am properly regarded as having had control over whatever harmful outcomes my action caused. I could, after all, have avoided those outcomes simply by not acting, and that is so whether I foresaw the outcomes or not, or even whether they were foreseeable or not. The libertarian account equates outcome-responsibility for a harm that I cause someone else (as opposed to a harm that I bring upon myself ) with a prima facie moral obligation to compensate the victim for her loss. As I suggested in section I, this is because libertarianism regards both the good and the bad outcomes that a person uniquely causes by her actions as an extension of her self-ownership.20 18. At a number of points, Honoré suggests a similar idea. Let me requote a sentence I quoted earlier from “The Morality of Tort Law”: “The argument for holding people responsible to others for harmful outcomes is that it is fair to make the person to whom the advantages will flow from an uncertain situation over which he has some control (or which he has chosen to enter into) bear the losses that may equally flow from that situation” (79, emphasis added). This is, of course, Honoré’s main argument for the risk principle, and for his particular conception of outcomeresponsibility in the social sense. But the idea of control which he introduces here is, I believe, the crux of outcome-responsibility in the personhood sense. 19. I discuss these at greater length in Perry, “Responsibility for Outcomes, Risk, and the Law of Torts”, above n. 17, at 81–90. 20. Perry, “Libertarianism”, above n. 9, at 363–73.

75

 .  The libertarian conception of outcome-responsibility gives rise to a number of problems. I have discussed these at length elsewhere,21 but let me briefly go over the two main difficulties. The first is that, in general, a harmful outcome to one person cannot plausibly be said to have been caused uniquely by the actions of someone else. Unintentional harm of a kind potentially subject to liability in tort is typically caused, on any plausible conception of causation, by actions of both parties; such cases involve a harmful interaction between persons rather than one person unilaterally acting upon, and thereby causing harm to, another. Ronald Coase made this point some time ago in connection with the economic analysis of externalities,22 but it is relevant to non-consequentialist theories of tort law as well. This fact about causation threatens any theory of absolute liability with serious indeterminacy. The second difficulty that the libertarian conception of outcomeresponsibility encounters is related to the first. The focus now, however, is on the appropriate understanding of control rather than on the appropriate understanding of causation. Libertarianism claims that a person has control over the harmful outcomes she causes, even if she could not foresee those outcomes, because she could have chosen not to act in the first place. She had, so to speak, a second-order choice about whether or not to become an active being in the world. This assumes there is a morally nonarbitrary distinction between activity and passivity that is marked by the concept of causation: one, active party unilaterally acts upon and causes harm to another, passive party. The problem is that this picture of human interaction is mistaken not only from a conceptual point of view—because it involves an erroneous understanding of causation—but also from a moral point of view. There is no morally meaningful choice about choice. Even a choice to do absolutely nothing, which no one can do for very long anyway, has potential consequences for others that render a purely temporal distinction between activity and passivity quite arbitrary. Harm that arises from an interaction between persons is typically caused, as I have already noted, by actions of both parties, and responsibility for the harm is not determined by the fact that one of those actions was later in time than the other. In a moral sense, all autonomous persons must be regarded as active beings at all times. There are, as I said earlier, other conceptions of outcome-responsibility for harmful outcomes to others. One family of such conceptions we can call 21. Stephen R. Perry, “The Impossibility of General Strict Liability”, Can J of L & Juris 1 (1988), 147; Perry, “Libertarianism”, above n. 9. 22. Ronald Coase, “The Problem of Social Cost”, J Law & Econ 3 (1960), 1.

76

     advertence-based. The basic idea is that I can only be said to have had control over an outcome if I was aware of, or adverted to, the possibility of its occurrence at the time that I acted. Only under those circumstances, the argument runs, could I have acted differently, and hence only under those circumstances can I have been said to have had control over the outcome.23 Intuitively, this seems much too narrow a conception of control, and hence much too narrow a conception of outcome-responsibility. Subjective awareness of the risk of harm is perhaps a condition of blameworthiness, and so perhaps an important element of action-responsibility. But it is too strong a condition for outcome-responsibility, because we have a strong sense that an individual can sometimes be responsible for harm even if she did not anticipate it. The best way to demonstrate that this is too strong a condition is to present an alternative conception of what it means to have control over an outcome, and, more particularly, to present a conception that does not require the agent to have adverted to the possibility of harm at the time of acting. Elsewhere I have argued for what I call the avoidability-based conception of outcome-responsibility,24 which rejects the claim that control over an outcome requires advertence to that outcome at the time of acting. According to the avoidability-based account, an agent is outcome-responsible for a harmful outcome if and only if he causally contributed to it, possessed the capacity to foresee it, and had the ability and opportunity to take steps, on the basis of what could have been foreseen, to avoid it. It is clear, I think, that both the libertarian and the advertence-based conceptions of outcome-responsibility take control over an outcome to reside in the agent’s ability to avoid the outcome. But libertarianism adopts too broad an understanding of avoidability, whereas the advertence-based approach adopts too narrow an understanding. The avoidability-based conception suggests, contrary to what libertarianism maintains, that both control and avoidability have an epistemic dimension. There is, to be sure, a sense in which a young child who turns off a tap avoids the outcome of the bathtub overflowing, even though he does not possess the cognitive capacity to appreciate the connection between what he has done and its consequence.25 If he avoided the outcome, then he must in 23. For an argument along these general lines see, Larry Alexander, “Reconsidering the Relationship among Voluntary Acts, Strict Liability, and Negligence in Criminal Law” Social Phil & Policy 7 (1990), 84 at 99–101. Coleman and Ripstein similarly assume that one can have control over a consequence only if the consequence was “intended or actually foreseen”. Coleman and Ripstein, above n. 11, at 127. Alexander uses this claim in order to argue for a subjective negligence standard. Coleman and Ripstein use it to argue for an objective standard that is not based on the notion of control. 24. See Perry, “Responsibility for Outcomes, Risk, and the Law of Torts”, above n. 17. 25. I owe this example to Jeremy Waldron.

77

 .  some sense have possessed the power to avoid it. This is not, however, the sense of avoidability with which outcome-responsibility should be concerned. At the same time, the avoidability conception does not goes so far as the advertence-based theories and insist that the epistemic aspect of avoidability involves actual awareness of the possibility that the outcome might come about.26 The power of avoidance depends, rather, on a general capacity to avoid. That general capacity is characterised, in turn, as a capacity to foresee an outcome and a related ability to take steps, on the basis of what could have been foreseen, to avoid it. The epistemic aspect of avoidability resides in the fact that the ability to take preventive steps depends in this way on the capacity to foresee. It depends, in other words, on the capacity to understand that in turning off the tap one is preventing the bathtub from overflowing. One does not possess the requisite ability to avoid an outcome merely because one has the power to act in a way that would physically stop the outcome from happening. It might be argued that the general capacity to foresee an outcome must be construed, once we have rejected an advertence-based approach, as the capacity of a reasonable person, and that the capacity of a reasonable person must be understood in terms that are independent of the individual agent’s abilities. There is another possibility, however, which is Honoré’s own characterisation of a general capacity. This is, in my opinion, his other great contribution to tort theory, in addition to the idea of outcome-responsibility itself. Honoré first introduced this idea in 1964 in his article “Can and Can’t”,27 and subsequently developed it further in “Responsibility and Luck”. The central claim is that “[g]eneral capacities can be measured by how people generally perform when they try to execute a given type of action, like shutting the door or crossing the 26. Arthur Ripstein has argued that my account of outcome-responsibility “is essentially a firstperson perspective on agency”, and that for that reason it cannot be reconciled with the objectivity of the foreseeability requirement in tort law. Arthur Ripstein, Equality, Responsibility, and the Law (New York: Cambridge University Press, 1999), at 100. He continues: “The root of the problem is that the idea of foresight relevant to outcome responsibility depends on its actual exercise. A person’s agency is expressed in her actual exercise of her capacities. That she, or some other, hypothetical person might have exercised them differently does not change what of her was expressed in the deed. A special case of this feature of outcome responsibility is that moral agency is expressed in what a person did know”.

Ripstein goes wrong, however, in construing my account of outcome-responsibility (and, for that matter, the general concept) as involving a first-person perspective on agency. Perhaps Honoré’s idea that outcome-responsibility is concerned with personal identity rests on such a perspective, but, as we have seen, that is not the only strand in Honoré’s thought. Outcomeresponsibility in the personhood sense is not best construed as an expression of one’s agency; it involves, rather, the determination of which outcomes should, and which should not, be attributed to one’s agency. See further n. 13 above. 27. Mind 73 (1964), 463, reprinted as an appendix in Responsibility and Fault.

78

     street or holing a six-foot putt” (38–9). Possession of a general capacity is thus compatible with failure to exercise it on a particular occasion. To say that one could have acted otherwise means that one generally acts otherwise in similar circumstances, not that one could have acted otherwise on this very occasion. As Honoré rightly emphasises, this approach makes the existence of responsibility compatible with determinism, but without ruling on the question of whether or not determinism is true. So far as the avoidability-based conception of outcome-responsibility is concerned, the basic idea is that a person possesses the general capacities to foresee and to avoid harm if she generally succeeds in doing so.28 Of course the characterisation of these general capacities is in part a normative matter. It is necessary to specify the description under which harm must be foreseen, and it is also necessary to identify the minimum level of general success in foreseeing harm that a person must attain before he or she can be said to possess the relevant general capacity. These are normative questions, and it is in answering them that we have recourse to the concept of the reasonable person. In fact, in this context, I believe that the notion of a “reasonable” person is more or less equivalent to that of an “ordinary” person, but that is not an issue that can be discussed here. The point to be emphasized is that individuals are not deemed by morality and the law to possess the capacities to foresee and avoid harm that are associated with an independent normative construct such as “the reasonable person”. Rather law and morality presume that individuals possess the general capacities to foresee and avoid harm that most ordinary persons possess, but the presumption is defeasible.29 Thus it can be shown, by demonstrating that a given individual does not succeed sufficiently frequently in foreseeing or avoiding harm to others, that that individual does not possess the requisite general capacities to foresee and avoid harm, and hence is not capable of being held outcome-responsible in particular cases. I have been discussing in this essay Tony Honoré’s notion of outcomeresponsibility and, to a lesser extent, his notion of general capacity. As will be evident, I do not follow Honoré’s own elaboration of outcome-responsibility in its details, and at a number of critical junctures we strongly disagree. But I hope it is also evident from the essay that these ideas are of fundamental importance to our theoretical understanding of tort law, and, beyond that, 28. I discuss and defend the claims summarised in this paragraph at much greater length in Perry, above n. 17, at 101–8. 29. This was Holmes’s view: “[I]t will now be assumed that . . . the law presumes or requires a man to possess ordinary capacity to avoid harming his neighbors, unless a clear and manifest incapacity be shown”. Oliver Wendell Holmes, The Common Law (ed. M. DeWolfe Howe, Cambridge, Mass.: Harvard University Press, 1963), 88.

79

 .  that they are important and deeply interesting contributions to moral and political philosophy generally. The subject of tort theory has been greatly illuminated and enriched by Honoré ’s work, and I, for one, very much look forward to the further contributions that I have no doubt will be forthcoming from him in the future.

80


5 RESPONSIBILITY AND FAULT: A RELATIONAL AND FUNCTIONAL APPROACH TO RESPONSIBILITY Peter Cane*

1. I

Amongst the most important propositions developed and defended in Tony Honoré’s rich and thought-provoking book, Responsibility and Fault,1 are these: that outcome responsibility is “the basic type of responsibility in a community”2 because it is essential to our identity as persons; that moral and legal responsibility are forms or versions of outcome responsibility; and that by justifying outcome responsibility we go a long way towards justifying strict legal liability. This essay is based on three main arguments about these propositions. First, in concentrating on the relationship between conduct and outcomes, Honoré gives insufficient attention to other important factors relevant to judgments of responsibility in various contexts; and in particular, the nature and quality of the agent’s conduct (which is central to ideas of responsibility in criminal law), the impact of that conduct on other people (which plays an important part in ideas of responsibility in civil law), and the interests that society has in the conduct of its members regardless of outcomes (or, put in a slightly different way, the relevance of social values to judgments of personal responsibility). In short, I want to argue that the idea of outcome responsibility is too focused on the relationship between agency and outcomes to provide a paradigm for responsibility in general. Secondly, I will argue that in treating moral and legal responsibility as versions of outcome responsibility, Honoré pays too little attention to the various functions of responsibility practices. According to Honoré’s account, outcome * I am very grateful to John Gardner and Jane Stapleton for penetrating comments on earlier versions. 1. Oxford: Hart Publishing 1999 (hereafter “RF”). 2. RF 27. RF is primarily concerned with what might be called “historic responsibility”; that is, responsibility for past events. Honoré only touches briefly (128–9) on what might be called “prospective responsibility”—as when we say of a person that they are “a responsible citizen”.

  responsibility is primarily concerned with allocating particular outcomes to particular agents (with “ownership of outcomes” we might say). By contrast, moral and legal responsibility are more concerned with the allocation of praise and blame for conduct and outcomes, with penalising bad conduct and with repairing bad outcomes. In order to understand the relationships between the various types of responsibility, it is necessary to take account of these various functions. My third argument is that Honoré’s project of justifying strict legal liability is compromised by his view that strict liability (like outcome responsibility) is a form of responsibility for bad luck; and by failure to advert to the different functions of outcome responsibility and strict legal liability respectively. Section 2 contains a restatement of Honoré’s theory of responsibility. Section 3 contains comments on various aspects of the theory in order to lay the groundwork for my own approach, which is explained in section 4. 2. T A  R ESPONSIBILITY

AND

F AULT

The essays in Responsibility and Fault contain a wealth of suggestive and complex ideas. I will first restate the parts of Honoré’s argument which are most germane to the topic of this paper. Inevitably, much of the detail will be lost, but I hope to capture the main features of his theory of responsibility. 2.1. Varieties of Responsibility

The focus of Honoré’s theory of responsibility is what he calls “outcome responsibility”. This he distinguishes from moral and legal responsibility on the one hand, and from fault-based and strict responsibility on the other. He regards moral responsibility and legal responsibility as species of outcome responsibility.3 Outcome responsibility is based on causal ideas:4 it is responsibility “for the good and harm we bring about by what we do”.5 Outcome responsibility is pre-moral in the sense that it is “natural” and “crucial to our identity as persons”.6 But it also has a moral basis in the principle of taking the rough with the smooth.7 3. 4. 5. 6.

RF 27. RF 1–2, 10. RF 14. RF 29–30. Honoré’s theory is concerned primarily with the responsibility of individuals. Except in the guise of vicarious liability, this paper does not tackle the difficult issues of corporate and group responsibility. 7. RF 9, 134.

82

   In Honoré’s scheme, moral responsibility is an “enhanced” form of outcome responsibility in the sense that it only attaches to outcomes that are the result of “fault”.8 Fault is conduct which “displays a bad disposition”,9 and “reflects badly on the character . . . of the agent”.10 Moral responsibility attracts “blame”,11 and may generate obligations of repair, but only if what the agent did was “unjustified”.12 Legal responsibility (or “liability”) is of two types—fault-based and strict. It is an enhanced form of outcome responsibility: people are never legally liable merely for having caused harm.13 Responsibility attracts legal sanctions only if what the agent did was “unjustified”.14 Fault can justify onerous sanctions and obligations of repair, provided they are proportional to the fault.15 By contrast, strict liability (i.e. liability “irrespective of fault”)16 is “morally objectionable” if it is too onerous.17 Honoré mentions four types of strict liability: liability for conduct that creates special risks;18 liability for creating risks in order to further one’s own interests;19 vicarious liability (i.e. liability for another’s conduct and its outcomes);20 and liability for failure to comply with an objective standard of conduct by a person who lacked the competence to comply.21 2.2. Responsibility and Fairness: The Distribution of Responsibility

Outcome responsibility rests on the idea that people can be expected to take the rough with the smooth. It is fair to expect a person to take the rough with the smooth if, overall, their conduct “normally”22 produces more good outcomes than bad outcomes.23 This is true of people who possess sufficient (“minimum”)24 capacity for rational decision and 8. RF 27. 9. RF 15. It “discloses a hostile or uncooperative disposition” (RF 27) and “an objectionable attitude to others” (RF 28). 10. RF 19, n 15. 11. RF 121. But see RF 127, 133; and 3.3.1.1. below. 12. RF 78. 13. RF 7. 14. RF 78. 15. RF 83. 16. RF 23. 17. RF 85–7. 18. RF 28. 19. RF 83–4. 20. RF 81. 21. RF 16–23. Concerning the meaning of the word “competence” see n. 34 below and text thereto. 22. RF 27. 23. Taking account of both merited and unmerited good outcomes: RF 28. 24. RF 26.

83

  behaviour.25 People with minimum capacity benefit from a system of outcome responsibility because under it, they have more good outcomes attributed to them than bad outcomes. So it is fair to them. In Honoré’s theory, then, minimum capacity is a necessary condition of fair allocations of outcome responsibility. It is also a necessary condition of fair allocations of (moral and legal) responsibility for fault26—but not a sufficient condition. This is because fault can justify not just attributions of responsibility for outcomes, but also the imposition of onerous penalties and obligations of repair in respect of those outcomes.27 Legal (and moral) concepts of fault require people to meet “objective” standards of conduct defined in terms of what “a hypothetical model person”28 would do or avoid. Some people (whom Honoré calls “shortcomers”)29 may possess the minimum capacity required for outcome responsibility, but lack the physical, intellectual or emotional competence30 needed to measure up to the standard of the hypothetical model person.31 A person who cannot comply with such a standard of conduct is not at fault for failing to do so, because this failure does not reflect badly on their character. To hold such a person responsible for the bad outcome of their failure to comply with the standard is to impose strict liability on them, not liability for fault. A person is at fault only if they cause a bad outcome that they were competent to avoid. According to Honoré, we treat a person as competent to comply with a standard of conduct if they usually succeed in doing so when they try,32 even if their efforts sometimes fail. This is because we know from experience that even competent people cannot exercise their abilities “invariantly and on every occasion”.33 But they have a “general competence”34 to succeed, even if they fail on 25. RF 76, 139. It is relevant that such people can also understand the principles according to which responsibility is allocated: RF 32. In Honoré’s view, we should, as far as possible, treat people as having this minimum capacity because it is crucial to our identity as agents rather than objects; and because even humans who lack this capacity benefit from a system of outcome responsibility: RF 134–5, 142. 26. RF 32. 27. RF 27. 28. RF 17. 29. RF 16. Honoré believes that there are very few such people: RF 29. 30. Honoré distinguishes competence from capacity: RF 17, 36. 31. Lack of competence may also affect a person’s ability to form an intention or to foresee consequences of their conduct. Such a person, although not relieved of responsibility, may be treated relatively leniently by the criminal law: RF 140–1 32. RF 15, 38. 33. RF 36–7. 34. Honoré tends to use the words “capacity”, “competence” and “ability” interchangeably. Thus, for instance, he speaks of “general ability” as well as “general capacity”. For the sake of clarity, I will use the term “capacity” only in the phrase “minimum capacity”. This is consistent with legal usage. I will use the phrase “general competence” to refer to the quality of having both

84

   particular occasions. Applying the principle of taking the rough with the smooth, a system of fault-based (moral or legal) responsibility is fair to people who possess general competence because “it is true by definition that, when they try, they usually perform up to their ability”.35 By contrast, shortcomers are people who, through lack of competence, cannot usually comply with the sorts of standards that people with general competence can meet. To treat shortcomers as being at fault for failure to comply with such a standard is unfair because their attempts to comply are more often unsuccessful than successful. Minimum capacity is a precondition of fair allocations of strict liability.36 Strict liability is an enhanced form of outcome responsibility. “This does not entail that whenever a harmful outcome is properly allocated to someone, this justifies imposing on him a strict liability to compensate for that outcome.”37 To justify adding an obligation to compensate to the allocation of responsibility there must be an “extra element”. One such element is the creation of “special danger”. Strict liability for creation of special dangers is justified as a way of “avoiding serious harm”.38 Vicarious (strict) liability is justified by the principle of taking the rough with the smooth.39 The reason why the creation of risks to further one’s own interests justifies strict liability is that even though not “morally bad”, such conduct is “not morally indifferent; conduct that may affect others cannot be that”.40 2.3. Justifying Responsibility Regimes

Honoré gives two justifications for outcome responsibility. The first is motivational. “To treat people as authors of and responsible for their actions is necessary if we value a society of individuals who respect themselves and one another”.41 A regime of outcome responsibility “makes for a better society

35. 36. 37. 38. 39.

40. 41.

minimum capacity and the competence to measure up, more often than not, to the standard of the hypothetical model person. A shortcomer is a person who possesses (minimum) capacity, but lacks general competence. RF 38. RF 32. RF 27. RF 28. RF 81. There may be an inconsistency in Honoré’s position here, in that he appears to believe both that strict liability is an enhanced form of outcome responsibility and that (strict) vicarious liability is not a form of outcome responsibility because it is not responsibility for what the responsible person has “done”: RF 126. Perhaps his position is that most (but not all) types of strict liability are enhanced versions of outcome responsibility. RF 84. Honoré does not spell out how imposition of objective fault liability on shortcomers can be justified. See 4.2 below. RF 11.

85

  because it encourages us to do well and to enjoy the credit that comes from doing well”.42 Secondly, Honoré suggests what might be called an “ontological” reason for recognising outcome responsibility, namely that it is “crucial to our identity as persons” and “the price to be paid for being a person”.43 He is less explicit about how to justify responsibility for fault. If a person causes a bad outcome that they had the general competence to avoid, they are at fault in relation to that outcome. If a person causes a bad outcome that they lacked the general competence to avoid,44 they are not at fault in relation to that outcome, and liability imposed on them to repair the bad outcome is strict.45 In criminal law, people with substandard competence are sometimes treated more leniently than those with (at least) average competence.46 Thus, in Honoré’s account, a person’s competence is relevant to their responsibility for bad outcomes. If outcome responsibility is the basic form of responsibility in a society, why should degrees of competence be relevant to responsibility? Why should it matter whether or not a person was at fault in producing the outcome? According to Honoré, “it is remarkably difficult to find a satisfactory answer” to this question.47 His suggestion is that life is a competitive struggle,48 a race to be run.49 Those who are handicapped in life’s race by lack of competence deserve to be treated less harshly for producing bad outcomes by their conduct than people of general competence. The conclusion seems to be that the justification for a system of responsibility for fault is that it provides people who possess general competence with an incentive to exercise their abilities to avoid bad outcomes by relieving them of (legal and moral) responsibility for bad outcomes which are not the result of fault.50 The ultimate justification for a system of responsibility for fault is instrumental – it promotes harm-reduction by encouraging people to use their abilities to avoid producing bad outcomes by their conduct. In order to justify strict responsibility, Honoré’s view is that we must first justify outcome responsibility. Once this is done, “it involves no great extension of principle for the law to impose strict liability”.51 42. RF 10. See also RF 125, 131–2 (it encourages good behaviour and discourages bad behaviour) and 140 (it encourages harm reduction). 43. RF 29. See also 125, 127 (it makes possible a sense of personal character and identity). 44. If, in other words, they are a shortcomer in relation to that outcome. 45. See n. 21 above and text thereto. 46. RF 140–1. 47. RF 140. 48. RF 12. 49. RF 140–2. 50. RF 38. 51. RF 40.

86

   3. S G-L C

In this part of the Article I will make a number of comments about Honoré’s approach to responsibility that lay groundwork for my own approach to responsibility, which is explained in section 4.

3.1. Determinism and the Lottery of Life

There are two strands in Honoré’s reasoning to which I have so far given little or no attention. One is causal determinism. Determinism is the idea that every event in the world, including human conduct, has an “antecedent sufficient cause”.52 Some people argue that causal determinism is incompatible with responsibility because it prevents us freely choosing what to do. Honoré’s view is that the truth or falsity of this argument is irrelevant to his theory.53 I agree. Even if causal determinism is true, it remains the fact that we experience freedom of choice; that in living our lives, we feel our conduct to be subject to varying degrees of external and internal limitation and constraint; that people differ in terms of their capacity and competence to achieve good outcomes and avoid bad outcomes;54 that people can change their behaviour patterns; and so on. Responsibility is a human construct; and such facts about human psychology and the way we experience our relationship with the physical world provide an appropriate basis for analysis and evaluation of our responsibility practices.55 It is important not to confuse determinism, which expresses a belief about the way things are, with fatalism, which is a psychological response to the human condition.56 Fatalism would be corrosive of our responsibility practices in a way that determinism by itself is not. A second prominent strand in Honoré’s reasoning which I have largely ignored so far is the idea that life is a sort of lottery: “a series of bets on our choices and their outcomes”.57 My view is that the lottery metaphor unnecessarily complicates the argument. As Honoré acknowledges, it is not clear why we should have to accept the results of a lottery that we did not choose 52. 53. 54. 55.

J.L. Mackie, Ethics: Inventing Right and Wrong (Harmondsworth: Penguin Books, 1977), 216. See esp. RF 135–8. RF 11, 12. Similarly: B. Williams, Making Sense of Humanity (Cambridge: Cambridge University Press 1995), 17–19, 243; T. Scanlon, “The Significance of Choice” in The Tanner Lectures on Human Values VIII (Salt Lake City: University of Utah Press, 1988) 165, 166–72, 207. 56. Similarly Williams, n. 55 above, 9–11. 57. RF 14.

87

  to enter.58 More importantly, it is intrinsic to a lottery that those who stand to win or lose have absolutely no control over the result. By contrast, it is central to the main thrust of Honoré’s position that most people have capacity and competence that they can exercise to achieve good outcomes and avoid bad ones. In fact, the lottery metaphor reintroduces the very determinism that Honoré convincingly argues is of no relevance to understanding our responsibility practices.

3.2. Outcome Responsibility

Let us now turn to a closer examination of the idea of outcome responsibility. I will first argue that Honoré’s account of the scope of outcome responsibility pays too little attention to the reasons why we attribute responsibility, and to the role of luck in our lives. Secondly, I will suggest certain difficulties in the idea that moral and legal responsibility are forms of outcome responsibility which arise out of the focus of the latter on the link between conduct and outcomes at the expense of considering the nature and quality of conduct and the impact of outcomes on “victims”. Thirdly, I will argue that in relation to bad outcomes, our responsibility practices perform at least three equally important (and related) functions concerned with the allocation, respectively, of ownership of outcomes, blame and penalties for bad conduct, and obligations to repair harm; and that outcomes play a different role in relation to each of these functions. 3.2.1. The Scope of Outcome Responsibility

For Honoré, outcome responsibility is responsibility for the good and bad outcomes of a person’s conduct, even those that are the result of good or bad luck;59 and it is fair to people with minimum capacity because, putting together the outcomes of all their conduct, the good outweigh the bad.60 Outcome responsibility, then, is based on the link between conduct and outcomes. Taken at face value, this account casts the net of responsibility 58. RF 29. 59. RF 9. 60. The empirical basis of the principle of taking the rough with the smooth is reasonably secure in relation to general competence. As a matter of observation, we can say that most people can usually comply with the standard of the hypothetical model person when they try. Its empirical basis is less secure in relation to minimum capacity. It is difficult to say whether, on balance, a person’s conduct produces a surplus of good over bad, because of problems of definition and measurement.

88

   extremely widely to embrace all events that can, as a pure matter of fact, be traced back to the conduct in question. However, it seems clear that this is not what Honoré meant; and so two fundamental questions arise: what count as the “outcomes” of a person’s conduct? and when can we justify holding people responsible (or allowing people to take responsibility) for the results of luck? 3.2.1.1. What are the “Outcomes” of Conduct?

A system of responsibility for the outcomes of conduct needs principles, commonly called principles of “causation”, to link particular “outcomes” with particular conduct.61 Indeed, it seems fair to say that Honoré’s idea of outcome responsibility is a causal concept of responsibility. A central proposition of Hart and Honoré’s analysis in Causation in the Law62 is that “what counts as the cause of an event depends on the purpose of the inquiry”.63 So, for instance, in legal contexts, the purpose of causal inquiries is typically to decide whether the defendant ought to be held liable for a particular event. For this reason, causation in the law is responsibility-laden. In non-legal as well as legal usage, the concept of causation has a large evaluative component, at least to the extent that it is used to allocate responsibility for outcomes. For instance, in picking “the cause” of an event from amongst its necessary conditions, human conduct will tend to be identified as its cause in preference to nonhuman (“natural”) occurrences. A natural causal factor will be treated as the cause of an event in preference to human conduct only if it was out of the ordinary. Other things being equal, a human agent is more likely to be identified as the cause of harm than its sufferer.64 Culpable conduct is more likely to be identified as the cause of resulting harm than non-culpable conduct; and more culpable conduct is likely to be identified as a cause in preference to less culpable conduct. A person who fails to prevent harm, or who provides an opportunity for harm to occur, is less likely to be treated as its cause than a person who acts to bring it about. And so on.65 This suggests that for the purposes of attributing responsibility, principles of causation, other than those concerned purely with matters of historical fact, 61. RF 1, 10. 62. H.L.A. Hart and Tony Honoré, Causation in the Law (2nd edn, Oxford: Clarendon Press, 1985). 63. RF 3. 64. There are some difficult problems of definition and attribution lurking here: P. Cane, “Retribution, Proportionality and Moral Luck in Tort Law” in P. Cane and J. Stapleton (eds), The Law of Obligations: Essays in Celebration of John Fleming (Oxford: Clarendon Press 1998), 150–1. 65. P. Cane, “Fault and Strict Liability for Harm in Tort Law” in W. Swadling (ed.), The Search for Principle: Essays for Lord Goff of Chieveley (Oxford: Oxford University Press, 1999), 194–7.

89

  form a subset of principles of responsibility, namely those that are concerned, in particular, with the connection between conduct and its consequences.66 For responsibility purposes, it is not possible to specify what are the outcomes of conduct independently of “principles of causal responsibility”. Such principles focus attention on the nature and quality of outcome-producing conduct, and on the relationship between the various causal factors which contribute to outcomes. In order adequately to justify a system of outcome responsibility we need to justify the principles of causal responsibility that underpin it. For the purposes of attributing responsibility, then, principles of causation are inextricably linked with principles of responsibility. And just as what counts as a cause depends on the purpose of the causal inquiry, so the answer to the question of what we are responsible for will depend on the purpose of the inquiry. The purposes of attributing outcome responsibility, as identified by Honoré, are to encourage us to do our best, and to “make possible a sense of personal character and identity”.67 We may have different reasons for attributing “moral responsibility” or “legal responsibility”; and so what count as the outcomes of our conduct may be different depending on whether we are concerned with our identity as persons, our moral obligations, or our legal liabilities. In short, the scope of outcome responsibility (as of other varieties of responsibility) depends on the purposes it serves. This conclusion should make us wary of Honoré’s contentions that outcome responsibility is the “basic” form of responsibility, and that moral and legal responsibility are “enhanced” versions of outcome responsibility. No doubt, these three forms of responsibility are related in various ways. I would not deny, for instance, that principles of moral responsibility contribute to our sense of personal identity. But to the extent that moral and legal responsibility perform different functions from outcome responsibility, we should not be too quick to conclude that the former are forms of the latter. 3.2.1.2. Outcome Responsibility and Luck

One of Honoré’s main concerns is to justify our practice of holding people responsible for the adverse outcomes of their bad luck.68 Put crudely, “luck” refers to events and outcomes that are outside our control. If causal determinism is true, there is a sense in which, from a human point of view, everything 66. J. Stapleton, “Perspectives on Causation” in J. Horder (ed.), Oxford Essays in Jurisprudence, Fourth Series (Oxford: Clarendon Press, 2000). 67. RF 125. 68. RF 9, 14.

90

   that happens is a matter of luck, because everything is ultimately outside our control. But if it be accepted that the value of our responsibility practices is not dependent on the truth or falsity of causal determinism, we can understand luck in a shallower way as referring to perceived constraints and limitations on our control over our own behaviour and over the circumstances of our lives. Even understood in this way, however, much of our behaviour and its consequences, both good and bad, are matters of luck. As Thomas Nagel puts it, “Whether we succeed or fail in what we try to do nearly always depends to some extent on factors beyond our control”.69 Honoré’s concern to justify responsibility for bad luck is not surprising given his view that people who (luckily) possess minimum capacity are outcomeresponsible for the outcomes of their conduct entirely regardless of the role of luck in producing them, and his view that “ownership” of outcomes of our conduct is crucial to our sense of personal identity. I agree with him that psychologically, the average person needs to be able to claim credit for good outcomes even if they are partly the result of good luck; and if we claim credit for good luck, we must also be prepared to accept responsibility for bad luck. If people could not claim, and did not have to accept, ownership of any conduct or outcome which was affected by luck, we would lose a secure sense of ourselves and others as agents whose conduct can have effects in the world; and we would end up seeing ourselves as “victims70 of circumstance”.71 On the other hand, entirely ignoring the role of luck in our lives would be equally damaging to our sense of personal identity. If people could take credit for all good outcomes of their conduct no matter how lucky, and were held responsible for all bad outcomes no matter how unlucky, we would lose the feeling of being able to act purposively and to influence events. For this reason, it is a virtue to acknowledge the role of luck in our own achievements and in other people’s misfortunes. Having a secure sense of our identity as agents whose conduct can have effects in the world requires us to distinguish between those factors outside our control that should be ignored in allocating responsibility for outcomes, and those factors outside our control that should be taken into account. Consequently, we can claim ownership of some of the lucky outcomes of our conduct, but not all; and we must accept ownership of some of the unlucky outcomes of our conduct, but need not accept all of them. It follows that a theory of outcome responsibility needs to strike a balance between sensitivity and insensitivity to luck if it is to perform its motivational 69. Mortal Questions (Cambridge: Cambridge University Press, 1979), 25. 70. Or passive beneficiaries. 71. RF ch. 4.

91

  and ontological purposes. It cannot simply aggregate all the good and bad outcomes of a person’s conduct regardless of the role of luck in producing them. This conclusion casts doubt on the argument that a system of outcome responsibility is fair to people with minimum capacity because, overall, their conduct produces more good outcomes and than bad outcomes. The fairness of a system of outcome responsibility will depend crucially on the fairness of the rules that determine how the burdens and benefits of luck are to be distributed. A system under which people were held responsible “for [all] the contributions of fate as well as for their own”72 would be just as unfair as a system that relieved us of responsibility whenever fate played a role.73 3.2.2. Outcome Responsibility, Agents and Victims

For Honoré, outcome responsibility is the basic form of responsibility, of which other forms of responsibility are “species” or “enhanced” versions.74 However, notions of responsibility found in the criminal law give reason to doubt this view. In the first place, outcome responsibility is based on the link between conduct and outcomes, and it pays little attention to the nature and quality of the outcome-producing conduct. By contrast, principles of criminal responsibility and of sentencing are much concerned with the nature and quality of the agent’s physical behaviour and mental states. Secondly, although the actus reus of the typical criminal offence is conduct-plusoutcome, there is an important set of crimes the actus reus of which consists of conduct regardless of outcome. Thus there are “victimless” and inchoate crimes; and attempting a crime is itself a crime—a “conduct crime” rather than a “result crime”. This is not to say that outcomes are unimportant to criminal responsibility, but only that their role does not justify viewing them as central to criminal responsibility in the way that they are central to outcome responsibility. Theories of moral responsibility, like principles of criminal responsibility, also give a role to the quality of the agent’s physical behaviour and mental states that Honoré’s concept of outcome responsibility apparently does not. Focusing on the link between conduct and outcomes at the expense of the nature and quality of the outcome-producing conduct makes it difficult, for instance, to explain the relevance to responsibility of differences in capacity and competence between agents.75 72. 73. 74. 75.

Nagel, n. 69 above, 31. See further 4.2. below. RF 27. RF 140–1 and 2.3. above.

92

   Consideration of principles of legal responsibility found outside the criminal law—in the law of tort and contract, for instance—also casts doubt on the view that outcome responsibility is basic. In Honoré’s account of outcome responsibility, outcomes are important by virtue of their link with conduct; whereas in contract and tort law, concern with outcomes focuses attention on victims as much as on the conduct that caused them.76 The purpose of this focus on victims is to justify the imposition of obligations of repair. But even at an ontological level of ownership of outcomes, victims should not be ignored. Responsibility for outcomes is crucial to our sense of personal identity not only as agents. Our identity as individuals depends both on our “achievements and failures”, and on “what has happened to us”.77 At least so far as concerns harm done by one person to another, responsibility is a two-sided affair: it is for something done to someone. If people could disown responsibility for harm done to others whenever bad luck played a part in producing it, this would not only destroy our sense of personal identity as agents, but also our sense of humanity as victims. On the other hand, responsibility to others for all the contributions of fate would not only debilitate us as agents, but also as victims. It is an important part of our identity as individuals to feel that we have some control not only over what others suffer at our hands, but also over what we suffer at the hands of others. It would damage our sense of personal identity to feel that all the harm we suffered was fated, as much as to feel that all the harm we caused was fated. It is not only agents who need to own outcomes. Victims do, too. To summarise: in Honoré’s account, the essence of outcome responsibility resides in the connection between conduct and outcomes. Neither the nature and quality of the conduct, nor the nature of the outcomes and their impact on others, is of much, if any, significance. Nor does the account find room for responsibility for conduct regardless of outcome. So there is good reason to be wary of the idea that moral and legal responsibility are versions of outcome responsibility.

76. In pointing out the importance in civil law of the impact of the outcomes of conduct on victims, I do not mean to suggest that it is unimportant in criminal law (on which see J Gardner, “Crime: in Proportion and in Perspective” in A Ashworth and M Wasik (eds), Fundamentals of Sentencing Theory: Essays in Honour of Andrew von Hirsch (Oxford: Clarendon Press, 1998), 31–52). But its importance is particularly obvious in civil law because civil law sanctions are victim-oriented (similarly Gardner, ibid., 49). 77. RF 76, n. 24 and text thereto.

93

  3.2.3. The Functions of Responsibility Practices

The conclusion to be drawn from the analysis in 3.2.2 is that however well-suited the idea of outcome responsibility may be for performing the motivational and ontological functions assigned to it by Honoré, it does not provide a good model for types of responsibility that are more directed to the performance of other functions. One such function is the allocation of punishment for bad conduct. This is the main function of legal responsibility for criminal conduct. Another function of principles of responsibility is to provide a foundation for obligations to repair, and to make amends for, bad outcomes. This function finds strong expression in the concepts of responsibility found in civil law. The importance of distinguishing between these various functions is illustrated by ambiguity in Honoré’s discussion of whether outcome responsibility attracts obligations of repair. He says that “outcome responsibility for harm to another does not by itself create a duty to compensate. The form that our responsibility for an outcome should take remains an open question. An apology or telephone call will often be enough. But outcome responsibility is a basis on which the law can erect a duty to compensate if there is reason to do so”.78

The reason why outcome responsibility “by itself ” does not create a duty to compensate is that the basic functions of outcome responsibility are motivational and ontological. Its value lies in the incentives it can provide to achieve good outcomes and avoid bad ones, and in “making possible” a sense of personal identity, not in justifying the imposition of obligations to compensate. However, if outcome responsibility by itself does not create an obligation to compensate, there is no reason to think that by itself, it creates any other sort of obligation of repair either. This point undermines Honoré’s project of providing a “moral basis” for strict legal liability which, he says, “involves no great extension” of the principle of outcome responsibility.79 If outcome responsibility does not, by itself, generate obligations of repair, the step from it to strict legal liability is large. It is one thing to say that bad outcomes of our conduct may help to define our individual identities regardless of our fault in producing them; or that being held responsible for bad outcomes regardless of fault may provide incentives to avoid such outcomes. But it is quite another to attach punishment or an obligation of repair to an outcome of conduct regardless of fault. I will return to this issue in 3.3.2.4 and 4.3 below. 78. RF 77–8, 27. 79. RF 40.

94

   3.3. Enhanced Outcome Responsibility

Honoré treats moral responsibility, legal responsibility, fault-based responsibility and strict responsibility all as species, or enhanced versions, of outcome responsibility.80 In 3.2.2 and 3.2.3 I suggested some reasons to question this approach. Against this background, in this section I explore in more detail the relationship between outcome responsibility and these various other forms of responsibility. 3.3.1. Outcome Responsibility, Moral Responsibility and Legal Responsibility 3.3.1.1. Moral responsibility

According to Honoré, moral responsibility for bad outcomes (unlike outcome responsibility) requires fault.81 In his view, to hold someone morally responsible for a bad outcome is to judge them to be blameworthy. This is a common starting point of many philosophical discussions of responsibility. Honoré cites Nagel’s statement that strict liability is “irrational as a moral position”.82 Mackie adopts what he calls “the straight rule of responsibility: an agent is responsible for all and only his intentional actions”.83 Applying this rule, he says, for instance, that we should not hold a person responsible for unforeseen consequences of “inadvertent negligence”.84 The force of the requirement of fault is that it justifies the allocation of blame. Unlike Mackie, Honoré finds no problem in treating responsibility for the unforeseen (but foreseeable) consequences of (inadvertent) negligence, on the part of people who possess general competence, as a form of fault-based responsibility. At first sight, this linking of moral responsibility and fault is attractive. It would seem unfair to “blame” a person for bad conduct or bad outcomes that are not the result of intention, recklessness or negligence on their part. On the other hand, if we focus on the allocation of obligations of repair rather than 80. He says, for example, that “civil liability in law, whether strict or based on fault, can be defended on the ground that it specifies a sanction to be imposed on a person who has anyhow lost a bet and will in consequence incur discredit”: RF 15. 81. RF 27. “Moral blame” is a “narrower notion” than outcome responsibility: RF 121. 82. Nagel, n. 69 above, 31; cited RF 23. See also H.L.A. Hart, Punishment and Responsibility: Essays in the Philosophy of Law (Oxford: Oxford University Press, 1968), 226. Honoré, as we have seen, finds a moral justification for strict (outcome) responsibility in the principle of taking the rough with the smooth. He also suggests moral justifications for strict legal liability. See generally 2.2. and 2.3. above. 83. Mackie, n. 52 above, 208. 84. Ibid., 211.

95

  on blame as a response to conduct that produces bad outcomes, it is less clear that moral responsibility (as distinguished from outcome responsibility on the one hand, and from legal responsibility on the other) is confined to outcomes resulting from fault. For instance, the conductor of an especially dangerous activity may be legally responsible, regardless of fault, for harm produced by the activity and, hence, required to compensate for it. Can we not also say that they may be morally responsible for such harm, and morally obliged to compensate for it, even if the conduct that produced the harm was not blameworthy? At one point, Honoré concedes as much. In most common law systems, parents are not vicariously liable for the torts of their children. Honoré, however, likens the moral responsibility of a parent to pay for the replacement of a window broken by their child to the strict vicarious liability of an employer for torts of an employee.85 Or suppose I “unavoidably” trip someone and cause them an injury. I may not be legally responsible for the harm, says Honoré, and I may not be blameworthy. But still I may “incur a moral obligation”—perhaps to apologise, or to help the person up, or to call for help.86 It seems, therefore, that despite Honoré’s initial linking of moral responsibility and fault, he also believes that there can be a moral obligation of repair even in the absence of fault—moral strict responsibility to make amends, if you like.87 How are we to resolve this apparent inconsistency? I think that the answer lies in rejecting the links between moral responsibility, fault and blameworthiness, and in adopting (as Honoré does)88 a conception of morality that treats it (so far as it is concerned with penalties and obligations) as a noninstitutionalised analogue of law—or, perhaps better, a conception that views law as an institutionalised analogue of morality. From this perspective, moral and legal responsibility are not distinguished from one another by the fact that the former invariably requires fault, while the latter does not. Both outside the law and within it, whether responsibility requires fault depends (partly at least) on the function(s) the responsibility judgment is serving. Both inside and outside the law, the link between fault and punishment is stronger than that between fault and obligations of repair. This helps to explain (for instance) why mental-state requirements (intention and recklessness) play a more prominent role in criminal law than in civil law; why strict liability is 85. RF 126. 86. RF 127, 133. 87. We may, for instance, acknowledge a moral obligation to help the starving in Africa regardless of whether their plight is in any sense our fault or, even, in any meaningful sense an outcome of our conduct. 88. RF 126, n. 14.

96

   more controversial in criminal law than in civil law; why in tort law a person may be vicariously liable for harm even if they were in no way at fault in relation to it; and why liability for breach of contract is typically strict. Similarly outside the law, a person may incur an obligation to repair harm even in the absence of fault in producing it; but blame and punishment regardless of fault are commonly viewed as problematic. I am not suggesting that it is wrong to link moral responsibility with fault —this is only a definitional matter. My point is that such linkage carries with it the danger of failing to notice certain important aspects of our extra-legal responsibility practices, certain functions of those responsibility practices, and certain analogies between our legal and extra-legal responsibility practices. 3.3.1.2. Legal Responsibility

Although he does not expressly say so, Honoré suggests that the particular problem in relation to legal responsibility is to justify the imposition of sanctions—punishment for breaches of the criminal law, and obligations to pay monetary compensation for breaches of civil law.89 For instance, he says that although strict responsibility for outcomes is not unfair to people with minimum capacity, it “should not automatically involve a legal duty to compensate”.90 On the other hand, as we saw in 3.3.1.1, it seems wrong to think that obligations to pay monetary compensation cannot exist outside the law. Honoré apparently agrees: if a child breaks a neighbour’s window, the parents should “pay for the damage” even if they are not legally liable for it.91 More generally, what primarily distinguishes legal and moral responsibility are not the penalties and obligations of repair that may be imposed in their name, but rather the grounds for penalties and obligations of repair that they respectively support.92 The law may make demands on us that morality does not, and vice-versa. In both realms, various sanctions are recognised. The appropriate sanction (if any) for any particular breach of a norm depends partly on the nature and quality of the conduct that breached the norm, and partly on the effect of that conduct on others. 3.3.1.3. Summary

To summarise, the law is not alone in recognising responsibility regardless of fault; nor does the law deploy sanctions unknown outside it. What distinguishes 89. 90. 91. 92.

Re criminal law see RF 67–8, 82–4; re civil law see RF 27, 77–8. RF 27. RF 126. And the fact that whereas the law must tell us clearly and univocally when obligations of repair arise, outside the law we tolerate a significant degree of disagreement about when moral obligations of repair arise.

97

  moral and legal responsibility, on the one hand, from outcome responsibility, on the other, is that the former are primarily concerned with the allocation of blame, punishments and obligations of repair in respect of bad conduct and outcomes, while outcome responsibility is primarily concerned with ownership of outcomes. This is not to say, of course, that these functions are logically independent of one another. As we have seen, blame, punishment and obligations of repair are related in complex ways. More importantly, perhaps, like outcome responsibility, much moral and legal responsibility is responsibility for outcomes. But responsibility for outcomes is not the same thing as outcome responsibility as Honoré explains it. Outcome responsibility is a distinctive form of responsibility for outcomes that focuses on the link between conduct and outcomes; and so it is misleading to call moral and legal responsibility “versions” or “species” of outcome responsibility. This is partly because not all legal and moral responsibility is responsibility for outcomes; but more importantly, because the prime functions of legal and moral responsibility are different from those of outcome responsibility. 3.3.2. Outcome Responsibility, Fault-Based Responsibility and Strict Responsibility 3.3.2.1. Strict Liability and Luck

In Honoré’s account, outcome responsibility “requires us to bear the risk of bad luck”. “This may extend to the risk of having to pay for harm that we inflict without fault.”93 This implies that strict legal liability is a form of liability for bad luck, as is liability, for failure to meet an “objective standard”94 of conduct, imposed on “that minority of shortcomers who cannot attain it”.95 This view is problematic to the extent that it suggests that strict liability is distinguishable by its embrace of bad luck. Even fault liability for harm is, in important respects, liability for bad luck. For one thing, whether a person has general competence (or, for that matter, minimum capacity) is itself, to some extent at least, a matter of luck. For another, a person with general competence may be said to have been at fault even if (and even though), on the occasion in question, they could not have exercised their abilities so as to avoid the harm.96 Indeed, in Honoré’s view, it is implausible to say, in 93. 94. 95. 96.

RF 9. RF 18–20. RF 22–4. RF 34–8 and ch. 7. For this reason, Honoré’s definition of fault, as conduct that reflects badly on a person’s character and displays a bad disposition, is problematic in light of his view that no-one can exercise their capacity and abilities “invariantly and on every occasion”. Fault may reflect badly on a person’s character, but need not.

98

   relation to any particular piece of conduct, that a person could have done otherwise than they did.97 Thirdly, general competence operates as a threshold condition of liability. Amongst those who have general competence, people may, as a result of luck, vary in the abilities and skills they enjoy, and hence in the size of the surplus of good outcomes over bad that they can achieve. But such (un)lucky variations in competence may not be fully reflected in judgments of responsibility. In the law of negligence, for instance, individual differences of skill and ability, amongst adults who possess minimum capacity, are generally ignored for the purposes of allocating liability. The fact of the matter is that all our conduct and its outcomes are, in some respects or other, outside our control. Therefore, it does not get us very far to say that the various forms of responsibility for bad conduct and outcomes are responsibility for, or in spite of, bad luck. Certainly, making clear the “contribution of fate” to our conduct and its outcomes does not provide a point of distinction between outcome responsibility (in Honoré’s sense), responsibility for fault and strict responsibility, because the contributions of fate are ubiquitous. The relevant issue about luck is when it should be ignored and when allowance should be made for it in allocating responsibility. And this is a question that arises in relation to all forms of responsibility for conduct and its consequences.98 3.3.2.2. Strict Liability and Fault

The concept of strict responsibility or strict liability has two different forms. One is responsibility/liability regardless of fault, and the other is responsibility/liability in the absence of fault. Both of these definitions of strict liability are negative. Neither tells us the basis on which strict liability may be imposed. One possible basis is causation of harm; although there are forms of legal strict liability that do not require harm, let alone causation of harm – tort liability for trespass to land, for instance. Honoré’s outcome responsibility is responsibility regardless of fault. Its basis is causation of good and bad outcomes. It is not responsibility in the absence of fault. In other words, absence of fault is not a precondition of outcome responsibility. The same is true of legal strict liability.99 Take vicarious liability as an example. An employer is vicariously liable for harm resulting from torts committed by its employees in the course of their employment, regardless of whether the employer was in any way at fault in relation to the 97. RF 35–36. 98. See further 4.2. below. 99. RF 23.

99

  occurrence of the harm. If the employer was at fault, it may be personally (i.e. non-vicariously) liable for the harm, but this does not prevent the employer being held vicariously liable as well, or instead. Philosophical discussions of responsibility often seem to assume that strict liability is responsibility in the absence of fault. However, legal regimes of strict liability are typically created in order to improve protection of the interests of potential plaintiffs, and so it would make no sense to build absence of fault into such regimes as a precondition of responsibility. Thus, an important justification for strict legal liability is to increase the chance that those at fault will be held liable in the face of difficulties of proof. Again, an important function of the legal concept of a “right” is to justify the imposition of obligations not to infringe the rightholder’s interests either by faulty conduct or otherwise. 3.3.2.3. The Distinction Between Fault-Based and Strict Responsibility

My argument so far is that the distinction between responsibility for fault and strict responsibility turns neither on the role of luck in our lives, nor on the presence or absence of fault. So what is the difference between them? It is this: responsibility for fault rests on a judgment that the responsible person failed to meet a specified standard of conduct, whereas strict responsibility rests on no such judgment.100 As Honoré puts it, a person may be strictly responsible for their conduct and its consequences even though their conduct was perfectly “permissible”.101 Suppose, for instance, that a statute creates liability for selling adulterated food. The liability will be fault-based if it depends on a judgment that the sale of adulterated food was a result of failure by the seller to meet some relevant specified standard of conduct; or that it would not have occurred if the seller had met all relevant specified standards of conduct. Conversely, the liability will be strict if it can be imposed regardless of whether the sale was the result of failure by the seller to meet specified standards of conduct; or of whether it would have occurred if the seller had met all specified standards of conduct. Fault liability is liability for conduct that fails to measure up to relevant specified standards of conduct. Strict liability is liability for conduct regardless of whether it falls short of specified standards of conduct. 100. Similarly: RF 32. 101. RF 23. For this reason, I reject the idea that strict (civil) liability is fair only if it is not too onerous: RF 85–7. This view treats strict liability as liability for a lesser form of fault, and arises from undue focus on the agent at the expense of the victim. The agent-focused nature of criminal responsibility helps to explain the importance attached in that context to proportionality between offence and penalty. Concerning fault and proportionality of obligations of repair in civil law see Cane, n. 64 above, 160–70.

100

   It follows that the imposition on shortcomers of responsibility for failure to comply with standards of conduct that they could not meet (because of lack of general competence) is not a form of strict liability.102 It is fault liability based on a judgment that the shortcomer failed to comply with a specified standard of conduct. The fact that the shortcomer could not have complied is irrelevant to the issue of fault, and raises a different issue about sensitivity to luck. Conversely, strict liability does not depend on a judgment that the person subject to it failed to comply with specified standards of conduct; and so, of course, the issue of whether they could have complied with specified standards of conduct is irrelevant to their liability. If a person unluckily lacks the capacity to comply with required standards of conduct, we may want to relieve them of strict liability as much as of fault liability. This is why, for instance, minimum capacity is a precondition of tort liability for harm, whether fault-based or strict; and why, in Honoré’s view, it is a precondition of the fairness of holding a person responsible for the outcomes of their conduct that they were able to understand the concept of responsibility. It also follows that the task of justifying strict responsibility is different from the task of justifying insensitivity to luck in the allocation of responsibility. The latter issue is discussed in 4.2. In order to justify strict liability, we need to justify holding people responsible for their conduct and its outcomes regardless of whether the conduct was in breach of any specified standards. This may be done by pointing to some other characteristic of the conduct, such as its impact on the interests of another person—perhaps that the conduct infringed that person’s rights, or that it subjected them to a “special risk of serious harm”. Or it may be done by reference to the role of the person held responsible—for example, as employer of the person whose conduct caused harm. In Honoré’s view, strict legal liability “falls typically on those who pursue permissible but dangerous activities”;103 and the justification for having a system of strict liability is instrumental—namely, to avoid serious harm.104 I am not saying that responsibility regardless of luck should not be called “strict responsibility”. This is only a matter of terminology. But we should be clear that strict liability in law has a quite different meaning. More importantly, it must be emphasised that in law (at least), fault liability is defined in 102. This is a welcome conclusion. The argument (implicit in “Responsibility and Luck”: RF 14–40) that a person who lacks the (general) competence which justifies the imposition of fault liability can nevertheless be held strictly liable because they have the (minimum) capacity necessary to own the outcomes of their conduct, is not attractive. 103. RF 23. 104. RF 28. He does, it is true, justify vicarious liability by the principle of taking the rough with the smooth (RF 81); but the principle is far too wide for this purpose.

101

  terms of failure to comply with specified standards of care, without reference to luck. The law’s sensitivity to luck is not expressed in the concepts of fault and strict liability, but in rules such as that which relieves of legal liability those who lack minimum capacity. 3.3.2.4. Summary

To summarise, the difference between fault-based responsibility and strict responsibility is that the former rests on a judgment that the agent failed to comply with a specified standard of conduct, while the latter does not. The question of whether the agent could have complied with specified standards is equally irrelevant to liability for fault and strict liability. The distinction between insensitivity to fault and insensitivity to luck seems to me to be as important to the concept of responsibility outside the law as within it. These conclusions have important implications for Honoré’s project of justifying strict legal liability. His basic argument is that strict liability (like outcome responsibility) is a form of responsibility for bad luck; that responsibility for bad luck can be justified by the principle of taking the rough with the smooth; and that it is a short step from justifying responsibility for bad luck to justifying strict legal liability. In 3.2.2. I argued that because the prime function of strict legal liability, but not of outcome responsibility, is to justify the imposition of penalties and obligations of repair, the justificatory gap between the two types of responsibility is greater than Honoré allows. In 3.3.2.1. and 3.3.2.3. I argued that because strict legal liability is no more (and no less) a form of liability for bad luck than fault-based legal liability, arguments designed to justify responsibility regardless of luck are unlikely to get us very far in justifying (strict) legal liability regardless of fault. In 4.3 and 4.5 I will argue that the justifications for strict liability are not to be found by focusing on the link between conduct and outcomes, but primarily on the impact of outcomes on victims and on wider social interests. 4. A F  R A  R

Many of the features of my approach to responsibility have already been outlined by way of comment on Honoré’s theory of outcome responsibility. In this section I bring them together and restate them more systematically and positively.

102

   4.1. The Functions of Responsibility Practices

In order to understand the nature and scope of, and to justify, our responsibility practices, we need to recognise that the various types of responsibility— outcome responsibility, moral responsibility, legal responsibility and so on— perform various different, though related, functions. Whereas the ontological function of responsibility (allocating ownership of conduct and outcomes) directs attention primarily to the link between conduct and outcomes, other functions of responsibility—such as the allocation of blame,105 punishment and obligations to repair bad outcomes—direct attention to other factors such as the nature and quality of the conduct, the impact of the outcomes on others, and the interests of society at large in the way its members live their lives.106 Patterns of legal responsibility provide a helpful starting point for exploring these elements of our responsibility practices. The focus of modern ideas of criminal responsibility is on the nature and quality of the offender’s physical behaviour and mental states, and the impact of crimes on their victims is of secondary importance.107 This is reflected in the fact that there are “victimless” and inchoate crimes, and that attempting a crime is itself a crime. Criminal sanctions are offender-oriented, and victims play a largely passive role in the criminal justice process. By contrast, responsibility in civil law is more two-sided. Victims play a central role in the civil justice process. Responsibility in civil law is always responsibility to someone; and all civil law sanctions are victim-oriented remedies. In civil law, the nature and quality of relevant outcomes and their impact on the victim are at least as important as, and often more important than, the nature and quality of the conduct that produced the outcome. So far as wider social interests are concerned, we will see in 4.5. that these play an important role in justifying strict legal liability. The general point is that the relative importance of the factors relevant to responsibility varies according to the various functions served by allocations of responsibility. 4.2. Responsibility and Luck

Because luck is ubiquitous, responsibility systems need to strike a balance between sensitivity and insensitivity to luck by distinguishing between luck 105. And praise. 106. Regarding the last see 4.4. 107. But see n. 76 above.

103

  that ought to be taken into account in allocating responsibility and luck that ought to be ignored. Thus, one function of the legal concept of causation is to define the acceptable scope of appeals to luck as a ground for avoiding responsibility. In tort law, for instance, tortfeasors must (as a general rule) take victims and the world as found, even though outside their control. In other words, agents must simply accept as given the background against and the circumstances in which their conduct occurs and takes effect, unless there is something quite out of the ordinary about it. This is reflected in the traditional distinction between the causa causans (or “effective cause”) of an event and the conditions sine qua non (or “necessary conditions”) of an event. Lack of control of the causal background of human conduct will not normally lead us to deny the status of effective cause to an agent whose tortious conduct was a necessary condition of harm unless there was something abnormal or extraordinary in the background.108 This is not the place to attempt a systematic discussion of the issue of (in)sensitivity to luck; but a few comments will serve to identify some important issues. One concerns the functions of responsibility practices. The balance struck between sensitivity and insensitivity to luck may depend on the purpose of the responsibility judgment to which it is relevant.109 For instance, luck may play a different role in judgments of outcome responsibility (concerned primarily with ownership of outcomes) than in judgments of legal responsibility (concerned primarily with allocation of punishments and obligations of repair). Various distinctions can be drawn between different types of luck.110 In the context we are presently considering, the most important distinction is between (what I have elsewhere termed) “dispositional” luck and “circumstantial” luck.111 Dispositional luck is a function of one’s personality, temperament, emotions, desires, abilities, and so on. Circumstantial luck is a function of circumstances external to ourselves in which our conduct occurs and takes effect. We have just noted that in the law of tort, for instance, principles of causation play a part in allocating the risk of circumstantial bad luck; while the so-called “objective, reasonable-person test of negligence” represents a complex set of rules and principles for the allocation of dispositional bad luck in failing to comply with standards of conduct. In Honoré’s theory, the concepts of minimum capacity and general competence function to allocate dispositional luck. 108. RF 6. 109. Similarly Scanlon n. 55 above, 187. 110. See, for instance, Nagel n. 69 above, 28.

104

   There is a considerable body of philosophical literature dealing with the issue of luck in the context of the allocation of resources, as opposed to the allocation of responsibility. Unfortunately for present purposes, the leading theorists opt either for total insensitivity to luck,112 or for total sensitivity to luck.113 Because luck is ubiquitous, responsibility systems need to strike a balance between sensitivity and insensitivity to luck. For this purpose, what are needed are criteria that cut across the distinction between luck and control. Dworkin draws a distinction between “handicaps” (a person’s “circumstances”) and “preferences and ambitions” (which are “features of body or mind or personality”).114 The former, he thinks, should be taken into account in the distribution of resources, while the latter should be ignored. However, it is not clear whether Dworkin sees this distinction as cutting across that between luck and control, or reflecting it. Nor is it clear what role it might play in relation to responsibility. The law provides a rich source of crosscutting criteria in its various principles of liability. Few of these principles directly address the issue of luck. What makes them salient to that issue is the ubiquity of luck. And it is precisely because of the ubiquity of luck, coupled with the apparent psychological and emotional inevitability and utility of responsibility practices, that the critical ethical question is not whether responsibility is compatible with luck, but whether our responsibility practices strike an acceptable balance between sensitivity and insensitivity to luck. Rawls thinks that whereas luck is irrelevant to the distribution of resources, it is relevant to the allocation of responsibility. Several writers have questioned this view.115 Certainly, from the perspective of the person “enforcing” the responsibility, as opposed to that of the person held responsible, legal responsibility, at least, can itself be viewed as a resource. For instance, when the House of Lords held, in Donoghue v. Stevenson,116 that manufacturers of products owe a duty of care to consumers, this gave consumers a legal and financial resource that they previously lacked. Putting the point more abstractly, it is a mistake to think that responsibility—legal responsibility, anyway—is purely a matter of “corrective”, as opposed to “distributive”, justice.117 If luck 111. 112. 113. 114.

Cane, n. 64 above, 144. E.g. J. Rawls, A Theory of Justice (Oxford: Oxford University Press, 1973), 310–15. E.g. G.A. Cohen, “On the Currency of Egalitarian Justice” Ethics 99 (1989), 906–44. R. Dworkin, “What is Equality? Part 2: Equality of Resources” Philosophy and Public Affairs 10 (1981) 283, 293. 115. E.g. S. Scheffler, “Responsibility, Reactive Attitudes and Liberalism in Philosophy and Politics” Philosophy and Public Affairs 21 (1992) 299, 306 n. 7. 116. [1932] AC 562. 117. P. Cane, “Corrective Justice and Correlativity in Private Law” Ox J Leg Stud 16 (1996) 471, 480–1.

105

  is relevant to the allocation of responsibility, it must also be relevant to the allocation of resources—or, at least, the resource of responsibility. This is obvious once we notice that if A causes harm to B (partly) as a result of factors outside the control of both of them, and on account of that bad luck, A is held not responsible, B must bear the cost of the harm even though the occurrence of the harm was, in some relevant respect, beyond B’s control as well as A’s. In such circumstances, to allocate responsibility is to distribute the costs of bad luck. It is for this reason that a crucial test of the acceptability of our responsibility practices is how they distribute those costs. As Ripstein argues, “The idea that a person’s life should depend only on the things he can control may make sense in the case of a particular individual if others are ready to devote their lives and resources to covering that person’s losses. But it cannot be made sense of in the case of a plurality of persons living together on terms of mutual respect”.118

In this light, reasonable-person tests of fault are best understood in terms of the balance they strike between our interests as agents and victims. Treating people as being “at fault” only if their conduct on any particular occasion failed to measure up to their own personal level of competence would put our interest in freedom from harm too much at the mercy of fate and the goodwill of others.119 It would also present considerable difficulties of measurement, because it would require us to assess competence against some agreed scale.120 Thus it is that in civil law at least, individual differences of competence are largely ignored in allocating responsibility.121 In criminal law, where the interests of victims receive less attention, differences in competence are taken into account to a greater extent in imposing penalties. The function of reasonable person tests of fault is to establish social standards of conduct that strike an acceptable balance between the interests in freedom of action and

118. A. Ripstein, Equality, Responsibility and the Law (Cambridge: Cambridge University Press, 1999), 269–70. 119. Ripstein, ibid., passim; helpfully discussed by J.C.P. Goldberg, “Rights and Wrongs” Michigan LR 97 (1999), 1828. 120. RF 19–20. It does not follow that (as Honoré argues: RF 24; but there is some equivocation at RF 19–20) reasonable-person tests of fault, like rules of strict liability, are designed, in part, to facilitate the imposition of liability on people who are “in fact at fault” but cannot easily be proved to be so. Reasonable-person tests define fault. In this context, there is no relevant concept of being “at fault” other than failure to comply with a reasonable person test. Reasonableperson tests are not a surrogate for a test of failure to exercise the abilities one actually possesses. 121. RF 22: “I rely on the competence of motorists in general, but need not suppose any particular motorist to be up to scratch in order to hold him to the objective standard.” See also the discussion of the capacity of children at RF 33–4. But Honoré’s explanation of why the law holds the inexperienced to the standard of the experienced is very different: RF 36, n. 41.

106

   freedom from harm that we all share. This implies that such tests are fair if, and to the extent that, they reflect what most people can achieve most of the time. Some people (“shortcomers”) “fall below a decent level in the gifts needed for social intercourse”.122 But such people are in the minority because social standards of conduct are typically pitched at a level that is achievable by most people most of the time.123 The interest of this minority of shortcomers, in freedom to act without a higher than average risk of incurring an obligation of repair, is sacrificed for the sake of affording greater protection to the interest we all have in freedom from harm. By contrast, in exempting from liability those who lack minimum capacity, the law tips the balance in favour of the agent’s freedom of action and against the victim’s freedom from harm.

4.3. Responsibility and Fault

The fault of the agent (i.e. whether and the degree to which their conduct deviated from specified standards of behaviour) is central to both criminal and civil legal responsibility, as well as to ideas of responsibility outside the law. It is difficult to explain why fault matters if the core of responsibility is located in the link between conduct and outcomes. In fact, judgments about the nature and quality of outcome-producing conduct are relevant to all of our responsibility practices. This is true even in relation to the rules of ownership of outcomes. Fault is one of the factors taken into account by the ideas of causation that underpin those rules: more-faulty conduct is more likely to be identified as the cause of an outcome than less-faulty conduct. Fault is also relevant to blame (other things being equal, the more at fault the agent is, the more blame they attract), to penalties (other things being equal, the more at fault the agent is, the more severe the appropriate penalty), and to obligations of repair (in tort law, for instance, only especially faulty conduct can attract an obligation to pay punitive damages). Fault is central particularly to ideas of criminal responsibility, the goals of which focus on punishment and deterrence of undesirable conduct. This is why mental-state requirements (intention and recklessness) 122. RF 24. 123. In a recent careful consideration of the legal concept of negligence, Andrew Simester has argued that in assessing the acceptability of the law’s approach we should distinguish, first, between desires and beliefs; and secondly, between different sources of false beliefs, such as inattention, defects of perception and recall, poor memory, and varying intelligence: “Can Negligence be Culpable” in J. Horder (ed.), Oxford Essays in Jurisprudence, Fourth Series, n. 66 above, 85–106.

107

  are so prominent in criminal law. Fault is less important to civil legal responsibility. For example, basic damages for breaches of civil law are measured by reference to the impact of the bad outcome on the victim, not according to the degree of the law-breaker’s fault. Responsibility in civil law is just as concerned with the victim as with the agent. This is why strict liability (liability regardless of fault) is less controversial in civil law than in criminal law. Why is fault central to criminal responsibility? The answer, I believe, lies in the importance of being able to disown as well as to own outcomes. As agents, we have an interest in freedom of action, in being able to act without incurring the serious penalties and blame that attach to criminal responsibility. This freedom has two aspects: one dependent on sensitivity to luck, and the other dependent on sensitivity to fault. Responsibility that was not tempered by sensitivity to luck would be far too onerous to form the basis of criminal responsibility; but so, too, would responsibility that was not tempered by sensitivity to fault. Fault is less central to ideas of civil legal liability partly because, on the whole, less stigma attaches to being held civilly liable. More importantly, one of the main purposes of the civil law is to acknowledge and protect our interest, as potential victims, in freedom from harm. Civil law attempts, in a way that criminal law does not, to strike a balance between our interest, as agents, in freedom of action, and our interest, as victims, in freedom from harm.124 I would argue that these two different ways of approaching responsibility—that embodied in the criminal law and that embodied in the civil law—illuminate our responsibility practices outside the law because they are also part of those practices. I do not say that legal and extra-legal responsibility practices are identical. But I do believe that the basic outlines of legal responsibility practices reflect responsibility practices outside the law.

4.4. Responsibility and Social Values

As well as taking account of our interests as individual agents and victims, legal responsibility practices also seek to accommodate wider social values and interests. This is obviously true in the case of criminal law, where the interests of the individual victims are subordinated to those of society at large. But it is also true of the civil law. For instance, one of the factors taken into account in tort law in deciding whether a person has been negligent is the social value of 124. I am not saying, of course, that the protection of our interest, as individuals, in freedom from harm is not one of the aims of the criminal law, but only that this interest is less important to ideas of responsibility in criminal law than to ideas of responsibility in civil law.

108

   the activity of which the allegedly negligent conduct was part. Considerations of “public policy” have led courts to protect participants in legal proceedings from civil liability; and to seek ways of limiting the tort liability of governmental bodies. Society’s interest in the maintenance of strong markets is recognised in the severe limits imposed on tort liability for infliction of economic loss in the course of competitive activities. Honoré accommodates this last example by saying that infliction of harm will attract a penalty or an obligation of repair only if there was no justification for its infliction.125 “Justification” is one of the concepts used in the law to give effect to a preference for protecting an interest of society at large over that of an individual sufferer of harm.126

4.5. Strict Responsibility and Balancing of Interests

Recognition of the role of our various interests as agents, victims and members of society in shaping our responsibility practices is especially important to an understanding of strict responsibility. Such responsibility plays an important part in the rules of ownership of outcomes—it helps to define the identity of individuals, and it gives people incentives to do their best to avoid bad outcomes and achieve good ones. Our sense of individual identity depends, to some extent, on owning outcomes regardless of whether the conduct that produced them was faulty; and a society in which people owned only those bad outcomes that resulted from their faulty conduct would not be a pleasant place to live. The role of strict liability in the allocation of penalties and obligations of repair is quite different. Strict responsibility in this context is not merely an extension or application of the concept of outcome responsibility. Rather, it is designed to give special protection to our interests as victims and members of society. Far from being agent-oriented, liability for penalties and obligations of repair that is independent of fault is oriented towards the interests of victims and of society. Remember that strict legal liability is not liability in the absence of fault, but liability regardless of fault. Often, a person held strictly liable will have been at fault, and an important function of strict liability is to increase the chance that those at fault will be held liable. A strict liability regime may reduce the incidence of unpenalised and uncompensated harm resulting from particular activities (whether faulty or not); and it may encourage the development of less harmful substitute 125. RF 78. 126. Similarly: J. Stapleton, Product Liability (London: Butterworths, 1994), 183.

109

  activities. The ultimate justification for strict responsibility will be found in the benefits it brings to individual potential victims and to society, not in ideas about agency and the ownership of outcomes.127

5. C

The analysis in this essay has been based on the argument that in order to understand our responsibility practices we need to pay careful attention to the various functions they perform; and the contention that we must not concentrate too much on the link between conduct and outcomes at the expense of the nature and quality of conduct and the interests of victims and of society. Honoré’s theory of outcome responsibility offers a valuable account of the ontological function of responsibility. However, his account of the scope of outcome responsibility is incomplete. And because he pays insufficient attention to other functions of responsibility, and to the relationships between agents and outcomes on the one hand, and victims and society on the other, Honoré’s claim that outcome responsibility is basic, and that other types of responsibility are forms or versions of it, is unconvincing. More specifically, I have argued that there is an important distinction between strict liability and liability for bad luck, and that Honoré’s failure to take account of this distinction undermines his claim to have found in the idea of outcome responsibility the key to justifying strict legal liability. The key to understanding the varieties of responsibility that together make up our responsibility practices lies in an appreciation of their various functions and the variety of relationships they regulate.

127. To this extent, I agree with Ernest Weinrib: The Idea of Private Law (Cambridge, Mass.: Harvard University Press, 1995), 179–183.

110

6 OBLIGATIONS AND OUTCOMES IN THE LAW OF TORTS John Gardner*

I

“Under strict liability”, according to Alan Brudner, “one has no right to act in ways that happen to injure another. Since, however, all action carries the risk of such injury, strict liability means that I have a right that you be governed in all your actions by concern for my welfare, and you have the same right over me. No doubt there is a mutuality of care here; but it is the mutual care of extreme altruists who, because they claim no worth as independent selves, can neither give nor receive effective confirmation of worth and hence can require no valid right to care. By contrast a fault [i.e. negligence] requirement establishes a reciprocity of care between selves”.1

Brudner casts his argument in terms of the potential plaintiff ’s rights, but (to avoid some distractions further down the line) I am going to reframe it in terms of the potential defendant’s obligations. True enough, P’s rights against D are not the same thing as D’s obligations towards P. A right is not an obligation; rather, it is the ground of (one or more) obligations.2 All the same, there is nothing that counts as the violation of a right other than a failure to perform (one or more of ) the obligations that it grounds. Thus the conditions under which D violates P’s rights—the conditions under which D wrongs P— are identical to the conditions under which D fails to perform (one or more of ) his obligations towards P. It follows that, if our interest is in what it takes for D to wrong P, and in the liability to which such wrongs may give rise in * I am grateful to the participants in the Columbia Law School workshop on Tony Honoré’s Responsibility and Fault for a fascinating discussion of my first attempt at the topic, from the ashes of which this rather different paper arose. Special thanks go to my commentator Niki Lacey and to Tony Honoré for giving me the benefit of his first reactions. Subsequent drafts were much improved by detailed comments from Peter Cane and Timothy Macklem. 1. Alan Brudner, The Unity of the Common Law (Berkeley: University of California Press, 1995), 190. For a similar line of thought, see Ernest Weinrib, The Idea of Private Law (Cambridge, Mass.: Harvard University Press, 1995), 182–3. 2. Joseph Raz, The Morality of Freedom (Oxford: Clarendon Press, 1996), ch. 7.

  law, the important differences between rights-talk and obligations-talk can largely be put on one side. It seems to me, accordingly, that we can faithfully recast Brudner’s argument in the following terms. When the law imposes strict liability on D for actions of D’s that injured P, says Brudner, it asserts that D had (and failed to perform) an obligation to take extreme care not to injure P—such extreme care not to injure P, indeed, that D would have effaced himself entirely in performing the obligation. In demanding that D have acted so entirely selfeffacingly for the sake of P, the law undermines its own position that P in turn owes the same obligation to take extreme care not to injure D. How could the existence of this reciprocal obligation be defended when, by the law’s own reckoning, D clearly counts for so little? This argument against the legal imposition of strict liability fails. I do not mean that it fails in its striking moral claims inspired by Kant and Hegel. Maybe Brudner is right that there comes a point at which one person takes such extreme care not to injure another that in the process she effaces herself entirely. Maybe Brudner is also right that there is something obnoxious about effacing oneself entirely in this way, or at least about being under an obligation to do it, or at the very least about being authoritatively held to be under an obligation to do it. And maybe that obnoxiousness comes, as Brudner suggests, of the fact that the self-effacing action is performed for reasons that cannot consistently be universalised to include the scenario in which D’s and P’s positions are reversed. Maybe Brudner is also right that, for this reason if no other, the law should not hold anyone to be under an obligation to take selfeffacingly extreme care not to injure another. Let all of this be true. The problem remains that it still does not add up an argument against the legal imposition of strict liability unless it is also true, as Brudner asserts right at the start of the argument, that in imposing strict liability on D for actions of D’s that injured P, the law holds D to have been under (and to have failed to perform) an obligation to take extreme care not injure P. And sadly this doctrinal premiss (as I will call it) is not true. In fact, it is an inversion of the truth. The truth is that when the law imposes strict liability on D for actions of D’s that injured P, it asserts that D had (and failed to perform) a straightforward obligation not to injure P. D would have performed this obligation not to injure P—and hence would have avoided the strictest of strict liabilities—if and only if she had not injured P. Now, not injuring P, and therein performing her obligation, is something that D might have done entirely fortuitously, without taking the slightest care, let alone extreme care, not to injure P. (Perhaps P just happens not to step off the pavement as D careers drunkenly and obliviously past in her car.) Conversely, injuring P, and therein 112

        failing to perform her obligation, is something that D might have done in spite of having taken the most extreme conceivable care not to do so, even to the point of total self-effacement that so worries Brudner. (Perhaps P has an unprecedented allergic reaction to all the cotton wool that D spent the whole day laboriously wrapping him up in). Under a regime of strict liability for D’s injuring P there could be no liability in the first case where no care at all is taken, whereas there could be liability in the second where every conceivable care is taken. When strict liability is at stake, in other words, the law does not give two hoots either way about the care that D took. The most inhuman lack of care does not count against her but by the same token the most superhuman investment of care does not count in her favour. It is this doubleedged obliviousness to D’s precautions that lawyers have in mind when they call liability by the evocative name “strict”. So strict liability is not, as Brudner suggests, a variant of negligence-based liability in which the standard of obligatory care has been cranked up to an awesome level. It is a quite unrelated mode of liability in which the law does not care about care-taking, and therefore does not treat the bestowing of care—any care at all—as having been obligatory. Correspondingly, no measure of self-effacement at all is called for in performing the obligation. You may object that in these remarks I am interpreting Brudner’s doctrinal premiss uncharitably by interpreting it too doctrinally. I am attributing to him the clearly mistaken view that D’s legal obligation, non-performance of which gives rise to D’s strict liability, is an obligation to take extreme, self-effacing care not to injure P. But what Brudner is actually suggesting (you may say in his defence) is that D’s declared legal obligation—admittedly a straightforward obligation not to injure P, for the performance of which taking care is admittedly neither necessary nor sufficient—effectively puts D under another (legally undisclosed) obligation as well, namely a morally obnoxious obligation to take extreme care not to injure P. To be exact, isn’t Brudner’s real point that taking the most extreme conceivable care, to the point of total self-effacement, is the only available means that D has to perform his straightforward legal obligation not to injure P? And that being so, doesn’t the very logic of practical rationality dictate that D also has a second, derivative obligation to take that most extreme conceivable care? To will the end, after all, is to will the means; and to will that the end (D’s not injuring P) be obligatory is surely, by the same token, to will the derivative obligatoriness of the only available means (D’s taking ever-such-extreme care not to injure P). It is the obnoxiousness of this latter derivative obligation that, on this more charitable reading of Brudner, casts shame on the former (parent) obligation, and hence casts shame on the strict liability that is based on the former obligation’s non-performance. 113

  I doubt whether this is really a more charitable reading. The proposed argument is open to challenge at virtually every step. To avoid tangential complications, however, let’s concede one premiss of the argument straight away, namely the one ascribed to the very logic of practical rationality. According to this premiss (if I may express it as a general formula) an action  is derivatively obligatory for a given agent D whenever another action  is obligatory for D and ing constitutes D’s only available means of ing. That premiss is relatively easy to grant, of course, because everything turns on what an only available means is. So what is it? The three words of the expression clearly signify three different conditions. To make the premiss a credible rendition of the logic of practical rationality, I propose the following as a minimal unpacking of the three conditions. For ing to be D’s only available means of ing it has to be the case that (a) D’s ing would contribute to D’s ing (it is a means); (b) D has the capacity and the opportunity to  in the circumstances (it is an available means); and (c) no other action available to D would contribute to D’s ing (i.e. it is the only available means). On this analysis, is taking ever-such-extreme care not to injure P indeed the only available means that D has of not injuring P? One may doubt whether it meets condition (c).3 But more importantly it fails to meet condition (a). Taking ever-such-extreme care not to injure P cannot possibly be the only available means that D has not to injure P because, on closer inspection, it is not a means of D not injuring P at all. When we read condition (a) we have to begin by grasping the relevant idea of a “contribution” to A’s ing. The basic idea we are looking for here is that of making it more likely that A s. After all, for Brudner’s purpose—the purpose of evaluating rival legal standards or rules—it is not enough to derive, as an object of evaluation, a one-off obligation on D to take ever-such-extreme 3. That D will not encounter P at all as he weaves his drunken way home in the car is, in general, a perfectly realistic prospect. So why isn’t this realistic prospect billed, in the proposed reconstruction of Brudner’s argument, as an alternative available means of D’s not injuring P, with the supposed result that D’s taking ever-such-great care not to injure P is not left as the only available means? The answer is that it is taken for granted that the set of available means of ing equals the set of available means of intentionally ing. By knocking out some of the competition in advance—namely all available means of unintentionally ing—this tweak of condition (c) makes it systematically easier for ing (an action performed in order to , such as “taking care to ”) to be declared the only available means of ing. The tweak therefore makes it easier than it would otherwise be to generate a derivative obligation to  (e.g. to take care not to injure) out of an existing obligation to  (e.g. simply not to injure). The conclusion of this paper will help to show why the tweak is illegitimate.

114

        care not to injure P as a means of not injuring P. What is needed is a standard or rule according to which D is to take ever-such-extreme care not to injure P as a means of not injuring P. So our interest is not in D’s possible one-off failures and successes in not injuring P by taking ever-such-extreme care not to. Our interest is in his prospect of failure and success, his tendency to succeed when he conforms to the advertised standard or rule, the probability of his not injuring P in the process. Thus the acid question facing Brudner’s rearguard defenders is: Is it really the case that (assuming D has the capacity not to injure P) D always continues to improve the probability of his not injuring P, the more care he takes not injure P? Is D’s taking ever-such-extreme care not to injure P in that sense making a contribution to his not injuring P? To that question the answer is a resounding no. In principle there always comes a point at which D’s taking further care not to injure P does nothing to improve the probability of his not injuring P. Indeed, in principle there always comes a (further) point at which D’s taking further care not to injure P is positively counterproductive. The further care D takes not to injure P, beyond this point, actually makes it more, not less, likely that D will injure P. Like other moral virtues, to put the point in familiar Aristotelian terms, the virtue of caution lies in a mean between deficiency and excess. Excessive caution, or at any rate one kind of excessive caution,4 is the caution that fails in its own terms, the caution that tends to precipitate the very eventualities that the excessively cautious person was taking such pains to avoid. In this vein, we are all familiar with the self-defeating menaces that are the overcautious driver, the overprotective parent, and the oversolicitous lover. Set alongside their all-too-careless counterparts these characters are of course genuinely admirable.5 But our admiration for them is tinged with pity and frustration. That is because (discounting the secondary success, the bittersweet moral victory, that lies in the mere fact of their being admirable) these too-careful characters are alas no more successful at what they do than their alltoo-careless counterparts—and that is in spite of, nay because of, all the care they take to do it. Their taking such care, put simply, is not a means to their actually doing what they are taking such care to do. On the contrary it is a means to their doing the opposite. So if they have a straightforward 4. Perhaps there are other kinds. Couldn’t one be overcautious either in being self-defeatingly cautious or in being, as Brudner emphasises, self-effacingly cautious? I wonder whether the latter fault is strictly speaking an instance of overcaution, as opposed to an instance of cravenness or slavishness (that being the virtue of humility taken to a self-defeating point). 5. I do not mean that they are admirable in spite of their fault. I mean that their fault is an admirable one. It does not follow that it is any the less a fault or that the conduct exhibiting it is any the less wrong or blameworthy. On the logic of admirable yet blameworthy wrongdoing, see Michael Stocker, Plural and Conflicting Values (Oxford: Clarendon Press, 1990), ch. 2.

115

  obligation not to hurt some other person (say), that obligation yields—on any credible view of the logic of practical rationality—no derivative obligation on them to take the extreme care that they irrationally imagine they should take. Why, it does not even yield a derivative reason to take such extreme care. What it yields, on the contrary, is a reason, and perhaps sometimes even an obligation, not to take such extreme care. It is a salutory reminder of the extent of reason’s cunning that a straightforward obligation not to injure P, far from yielding an obligation to take ever-so-much care not to injure P, could in principle yield a derivative obligation to do exactly the opposite, i.e. to proceed without taking very much care, or in extreme cases maybe even any care at all, not to injure P.6

II

I mention such dramatic and hard-to-envisage cases mainly to bring home this message. The highest level of productive care (as we might call that point beyond which taking further care makes one no more likely to do what one is taking care to do) is not a constant. It varies from activity to activity. If one must walk across a bed of hot coals it is probably best, as a rule, to be entirely careless where one places one’s feet, since any care one takes (e.g. to avoid sharp coals) will, as a rule, just slow one’s progress and intensify one’s pain. Here the highest level of productive care is on the low side. On the other hand, if one is walking a tightrope very great care in placing one’s feet is as a rule exactly what is called for, so I am told, and the highest level of productive care is very high. Thus the overcautious types we mentioned might be, in that respect, good tightrope-walkers but bad hot-coals-walkers. There are, in other words, very dramatic variations in where the mean of caution lies, as we shift from activity to activity. In principle, however, a highest level of productive care always lurks somewhere. Even in tightrope-walking there comes a point at which one’s ever more elaborate precautions against falling (lessons in concentration, extra checks on the rope, intensive practice sessions all afternoon, etc.) tend to stop helping one walk the rope, and soon enough (thanks to the extra stress and tiredness they bring on, etc.) they become the very things that precipitate one’s fall. So a straightforward obligation on D not to fall from his tightrope could not—pace Brudner’s would-be charitable interpreters—yield a derivative obligation on D to take ever-such-extreme care not 6. The most important modern study of this phenomenon of self-defeating endeavour is Derek Parfit’s Reasons and Persons (Oxford: Clarendon Press, 1984), part 1.

116

        to fall from his tightrope. What it might yield is a derivative obligation on D to take all reasonable (i.e. productive) care not to fall from the tightrope, where this is contrasted with the unreasonable (i.e. excessive) care not to fall from the tightrope that would actually hasten D’s falling. When I say “reasonable care” here, do I mean reasonable care as this is understood by lawyers—namely the care that all and only the negligent fail to take? Something close. What I have been calling the highest level of productive care is indeed one of the central determinants of what counts, in law, as the negligent pursuit of this or that activity. To be sure, it is not the only determinant. Even an extra measure of productive care is not regarded as an extra measure of reasonable care, and so is not required to avoid negligence-based liability, if it is too difficult or costly for D to lay on. But Brudner’s argument plays up the difficulties and costs to D of laying on additional measures of care, without noticing that there is always, at the highest level of productive care, an independent cap on the rationality (and hence the derivative obligatoriness) of D’s incurring those difficulties and costs. Because of this cap, there is no possibility of a straightforward standardised legal obligation on D not to injure P yielding a derivative standardised obligation on D to take eversuch-extreme measures of care not to injure P. If the obligation not to injure P yields any derivative obligation to take care at all, it is at most an obligation to take all productive measures of care—all care that contributes by making it more likely that one will avoid what one cares to avoid—and that often means, pace Brudner, only very modest measures of care. In principle, as I said, it might sometimes mean no measures of care at all. These remarks were supposed to help bring out the rational relationship between negligence-based (or fault) liability on the one hand and strict liability on the other. Essentially, the obligation that one is held to have under a regime of fault liability is an obligation to take productive care to do that which, under a regime of strict liability, one would merely have a straightforward unadorned obligation to do. As well as bringing out the connection between the two modes of liability, however, this last formulation also helps to bring out the simple ineliminable qualitative contrast between them. The contrast is but one instance of a larger contrast that turns out to be at stake in many of the deepest puzzles of moral and political philosophy. Elsewhere,7 I have coined some terminology to mark the larger contrast. I have called it the contrast between obligations to try and obligations to succeed. An obligation to succeed is a straightforward obligation to , for the performance of which only one’s actually ing matters, never mind what steps one takes (if any) with 7. In “The Purity and Priority of Private Law”, U Toronto LJ 46 (1996), 459 at 486.

117

  a view to ing. An obligation to try is the converse case: only the steps one takes with a view to ing are relevant to whether one performs the obligation, never mind whether one actually s thereby. On this account, a straightforward obligation not to injure P—the obligation that D is held to have been under (and to have failed to perform) when she is held strictly liable for injuring P—is a straightforward example of an obligation to succeed. By contrast what the law calls D’s “duty of care” to P—the obligation to take care not to injure P, nonperformance of which puts D at fault and grounds D’s fault liability—is a straightforward example of an obligation to try. Or maybe not so straightforward? Some lawyers, in my experience, are resistant to the suggestion that an obligation to take care is an obligation to try. On the one hand, it is said, no amount of trying is logically sufficient to perform one’s obligation to take care in law, for the legal standard of negligence is an impersonal (or “objective”) standard and this means that, for at least some people, try as they might they will not perform their legal obligation. On the other hand, it is said, no amount of trying is logically necessary to perform one’s obligation to take care in law. One takes reasonable care not to injure in the relevant sense just as long as one acts in ways that actually reduce the risk of injury one poses to an acceptable level, even if one was not at all concerned, or motivated, to reduce the risk. These objections combine to portray the legal obligation to take care as just another obligation to succeed—to be exact, an obligation to succeed in not behaving too riskily, never mind what one was trying to do—and hence they purport to eliminate the “ineliminable qualititative contrast” that I drew between strict liability and fault liability. Both of the objections are confused, however, and my contrast stands uneliminated. Regarding the first objection: If some people do not perform their obligation to take care however hard they try, this does not go to show that it is not an obligation to try. It only goes to show that it is an obligation to try harder (more assiduously) than they are capable of trying. True, an obligation to try harder than one is capable of trying is a problematic obligation. In particular, it runs up against what I set out as condition (b) above, namely the “availability” condition for establishing an obligation. But recall that the availability condition, like the other conditions I mentioned, was only a condition for establishing an obligation by deriving it from another obligation via the “only available means” argument. And recall that the suggestion that a legal obligation to take care might be an obligation derived from another obligation by the “only available means” argument was floated only as part of an ad hominem bid to salvage Brudner’s attack on strict liability. If the legal obligation to take care is not after all an obligation derived from another obligation, 118

        or is derived from another obligation only by some argument other than the “only available means” argument, then there is no reason that I can think of to assume that the obligation to take care exists only subject to availability condition—no reason to assume, in other words, that one has it only on condition that one has the capacity to perform it.8 But even if there is reason to make this assumption regarding obligations more broadly, there is no reason to think that the assumption applies any the less to obligations to succeed than it does to obligations to try. The point is that the distinction between obligations that one has only on condition that one has the capacity to perform them and obligations that are not subject to this condition is a distinction that cuts across the distinction between obligations to try and obligations to succeed.9 That one lacks the capacity to perform one’s obligation does not entail that one has an obligation to succeed rather than an obligation to try. Rather, it remains an obligation to try if the explanation of why one lacks the capacity to perform it is that one lacks assiduousness, i.e. the capacity to try hard enough. As for the second objection: According to this objection, negligence is a matter of the injurious tendency of one’s action, and in assessing that injurious tendency intended effects (those one was trying to achieve) are regarded, through the lens of the negligence test, as being completely on a par with sideeffects. All are relevant according only to their foreseeability, not according to whether or not the agent literally “cared” (here meaning “tried”) to bring them about or avoid them. Trying assiduously enough to avoid injuring is admittedly one way to avoid being negligent; but one can also avoid being negligent by (fortuitously) posing only acceptable risks of injury by one’s conduct. So we are often told by, for example, those who have an economistic interpretation of the law. But is this the real legal position? I think not. Let’s concede— although the matter is more complicated than it looks—that the law doesn’t locate D’s negligence in the fact that he tried to do what he tried to do.10 But it nevertheless clearly does locate D’s negligence in what he meanwhile didn’t try to do. The whole point is that D didn’t take sufficient care to avert (limit, 8. That “ought” implies “can” is commonly given as the reason. But that is not a reason—that is a restatement of the same misguided assumption. The assumption is often ascribed to Kant but much of Kant’s work was devoted to refuting it in the sense in which it is usually ascribed to him and widely but mistakenly assumed to be true. Cf. n. 42 below. 9. Those who think that “obligations to succeed that I have the capacity to perform” is inevitably an empty set are confusing the capacity to perform with the capacity to guarantee performance. Cf. Michael Moore, “Authority, Law, and Razian Reasons”, Southern California LR 62 (1989), 827 at 875. 10. The complications are brilliantly exposed in John Finnis’s “Allocating Risks and Suffering: Some Hidden Traps”, Cleveland State LR 38 (1990), 193.

119

  reduce, control) the injurious side-effects of his endeavours. The words “to avert” here mean “in order to avert”, “with a view to averting”, or (in other words) “intending to avert”. What else could be meant by them? Taking care is an essentially intentional action. One cannot take care not to  without trying not to . That is also the law’s position. Negligence in law is a failure to try assiduously enough to avert (limit, reduce, control) the unwelcome side-effects of one’s (otherwise valuable) endeavours. It follows that the obligation that one fails to perform when one acts negligently is indeed an obligation to try. The nonperformance of an obligation to try is what gives rise to fault liability in law, just as the nonperformance of an obligation to succeed is what gives rise to strict liability. The contrast between the two is as basic as they come.

III

Brudner’s failed objection to the legal imposition of strict liability was notable for having virtually turned on its head the most common objection to strict liability in the law. Brudner’s objection proceeded from the thought that to avoid incurring strict liability, one needs to exercise the most extreme conceivable care. By contrast, the common objection focuses on the fact that, when strict liability is at issue, the law does not give two hoots how much care one takes. This common objection in turn comes in two variants. I will call them the moral intelligibility variant and the institutional fairness variant. According to the institutional fairness variant, it is unfair (or contrary to the rule of law, or something along those lines) to ground legal liability in nonperformance of an obligation to succeed, since this makes it impossible even in principle for people to plan their lives according to whether their actions will or will not attract legal liability. For they cannot possibly know this until they know how their actions turn out—as successes or failures—by which time it is too late to reconsider whether to perform them. Unfair? You may think so. But if you sign up to the other variant of the objection—the moral intelligibility variant—you do not even get as far as asking whether the legal enforcement of obligations to succeed is unfair. For what you deny, in a sense, is the very possibility of there being obligations to succeed, be they enforced or not. Of course, this is not to deny that the law asserts the existence of unperformed obligations to succeed whenever it imposes strict liability. That cannot be denied. Nor is it denied that the law may have its reasons for imposing the liability and hence for asserting the existence of the underlying obligations. Naturally it may. What is denied is that the law’s assertion of the 120

        existence of the obligations can be made morally intelligible. For morally speaking—i.e. apart from the law—there is and can be no such thing as an obligation to succeed. Or so the objection goes. Strict liability, as Thomas Nagel famously put it, “may have its legal uses but seems irrational as a moral position”.11 At first sight this second variant of the common objection seems to miss the target completely. If the assertion of an obligation to succeed “has its legal uses”, why is it an objection to those legal uses that obligations to succeed could not possibly exist apart from the law? Why should the law care whether its assertions are morally intelligible so long as they are legally useful? The answer, of course, is that the law by its very nature claims to bind people morally; it purports to tell them what they really ought to do, not merely what they ought to do according to law.12 Legal obligations, in short, are would-be moral obligations. What this means is that, if they are to be defended as legal obligations, they have to be defended as would-be moral obligations, i.e. inspected in a moral light. It does not follow, of course, that defensible legal obligations all need to be institutionalisations of moral obligations that already exist apart from the law (the nonperformance of which would be malum in se). Defensible legal obligations are often new moral obligations created by law (the nonperformance of which is malum prohibitum), and in principle the creation of these new moral obligations can be defended perfectly adequately by relying on the usefulness (the consequential advantages) of the liability regime thereby brought into being.13 But the creation of such legal obligations always remains subject to the following constraint: On pain of indefensibility as legal obligations, they must be obligations that could conceivably have existed apart from the law. On pain of indefensibility, it must always make sense to say of any legal obligation “this is your moral obligation”. Otherwise it is impossible to judge legal obligations as would-be moral obligations. So if, as many are wont to argue, it never makes sense to say of an obligation to succeed “this is your moral obligation”, then the assertion of obligations to succeed in law is always 11. Thomas Nagel, “Moral Luck”, Proc. Arist. Soc. Supp. Vol. 50 (1976), 137; reprinted in his collection Mortal Questions (Cambridge: Cambridge University Press 1979), 24 at 31 (to which page references below refer). 12. For two largely complementary defences of the thesis that such claims form part of the very nature of law, see Raz, The Authority of Law (Oxford: Clarendon Press 1979), chs 1 and 2, and Finnis, Natural Law and Natural Rights (Oxford: Clarendon Press 1980), chs 1 and 11. 13. Thus when I say “moral” here I do not mean “moral” in the narrow technical sense often favoured in recent tort theory, which is by way of contrast with “economic” or “policyoriented”. Contrast, for example, Ernest Weinrib’s anti-mala-prohibita invocation of “moral” in “Towards a Moral Theory of Negligence Law”, Law and Philosophy 2 (1983), 37.

121

  indefensible, whatever their “legal uses”. Although rarely spelt out, this is the moral intelligibility variant of the common objection to the legal imposition of strict liability. These lines of argument are invoked to discredit strict liability. But as I expressed them it seems that they also bite, in the process, against some other kinds of legal liability. Take, for example, liability for the tort of negligence at common law. In spite of its name, the tort of negligence is not the simple tort of acting negligently towards P. It is the more complex tort of injuring P by acting negligently towards P. To put it another way, although one has a legal obligation to take reasonable care not to injure P (the “duty of care”), breach of this straightforward obligation to try is not a tort. It is not a legally actionable breach of obligation. What is legally actionable, and is a tort, is the breach of a more complex obligation, namely the obligation not to injure P by not taking reasonable care not to injure P. This is not, as you can see by the italicized words, a straightforward obligation to try. Rather it is a kind of hybrid trying–succeeding obligation. More than one kind of hybrid trying–succeeding obligation is, of course, logically possible. One kind of hybrid would be an obligation to succeed-by-trying—an obligation, for instance, to take reasonable care not to injure P and (thereby) to avoid injuring P. This is obviously a tall order. One would fail to perform this hybrid obligation if either one did not take the correct measure of care not to injure P or one did in fact injure P. But the hybrid obligation, nonperformance of which constitutes the tort of negligence, is not a hybrid of this “tall-order” kind. It is the converse case. One commits the tort of negligence only if one both failed to take the correct measure of care not to injure P and one did in fact injure P. The obligation at stake is not an obligation to succeed by trying but an obligation not to fail for want of trying. We might call this a “shortorder” as opposed to a “tall-order” hybrid. It is more readily performed than either a straightforward obligation to try or a straightforward obligation to succeed, all else being equal, since there are always two alternative ways of performing it: either by trying to the requisite extent (whether successfully or not) or else by succeeding (whether or not by trying). The fact that it is more readily performed than either a straightforward obligation to try or a straightforward obligation to succeed does not entail, however, that the presence of such a hybrid obligation in the law is also easier to defend. True, one can quickly nuance the institutional fairness argument so that it cuts against the legal enforcement of straightforward obligations to succeed (i.e. strict liability) but not against the legal enforcement of shortorder hybrids (as in the action for negligence at common law). The institutional fairness argument is indeed most often ventured asymmetrically, with 122

        the focus on helping those who want to be sure in advance that they will not incur any liability, rather than on helping the perhaps more eccentric types who want to be sure in advance that they will incur it. Since taking the legally specified measure of care is logically sufficient to extinguish liability for the tort of negligence, one can plan to steer clear of that liability just as readily as one could plan to steer clear of liability for non-performance of a simple obligation to take the specified measure of care (i.e. a straightforward obligation to try). True, taking a legally specified measure of care is not logically necessary to extinguish liability for the tort of negligence, since that liability could also be extinguished by the happy fortuity that one injured nobody by one’s carelessness. But naturally this possibility is no skin off the nose of people who merely want to plan safely to avoid the liability, for they only need to know what investment of care would be sufficient to do so. If these people are the only ones that the institutional fairness argument seeks to protect, then that argument cuts against strict liability (the legal enforcement of straightforward obligations to succeed) but not against liability for the tort of negligence (the legal enforcement of short-order hybrid obligations not to fail for want of trying). Matters are not so clear-cut, however, when we come to the moral intelligibility objection. If straightforward obligations to succeed (and hence strict liability torts) turn out to be morally unintelligible, could obligations not to fail for want of trying (and hence the tort of negligence at common law) nevertheless turn out to be morally intelligible?14 The final answer to this question depends, of course, on what exactly it is that is said to make obligations to succeed morally unintelligible. For the time being, however, our interest is not in the answer. Our interest is in the question itself. As followers of modern policy debates about the law of torts, especially those that rage in the North American law schools, we are all accustomed to think of the history of modern tort policy as a struggle for supremacy between the reigning tort of negligence on the one hand, and, on the other hand, the upstart strict liability torts (on the Rylands v. Fletcher15 model) that occasionally surface, or are promoted, as pretenders to its throne (nowadays, for example, in the field of products liability). In the mentality of modern tort law and tort theory, to defend the one is implicitly to attack the other. But our question puts a completely different complexion on the relationship between them. It shows that in another dimension the tort of negligence may on the contrary have its moral fate bound up with that of its strict liability rivals. For the tort of 14. See Nagel, “Moral Luck”, above n. 11, 28–9. 15. (1866) LR 1 Ex 265; [1868] LR 3 HL 330.

123

  negligence and its strict liability rivals on the Rylands v. Fletcher model stand together in opposition to another little-noticed but increasingly powerful pretender, a model of tort liability in the incurring of which it matters not one jot whether D actually injured P, but only how much care D took that P not be injured. The really stark contrast is, of course, the contrast between this kind of “bare negligence” tort (which consists in nonperformance of a straightforward obligation to try, never mind one’s success) and a strict liability tort (which consists in nonperformance of a straightforward obligation to succeed, never mind whether or to what extent one tried). In this contrast, the tort of negligence at common law is logically non-aligned. It consists in the nonperformance of a hybrid trying–succeeding obligation (to be exact, an obligation not to fail for want of trying), and hence shares as much and as little of its logical structure with strict liability torts as it does with bare negligence torts. But is it also morally non-aligned? There are two dramatically contrasting ways of thinking about the moral intelligibility, and hence the defensibility, of the tort of negligence at common law. It is possible to take compromise positions intermediate between them but they represent the two magnetic poles of the debate, the first of which has lately proved more magnetic than the second. The first interpretation—the one which has come into the ascendant in the late-twentieth-century literature—would have it that the tort of negligence is essentially a variation on the theme of a bare negligence tort. The essence of D’s wrong, morally speaking, lies in his failure to take adequate care not to injure P. The extra condition that P actually have been injured by D’s failure to take care is regarded, correspondingly, as morally secondary. Its defence is accordingly apt to be a parasitic defence, i.e. one that presupposes that the real moral wrong has already been done to P. Thus the extra condition is collaterally defended as, for example, a locus standi condition calculated to optimise the incidence of litigation in respect of the wrong,16 or as a way of ensuring that compensation for the wrong will in principle be quantifiable without resort to excessive speculation about probabilities. But by the same token the door is left open to allowing the extra condition that P be injured by D to be eliminated or watered down while leaving the central moral theme of the tort in place. On this interpretation, one can imagine modifying the tort such that in some cases (e.g. mass injuries arising from multimanufacturer pharmaceuticals or industry-wide environmental hazards) it is no longer a condition of D’s liability for the tort of negligence that D actually 16. See, most obviously, Richard Posner’s influential “A Theory of Negligence”, J Leg Stud 1 (1972) 29 at 46–52.

124

        injured P, but only (say) that P was injured and D was one of a finite group of careless operators who may well have injured her, or even (say) that P was not injured at all but only left insufficiently protected against injury by D and his fellow operators. These are regarded, on the first of the two contrasting interpretations that I have in mind, as relatively marginal adjustments to a tort, the moral essence of which lies in nonperformance of a straightforward obligation to try, viz. the common law’s famous duty of care. But on the other interpretation that I have in mind these adjustments are far from marginal. They represent the most dramatic possible abrogation of the moral foundation of the tort of negligence. For in spite of the tort’s misleading name, the moral essence of D’s tort in a negligence case is really just that he injured P. In other words, the tort of negligence at common law is morally speaking a variation on the strict liability model of a tort, in which what is of the essence is what one actually does (injures P), never mind what one merely tries to do (one’s not taking adequate care not to injure P). The extra condition that D not have taken adequate care not to injure P is regarded, correspondingly, as morally secondary. Its defence is accordingly apt to be a parasitic defence, i.e. one that presupposes that the real moral wrong has already been done to P. The negligence condition is defended, for example, as one that is needed merely to meet the institutional fairness objection. The moral wrong one does is essentially just to injure P, but it is unfair to hold D liable for that wrong without providing a way in which D could have taken steps to guard himself against incurring the liability. Thus the extra condition of fault is grafted on to the tort for institutional reasons. Or some such parasitic, morally secondary reasons. Correspondingly the negligence condition could in principle be removed in appropriate cases without eating away at the moral foundations of the tort. Where someone embarks on ultra-hazardous activities, for example, they may be regarded as having been put on fair warning of their liability for the injuries they do merely by the fact that their activities were ultra-hazardous. They could plan not to incur the liability simply enough, by giving up the blasting or the chemical processing that is creating the hazard. So under some conditions—e.g. those prevailing in Rylands v. Fletcher—the extra negligence condition may defensibly be dropped, and then we are left with the moral essence of the tort, which emerges as the simple unadorned wrong of D injuring P.17 Or so goes the rival—and nowadays markedly contrarian—interpretation of the tort of negligence. 17. We can see here one reason why Brudner may have thought that strict tort liability demands of D that he be entirely self-effacing. It is not because tort liability is strict but because in modern legal systems it is typically strict and conditional, i.e. it is a strict liability that arises only when one is engaged in certain pursuits, such as blasting and manufacturing consumer products. These

125

  IV

We owe to Tony Honoré the most important modern attempt to rehabilitate this contrarian interpretation of the tort of negligence at common law.18 When we put the tort of negligence side by side with modern strict liability torts, argues Honoré, the first feature that calls for our defensive or critical attention is not the much-discussed feature that divides the two (“fault” v “nofault”), but the less-widely-remarked-upon feature that they both have in common, namely what Honoré calls their shared element of outcome responsibility. Depending as it does on the moral defensibility of outcome responsibility, the tort of negligence is in the same boat, morally speaking, as a strict liability tort. The tort of negligence is essentially a variation on the theme of a strict liability tort, in which the basic wrong lies in D’s actually injuring P. Accordingly, those who want to defend the tort of negligence at common law had better begin by defending, not attacking, the morality of strict liability. There is plenty of time to part company with enthusiasts for strict liability later, when one comes to the question of what further conditions must be met before legal liability is justified. In answer to that question, enthusiasts for strict liability may say “anything but a fault condition” while enthusiasts for the tort of negligence may say “a fault condition”. That polarisation is for later. Before that, according to Honoré, comes the morally more basic problem of understanding the common ground that unites the two competing sets of enthusiasts against everyone else, namely the moral significance that they agree in attaching to D’s actually injuring P—the element of outcome responsibility. extra conditions are needed to meet the problem of institutional fairness. To guarantee avoiding the strict liability, it is logically sufficient (but notice: not necessary) that one give up the pursuit in question. Now surely that would be a huge (self-effacing) burden for one to bear? True, but one cannot plead this as an objection to the liability’s being strict. Suppose, in response, the law called one’s bluff. Suppose it removed the condition and left one with an unconditional strict liability instead, applicable as much to car-driving and hairdressing as to blasting and consumer manufacturing. Then the law would have met one’s objection head-on by removing one’s reason to give up the blasting in favour of, say, hairdressing. Yet rather than eliminating strict liability the law’s reaction would have expanded it to take in hairdressing as well. The point I am making is that if strict liability in modern tort law seems to require that one efface oneself by taking extreme measures to avoid liability, that is usually because there is not enough of it around to make the self-effacement utterly fruitless. 18. Largely in the papers collected in his Responsibility and Fault (Oxford: Hart Publishing, 1999), although anticipated in the preface to the second edition of Hart and Honoré, Causation in the Law (Oxford: Clarendon Press, 1984). In the following footnotes, all references to papers reprinted in Responsibility and Fault use the pagination of the reprint.

126

        It should be said that Honoré’s first articulation of this contrarian line of thought suffered from an equivocation. At first he made the wrong connection between strict liability and liability for the tort of negligence. He said that it was the objective standard of care in the tort of negligence which gave that tort an affinity with strict liability torts. “For the objective standard of competence,” he wrote in his first major venture into the subject, “imposes a form of strict liability on that minority of shortcomers who cannot achieve it”.19 We have already seen that this is a mistake. To defend strict liability is to defend, minimally, the moral intelligibility of straightforward obligations to succeed. An obligation to try harder than one is capable of trying, of the kind that is created when “shortcomers” encounter the objective standard of care, is not an obligation to succeed. It remains an obligation to try. As we saw already, the distinction between obligations to try and obligations to succeed cuts across the distinction between obligations that one has only to the extent that one has the capacity to perform them and obligations (if there be any) that are not subject to this condition. The problem of strict liability correspondingly cuts across, rather than tracking, the problem of the objective standard of care in the tort of negligence. By lumping the two together Honoré equivocated about which of the two problems he was really trying to tackle. Honoré was led astray here by his instinct to interpret the common objection to strict liability as an objection to the intelligibility (or, on its other variant, the fairness) of the law’s exposing people to “moral luck”, i.e. luck in whether they come up to scratch in what they do and hence in what judgments they are properly open to. Surely, he reasoned, not only strict liability but also the objective standard of negligence can be objected to on this ground? True enough—if the objection makes sense. But, as Nagel established, it does not. There can be no such thing as a coherent general objection to our being exposed to moral luck. Attempts to explicate such an objection are an object lesson in the hazards of argumentative overkill. For what counts as luck is always, Nagel shows, luck only relative to some baseline or other. Whenever something is held to be luck, there is necessarily something else that is held not to be luck, and it is only relative to this second thing that the first counts as lucky or unlucky. The problem with a general objection to our exposure to moral luck is that everything we do is entirely a matter of luck relative to some baseline or other. Even when I maliciously attempt to do away with my professional rival—on any credible view an action that is open to 19. “Responsibility and Luck”, Law Quarterly Review 104 (1988), 530, reprinted in Responsibility and Fault, 14 at 22.

127

  some moral judgment—my doing so is luck relative to some baselines (e.g. the baseline of my genetic make-up, which pre-programmed no such behaviour and was compatible with my leading a life without any such attempt). It follows that to object to moral luck tout court is to object to morality tout court. Indeed it is to object to judging people’s actions by any standards at all. As Nagel himself puts it: “The area of genuine agency, and therefore of legitimate moral judgment, seems to shrink under this scrutiny to an extensionless point”.20 Misreading Nagel’s tone here, some took it that they were supposed to endorse the conclusion of this sentence, and to regard morality itself as having been discredited by Nagel’s demonstration that there is no place for morality to hide from luck. But in fact Nagel’s argument is a classic reductio. The conclusion is absurd—agency does have some reach and moral judgment does have some area of application—so something must have gone wrong in the premisses.21 What went wrong, Nagel showed, was the instinct to formulate as a general objection to the intelligibility (or fairness) of exposing people to “moral luck” what were really different people’s quite different and unrelated objections to the use of different baselines as the baselines for our moral judgments. Some people think that the question of whether we come up to scratch in what we do should be relativised (baselined) to our own personal capacities, and in the light of that view they rebel at objective standards of care as used in the law. Some think—an overlapping constituency, I suspect —that the question of whether we come up to scratch in what we do should be relativised (baselined) to the possible limits of trying, and in the light of that view they rebel at strict liability. Each of these suggested baselines needs to be defended, and for that matter objected to, on its own merits. Maybe refusing to acknowledge one or other or both of these baselines as fixing the proper boundaries of moral judgment is morally unintelligible or institutionally unfair. But the observation that in refusing to acknowledge them we expose people to “moral luck” does absolutely nothing to tell us what it is that would make such reliance either unintelligible or unfair. For (as Nagel shows) we are automatically exposed to moral luck, against some baseline, just in virtue of the fact that our actions are held up to judgment at all. The so-called “problem of moral luck” (like the so-called “problem of free will” with which it is supposedly connected) is therefore a pseudo-problem.22 It represents the 20. “Moral Luck”, above n. 11, 35. 21. Even if you don’t accept the assertions, Nagel and Honoré do. So maybe I should say that the argument is a classic reductio for their purposes. 22. An irony: Bernard Williams and Tom Nagel conjured up “the problem of moral luck” as a topic for their joint Aristotelian Society seminar in 1976. Both of them showed, in their different

128

        bundling together of various different problems about the grounds of moral judgment that have nothing at all in common save that they are all problems about the grounds of moral judgment. Two of these problems—and they are as distinct from each other as it is possible for two problems to be—are the problem of the objective standard of care (can there be moral obligations to do more than one is capable of doing—whether in the way of succeeding or trying—and if so should they be legally enforced?) and the problem of strict liability (can there be straightforward moral obligations to succeed—whether within or beyond the limits of one’s capacities—and if so should they be legally enforced?) Fortunately it did not take Honoré long to recover from the effects of this distracting conflation. A few pages later he puts it behind him and trenchantly captures the real respect in which the tort of negligence is morally aligned with strict liability torts of the Rylands v. Fletcher variety, and to that extent stands or falls with them. It turns out to have nothing to do with the objective standard of care. Rather, explains Honoré, “[s]trict liability is one species of enhanced responsibility for outcomes. This does not entail that whenever a harmful outcome is properly allocated to someone, this justifies imposing on him a strict liability to compensate for that outcome. . . . [R]esponsibility for a harmful outcome should not automatically involve a legal duty to compensate. An extra element is needed to ground the legal sanction. Sometimes [as in the tort of negligence at common law] the extra element is fault. . . . For strict liability [as under Rylands v. Fletcher] the extra element is usually that the conduct of the harm doer carries a special risk of harm of the sort that has in fact come about”.23

There admittedly comes a point at which one needs to decide what further conditions, if any, one will insist upon for legal liability: maybe (objective) fault, as in the tort of negligence, maybe not, as under Rylands v. Fletcher. But that little internecine squabble is for later. First, in the order of moral argument, comes the question of what kind of standard these further conditions of liability are supposed to be grafted onto, and why. In the tort of negligence and Rylands v. Fletcher torts alike, argues Honoré, the underlying standard is a standard of outcome responsibility, a standard which attaches a person (D) to the way his actions actually turn out (say, with P’s being injured). ways, that there is no such problem. Yet strangely the problem took off while its agreed dissolution was forgotten. Williams’ eponymous paper appears before Nagel’s in the original periodical printing (above n. 11), and is reprinted in Williams, Moral Luck (Cambridge: Cambridge University Press, 1981). 23. “Responsibility and Luck”, above n. 19, 27.

129

  I think Honoré chooses the label “outcome responsibility” to designate the issue that he is interested in here because the label’s intriguing ambiguities correspond to some intriguing ambiguities in the issue itself, and these Honoré understandably wants to keep alive for investigation. For a start, talk of an action’s “outcome” equivocates nicely between a reference to an action’s consequences and a reference to (what some philosophers call) its results.24 The consequences of an action are what they sound like: they are eventualities that follow an action and that are also causally connected to it. Results are different. They are the causal constituents of actions, i.e. they do not follow the action but form part of it.25 P’s death, for example, is a consequence of D’s trying to kill P. On the other hand P’s death is a result of D’s actually killing P, because the action of killing P (unlike the action of trying to kill P) is partly constituted by P’s dying. Until P is dead, D hasn’t killed him but has only tried to. Death is a consequence of the one action and a result of the other, but it is perfectly natural to say, in the intentionally ambiguous terminology chosen by Honoré, that it is the “outcome” of them both. As for “responsibility”, this is a notoriously multivocal term.26 For Honoré’s purpose, the interesting ambiguity that it harbours is this one. In some contexts “my responsibility” means something very close to “my liability”. Being responsible in this sense means facing the adverse normative consequences of having failed to perform one’s obligations (such as being subject to a power of punishment or being subject to a new obligation to pay compensation or apologise). But on other occasions “my responsibility” refers to something else: it refers to the obligations themselves rather than to the normative consequences of their nonperformance. Thus failing in one’s legal responsibility (obligation) may have, as one of its legal consequences, legal responsibility (liability). Actually, to be strict, a responsibility in the “obligation” sense is not exactly the same as an obligation, so the “” in the first set of parentheses here is slightly misleading. Rather a responsibility in this sense, like a right, is the ground of (one or more) obligations.27 Often, indeed, D’s 24. This useful terminology is G.H. von Wright’s, from Norm and Action (London: Routledge and Kegan Paul, 1963), 39ff. The distinction is interestingly nuanced in Anthony Kenny, Will, Freedom and Power (Oxford: Basil Blackwell, 1975), 54ff. 25. I call a result a “causal” constituent because the action of which it forms part is an action of causing that result (or occasioning it, or provoking it, or inducing it, or standing in some other causal relationship to it). 26. The best study of the various concepts of responsibility and the relationships among them remains H.L.A. Hart’s “Varieties of Responsibility”, Law Quarterly Review 83 (1967), 346. 27. This explains why, as Hart notices but does not satisfactorily explain (ibid. at 347), some obligations are not associated with any responsibilities, while others are associated with more than one. The same goes for rights.

130

        obligations to P are grounded in the combination of P’s rights and D’s responsibilities. But just as we did with rights, we may put this subtlety on one side for present purposes. That is because just as there is nothing that counts as D violating P’s rights other than D failing to perform her obligations to P, so there is nothing that counts as D failing to fulfil her responsibilities to P other than D failing to perform the associated obligations. Thus if we are interested in the ingredients of D’s wrongs—for instance, her torts or breaches of contract—the difference between an obligation and the responsibility that grounds it need not concern us. Nor, for that matter, need we concern ourselves with the similarly tangential conceptual wedge that can be driven between liability and responsibility-in-the-liability-sense. We have grasped the pertinent aspects of the two concepts of responsibility that bear on Honoré’s discussion if we think of them as responsibility (liability) and responsibility (obligation). These twin ambiguities – outcome (consequence? result?) and responsibility (liability? obligation?) – combine nicely to provide Honoré with the ambiguity in the expression “outcome responsibility” that he wants to preserve for the purposes of his discussion. One of the issues that he is interested in under this heading is the issue of whether it is fair to hold people responsible (liable), whether in law or in other social settings (e.g. games, professional codes of practice, traditional codes of honour, etc.), for the actual (as opposed to intended or projected) outcomes (consequences or results) of their actions. The other issue he is interested in under the “outcome responsibility” heading is the issue of whether we can really have responsibilities (obligations) to perform actions defined in terms of their outcomes (results). It does not take long to see that these two issues correspond to the two variants I mentioned of the common objection to strict liability. The first corresponds to the institutional fairness variant, which was an objection to (legal or other) liability based on non-performance of obligations to succeed. Honoré notices that if the institutional fairness variant of the objection is successful, it bites not only against forms of liability that depend on what one did-including-results (i.e. one’s success) but likewise against forms of liability that depend on what happened when one did it (the actual consequences of one’s actions). Thus he rightly reads the institutional fairness objection as an objection to outcome (results or consequences) responsibility (liability). The other issue he is interested in is, however, the deeper one of the two. It is the issue of whether it can ever be one’s responsibility (obligation) to perform actions that have a certain outcome (result). Only if it can be do we ever get to the further question of the fairness of making us legally (or otherwise) liable when we fail to perform such obligations. This deeper issue is the 131

  very same one that is raised by those who deny the moral intelligibility of obligations to succeed. To deny the moral intelligibility of obligations to succeed is to deny the very possibility of outcome responsibility in Honoré’s second sense, never mind its fairness. It is to deny that our responsibilities (obligations) really do extend beyond our merely trying to do things so as to include our actually succeeding in doing them (i.e. acting with specified results). Hence it is to deny that we can make morally intelligible a law that asserts that our responsibilities do so extend. Corresponding to these two different explanations of what outcome responsibility is,28 Honoré offers two quite different defences of it. One, designed to meet the institutional fairness objection, is his well-known and carefully crafted “betting” argument. Action is by its nature a gamble, he says: some you win, some you lose. Facing liability when you lose is only half of the story of outcome responsibility; the other half includes all the positive normative consequences (admittedly mostly laid on extra-legally) that flow from doing what you do successfully. So long as the mechanism really does cut both ways in social life taken as a whole, one has no complaint of unfairness merely because the law concentrates on the down side.29 Although I agree that this argument is sufficient to its task, I also tend to think that it is unnecessary. In my view, one can overcome the institutional fairness objection to (at any rate some pockets of ) strict liability in the law with far less elaborate manœuvres

28. Elsewhere in this volume, Stephen Perry calls them respectively the “social” and the “personhood” senses of outcome responsibility. 29. “Responsibility and Luck”, above n. 19, 24–9. Let me mention just one point about the “betting” argument in passing, because it supplements some remarks I have already made about the structure of Honoré’s enterprise. In pursuing the “betting” argument, it seems to me, Honoré gradually slips back into his initial mistake of confusing objections to strict liability with objections to the objective standard of care. As the “betting” argument proceeds he starts to engage with an imaginary objector who says: some lose more than they win and some win more than they lose. How is that fair? Honoré responds with an (implausibly) optimistic view of ordinary people’s failure rates that is supposed to marginalise the problem of net losers (at 28). He should have responded much more robustly by saying that the new imaginary objector was changing the subject and alleging a totally different unfairness. The new imaginary objector was objecting to the unfairness of the law’s failure to relativise legal liability to people’s varying capacities. Honoré’s “betting” argument was not, however, a defence of the law’s failure to relativize to capacities. It was a defence of the law’s refusal to ignore the importance of success as well as the importance of trying. All that one can ask of the betting argument is that it eliminate the unfairness that it was devised to eliminate, viz. the unfairness of outcome responsibility. That it leaves another supposed unfairness untouched is not, so to speak, its problem. I say “supposed” unfairness because I believe that, barring special cases, there is nothing unfair about people being held up to standards that personally they are unable to meet. The robust (and correct) answer to this complaint of unfairness is that, barring special cases (e.g. young infants and the seriously mentally ill), people should be able to meet the standards in question and have no complaint if they are judged by them when they cannot.

132

        that leave fewer hostages to fortune.30 But I am not going to pursue this matter in what remains of this paper. Instead I am going to focus my attention on Honoré’s second argument. This second argument is offered not as a rebuttal of the institutional fairness objection but rather as a response to the moral intelligibility objection. Can one even make sense of the idea that people have obligations to succeed? Can one make sense of the idea, in other words, that they are outcomeresponsible in the second of Honoré’s two senses? Not only can one make sense of this idea, answers Honoré. More to the point, one cannot manage without it. It is not the presence of outcome responsibility that makes no sense, but rather its absence: “[O]utcome allocation can be defended on grounds deeper than the overall balance of benefit over burden; and so, in its wake, can strict liability. For outcome allocation is crucial to our identity as persons; and unless we were persons who possessed an identity, the question of whether it was fair to subject us to responsibility could not arise. If actions and outcomes were not ascribed to us on the basis of our bodily movements and their mental accompaniments, we could have no continuing history or character. There would indeed be bodies, and associated with them minds. Each would possess a certain continuity. They could be labelled A, B, C. But having decided nothing and done nothing these entities would hardly be people”.31

This approach to the problem is exciting and unusual. The approach is to fight fire with fire. If the absence of outcome responsibility really makes no sense then its presence can’t but make sense, and those who think it unintelligible must therefore be thinking fallaciously. Honoré’s remarks therefore hold out the tantalizing promise of a philosophical role reversal, with “moral intelligibility” objectors to the legal enforcement of obligations to succeed finding themselves, for once, in the defensive position, with their own challenge thrown back at their feet. Nor is the challenger content with rescuing obligations to succeed from oblivion; he also aims to elevate them, in the process, to a kind of moral priority. Remember Brudner’s view that a regime of strict liability takes us to the point of total self-effacement, to the point at which, as agents, we are obliterated from the world? On the contrary, according to Honoré. A regime of strict liability represents the starkest possible reaffirmation of our agency and its importance in the world, because the simple 30. In my view one overcomes the objections simply by putting potential Ds on fair warning that they are embarking on an activity (e.g. blasting) in respect of which strict liability applies to them. Cf. n. 17 above on the “ultra-hazardous activities” condition as a fair warning condition. 31. “Responsibility and Luck”, above n. 19, 29.

133

  idea at the heart of strict liability—the idea of outcome responsibility—is the idea that we leave traces of ourselves forever imprinted on history, in the form of the countless welcome and unwelcome events that were (as Honoré puts it elsewhere) “unequivocally our doing”.32 In what we merely try to do this imprint is lacking, and the power of our agency is therefore but meanly represented. In that sense, trying to  is secondary. Success—actually ing—is primary. “It is only this primary outcome responsibility”, observes Honoré, “that can explain why we (rightly) judge murder more severely than attempted murder, and causing death by dangerous driving more severely than dangerous driving”.33 And it is only this primary outcome responsibility—he could have added—that puts the requisite moral distance between someone who merely fails to take an adequate measure of care not to injure P, but fortunately leaves P uninjured, and someone who fails to take that same adequate measure of care, by the same margin, and actually injures P in the process. For first and foremost, the deeper argument goes, we are what we do—complete with results.

V

This deeper argument of Honoré’s is never developed to the same level of specificity as his “betting” argument. In subsequent writings, indeed, Honoré has preserved the speculative and exploratory tone of the words just quoted.34 It is the tone of a philosophical promissory note. For this reason the force of Honoré’s remarks—the decisive argument that they tantalisingly promise— has never been fully brought home to those who doubt the moral possibility, let alone the moral priority, of straightforward obligations to succeed. Yet Honoré’s remarks do alert us to the basic steps of just such a decisive argument. True, the argument does not quite establish the moral intelligibility of straightforward obligations to succeed, let alone their moral priority over obligations to try. What it does establish, when fully spelt out, is the moral intelligibility of reasons to succeed, and one important sense in which those reasons have moral priority over mere reasons to try. With this much established, the most important source of doubt about the possibility of straightforward obligations to succeed—namely scepticism about the independent 32. Causation in the Law, above n. 18, at lxxxi. 33. “Responsibility and Luck”, above n. 19, 31. 34. For instance, in “The Morality of Tort Law: Questions and Answers” in D.G. Owen (ed.), Philosophical Foundations of Tort Law (Oxford: Clarendon Press, 1995), also reprinted as ch. 4 of Honoré, Responsibility and Fault, above n. 18, at 76–7.

134

        rational salience of success—is roundly despatched. Of course, at this point the question may still in principle be raised of whether these reasons to succeed with their special moral priority can ever be categorical mandatory reasons to succeed—that is to say, obligations.35 But once we see that there are indeed moral reasons to succeed and that they are in one way more basic than moral reasons to try, the available arguments for doubting that they can be categorical and/or mandatory reasons are few and unconvincing. Maybe more convincing ones could be found. But the challenge to find more convincing ones is at any rate decisively thrown at the feet of the doubters, as Honoré promised us that it would be. To see the real strength of Honoré’s argument, one needs to focus on nothing so much as its apparent weakness. “Having decided nothing and done nothing”, concludes Honoré briskly, “these entities would hardly be people”. Surely this conclusion goes much too far, much too fast? In the first place, for the purpose of telling the story of our lives we can surely insist on the significance of the events out in the world that were “our doing”, without allowing that this significance was necessarily a moral significance? Can’t things be part of the story of what we did, in an autobiographically pertinent sense, without being part of the story of our rightdoing and our wrongdoing? Couldn’t what we do full stop include our successes (and failures) and yet what we do wrong or what we do qua moral agents only extend as far as our attempts (and neglects)?36 But even if we postpone this first worry for a moment—which is exactly what I propose to do—there is a second, and possibly more alarming, one. Honoré seems to be claiming that if one excises results from the story of our rightdoing and wrongdoing, it is not only the case that there was nothing we did (in the sense of nothing out in the world that was our doing) but also nothing we decided. Surely, on the contrary, what we decided is among the things that are left over when we excise from the scope of our moral agency those events in the world that were our doing in Honoré’s sense? Deciding, to put it another way, belongs to the “trying” side of the trying/succeeding divide. That reminds us that there is after all a story left over when results are excised from the story of our moral agency. It is the story of our trying, a story of our endeavour rather than our achievement. And our deciding belongs to that very story. So how can it be said, as Honoré says, that in the absence of outcome responsibility we 35. On obligations as categorical mandatory reasons, see John Gardner and Timothy Macklem, “Reasons, Reasoning, Reasonableness”, in Jules Coleman and Scott Shapiro (eds), The Oxford Handbook of Jursiprudence (Oxford: Oxford University Press, 2001, forthcoming) 36. This is the gist of the response to Honoré proposed by Arthur Ripstein in his contribution to this volume. Ripstein distinguishes the undoubted “first-person” importance of outcomes (results, successes, achievements) from their possible “third person” (or moral) irrelevance.

135

  would have decided nothing as well as done nothing, so that there would be no story left of us as human agents, as opposed to a different one? Let me explain how exactly it can be said. Deciding, and for that matter trying, are actions of a logically parasitic type. One does not merely decide full stop or try full stop. Necessarily, one decides to  or tries to , where ing is another action. So necessarily there exists, whenever one tries or decides, some further action  such that one decides or tries, as the case may be, to perform it. What is more, the kind of parasitism involved here is a distinctively rational kind of parasitism. The “to” in the expressions “trying to” and “deciding to” (like that in “taking care to”) is the familiar intentionimplicating “to” that we also find in the expressions “with a view to”, “in order to”, and “intending to”. To be exact, trying to  is acting with a view to ing, while deciding to  is (one way of ) preparing oneself, with a view to ing, to act with a view to ing (roughly, it is trying to make it the case that one will try to ). In these characterisations, acting “with a view to ing” means, in turn, acting for the reason (inter alia) that one’s action will (supposedly) contribute to one’s ing. That one’s action will contribute to one’s ing is a reason for that action, however, only if one also has a reason to . Thus to act for the reason that one’s action will (supposedly) contribute to one’s ing is possible only if one is prepared to regard or treat oneself, at least for present purposes, as having a reason to . This is not an optional extra. Someone who really thinks that they have no possible reason not to injure P—i.e. no possible reason to succeed in not injuring P— cannot conceivably try not to injure P or decide not to injure P, for they cannot conceivably act or prepare to act for the reason that what they do will not injure P. Accordingly, if it is impossible to make sense of the idea of a reason to —where  signifies the successful action which one is trying or deciding to perform—it is also impossible to make sense of the idea of a reason to try to  or to decide to . Honoré is right, then, to think that if our success turns out to be rationally insignificant across the board, then our trying and deciding (etc.) cannot but be rationally insignificant too. Assuming, then, that his conclusion is about rational significance—about what belongs to the story of our lives as rational agents—he is spot on. To deny that success can have independent rational significance is to leave us without any story of our lives as practical reasoners. It is not merely to leave us with a story of our lives as practical reasoners that omits our successes and failures but includes our attempts, decisions, precautions, neglects, etc. For the latter story depends for its intelligibility on our granting the intelligibility of a more complete—or dare I say morally richer?—story in which our successes and failures are registered as rationally significant too. 136

        An unargued assumption of many moral philosophers as well as almost all economists and decision theorists—and I suspect also of most lawyers—is that a reason to  is the very same thing as a reason to try to . Indeed the statement “D has a reason to ” is often promptly interpreted, without explanation, as “D has a reason to try to ”. But the above remarks already show that this is a serious mistake. A reason to try to  is not a reason to  but a reason to act for the reason (inter alia) that one’s so acting will (supposedly) contribute to one’s ing. This difference makes possible a variety of important asymmetries between one’s reasons to  and one’s reasons to try to . I will mention three. (i) Where acting with a view to ing would not contribute to one’s ing, one has reasons to  (i.e. reasons to succeed) without corresponding reasons to try. I am on a clifftop miles from anywhere looking down helplessly on a man drowning in the stormy sea below.37 Because no amount of trying would bring me closer to success, my reasons to save the man do not yield any reasons to try to save him. Yet I still have the same reasons to save him that I would have if doing so were perfectly straightforward. That is why the situation is so horrifying. If my reasons to save him were eliminated by the impossibility of my doing so then my not saving him would be nothing to me; it would leave no trace on my life as an agent; I could walk away without compunction. As it is I am merely blocked from doing as these reasons would have me do because there is no contributory action—including trying to save— that I have any (derivative) reasons to perform. The situation is an unusually stark variant of a common one in which one has more reason to succeed than one has (derivative) reason to try, thanks to the fact that one’s prospects of succeeding by trying are limited.38 (ii) I add “derivative” in parentheses here because I do not mean to deny that one may have additional reasons to try that do not correspond to reasons to succeed. This is an asymmetry in the opposite direction. Just as one may have less reason to try than to succeed, so one may have more reason to try than to succeed. The extra reasons to try in such a case are non-derivative reasons to try: they are reasons to try quite apart from the contribution that trying makes to success. Suppose that it is my daughter or my brother drowning in the sea below. Then I have additional reasons to try to save her or him, since such an 37. For more on this case see my “Justifications and Reasons”, in A.T.H. Smith and A.P. Simester (eds), Harm and Culpability (Oxford: Clarendon Press, 1996). 38. I mean one’s prospects of succeeding by any amount of trying. The problem under scrutiny here is different from the problem of self-defeating endeavour encountered earlier, in which the problem was that of how much one tried, granting that one had every reason to try.

137

  attempt may be an expression of love even if doomed (or perhaps: even more of an expression of love because it is doomed). Naturally it is possible to act for these additional reasons, and possibly, in some cases, one does not express what those reasons would have one express by one’s doomed attempt unless one actually acts for them. But it is not possible to act for these additional reasons alone, because unless one also aimed at success (i.e. also acted for the derivative reason that one’s trying would supposedly contribute to one’s succeeding) one just wouldn’t be trying—and so, obviously, one wouldn’t be doing what one’s non-derivative expressive reasons to try would have one do. This was why I included the parenthetical “inter alia” in my characterisation of reasons to try: A reason to try to  is not a reason to  but a reason to act for the reason (inter alia) that one’s so acting will (supposedly) contribute to one’s ing. There are other reasons to try apart from those based on the contribution that trying makes to succeeding, and perhaps some of these other reasons to try are obligations. One can try for these additional reasons. But the (supposed) contribution that one’s trying makes to succeeding must be among one’s reasons for trying or else one just isn’t trying. (iii) Does it follow that there are no possible reasons to try unless it is the case that there is at least one reason to succeed, and that trying would contribute to one’s success? You may think that this follows, but it doesn’t quite. What follows from the foregoing is that, to have any reason to try, the agent must believe or accept that she has a reason to succeed and must believe or accept that trying would contribute to succeeding. That was why I included the parenthetical “supposedly” in my characterisation of reasons to try. I could try to save the drowning man if I mistakenly thought that I could just possibly save him and that I had a reason to do so. In this case, I could have reasons to try to save him. Of course I don’t have the particular reason that I take myself to have, viz. the derivative reason. About that I am mistaken. But I might still have the non-derivative reasons, e.g. that of expressing my love for him, and I can do as these reasons would have me do by acting for what I mistakenly take to be the derivative reason. These are just a few of the many complexities that figure in the relationship between reasons to try and reasons to succeed. I mention them, in admittedly brief outline, to bring out the sense in which reasons to succeed are “primary” (to use Honoré’s word) as well as the sense in which they are not. Reasons to succeed are primary in the sense that the intelligibility of reasons to try depends on the independent intelligibility of reasons to succeed. If there can be no such thing as a straightforward non-derivative reason to succeed then there can be no such thing as a derivative reason to try. From this 138

        it follows that there can be no such thing as trying (or deciding, intending, aiming, etc.), and thus there can be no such thing as a non-derivative reason to try either. But once we have the possibility of derivative reasons to try on the table (by admitting that there are reasons to succeed from which those reasons derive), non-derivative reasons to try also become intelligible. Thus the case for trying is not by any means exhausted—it is not necessarily even dominated—by the case for succeeding. In that sense the story of my life as a rational agent, told as a story of endeavour, is not merely a pale shadow of the story of my life, told in terms of achievement. Both aspects have some independent significance and the full story of my life as a rational agent is the story that has room for both: it is the story of my trying (or neglecting) and my succeeding (or failing), including sometimes (double triumph) my succeeding-by-trying and sometimes (double trouble) my failing for want of trying. The last notably discouraging case is the one that the tort of negligence latches onto: the case of someone who fails for want of trying, whose action is doubly deficient in its conformity with reasons, for not only did she not succeed; judged by the applicable measure of assiduousness, she didn’t even try. This line of argument almost fulfils Honoré’s promise, I believe, and does so (I hope) in the spirit in which the promise was originally intended. What Honoré was promising was an argument according to which, firstly, our not having straightforward obligations to succeed would be unintelligible (so that our having them couldn’t possibly be unintelligible) and according to which, secondly, our obligations to succeed would have some kind of argumentative primacy over other obligations, such as obligations to take care. The argument just ventured meets these conditions readily, save only that it softens “obligations” to “reasons”. So can the argument be replicated, mutatis mutandis, with “reasons” hardened back up to “obligations”? Not quite. That is because the substitution of “obligations” makes for additional asymmetries on top of those I labelled (i), (ii) and (iii). It is perfectly possible that obligations to try might derive from non-obligatory reasons to succeed, or that nonobligatory reasons to try might derive from obligations to succeed. Thus there being obligations to try without obligations to succeed is not made unintelligible by the same knock-down manœuvres as made it unintelligible that there are reasons to try but no reasons to succeed. Obligations to try to  could, after all, be obligations to act for the (itself non-obligatory) reason (inter alia) that one’s so acting will (supposedly) contribute to one’s ing. Still, after the foregoing considerations the ball is now firmly in the court of those who deny that we can be subject to obligations to succeed. If there are indeed reasons to succeed—and there are—is there any possible reason to 139

  doubt that they could be obligations to succeed, such that failing to do as they would have one do would be wrong and hence could properly be made tortious in law? An obligation is no more and no less than a categorical mandatory reason. It is categorical in the sense that it applies to people independently of their prevailing personal goals. It is mandatory in the sense that it is a reason that operates, on at least some occasions, to the partial or total exclusion of at least some countervailing reasons. Why would anybody think that these particular properties—being categorical and being mandatory—could not be possessed by reasons to succeed, now that the possibility of reasons to succeed has been established? Commonly, in my view, the following mistake tends to steer lawyers’ thoughts in that direction. They think that a reason to do something counts as mandatory if and only if we would be justified in attaching adverse normative consequences (legally or otherwise) to its nonperformance. Since attaching adverse normative consequences to non-performance would be unfair in the case of a straightforward obligation to succeed (the thinking goes), it follows that there can be no such obligation. The problem with this line of thinking is that it wheels out the problem of unfairness much too early. Once we have established that an action is obligatory, it remains to be discussed whether the obligation should be enforced, or more generally whether people should have to bear any adverse normative consequences of its nonperformance. Possibly, as Honoré points out, further conditions have to be met before such measures would be justified, including measures to overcome the institutional fairness objection. We always admitted that there would be further internecine squabbles to come on this subject: what some would regard as sufficient to overcome the institutional fairness objection (e.g. the Rylands v. Fletcher condition of an ultrahazardous activity) others would regard as insufficient to meet that objection. But it is jumping the gun to use the institutional fairness objection as an objection to the very possibility of the obligation, when it is not yet a foregone conclusion that those who fail to perform it will be subject to any adverse normative consequences at all. To put it simply, the institutional fairness objection is one objection, and the moral intelligibility objection is another. One cannot have two bites at the institutional fairness objection by saying that obligations to succeed are morally unintelligible because if only they existed they would have unfair normative consequences. The proper response to this alleged unfairness is to detach the normative consequences, not to deny the intelligibility of the obligation. Those who say that they cannot detach the normative consequences because they are built into the very idea of mandatoriness have simply misunderstood the idea of 140

        mandatoriness.39 The mandatoriness of a reason lies in the fact that it operates to the exclusion of at least some countervailing reasons. Whether one is subject to adverse normative consequences in the event that one does not do as the reason would have one do is a separate—detachable—matter.

VI

To the best of my knowledge, only one serious (philosophically credible) objection has ever been raised to the proposal that reasons to succeed can be obligatory reasons. It is Kant’s famous objection. According to Kant: “A good will is not good because of what it effects or accomplishes—because of its fitness for attaining some proposed end: it is good through its willing alone—that is good in itself. . . . Even if, by some special disfavour of destiny or by the niggardly endowment of stepmotherly nature, this will is entirely lacking in power to carry out its intentions; if by its utmost effort it still accomplishes nothing, and only good will is left (not admittedly as a mere wish but by the straining of every means so far as they are in our control); even then it would still shine like a jewel for its own sake as something which has full value in itself. Usefulness or fruitlessness can neither add to, nor subtract from, this value”.40

These remarks furnish the first and second premisses of Kant’s curious argument against the moral intelligibility of obligations to succeed. It is an argument that dwells on the fact that obligations are categorical reasons (not conditional on the agent’s prevailing personal goals) rather than the fact that they are mandatory reasons; thus it leaves fully open the possibility of mandatory noncategorical reasons to succeed. The argument, which is completed by Kant with the addition of a third premiss several pages later,41 goes something like this: (1) the only source of unconditional (a.k.a. moral) value in our actions is the good will; (2) the good will infects not the whole of what we do but only that part of it that consists in our trying to do good; (3) performing one’s obligations is of unconditional (a.k.a. moral) value; (4) thus there can be no obligations to succeed but only obligations to try. 39. To be exact they are still in the thrall of the crudest “sanction theories” of obligation that were decisively discredited by Peter Hacker in his famous “Sanction Theories of Duty”, in A.W.B. Simpson (ed.), Oxford Essays in Jurisprudence: Second Series (Oxford: Clarendon Press, 1973). 40. Groundwork of the Metaphysic of Morals (trans. Paton, New York: Harper and Row, 1964), 62. 41. Ibid. at 68.

141

  This objection is sometimes confused with what I earlier dismissed as the pseudo-objection to “moral luck”. But in fact Kant’s argument is much more carefully targeted and hence withstands much more critical attention. Unlike the “moral luck” pseudo-objection, for instance, Kant’s objection has no quarrel with the objective standards of trying that are at work in the legal criterion of negligence. Indeed in premiss (1) such an objective standard is explicitly set up, namely the standard of the good will, the standard of perfect moral virtue. Not everyone, Kant agrees, is capable of meeting this standard. Some people, indeed, are moral degenerates who would not recognise a moral consideration if it slapped them in the face.42 Relative to the baseline of their attenuated moral capacities it is a stroke of bad luck that they have the moral obligations that they have, and which they are doomed by their own degeneracy to violate. So Kant’s is not a version of the “moral luck” pseudo-objection. Rather it is an objection carefully targeted against obligations to succeed. And even regarding obligations to succeed its impact turns out to be highly selective. On closer inspection Kant’s third premiss calls for a modification even by the lights of his own views. In the end he only stands by the more limited claim that it must be possible for performing one’s obligation to be an act of unconditional value. Thus possession of a good will must be sufficient, even if not always necessary, for performance of one’s duties.43 This rules out straightforward obligations to succeed, and what I called “tall-order” hybrid obligations (obligations to succeed by trying), but it does not rule out the short-order hybrid obligation (the obligation not to fail for want of trying), nonperformance of which constitutes the tort of negligence at common law. For as we saw trying (to the legally specified degree) is sufficient, but not necessary, to perform that obligation. Thus, thanks to an important Kantian concession on the third premiss, the tort of negligence is morally intelligible to Kant. On the other hand, a strict liability tort of the Rylands v. Fletcher type still is not. I do not propose to tackle Kant’s argument here. Showing that it collapses —which it does, spectacularly—is a task for another paper. Here I merely 42. Thus it is an error to associate Kant with the thesis that “ought” implies “can” in the way in which this thesis is normally read. In the sense in which it is normally read it is taken to mean that those who lack certain moral capacities lack the corresponding obligations. Kant believed that the doctrine worked in the opposite way. He believed that since (necessarily) everyone has the obligations it follows that fundamentally they have the moral capacities as well. Qua human they have it in them to be less incapable than they are. As Kant spells it out: “Ethical duties must not be determined in accordance with the capacity to fulfil the law that is ascribed to human beings; on the contrary their moral capacity must be estimated by the [moral] law, which commands categorically.” (Kant, The Metaphysic of Morals (trans. Gregor, Cambridge: Cambridge University Press, 1996), 164.) 43. See the Groundwork, above n. 40, at 65–6.

142

        leave the argument on the file, for it seems to me that by developing Honoré’s sketchy thoughts, we have now done the most important work of turning the moral intelligibility objection to strict liability back against its supporters. To achieve this what we needed to do was to bear out Honoré’s claim that a story of our lives as rational agents that refuses to admit the rational salience of our successes (and failures) as well as our tryings (and neglects) is an unintelligible story. It cannot even accommodate the rational salience of our tryings (and neglects) and so, in the end, comes down to no story at all. You may say, reprising an earlier worry that we postponed, that it is one thing to establish the rational salience of success and another to establish its moral salience. Maybe there are reasons to succeed but surely they need not be moral reasons? Possibly in some senses of the often-unhelpful word “moral” this is true. For example, in Kant’s rather technical sense, only non-derivative reasons to try could possibly be moral reasons, whereas derivative reasons to try and the reasons to succeed from which they are derived would instead be set aside as “prudential” reasons. Such matters of classification need not detain us here. For the notion of the morally intelligible that was implicated in the moral intelligibility objection was not, you will recall, a narrow one in which moral considerations could be contrasted with prudential ones or aesthetic ones or economic ones, etc. The notion of the morally intelligible that we had in mind was merely the notion of what is intelligible apart from the law (or similar institutional arrangements) so that the law could intelligibly claim to be binding people from something other than a narrowly legal point of view. Our argument showed that reasons to succeed are indeed intelligible apart from the law—i.e. morally intelligible in the relevant sense—and that reasons to try are not morally intelligible, in the relevant sense, without them. From here it is, in my view, but a relatively short step to the conclusion that at least the first half of this conjunction still holds true when the reasons in question are obligatory. There is no special problem, in my view, with mandatory categorical reasons to succeed. But instead of making that short step here let me just leave it in the hands of others, including those who remain attached to Kant’s curious argument, to try and block it. That, I think, is the way best to honour the spirit of Honoré’s ground-breaking contrarian contribution to the philosophical study of the law of torts, and indeed the philosophical study of the human condition.

143


7 UNPACKING “CAUSATION” Jane Stapleton*

The central insight of Hart and Honoré’s Causation in the Law1 is that we can understand more of what we mean in this area if we consider how we use causal language. Their own linguistic study led them to construct an atlas of causal usage which is still the essential starting point for the excavation of such underlying meaning. In this study I suggest that a consideration of the diverse uses to which causal language is put leads to some quite different conclusions from those drawn by Hart and Honoré. Using the law of torts as the exemplar, Part 1 of this essay considers the core claim of their book:2 that the law has not in the past adopted, nor should it in the future adopt, a special “legal” meaning for “causation”. I argue that if we look at how we use causal language we see that sometimes it expresses views about the factual question of whether a factor was historically involved in a particular outcome, and sometimes it expresses a view about whether a factor that was historically involved is relevant to the specific purpose we have for the enquiry. I then argue that since, in “ordinary” use, causal language serves many different functions and can mean different things, “common sense causation” is not a single elemental concept. Because of its ambiguous * I am grateful to the Editors for many helpful comments on drafts of this article. Especial thanks are due to Professor Tony Honoré and Professor Richard Wright. Our many conversations helped to clarify for me important differences between their well-known approaches to this area and my own. I am most grateful for their kindness, patience and intellectual generosity. 1. Hart and Honoré, Causation in the Law (Oxford: Oxford University Press, 1st edn 1959, 2nd edn 1985). 2. The most important analysis is by Wright, “Causation in Tort Law”, Cal L Rev 73 (1985), 1735; “Causation, Responsibility, Risk, Probability, Naked Statistics and Proof: Pruning the Bramble Bush by Clarifying the Concepts” Iowa Law Rev 73 (1988), 1001. See also reviews of the second edition: Northern Ireland Legal Quarterly 38 (1987) 214 (Robert Moles); Peter Lipton, “Causation Outside the Law” in Hyman Gross and Ross Harrison (eds), Jurisprudence: Cambridge Essays (Oxford: Oxford University Press, 1992) 127; Ethics 97 (1987), 664 (Lawrence Becker); Loyola Law Review 34 (1988), 463 (Henry Gabriel); Cambridge Law Journal 44 (1985), 477 (Tony Weir); Yale Law Journal 96 (1987), 1389 (David Howarth) – footnote 2 of which helpfully lists most of the reviews of the first edition. For law students a particularly useful short review of the first edition is Patrick Kelley, “Causation and Justice: A Comment” Washington University Law Quarterly 4 (1978), 635.

  nature, it cannot fulfil the needs of the law and that this is the source of current doctrinal difficulties in this area: sometimes “causal tests in law” are addressing disputes about whether a factor played a role in the history of an outcome;3 sometimes they are addressing issues of the relevance of an historical factor to a particular purpose. I conclude that law should restrict the deployment of causal language to the historical involvement enquiry: in other words, that law needs to embrace a special meaning for “causation”. Part 2 then unpacks “causation” into these two distinct underlying enquiries that causal language is used to express but often obscures, historical involvement and purposive relevance. Once that separation is in place it is possible to appreciate why the “necessary element in a sufficient set” (“NESS”) notion formulated by Hart and Honoré as a test of “causally relevant condition”4 and elegantly popularised by Wright as a test of “causation”,5 is in fact only relevant to the historical involvement enquiry. In relation to that enquiry it is indeed the correct approach. I suggest, however, that the NESS approach should be crystallized in a “targeted but-for” test, which allows us to search for the perspective from which the targeted factor can be seen to have played a role in the history of the outcome. We must ensure that we apply the “targeted but-for” test separately to every element constituted or introduced by the candidate factor. We must also ensure we apply it to all candidate factors. We must resist the temptation to target only one effect of the targeted factor. We must also resist the temptation to target only one factor and so treat all other factors as part of a “set-stage”: what I call, the single “snap-shot” view of history. Unless we consider separately each effect of a targeted factor and consider every possible factor, the continuous interrelated nature of historical processes will not be revealed and we will fail to identify all the factors that played a role in the history of an outcome. P 1: L N  S M  “C” A. Contributions to Understanding

At the outset it is important to note significant contributions to our understanding that flow from the work of Hart and Honoré. One is their insistence 3. I name the elemental concept underlying our understanding of why the story of the double hit hunters’ case would be incomplete without the careless conduct of each hunter, “playing a role in the history of the outcome”: see text at n. 109 below. 4. Causation in the Law, 2nd edn, above n. 1, at 113. 5. That captures “the essence of causation”: Wright, “Causation, Responsibility, Risk, Probability, Naked Statistics and Proof: Pruning the Bramble Bush by Clarifying the Concepts”, above n. 2, at 1019.

146

  that it is legal doctrine that determines the threshold issue of what constitutes the elements of the particular tort.6 So, for example, in the tort of negligence it is legal doctrine that determines that a doctor owes a duty of care to the patient, the standard against which breach of the duty is judged, that damage is the gist of the action7 and that the tortious conduct must have played a role in the history of the damage for which recovery is sought.8 Another contribution was the way they successfully refuted the view of Realist scholars such as Leon Green9 that the reason we do not say, for example,10 that the decision of the great-great-grandmother of Lee Harvey Oswald to have children was a “cause” of President Kennedy’s death in Dallas, can be adequately captured by the term “legal policy”. Hart and Honoré’s work convincingly suggests to us that where the law refuses to impose liability on a person for a consequence of their conduct, the law will have taken into account a range of legal concerns not all of which can be dismissed as pragmatic “policy”: some are far more accurately described as concerns of justice, fairness, blame and moral responsibility. Another significant advance is Honoré’s thesis11 that it is essential to the perception of self and human identity that the individual be identified with some of the consequences of his or her conduct. Among other reasons, this thesis is important because in its corollary, that it is essential to human identity that the individual be identified with only some of those consequences, it points to a concern, neither policy-generated nor specifically “legal”, that operates when as citizens we do not blame a person for the endless stream of consequences of his or her conduct, or when as jury-persons or judges we truncate the defendant’s scope of liability for the consequences of his or her tortious conduct.12 6. Including what constitutes the tortious conduct, whether the tort is actionable per se, and what classes of damage are recoverable under the particular tort (for example, the injury of disruption by noise is recoverable in the tort of nuisance but not in the tort of negligence). Causation in the Law, 2nd edn, above n. 1, at lix–lx and lxii. 7. Stapleton, “The Gist of Negligence, Part 2: The Relationship between “Damage” and Causation” Law Quarterly Review 104 (1988), 389. 8. Honoré, Responsibility and Fault (Oxford: Hart Publishing 1999) at 100: “the law determines the way in which the causal inquiry is framed”. 9. The Realists promoted an analytical configuration that separated the factual question of historical involvement (which they called cause-in-fact) from the issue of the appropriate scope of liability for the consequences of tortious conduct (termed “proximate cause”) which they said was governed by “policy”. It was in the latter context they would have said that Mrs Oswald was not a proximate “cause”. 10. This is the example I used in: Stapleton, “Perspectives on Causation” in J Horder (ed.), Oxford Essays in Jurisprudence, Fourth Series (Oxford: Oxford University Press, 2000) 61 at 67. 11. Causation in the Law, 2nd edn, above n. 1, at lxxx; Honoré, Responsibility and Fault, above n. 8, ch. 2 (reviewed at Law Quarterly Review 116 (2000) 321, by Roderick Bagshaw). 12. Though, of course, there are other concerns at work here too. In general see Cane, “Responsibility and Fault: A Relational and Functional Approach to Responsibility”, in this volume.

147

  A final core contribution was that by attempting to map the contours of causal usage, Hart and Honoré provoke us to ask whether those patterns “mean” something and, if so, what it is that they mean. To set in context why a consideration of causal usage might lead to different conclusions than those Hart and Honoré drew, certain general claims of Causation in the Law must be set out. B. The Claim that Ordinary Causal Usage
Legal Causal Usage13

Hart and Honoré argued that “it is the plain man’s notions of causation (and not the philosopher’s or scientist’s) with which the law is concerned”14 and should be concerned.15 They argued that these notions have very deep roots in all our thinking and that the delineation of the causal concepts which pervade ordinary thought, which they call “common-sense notions of causation”, is an indispensable preliminary to legal judgments of responsibility even if the end conclusion is that they are outweighed by other considerations.16 Ordinary Usage in which Context?

Though in parts of the book Hart and Honoré rightly emphasize that causal usage may be affected by the context and purpose of the enquiry,17 most of the analysis seems to focus on those non-legal contexts in which “ordinary people” are using causal language to attribute moral responsibility.18 Yet it is undoubtedly the case that “ordinary people” use causal language to express themselves in a whole range of enquiries, some of which are not concerned with moral attribution at all. For example, people frequently use causal 13. Lipton, above n. 2, accepts this “shared concept” claim, but Moles, above n. 2, argues that “to treat the language of the law as an example of ‘ordinary language’ is to refuse to come to terms with the point which is made at the beginning of any introductory sociological analysis of the topic . . .” (at 215) and Howarth, also n. 2, at 1404 concludes the “authors are advocates, not sociologists”. 14. See Causation in the Law, 2nd edn, above n. 1, at 1. See also at xxxiv (“it is the ordinary man’s conception of cause that is used by the law”) and xliii (“customary legal language which is also that of ordinary speech”). 15. This followed from their embrace of the “principle, inherent in the rule of law, that legal responsibility should depend on criteria which ordinary people can apply, even if they cannot explain the principles on which they do so” (emphasis added): Honoré, “Herbert Lionel Adolphus Hart 1907–1992” Proc Brit Acad 84 (1994), 295 at 304. See also n. 66 below. 16. Causation in the Law, 2nd edn, above n. 1, at 1–2. See also at lxvi (“common sense causal criteria . . . are deeply ingrained in the thought of both ordinary people and lawyers . . .”). 17. Ibid. at 28–41: “normality” depends on context; and at 110–11: “bifurcation of causal questions”. 18. Or at least identify certain consequences of a person’s conduct with that person. See n. 11 above.

148

  language to describe factors that make up the “recipe” that led to an outcome. This is particularly true of scientists:19 Newton may well have expressed his famous “apple” observation as “gravity is a cause of an apple falling to earth”. Sociologists may describe as a “cause” of the falling crime rate the decisions of socio-economically deprived women to abort their pregnancies.20 Lawyer– economists may say that a “cause” of the diethylstilbestrol (DES) injuries was the presence of the foetus in the womb when the drug was ingested. My favourite example comes from a documentary about car accidents: drinkers in a pub said that one of the “causes” of the high road toll were cement telegraph poles! Hart and Honoré do note, however, that the plain man’s notions of causation “have very deep roots in all our thinking and in common ideas of when it’s just or fair to punish or exact compensation”21 so it seems that by “the plain person’s notions of causation” Hart and Honoré mean the plain person’s use of causal terms only in the context of the allocation of something like blame or individual responsibility for outcomes. Causal Usage “Patterns” reflect Consensus

Even if the analysis is limited to this context, many22 would argue that in situations where the facts are clear, what we are seeing in the causal usage Hart and Honoré describe are conclusions based on unexpressed and complex responsibility23 judgments, and that only where there is a consensus about 19. See text at n. 114 below. 20. For example, when two academics released their research showing a correlation between the increased use of abortion following the liberalization of its availability in Roe v Wade and the recent reduction in the US crime rate, the press reported that the goal of the researchers had been to “shed light on the causes of crime” (Washington Post, 22 August 1999. See also The Guardian, 10 August 1999: to “determine what causes reduced crime”) and that their conclusion was that “legalizing abortion in the 1970s may be a leading cause of plummeting crime rates in the 1990s” (San Jose Mercury News, 10 August 1999). Contrast other contexts, such as that discussed in relation to Mrs Oswald, where we refuse to use causal language to describe the link between the reproductive decisions of ancestors and the crimes of their descendants. 21. Causation in the Law, 2nd edn, above n. 1, at 1. But contrast the triple function of the concept of “causation” suggested by Honoré, n. 60 below. 22. This is a general complaint reviewers made of the book: see e.g. Kelley, above n. 2, especially at 640: “common sense notions of justice are at work here, not common sense notions of causation”. 23. Even in a context of the attribution of responsibility we may use causal language to express our common conclusion about the prior historical involvement question, which is a pure factual issue. We would all agree that the Texan Governor’s invitation to President Kennedy to visit Dallas played a role in the history of the death of Kennedy and when expressed in causal terms this will produce a uniform pattern of usage: Governor Connally’s invitation was a “cause” of the death. But what this common pattern reflects is our agreement about a fact, the fact of historical involvement. The common pattern of usage in relation to Mrs Oswald does not reflect agreement about a fact but our agreement about attribution of responsibility.

149

  these underlying judgments were Hart and Honoré able to find a consensus in causal usage. For example, it may well be that, in the context of the allocation of moral responsibility, no one would say that a driver who had carelessly injured a pedestrian was a “cause” of the latter’s death when the ambulance conveying her to hospital is struck by lightning. They might even say that the lightning broke something they call the chain of causation between the bad driving and the death. The question is, what do these patterns and metaphors tell us? Hart and Honoré seek to “translate [such] bewildering metaphors into factual terms”,24 which elsewhere they call “common sense causal criteria”, “causal principles” or “common sense principles”.25 For example, they deduce from cases such as this one about lightning that the abnormal conjunction of events intervening between the careless driving and the death negatived a particular type of “causal connection” they call “causing harm”.26 No Such Consensus in Forensic Disputes

Even if we put aside for the moment questions of whether such “principles” are “factual”, internally coherent or analytically useful, it is hard to see how they could be of forensic assistance in those many legal disputes where the parties agree about the facts but nevertheless couch their remaining disagreement in causal terms. In such cases the disagreement about causal language merely reflects an underlying dispute about responsibility. As Lord Hoffmann has said about appellate cases that are presented in terms of “causation”, “the argument is almost always an argument over the law . . . it is an argument over the true scope of the rule which imposes liability”.27 Unless we explicitly address these underlying responsibility disputes and the conflicting concerns that drive them, we fail to address the real nature of the dispute between the litigants. Take the case of a golfer who makes a negligence claim against a golf course, which provided no weather shelters, for being injured by lightning.28 The 24. 25. 26. 27.

Causation in the Law, 2nd edn, above n. 1, at xxxiv. Ibid. at lxvi, 49 and 170 respectively. Ibid. at 163. Lord Hoffmann, “Common Sense and Causing Loss”, lecture delivered to the Chancery Bar Association, 15 June 1999. 28. See the discussion of lightning and “freakish/indirect outcomes” in Stapleton, “Legal Cause: Cause-in-Fact and the Scope of Liability for Consequences” Vanderbilt Law Journal 54 (2001), 941. See also the discussion by Hart and Honoré of the “fumes in the barge” case of Johnson v Kosmos Portland Cement Co (1933) 64 F.2d 193 (6 Cir.) (cf. Causation in the Law, 2nd edn, above n. 1, at xlv and 201).

150

  plaintiff must assert that the golf course’s alleged carelessness in not providing any shelters bore a causal relation to his injury. The golf course may well concede that it was careless in not having provided rain shelters, but may try to defend itself by saying that, in the circumstances, there was no causal relation between this tortious conduct and the injury, because the lightning broke the “chain of causation”. What is really at stake here, given the facts are agreed, is the normative evaluation of whether a golf course should have provided shelters that were resistant to lightning. This is a dispute about responsibility, and it is one on which reasonable minds might differ for a variety of moral as well as policy reasons. Hart and Honoré’s elaborate differentiation of types of “causal connection” seems not only to obscure the nature of the problem, but to ignore the conflicting causal assertions of the parties to the litigation. Hart and Honoré insist that their approach can give important forensic assistance. Though they concede that “common judgments” of “ordinary people” provide only a general guide, they claim they still give an “important indication of what are the relevant factors” for courts.29 So while they concede that there are borderline cases where “owing to the character of the intervening events” we may disagree, they argue that “use can still made of common sense notions . . . [because] decisions here may be controlled not simply by considerations of what would be just, or expedient, or on the whole best . . . but also by considerations of how like or unlike the standard clear case the present debatable case is”.30 Yet it is not clear what the “relevant factors” are or how this analogizing works31 except for the “character of the intervening events”.32 The Ambiguous References to “Causal” Notions as Issues of “Fact”

It is also not clear what Hart and Honoré mean when they say that they are translating causal metaphors into “factual terms”,33 that common notions of causation are “questions of fact”34 and so on. If they mean only that, for the 29. 30. 31. 32.

Causation in the Law, 2nd edn, above n. 1, at 75. Ibid. at 92. See e.g. Kelley, above n. 2, at 638–9. Elsewhere the authors seem to qualify to the point of repudiation their claim for forensic usefulness Causation in the Law, 2nd edn, above n. 1, at 5 (emphasis added): “The causal language used by the courts . . . has seldom been a mere disguise for judgments of policy or expediency or judicial intuitions of what is just: though since the causal notions latent in ordinary thought . . . have aspects which are vague and indeterminate, decisions involving them outside the central area of simple cases have been powerfully and properly influenced by judicial conceptions of policy or justice.”

33. Ibid. at xxxiv. 34. Ibid. at 92: that English law treats questions involving “common notions [of causation] . . . as questions of fact . . . distinguishing them from general questions of . . . justice . . .”. Compare 72: “the distinction between cause and conditions is not a ‘scientific’ one . . .”.

151

  purposes of attributing moral responsibility, no one would say Mrs Oswald was a cause of Kennedy’s death, this would be unproblematic because this is a true empirical statement regarding the facts of how language is used in this context. But Hart and Honoré seem to claim more than this. Another possible construction of their references to the “factual” nature of the “causal” issue is merely the observation that in law questions labelled “causation”, “proximate cause” and so on are issues to be decided by the jury (where juries sit), and are therefore subject to a very high threshold for appellate review. The traditional shorthand way to describe this distribution of power between judge and jury was to say that such jury issues were “questions of fact”. Hart and Honoré certainly approved this pro-jury distribution of issues and power but, as we will see,35 it was unnecessary to express that support using this “factual” characterisation. Another possible reading of Hart and Honoré’s references to the “factual” nature of causal issues emerges from the following passage: “The traditional view . . . [was] that particular questions arising in particular cases as to whether a defendant’s act was the cause of harm are not themselves questions of policy but questions of fact, and this is not confined to questions of whether the defendant’s act was a sine qua non of the harm . . . the traditional insistence that questions of causation are questions of fact did not merely mean that so long as there were juries they should answer them . . . The traditional contention meant that the criteria for deciding such questions were not inventions of the law but were to be found outside the law in what was assumed, rightly or wrongly, to be part of the ordinary man’s stock of general notions. Hence the insistence that these questions were questions of fact often went together with the insistence that the criteria for deciding them were to be drawn from common sense”.36

If all that is meant here is that some reasons why we would say a factor is or is not a “cause” do not relate to narrow legal concerns, such as smooth court administration, but rest on wider concerns about morality, society and the economy that exist “outside the law”, it would not be controversial. But the 35. See below note 67 and accompanying text. 36. Causation in the Law, 2nd edn, above n. 1, at 91. See also at lii–liii: “. . . [causal minimalists] amalgamate the issue of the scope of the law of negligence with ordinary causal issues. The court or jury must then deal, under the rubric of remoteness or proximity, both with the appropriate scope of the law of negligence and with the question whether on this occasion defendant’s breach of duty caused the harm complained of. This amalgamation is undesirable, because the scope issues raise questions of law, and the causal issues questions of fact. The former are more appropriate for the judge to decide, the latter for the jury or trier of fact . . . [as a result of] the amalgamation of causal and noncausal issues under headings such as ‘remoteness’ and ‘proximity’ . . . causal limitations may be seen as just another set of policy limitations, which incorporate some intuitive notion of the reasonable limits of responsibility” (emphasis added).

152

  Hart and Honoré thesis is troublesome because it seems implicitly to suggest that these concerns, these “criteria . . . outside the law”, though masked by causal usage, are somehow crystallised within the patterns of use of causal language in such a way that these patterns can be used as a short-cut to the appropriate attribution of consequences to conduct, without the need to unpack the underlying “stock of general notions” on which that usage rests. Specifically, the authors’ claim that causal usage is different in cases where consequences follow conduct without the intervention of voluntary human conduct (or intervention of abnormal or coincidental events), from usage in cases where there is no such intervention.37 They construct crystallised “causal principles” from this usage and state that these principles simply concern “what happened on a given occasion”.38 The implication is that their causal principles describe a fact rather than a normative evaluation of facts, and that there is no need to unpack these principles any further. An Incomplete Topology of Attributive Judgments

The fact of intervention of a third party is certainly relevant to our enquiry if the enquiry’s purpose is the legal attribution of a consequence to a tortfeasor. But it is not clear from Hart and Honoré why it is relevant. In that sense they produced a topology, not a geology of the meaning behind causal usage in situations. Moreover, Hart and Honoré focus on only two factual details: intervening voluntary action of a third party and an intervening coincidence or abnormality. Yet other facts can be relevant to attribution and therefore possibly to causal usage, such as: • Where there is intervening wrongful conduct of another party (which is also an historic factor in relation to the outcome): its relative wrongfulness; the nature of the relationship, if any, between the defendant and the other party; and the nature of the relationship, if any, between the defendant and the victim. • The degree of wrongfulness of the defendant’s tortious conduct, for example whether the defendant was reckless or intended harm to the plaintiff. • The seriousness of any intended harm. 37. Honoré, “Herbert Lionel Adolphus Hart 1907–1992”, above n. 15, at 304: “J.L. Austin . . . urged Hart to study the principles on which in extra-legal contexts causes are picked out from mere conditions. Here voluntary action and what was abnormal in the context emerged as the leading criteria.” 38. Causation in the Law, 2nd edn, above n. 1, at 307.

153

  • Any prior dealings between the parties, especially any specific explicit or implicit undertakings by the defendant or the plaintiff. • Whether the defendant was acting in pursuit of commercial profit. • The nature of the injury (in other words, the nature of the interest of the victim affected by the defendant’s tortious conduct). • Whether the defendant had a reasonable opportunity of limiting his liability for the consequence by an agreed term. • The degree to which, at the time of the defendant’s tortious conduct, it was foreseeable to a prudent party in his/her position that the kind of risk that actually occurred would eventuate. Such factual elements in a case, and their interrelationships, throw up a diverse range of concerns, all of which may plausibly bear on the ordinary person’s judgment of whether a particular consequence should be attributable to the tortious conduct.39 By focusing on only two elements in the factual matrix, Hart and Honoré neglect the rich diversity of concerns that influence our attributive judgments which we may express in causal language. Their topology of causal language is incomplete, and reveals only a very few of the “relevant factors”.40 By separating these two elements from other relevant factual details and describing them as “causal” limits on legal responsibility in contrast to other “non-causal” limits,41 Hart and Honoré accord them a pre-eminence that seems unjustified. Suppose that D1 cuts V’s finger without third-party intervention with the consequence that V freakishly bleeds to death. Compare this to a case where D2 hands a knife to D3 and urges D3 to cut the finger of V2 who freakishly bleeds to death. Hart and Honoré’s “causal principles” focus on whether an intervention such as that of D3 was present. But what if, when we look at all the facts, we find D1 was a well-meaning and careful surgeon, compared to D2 who wanted V2 to bleed to death and had paid a large bribe to D3 to do the cutting? Why is the type of physical sequence, namely 39. And therefore any expression of this conclusion in causal terms. Though Hart and Honoré seem to acknowledge the contextual dependence of causal statements, they do so with a focus on their two factors: Causation in the Law, 2nd edn, above n. 1, at lxxvi “common-sense causal judgments . . . are highly contextual. They focus on those factors which in a particular situation appear abnormal or unusual”. 40. See quote in text at n. 29 above. 41. Hart and Honoré deciphered only 5 very broad limits. Two they called “causal”: the necessary element in a sufficient set requirement, the NESS “test” which excluded factors that were not causally relevant conditions (i.e. their version of a test of historical involvement: see Part 2 below); and the 2 types of intervention. Three they called “non-causal”: probability; scope of the rule; and equity of the parties. See Causation in the Law, 2nd edn, above n. 1, at xlvii–xlviii, 92 and 304ff. See also Howarth, n. 2 above, at 1412–19.

154

  whether the process involved the intervention of D3, worthy of greater emphasis than these other facts?

C. The Claims (i) that Ordinary Causal Usage Reveals Three Distinct Issues; and (ii) that Law should deal with Two of these under “Causation” The Analytical Distribution of Issues Promoted by Realists

Realists reacted against conservative judicial manipulation of legal rhetoric, and wanted to strip away from judges the comfort of opaque terms to force them to address the underlying normative issues driving their judgments.42 This led to a critique of the two traditional analytical categories that, combined, address the issues of whether the defendant’s tortious conduct was involved in the history of the outcome and whether, if that were so, the outcome should be judged as within the appropriate scope of liability for the tort. In the United Kingdom and Commonwealth these two categories have been known as “causation” and “remoteness of damage”. In the US they have been known by various terms, but the dominant terminology seems at present to be “cause-in-fact” and “proximate cause”.43 Realists objected to judges denying a claim, where the defendant’s tortious conduct was clearly part of the history of the outcome, on the mere assertion that it was not “a cause” of that outcome. So Realists such as Leon Green believed that the legal category of “causation” should correspond solely to the first enquiry, that of historical involvement.44 Since this was an issue of pure fact, Green’s argument was equivalent to an argument that, in law, the issue of causation should be entirely factual unlike its more diverse roles in ordinary language.45 All else would be relegated to analysis under the second category, the one I would rename the “appropriate scope of liability for the consequences of the 42. In tort and elsewhere. See Goldberg and Zipursky, “The Moral of MacPherson” U Penn Law Rev 146 (1998), 1733 and Powers, “Reputology” Cardozo Law Rev 12 (1991), 1941 at 1951–2 on the eclipse of the earlier pseudo-scientific metaphors of legal/proximate cause. 43. Stapleton, “Legal Cause: Cause-in-Fact and the Scope of Liability for Consequences”, above n. 28. 44. Hence they are known as “causal minimalists”. Compare Hart and Honoré who say that “the distinction between cause and conditions is not a ‘scientific’ one . . .” Causation in the Law, 2nd edn, above n. 1, at 72, but see the text at nn. 33 to 38 above. 45. Save for some normative rules to cover evidentiary gaps. Green was, however, unable adequately to accommodate cases of over-determination which seemed to provoke policy-driven exceptions to what was at that time, wrongly (see part 2 below), the generally accepted test of cause-in-fact, namely the traditional but-for test.

155

  tortious conduct”.46 Under this heading, the non-factual nature of the remaining enquiry should not be masked by causal language. The Realists hoped that the clarity produced by this division of issues would justify the law using a special and precise meaning for “causation”, much narrower than the diverse meanings expressed by causal language in everyday life.47 There is much wisdom in this approach, but Realists made two errors. First, they were confounded by the difficulties of over-determined events and their attempts to accommodate them within their model failed. As we will see, it was Hart, Honoré and Wright who resolved these difficulties. Secondly, Realists described the concerns that would determine the scope of liability for consequences merely as “legal policy”. This was rightly attacked by Hart and Honoré who emphasised that the reason why we would not say, for example, Mrs Oswald was a “cause” of Kennedy’s death is not adequately captured by the term “legal policy”.48 To think that the deployment of causal terms in such cases is “a mere disguise for arbitrary decision or judicial policy . . . is a blinding error”.49 This insight might have been accommodated within the Realist model by an argument that the concerns that would determine the second analytical category (the scope of liability for consequences) were more diverse than “legal policy”. But Hart and Honoré were advocates of linguistic analysis. They claimed that causation in law tracked and should track ordinary usage. In other words, they claimed that the sorts of non-policy concerns reflected in our attitude to Mrs Oswald were reflected in causal usage and that patterns of such usage had been and should be used by courts when determining issues falling under the first of the two legal analytical categories, namely “causation”.50 The Analytical Distribution of Issues Defended by Hart and Honoré

In effect Hart and Honoré recognised three distinct issues, two of which they described as “causal” and one “non-causal”. They argued that it is critical to keep each of these issues separate.51 46. Stapleton, “Legal Cause: Cause-in-Fact and the Scope of Liability for Consequences”, above n. 28. 47. Howarth, n. 2 above, at 1404 raises the question why this difference between “causation in law” and “causation” in non-legal contexts would be objectionable and concludes that Hart and Honoré provide no explanation. See also n. 60 below. 48. Causation in the Law, 2nd edn, above n. 1, at lxxviii: “the non-arbitrary character of the distinction between a mere condition sine qua non and a cause . . .”. See also the quote at n. 36 above. 49. Causation in the Law, 2nd edn, above n. 1, at 3. 50. Ibid., at 305. 51. Ibid., at xlvii; cf. at 305 (“both the administration and the understanding of the law will be served by pressing this distinction rather then by their fusion in an omnibus rubric of

156

  1. “Causally relevant condition”: Hart and Honoré noted that before a defendant’s tortious conduct could be shown to satisfy the requirement, “causation in law”, the claimant had to show that that conduct had been a “causally relevant condition” of the consequence.52 This is a question of fact and is the one that I describe as “a factor being involved in the history of the outcome”. Hart and Honoré describe this factual question as one of two “causal questions”.53 Their suggested approach to this issue in terms of a “necessary element in sufficient set” (“NESS”) constituted a major breakthrough in understanding, as Richard Wright’s work makes clear.54 2. “Causal Connection”: To satisfy any legal requirement of “causation” Hart and Honoré said the plaintiff also had to satisfy another “causal” requirement: namely, that there was what they called an appropriate “causal connection” between that conduct and the outcome. It is here that the patterns of causal usage of ordinary people came in. For example, gravity and Mrs Oswald were “causally relevant conditions” of Kennedy’s death. They clearly played a role in the full history of his death. However, Hart and Honoré argued that ordinary people (at least if their purpose was explaining the outcome,55 identifying the outcome with an actor, or attributing individual responsibility for it) would not say that either of these factors bore any “causal connection” to the death. They would not say that gravity was a “cause” or that Mrs Oswald was a “cause”. Hart and

52.

53. 54. 55.

‘proximate cause’ ”) and at 306–7. Oddly this often seems to run counter to their descriptive claims, because they seem to accept that their notions of causal connections are relevant to US issues of proximate cause. Ibid., at lvi, lxiv. There are some odd passages where the proof of “causal connection” is said to hinge on factors that seem more accurately described as ones on which proof of “causally relevant condition” hinges. See e.g. the discussion (at liii–liv) of market share liability, a doctrine that allows liability to be imposed when it has not been shown that the defendant’s tortious conduct played any role in the history of the plaintiff’s injury. See also the point (at lviii) that verifying the existence of “causal connection” involves counterfactual speculation, whereas this is the technique of establishing “causally relevant condition”. See also Honoré, “Causation and Disclosure of Medical Risks” Law Quarterly Review 114 (1998) 52 at 53–4: “in tort law a causal issue is . . . an issue of fact to be proved by the plaintiff . . . sometimes, however, for reasons of policy the burden of proving causal connection is shifted . . . an example is Cook v. Lewis [1951] 1 S.C.R. 830 . . .”. What was shifted in Cook’s case was the burden of proof in relation to whether the defendant had played any role in the history of the injury at all, that is whether it was what Hart and Honoré call a “causally relevant condition”. Causation in the Law, 2nd edn, above n. 1, at 110. See lviii: “we do not treat every ‘causein-fact’ (causally relevant condition) as a cause of the harm in the full sense . . .”. On the difference between “causal” and “non-causal” notions, see n. 41 above. See Wright’s works cited in n. 2. The authors would say that, in explanatory contexts, this is because neither gravity nor Mrs Oswald is an intervention in the world that explains why the event occurred: Causation in the Law, 2nd edn, above n. 1, at 33.

157

  Honoré correctly noted that this usage was not driven by policy concerns. For this reason they also argued, unconvincingly in my view, that it was more appropriately separated from such issues. However, Hart and Honoré did not resolve the ambiguity noted earlier:56 was the task of establishing that a causally relevant condition bore a suitable “causal connection” to the harm that of establishing a fact; or was “causal connection” a conclusion arising from normative “ideas of when it’s just or fair to punish or exact compensation” based on the facts of the case.57 3. Consequence is “proximate”/“not too remote”: According to Hart and Honoré, once causal connection was established, the claimant then moved on to the second analytical requirement, namely showing that the outcome was within the appropriate scope of liability for consequences of the tortious conduct. Here Hart and Honoré seem sanguinely to have accepted the Realists’ assertion that this issue was and should be governed solely by policy concerns.58 Why the Realists’ Approach is Preferable

The crux issue is whether we agree with the implicit argument of Hart and Honoré that we should pack into the legal category of “causation”, rather than the remoteness or proximate cause category, the sort of concerns that drive causal usage in explanatory and attributive contexts: that 1 and 2 belong together. Certainly this seems traditionally to have been a popular division of issues in Commonwealth jurisdictions. Alternatively we might agree with the Realists that the legal category of “causation” should be reserved for the factual enquiry of whether the defendant’s tortious conduct played a role in the history of the outcome: that 2 and 3 belong together. The danger for the traditional way of dividing issues between legal categories, is that by packing normative concerns into the “causation” category they may go unenunciated. This danger is compounded by Hart and Honoré’s argument that “causation in law” does and should track “ordinary /common sense” causal usage. The problem here is that we sometimes use the same causal language to express the fact of historical involvement59 and 56. See text at nn. 33 to 38. 57. Causation in the Law, 2nd edn, above n. 1, at 1. See also “the distinction between cause and conditions is not a ‘scientific’ one . . .” at 72. 58. Ibid., at 305–7. 59. This is what Hart and Honoré seem to term the “recipe” function of “causation”: see Causation in the Law, 2nd edn, above n. 1, at 31. See also n. 21 above.

158

  sometimes to express a judgment about the relevance of an historical factor to the purpose of our overall project, which purpose might be explanatory, identificatory, attributive and so on. The authors themselves do this too: for example asserting that “a cause is basically like an element in a recipe” while at the same time asserting that “a cause is . . . an intervention in the existing or expected course of events . . .”.60 Yet, though Hart and Honoré did not clearly and consistently61 acknowledge it, these are two quite distinct types of causal usage, expressing responses to the two quite distinct enquiries. Since in “ordinary” use causal language can mean such different things, “common sense causation” is not a single elemental concept. Take the role of gravity in a car accident. It qualifies as a “cause” under the recipe definition but does not qualify as a “cause” under the “intervention” definition. Not only are there two distinct underlying enquiries, one historical and one purposive, but the disputes to which they give rise are of a fundamentally different nature. If under the rubric of “causation in law” courts decide not only issues of historical involvement but also attributive questions, there is a danger that the factual nature of the former will be used by counsel to suggest that the attributive question is one of fact that requires no supporting normative argument, a type of “you’ll know it when you see it”62 assertion. Unwittingly Hart and Honoré’s book, published in 1959, seems to have encouraged this evasion of the duty to provide reasons in many Commonwealth cases.63 Certainly it has 60. Honoré, “Medical Non-Disclosure, Causation and Risk: Chappel v. Hart” Torts Law J 7(1999) 1 at 5. See also Causation in the Law, 2nd edn, above n. 1, at 29 (a cause as a difference from the normal course) and 31 (a cause as an element in a recipe). Honoré sometimes refers to a triple function for the concept of “causation”: “causation is used mainly for three purposes: to provide recipes, to explain events and to allocate responsibility”: Responsibility and Fault, see n. 8, at 108 and 120. In my view, it is impossible for a unitary concept of “causation” to straddle coherently these 3 very different types of enquiry. In the non-causal terminology I promote in Part 2, these three enquiries are respectively: the factual issue of historical involvement in a transition; the purposive enquiry where the purpose is explanation; and the purposive enquiry where the purpose is attribution of responsibility. 61. See Causation in the Law, 2nd edn, above n. 1, at 24 where there is a distinction drawn between explanatory contexts and attributive contexts. With respect, however, the fundamental distinction is that between the factual enquiry of historical involvement and the non-factual purposive enquiry which selects as relevant to that purpose only certain historically involved factors. The purpose might be explanatory, morally attributive, legally attributive, regulatory etc. 62. This was encouraged by statements such as: the “principle, inherent in the rule of law, that legal responsibility should depend on criteria which ordinary people can apply, even if they cannot explain the principles on which they do so” (Honoré, “Herbert Lionel Adolphus Hart 1907–1992”, above n. 15, at 304, emphasis added). Compare Causation in the Law, 2nd edn, above n. 1, at 26, to the effect that common sense “causal notions [rest] on stateable principles; though the ordinary man who uses them may not, without assistance, be able to make them explicit”. 63. In contrast to the situation in the US where the book has only occasionally been cited by courts, see review by Henry Gabriel, n. 2 above, at 464.

159

  not been unusual for courts to cite the book for the proposition that causation in law tracks the “common sense” meaning of causation used by ordinary people. But, as Lord Hoffmann notes,64 such courts have often done so merely as a substitute for the careful enunciation of relevant concerns governing the responsibility dispute that had arisen from the agreed facts. In short, it is the traditional mixing of factual and evaluative issues under the legal issue of “causation” that is at the root of doctrinal confusion and unarticulated judgments in the area in Commonwealth jurisdictions. Respect for Citizens/Transparency of Law/Rule of Law

Hart and Honoré respected jury determinations and embraced the broader “principle, inherent in the rule of law, that legal responsibility should depend on criteria which ordinary people can apply”.65 A theory which held that the legal concept of causation was something equivalent to what the jury person would have used outside the law was attractive because it supported the claim that a jury could be trusted to do the right thing guided by undeciphered code words of “causal connection”. By dividing the issues remaining after historical involvement has been established, into “causal issues” and “non-causal, policy issues”, and allocating the “causal issues” to the analytical category called “causation”, Hart and Honoré could more easily characterize them as “questions of fact” which were appropriate for the jury to decide.66 But the relevance of, say, third party intervention is a normative issue not a dispute about what the facts were. Moreover, it was not necessary to use this confusing characterisation in order to reserve such issues for the jury: it is well accepted that juries decide purely normative issues, as every case that goes to the jury where the facts are agreed attests.67 64. Lord Hoffmann, “Common Sense and Causing Loss”, n. 27 above, noting that the numerous judicial citations of “common sense” in relation to “causation” often “conceal, or perhaps reveal, a complete absence of any form of reasoning . . . in the best tradition of English antiintellectualism”. 65. Honoré, “Herbert Lionel Adolphus Hart 1907–1992”, above n. 15, 295 at 304. 66. Which they “can handle with a minimum of guidance”: Causation in the Law, 2nd edn, above n. 1, at lv. See also 306–7 arguing for a division of issues: “The genuinely causal limitations present issues which, however vague the criteria for their solution, are to be solved by the use of common-sense notions or analogies to them and are eminently suitable for submission to a properly instructed jury.” 67. In the US the jury decides both cause-in-fact and proximate cause. There are very few “rules of law” governing proximate cause (that is, issues decided by the court concerning the appropriate scope of liability for the consequences of tortious conduct). Most are “questions of fact” which is mere code for something the jury is allowed to decide: Stapleton, “Legal Cause: Cause-in-Fact and the Scope of Liability for Consequences”, above n. 28. Compare Causation in the Law, 2nd edn, above n. 1, at xxxiv.

160

  Ironically, one effect of the cryptic characterization of disputes such as those about the normative relevance of third-party intervention as “causal” disputes, is that it can itself distance the world of law from the understanding of ordinary people by adding a mysterious layer of indeterminate terminology.

D. The Claim that there is a Constellation of “Causal Connections”

Hart and Honoré claim that the way “ordinary people” use causal language reveals various types of “causal connection”.68 The central type is said to be that of the initiation of physical sequences, the physical manipulation or intervention in the world that brings about consequences without voluntary interventions of others or abnormal contingencies (“causing harm” which I will call “causal connection type 1”).69 Around this is said to cluster a group of concepts, which in a broad sense Hart and Honoré term “causal” concepts,70 such as providing reasons for people to act (“inducing wrongful acts”, which I will call “causal connection type 2”) or providing opportunities for people to act or occasions for events to occur (“occasioning harm”, which I will call “causal connection type 3”). Hart and Honoré claim that causal connection type 1 is not traced through an intervening voluntary action by a third party71 or abnormal contingencies.72 In contrast it is said that causal connection type 2 may be traced through intervening voluntary action, and that causal connection type 3 may be traced through intervening voluntary action and intervening coincidence.73 “Causal Connections”: Phenomena Reflected in Ordinary Usage; or “Convenient” Analytical Constructs?

In deducing these “causal connections”, it is not clear what causal usage Hart and Honoré are relying upon. Although the fundamental project is to study the language patterns of ordinary people, at a critical stage they argue that: 68. Causation in the Law, 2nd edn, above n. 1, at xxxiii. 69. Also described as “the central type of causal inquiry”: Causation in the Law, 2nd edn, above n. 1, at lxxx, lviii, lxxviii, 28–9, 43, 71. 70. Also described as “causal notions” (ibid., xxxiii), “types of causal relationship . . .” (xxxiv) and “quasi-causal bases of responsibility” (lxxx). 71. Such as a deliberate act: Causation in the Law, 2nd edn, above n. 1, at 43. 72. Honoré has described such actions and events as being used to “distinguish causes from mere causally relevant conditions”, suggesting that it is only causal connection type 1 that corresponds to a “cause”: Honoré, Responsibility and Fault, above n. 8, 102 (emphasis added). Compare Causation in the Law, 2nd edn, above n. 1, at 135. 73. Causation in the Law, 2nd edn, above n. 1, at 186. For a fine attempt at summarizing Hart and Honoré’s theory of when “causal connection” is defeated see Lipton, above n. 2, at 132–3.

161

  “it is convenient to refer to cases [where the connection between the relevant conduct and the harmful consequence does not depend on an intervention] as “causing harm”, and to refer to cases where harm is the consequence of one person providing another with reasons or opportunities for doing harm as cases of “inducing” or “occasioning” harmful acts”.74

But say some ordinary people would use the term “cause” to express their attribution of a consequence to the tortfeasor even where such an intervention had occurred? For example, say a car driver carelessly knocks over a pedestrian, breaking her right leg, and at hospital a doctor wrongly amputates the left leg. The legal question of whether the driver is liable for the injury to both legs or only the right one is a question about the appropriate scope of liability for consequences of the tort that has been committed. An ordinary person, expressing her view that the driver should be liable for the injury to the left leg might say the driver was a “cause” of it. Another might, in line with what the authors find convenient, say he merely “occasioned” it.75 It seems that the reason the authors find it “convenient” to ignore ordinary usage here is because it can be so diverse. How does this square with their claims that linguistic analysis of ordinary speech has something to offer us in addressing legal disputes about responsibility such as this one about liability for the amputation of the left leg? If Different “Causal Connections” are just Different Fact Patterns, What is their Analytical Utility?

Other scholars have analysed difficulties in Hart and Honoré’s descriptions of “causal connections”76—what counts as an intervention? By what criteria are 74. Causation in the Law, 2nd edn, above n. 1, at 71, emphasis added. See also at 2: that this is where the “language of ‘cause and effect’ sits most happily”. 75. In places even the authors seem to use the terms interchangeably: Causation in the Law , 2nd edn, above n. 1, at 195 reads “The ‘causal connection’ between a defendant’s act and the harm may be succinctly described by saying that he has ‘occasioned it’ The use of this notion in the law is an extension of the general idea, common in non-legal thought, that the neglect of a precaution ordinarily taken against harm is the cause of that harm when it comes about”. (emphasis added).

At 307 the authors state that the causal issue (including all 3 types of causal connection that they identify including mere “occasioning”) can be separated from the non-causal issues limiting the scope of liability for consequences by the question: “Was the harm to A caused by B’s act?”. Elsewhere (lxxviii–lxxix) they state that if A hits someone so that he lies unconscious and B takes advantage of the opportunity to stab B, “A does not cause B’s action or its outcome; though it is, of course, a sine qua non of that outcome”. See also Howarth, above n. 2, at 1402. 76. See for example Kelley, above n. 2, at 638–9 who convincingly argues that Hart and Honoré’s 2 “intervention” bases for distinguishing “cause” from mere condition both depend on noncausal evaluative notions. See also Howarth, above n. 2, at 1399, 1403, 1414–15 and the important study by Lipton, above n. 2.

162

  voluntariness, abnormality and coincidence judged? A more central problem emerges from the main “principles” Hart and Honoré offer with regard to these causal connections. For example, causal connection type 1 is described as one where the defendant did something that was sufficient to produce the consequence without the intervention of other persons or of contingencies.77 Yet the major “principle” about causal connection type 1 that we are offered from the review of causal usage seems to be just the corollary of this definition: that it is “negatived” by the intervention of third parties or of contingencies. The three types of causal connections seem to reduce simply to a way to redescribe three types of factual sequence. So how does the terminology of “causal connections” help us? It is true78 that the phenomenon of third-factor intervention may provide a moral or policy reason for not attributing responsibility for the later consequence to the tortious conduct. Other things being equal (i.e. ignored), we do tend to allocate more responsibility to the stabber than the person who gave the stabber a reason to stab. But there is no mystery here: it is not only possible to expound why such cases are differentiated and do it in evaluative non-causal language, but it is also essential for clarity of legal reasoning to unpack these reasons.79 For example, those explanations may well be in deterrence terms: for a person has more control over whether he stabs than over whether a person he gave a reason/opportunity to stab will do so.80 An explanation may well also be expressed in moral terms: that there is often more blame to be attached in respect of a foreseeable consequence that flowed without the intervention of others or contingencies. Hart and Honoré say that there is a “causal problem . . . raised by the presence of the [intervening] third factor: the law must decide whether or not the third factor negatives causal connection” (emphasis added).81 It is not clear what these causal terms add. They may also distort the enquiry. At the end of the day the importance of the fact of intervention can only be judged in the light of all the facts of the case and the concerns that they raise. For example: we allocate less blame to the person who innocently gave a third person a reason to stab a victim than to one who gave it maliciously; we allocate less blame to a careless motorist when his victim is treated and injured further by a drunken doctor than by an inadvertent one; and we allocate less blame for failing to intervene in a nightclub brawl to a customer than to the club’s security personnel because there 77. 78. 79. 80. 81.

Causation in the Law, 2nd edn, above n. 1, at lxxx; 28–9; 43; 71. As we saw earlier, see text accompanying nn. 40–1. See n. 86 below. Causation in the Law, 2nd edn, above n. 1, at lxxix–lxxx. Ibid. at 134.

163

  was no implicit undertaking by the customer to restrain such conduct. Hart and Honoré’s focus on the factual feature of intervention and their characterisation of this part of the normative question as about whether “causal connection” was negatived, deflects attention from the explication of all the normative considerations that are in play. Only by looking at the whole context, without making intervention the central focus, will we appreciate why it is that the law draws many of its important distinctions. Examples here include: why injury from lightning may be within the scope of liability for a golf course but not a public park; why commercial hosts in the US may be liable for the injuries their drunken patrons inflict on others but why social hosts are not liable for such injuries; and why fraudulent defendants may be liable for the consequences of an unrelated and coincidental fraud of a third party.82 The Cultural and Temporal Dependence of Judgments Concerning the Relevance of Intervention

That re-describing cases in terms of different types of causal connection is both unhelpful and can obscure the normative nature of the task at hand can be illustrated by cases where the facts are the same but the relevance to liability of an intervention is judged differently. Take the example of a customer of a store who is raped in the store car park while the parking lot attendant was sleeping. Assume that had the attendant been alert the crime would have been prevented; so the sleeping of the attendant undoubtedly played a role in the history of the rape. In 1950 such facts may well have supported findings that the store owed the customer a duty of care and that the sleeping attendant had breached that duty. But it would have been highly unlikely, at that time, for liability for that breach to have been judged to extend to the relevant consequence, rape by a third party. It is irrelevant that this might have been expressed by one citizen in terms that the negligence was not a “cause” of the rape and by another in terms that, even though it was a “cause”, the attendant should not be legally “to blame” for the rape. Indeed, a focus on any such different causal usages will fail to reveal the common responsibility judgment of these citizens and deflect attention from the task of explicating the reasons for that judgment. Today, in certain jurisdictions, no liability would still be the forensic result. But in certain States in the US it is likely that the rape would be judged to fall within the appropriate scope of liability for consequences for 82. See Stapleton, “Legal Cause: Cause-in-Fact and the Scope of Liability for Consequences”, above n. 28; see also below n. 105.

164

  the tortious conduct. This shift in result over time poses problems for Hart and Honoré because their introduction of the notion of different “causal connections” produces a focus on the unchanged factual feature of intervention (here the conduct of the car-park rapist), at the expense of other facts in relation to which moral and policy attitudes may have shifted over time such as the pursuit of profit by the defendant-store. Perhaps Hart and Honoré would explain such shifts in terms of the law choosing to change what they describe as the “grounds of responsibility”83 from a set including a requirement of causal connection type 1 to a set including a requirement broad enough to be satisfied by type 3 (providing opportunity for the rapist). They might then explain that shift in grounds of responsibility in terms of shifts in “common ideas of when it’s just or fair to punish or exact compensation”.84 But this seems a roundabout way of expressing the underlying phenomenon: namely the application, to a case of intervention, of values that are temporally and culturally dependent. The loop of reasoning through “causal connection” seems unnecessary and distracting. Relevance of Intervention: Duty

It is important to emphasise that this criticism of the “convenience” of the “causal connection” device applies equally if the legal dispute concerning a case of intervention is framed, as it very often is today in Commonwealth negligence claims,85 as one about the recognition of the duty of care rather than as one concerning the appropriate scope of liability for consequences of tortious conduct. Professor Honoré has conceded that the issue of whether the defendant owed a duty to control the intervening third party is a question of law, and states that it is this duty that provides the reason why the intervention does not negative causal connection.86 But does this not mean we have come full circle? 83. Causation in the Law, 2nd edn, above n. 1, at xliii–xlv; liv. Compare (a) the statement (at lxvii) that even after the Privy Council decision in Overseas Tankship UK Ltd v. Morts Dock & Engineering Co Ltd [1961] AC 388 (Wagon Mound I ), the causal connection required by negligence was the one that “requires that there should not intervene between the negligent conduct and the harm the sort of factor . . . which counts as a ‘superseding cause’ ”, in other words causal connection type 1; with (b) statements (at 64, 133 and xliv–xlv) that the “causal connection” required for negligence encompassed not only causal connection type 1 but also that of occasioning harm. 84. Causation in the Law, 2nd edn, above n. 1, at 1. 85. Compare the US, see Stapleton, “Legal Cause: Cause-in-Fact and the Scope of Liability for Consequences” above n. 28, at footnote 30. 86. Honoré, “Suicide in Police Custody: Reeves v. Commissioner of Police of the Metropolis” Torts Law J 8 (2000), 1 at 3. It is a legal normative question whether D has a duty to control the third party and not provide him with an opportunity to stab V. It is true that courts require more explicit reasons to impose such an affirmative duty than they require before they impose the

165

  Do we not find that whether “causal connection” is or is not negatived turns on whether a duty was owed to restrain the intervention, which in turn we know is determined by a range of normative concerns?87 Why cloud this nature of the adjudicative task, one about the appropriate weight of such normative concerns, with a loop of legal reasoning involving the artificial construct of “causal connection” with its misleadingly factual flavour? Today plaintiffs are increasingly making negligence allegations, such as the rape in the car park example, to the effect that the defendant had carelessly “provided opportunities” for third party intervention. So it is important that lawyers unpack the normative concerns influencing judgments about when merely providing an opportunity for a third party to intervene brings that consequence within the appropriate scope of liability. This can only be done in the light of all the factual features of the case, the specific policy and moral concerns that they raise and the particular weight accorded them by the factfinder at the particular date and in the particular culture. Hart and Honoré’s insertion of the question of which causal connection is required by the “grounds of responsibility” obscures this task. P 2: U C   T U E

I argued in Part 1 that the source of doctrinal instability in relation to “causation” in law is the widespread assumption that there is some single elemental “concept of causation”, and the failure to give adequate weight to the fact that causal language is used to express answers to two separate questions. We need to abandon the notion that “causation” is a fundamental, elemental concept with a single meaning. The way forward is to cut through the ambiguity of causal terms and to map in completely non-causal language the terrain over which they wander so confusingly. Part 1 noted that this terrain comprises the two principal and distinct enquiries that can be made about a transition to an outcome and that help us navigate the world: historical involvement and relevance to purpose. The argument was then made that clarity in legal duty on D not to stab V directly. But the difference lies in the fact that there is such a wide normative consensus about the reasons for the latter duty that these reasons tend to go unstated: but they do exist, see text at n. 79 above. Normatively it is more controversial to impose affirmative duties so we tend to demand explicit reasons. To characterise the difference here as one of different “causal connection” requirements deflects attention from the need to explicate the reasons for the difference in our normative reactions. 87. On “unpacking” the concerns underlying the recognition of duty see Stapleton, “Duty of Care Factors: a Selection from the Judicial Menus” in Cane and Stapleton (eds), The Law of Obligations (Oxford: Clarendon Press, 1999), 59.

166

  reasoning would be enhanced if legal analysis keep these enquiries separate. In particular, the ambiguity of causal terms should be eliminated by the law confining “causation in law” to the factual issue of historical involvement. In Part 2, I examine more closely these two separate enquiries underlying causal language. I argue that the “necessary element in a sufficient set” (“NESS”) notion formulated by Hart and Honoré as the test of “causally relevant condition” and popularised by Wright as the test of “causation”, is in fact only relevant to the historical involvement enquiry. In relation to that enquiry it is indeed the correct approach. Finally, I suggest that the NESS approach should be crystallised in a “targeted but-for” test which allows us to search for the perspective from which the targeted factor can be seen to have “made a difference” in the history of the outcome. I argue that Wright and Honoré failed to apply the NESS approach correctly to the desert traveller case and I use this to illustrate the importance of two aspects of the targeted but-for test. First, it is critical that we apply that test separately to every element constituted or introduced by the targeted factor. Secondly, we must apply the NESS approach separately to all candidate factors and resist the temptation to treat all other factors as part of a “set-stage”: what I call, the single “snap-shot” view of history. Unless we consider separately each effect of a targeted factor and target every possible factor, the continuous interrelated nature of historical processes will not be revealed and we will fail to identify all the factors that played a role in the history of an outcome. A. The Two Underlying Enquiries 88

The first enquiry is that of “historical involvement”. Did the factor in question play a role in the history of the transition to the relevant outcome? This is a factual question (with an infinite list of factors that satisfy it),89 though there are choices about how we define the factor and the outcome.90 Later I define this fundamental concept in more detail.91 88. The sketch that follows builds on Stapleton: “Perspectives on Causation”, above n. 10, 61–84; and “Legal Cause: Cause-in-Fact and the Scope of Liability for Consequences”, above n. 28. 89. This is particularly obvious given that we can define the factor as an omission . . . save for every person in Dallas failing to kill Lee Harvey Oswald the day before Kennedy’s death, the President would not have died. The infinity of factors that might have happened, but did not, forms the backdrop against which the processes of history are identified. 90. I leave aside these two variables here. In law both are determined by legal rules such as who owes a duty of care, what conduct is required of that duty, what losses are actionable types under the relevant cause of action and so on. See Honoré, Responsibility and Fault, above n. 8, 100–6; Stapleton, “The Normal Expectancies Measure in Tort Damages” Law Quarterly Review 113 (1997), 257. 91. See text at nn. 109 to 110 below.

167

  The second enquiry addresses the relevance of the role that the targeted factor played in the history of the outcome to the purpose of the overall project.92 A crude analogy is how, viewed through a camera, which features stand out from the background depends on the focus setting we choose. The purpose of our project provides the focus of the second enquiry. Clearly the focus will be different according to whether our purpose is merely explanatory,93 or we are trying to allocate moral responsibility, or judging the scope of liability for consequences of tortious conduct, or trying to find the most economically efficient way of regulating an activity and distributing resources. For example if the purpose of the project is the allocation of liability in a particular tort and the defendant’s tortious conduct satisfies the first enquiry (that is, it was historically involved in the history of the relevant outcome), the second enquiry is whether that outcome should be within the scope of liability, given the reasons why the law has recognised that cause of action. The nature and shape of the obligation determines the shape of the first historical enquiry because it defines the targeted factor: whether the defendant’s particular tortious conduct was part of the history of the victim’s actionable loss. Clearly it also determines the shape of the second purposive enquiry about the allocation of legal responsibility for the consequences of tortious conduct. Disputes can arise in relation to how the law should structure both enquiries. In relation to historical involvement these disputes relate to how the law should handle evidentiary gaps concerning past facts and evidentiary gaps concerning the hypotheticals used when answering the question of historical involvement. In the second enquiry there may be disputes about whether the law should require the historic factor to be necessary for the outcome in the particular circumstances,94 as well as other types of dispute about the purposive relevance of various historic factors. Appellate courts, called upon to address such legal disputes, should emphasise that they are not disputes about “fact” in any meaningful sense and insist that the parties argue 92. For example, in the history of most terrestrial processes gravity plays a role. Yet it is rarely of purposive relevance. In the law of obligations the purpose is to allocate legal responsibility for certain conduct and outcomes. How important we judge a factor is to that purpose will depend on our individual view about the appropriate incidence, content and extent of legal responsibility in the context. This view is both temporally dependent (see the rape in the car-park example) and culturally moulded. The latter is illustrated by cultures where if S saves life of V, even if coincidentally, this outcome is not only “identified” with S, but also S is regarded thereafter as “responsible” for the fate of V. 93. Even experimental scientists set themselves purposes that provide such a focus as Lipton’s example of oxygen shows (see n. 2 above at 130–1). See also Lipton’s distinction between interest relativity and situation relativity. 94. See Stapleton, “Legal Cause: Cause-in-Fact and the Scope of Liability for Consequences”, above n. 28.

168

  directly and in non-causal terms about their competing normative perspectives on the dispute. B. Risk, Probability, Coincidence and Foreseeability

Notions of risk, probability, coincidence and foreseeability illustrate how illuminating it is to discard the language of causation and unpack the substantive content hidden by its shifting use into the two distinct underlying enquiries, historical involvement and relevance to purpose. Take the following cases: • D1 puts her baby on the grass to play, then a meteorite falls on the child killing it; • D2 unwittingly puts his baby on the grass to play next to a concealed vessel of sodium (an element that explodes on contact with water), then it starts to rain and the sodium explodes killing the child; • D3 carelessly puts his baby on the grass to play next to a vessel of sodium knowing the contents and dangers of the vessel, then it starts to rain and the sodium explodes killing the child. Hart and Honoré deal with such cases in terms of what it takes to negative “causal connection”, invoking notions such as abnormality, coincidence and whether the defendant’s conduct did or did not “substantially increase the risk” of the injury.95 So long as it did not increase that risk, they state, the tortious conduct is not recognized as a cause of the accident. But compare the approach I advocated in Part 1,96 one where the two analytical legal categories of “causation” and the “scope of liability for consequences” are directly aligned with the two underlying enquiries that can be made about a transition in the world: historical involvement and relevance to purpose. We then see that in all the above baby cases the placing of the child on the grass was part of the history of its death and that notions of risk, probability, coincidence and foreseeability are completely irrelevant to answering this first, historical involvement, enquiry. With hindsight we know that that conduct did not merely pose a risk of that outcome but completed the conditions that brought it about. Let me spell this out. 95. Causation in the Law, 2nd edn, above n. 1, at 168–70 and xxxviii–xxxix. See also Honoré’s use of the idea of “increased risk”: “Medical Non-Disclosure, Causation and Risk: Chappel v. Hart”, above n. 60; “Suicide in Police Custody: Reeves v. Commissioner of Police of the Metropolis”, above n. 86. 96. An approach that makes comprehensible the diverse packaging of the “no liability” response of the law here, sometimes being couched in terms of “no-breach”, sometimes as “no-cause” and sometimes as the outcome being “too-remote”.

169

  “Risk” is a term that tends to be used ambiguously, sometimes to mean only foreseeable risk, and sometimes to mean probability of an outcome even if at the time this outcome was not foreseeable.97 The term is best avoided unless the sense in which it is used is made explicit. In contrast, the notion of “probability” is illuminating. By “probability” I mean the general likelihood of an outcome judged in the light of all prior facts, but only prior facts.98 For example, one effect of the failure of ancient Egyptians to wash hands before operating on a patient was to increase the probability of infection, though this effect was not foreseeable at that time. To put this in terms of the first baby case: an all-seeing scientist, knowing all about the physical processes before the moment when D acted, would be able to say whether D’s conduct had increased the prospect of death by meteorite, the “probability” of that fate being judged at that time and assuming that previous processes went on at the same rate as before.99 No doubt the answer would be that D1 had not increased the probability. (Similarly, to use a well-known example, speeding would not be judged to increase the risk of a passenger being injured by a tree falling on the car just at the moment it speeds by).100 One way we sometimes state this is to say that the later conjunction of such historically involved factors is a coincidence.101 But notions of probability 97. Compare Honoré, “Suicide in Police Custody: Reeves v. Commissioner of Police of the Metropolis”, above n. 86, at 3 “the increase in risk has of course to be judged in advance, not with the aid of hindsight”. Elsewhere I have tried to capture the notion of probability with the term “expectancy/expectancies”: “The Normal Expectancies Measure in Tort Damages”, above n. 90. 98. In judging “probability” we do not take into account all facts known at trial. (If we did, then, except in cases of over-determination, each factor that played a role in the history of an outcome would have increased its probability from zero to one!) In contrast, in judging whether a piece of conduct “played a role in the history of an outcome” we do take into account all facts known at trial. 99. For example, gravity did not change; rain occurred at a general rate, etc. Lipton, above n. 2, at 137, puts this point somewhat differently. This notion of probability is independent of whether the risk was foreseeable at the time. 100. Contrast Causation in the Law, 2nd edn, above n. 1, 165 and 168. The same applies where the relevant risk is quite foreseeable, but the outcome is a coincidence in this sense, see: the tripping on the pavement example given by Mahoney JA in Alexander v. Cambridge Credit Corp Ltd (1987) 9 NSWLR 310 at 333; the handing over of a loaded gun to a child who drops it on its her foot; Chappel v. Hart (1998) 156 ALR 517, [1998] HCA 55. See Stapleton: “Legal Cause: Cause-in-Fact and the Scope of Liability for Consequences”, above n. 28; “Perspectives on Causation”, above n. 10. 101. Though we tend to say this only where the background probability is very low, the point applies whenever the conjunction of factors was no more probable to occur at the relevant time and place than it would have been at a later time, including cases where the probability was significant: Causation in the Law, 2nd edn, above n. 1, xxxix and 78. Good examples of such significant probability include: that one will trip on a pavement; or that a child will drop an object on its foot (see foregoing footnote). Contrast Lipton, above n. 2, who argues that central to the distinction between causes and mere conditions is the issue of whether the factor increases the likelihood of the effect, his “likely difference condition”.

170

  and coincidence play no role in the first enquiry of playing a role in the history of an outcome: freakish, coincidental things do happen. That a factor did not increase the probability of an outcome does not prevent it playing a role in the history of that outcome. In relation to D2 (and D3), the scientist would say that D2 had increased the probability of death by sodium explosion because in the light of prior experience we know sodium explodes on contact with water and that rain is not unusual. But again, the fact that D2 and D3 increased the probability of the outcome is irrelevant to the issue of playing a role in history. Finally, the scientist would say that, though D1 and D2 had not foreseeably increased the probability of death by sodium explosion, D3 had. Again, none of this is relevant to the issue of playing a role in history.102 If we continue to ignore causal language and move onto the second, purposive, enquiry we find, by contrast, that notions of probability, coincidence and foreseeability can often be relevant to our purpose, and we can then begin to expose the reasons why. For example, say the purpose of that second enquiry was merely explanatory. Then phenomena such as coincidence, or that a factor increased the probability of the outcome, may well be relevant to explanation in the context of an implied frame of reference.103 Next, say the purpose of the second enquiry was attribution of legal responsibility in the tort of negligence. Then, because the conjunction of the conduct of D1 and the fall of the meteorite was a coincidence, and because D2 only unforeseeably increased the probability of the outcome that occurred, the conduct of D1 and D2 are treated as irrelevant to the purposes of this tort, and liability is not imposed. This result might be expressed in terms of “no-duty” or “no-breach” or the outcome might be said to be “too remote” (or not a “proximate cause”). But, however the no-liability result is formulated we can “unpack” the reasons on which it is based: namely, that, given the features of the case, a reasonable person would not have taken any more steps to avoid the meteorite/sodium than D1 and D2 took, which here was none. The consequence would have happened even if the defendant had acted reasonably, so the tortious conduct of the defendant was not relevant 102. Moreover, as these examples show, to play a role in such history a factor need not affect the existence, probability or foreseeability of another historically relevant factor. So, for example, it is irrelevant that the speeding did not increase the risk of lightning hitting the tree. It is true, however, that sometimes one way a factor becomes historically involved is by increasing the probability of such a factor such as a rapist being encouraged to strike by the fact that he sees that the car-park attendant is asleep. 103. In contrast, in the usual frame of reference, gravity, though it plays a role in the history of virtually every process we observe with our unaided senses, is rarely relevant when our purpose is explanatory.

171

  to the purpose of the cause of action (sanctioning of certain unreasonable conduct). The choice of whether this is expressed in terms of “no-breach” or as a finding that the outcome was outside the “appropriate scope of liability for consequences” depends on other factors such as whether the defendant, in placing the child on the grass, was carelessly running another risk, for example the risk of the child being mauled by a vicious dog that was present.104 But simply saying there was “no causal connection” masks the underlying normative reasons for the no-liability result which can and should be enunciated. In the third case the careless conduct foreseeably increased the probability of death by sodium explosion, so it could well be relevant to the second, attributive, enquiry in negligence, and so D3 might be held liable. Finally, “double fraud” cases illustrate well that it is the purpose of the second enquiry that determines the attitude taken to factors such as coincidence. Say the purpose of the second enquiry is the attribution of responsibility in the tort of deceit. Take the case where, due to the defendant fraudulently misrepresenting their value, the plaintiff pays more for company shares than they were worth (the first loss). Later an unassociated internal fraud within the company is exposed and as a result, the value of the shares falls further (the second loss). Assume that, but for the defendant’s fraud, the sum would have been invested in such a way that the second fraud would not have affected it. This establishes that the defendant’s fraud played a role in the history of the entire loss. This brings us to the second enquiry where it is open to the court to hold that this entire loss is within the scope of the liability of the defendant for the consequences of his fraud. This could be done on the sound normative ground that this extent of liability is justified as a deterrent to frauds of the first type. This could be the result even though the first fraud did not increase the probability of the second fraud, and the conjunction of the first and second frauds might be described as a coincidence.105 By analysing the case in non-causal terms, and dividing the issues between historical involvement and purposive relevance, we are able to expose the normative reasons for holding D liable for the entire loss

104. In which case the traditional presentation would be “breach but the outcome is too remote”. See also the “tripping on the pavement” and the “handing over of the gun” examples in n. 100 above. 105. See the House of Lords decision in Smith New Court Securities Ltd v. Scrimgeour Vickers (Asset Management) Ltd [1996] 3 WLR 1051. On which see J. Stapleton, “The Normal Expectancies Measure in Tort Damages”, above n. 90, at 273–4; also Stapleton, Torts Law J 7 (1999), 296 at 305–6, Cambridge L J 56 (1997) 17.

172

  despite the coincidental nature of the second fraud.106 In contrast, such cases present complex difficulties to “causal” theorists who have constructed a notion of “cause” that hinges on requirements such as the absence of a later abnormal intervention (Hart and Honoré)107 or that the factor increased the likelihood of the outcome (Lipton).108 To sum up: by digging below the mask of causal slogans, abandoning the search for “causal principles”, and looking below notions of risk, coincidence and unforeseeably increasing the probability of the outcome, we expose the normative concerns that can give those notions relevance in law. Causal language merely obscures these concerns. All normative concerns are better dealt with under a normative analytical category such as “breach” or the “appropriate scope of liability for consequences of tortious conduct”. This then limits “causation” in law to the first enquiry: whether the defendant’s tortious conduct played a role in the history of the outcome, a purely factual issue.

106. Similarly the liability for coincidence allowed by the eggshell skull rule is better understood in terms of expressible legal concerns (summed up in the normative statement that the tortfeasor should take his victim as he finds him) than in terms of opaque “causal” exceptions and “asymmetrical treatment of abnormalities”. Contrast Lipton, above n. 2, at 139: “in both common sense and law . . . a subsequent abnormal condition defeats causal connection, but a preexisting one does not”. 107. An example of how “causal” terminology can hamper understanding is the approach taken by Honoré to coincidence in “Medical Non-Disclosure, Causation and Risk: Chappel v. Hart”, above n. 60. On the assumption that Dr Chappel’s failure to warn did not increase the “risk” (i.e. the probability of the harm), Honoré argues that the finding of liability offended “causal principles” because the doctor’s failure was “not a cause of the injury . . . though it was certainly a but-for condition”. There was an “absence of causal connection”. (Emphasis added). This followed, he argued, from his definition of cause: “a cause is . . . an intervention in the existing or expected course of events” (ibid. at 5). Conduct that did not increase the “risk” of the outcome could not therefore be a cause. This approach does not capture the clear role that Dr Chappel played in the history of the injury nor expound the reasons why, nevertheless, many commentators have argued that under the law of negligence that coincidental outcome should fall outside the appropriate scope of his liability for his breach. It also jars with the other definition of cause that Honoré gives: “a cause is basically an element in a recipe . . . sufficient with others to bring about a certain outcome” (ibid. at 5). Dr Chappel’s conduct was clearly part of the recipe that produced the injury to Hart. 108. Lipton, above n. 2, also provides an example of the unnecessary complexities that a causal treatment of such cases generates. Lipton’s construct of “causation” contains the requirement that the factor “increase the likelihood” of the outcome, so he embraces the idea that “an actor’s foreknowledge affects our judgments of likelihood” and therefore our judgments of causation, a result he concedes is “counterintuitive” (139–40). The strain inherent in requiring such a “causal” construct to straddle the two quite separate enquiries of historical involvement and relevance to purpose is again evident in his conclusion that foreknowledge does not affect “the causal story from a physical point of view, but it does affect our judgement of likelihood and so our judgement of whether the [factor] was a cause or mere condition”. (140).

173

  C. Historical Involvement

Realists wisely wanted to reserve any legal requirement of “causation” for factual issues. But they did not grasp that the only factual issue here was the fundamental notion of a factor’s involvement in the history of a transition. They could not penetrate beneath causal rhetoric, and so failed to grasp the lesson that cases of over-determination teach us. Take a standard case of “over-determination”: two hunters, X and Y, carelessly shoot into a wood, both bullets hitting the victim (V), either being sufficient to kill (a case I will the “double hit hunters’ case”). Why does this case trouble lawyers?109 It is because they try to understand it in terms of “causation”, an ambiguous artefact of human discourse, incapable of unambiguously pinpointing any underlying factual enquiry. But there is something we all understand about the “double hit hunters’ case”. We know that in some way the story would be incomplete without the careless conduct of each hunter. I call this elemental concept that we understand so well in the double hit hunters’ case, “playing a role in the history of the outcome”. How then do we capture this understanding of “playing a role in history”, according to which over-determined events are untroublesome? How can we “test” whether a factor, including both “interventions” in the world such as careless shooting and “states” such as gravity, played a role in the history of an outcome? To begin, our attitude to the double-hit hunters’ case shows that our understanding of playing a role in the history of an outcome does not hinge on the idea of necessity: the traditional “but-for” test of necessity notoriously fails to identify the role of either hunter! Our understanding hinges on a hindsight sufficiency notion, which is correctly captured by Hart and Honoré’s notion of “a necessary element of a sufficient set”, the “NESS” test later elegantly developed and popularized by Richard Wright.110 Though still labelled in ambiguous “causal” terms—specifically as a test of “causally relevant condition” by Hart and Honoré, and as a test of “causation” by Wright—the NESS approach correctly pinpoints the essence of historical involvement, and accommodates our understanding of over-determined events such as the double hit hunters’ case. 109. Causation in the Law, 2nd edn, above n. 1, at xli–xlii. 110. Causation in the Law, 2nd edn, above n. 1, at 111–17, 122–5, 128–9 and 235–53. Wright’s main treatments of the NESS test appear in: Causation in Tort Law, above n. 2, at 1788–1803; and Causation, Responsibility, Risk, Probability, Naked Statistics and Proof: Pruning the Bramble Bush by Clarifying the Concepts, above n. 2, at 1018–1042.

174

  It is surprising, therefore, that Wright and Honoré did not apply the NESS approach correctly to the desert traveller case; but it is also illuminating. In the following sections I argue that problems in their approach can be avoided if: we restate the NESS approach in non-causal terms; we restate it explicitly as a test of historical involvement; crystallize it in a “targeted but-for” test; we apply that test separately to every element constituted or introduced by the targeted factor; and we apply that test to every candidate factor in turn, in order to resist the temptation to treat other factors merely as a “set-stage” against which a single “snapshot” view of history is taken.

D. Searching for Historical Involvement with the Targeted But-For Test

To be an historical factor a factor must constitute (like gravity) or have introduced into the scene (like shooting), an element that was a necessary member of a set of conditions that was, at least later, actually present and was sufficient to produce the outcome that occurred. In effect, this requires that there is some perspective, on those factors that were present when the transition to the outcome occurred, in which the targeted factor can be seen to have “made a difference” in the history of the outcome. It is important that we have a way to search for this perspective. As the search engine I suggest the following “targeted but-for test”: Take all factors existing at the time of the transition to the outcome in question, including the factor that we have chosen to target, say the tortious conduct of hunter X in the double hit hunters’ case. If there is a notional sequence of removing factors (except any that preempted the relevant effect of the targeted factor)111 from that set such that (I) a stage is reached where, given the remaining factors, even those that we might call coincidental or freakish, the actual transition to the outcome might112 still have occurred, but that 111. A factor can only be notionally removed not misdescribed. It is also important to note that, formally, the factor removed in, say, the double-hit hunters’ case is not the presence of one of the hunters, but his negligent conduct. So we consider what would have happened had he shot carefully, not had he omitted to shoot at all: see Honoré, Responsibility and Fault, above n. 8, 103–5; Stapleton, “The Normal Expectancies Measure in Tort Damages”, above n. 90. 112. Of course, this state of affairs never existed, so we often cannot be sure that it would still have occurred. For example, say D1 plants a car bomb in V’s car timed to go off at noon but X triggers the car’s thief alarm so that the bomb is discovered. We might conclude X “saved V’s life” but that is speculation because if X had not existed it is still possible that V would have driven through a puddle before noon thereby disarming the bomb. See also the uncertainty of the historic role of Enemy No.2 in the desert traveller case: n. 122 below.

175

  (II) the further removal of the targeted factor leaves a set which would not, in the course of things we now know happened, have produced the transition, then the targeted factor played a role in the history of the transition to the outcome. This analytical approach focuses on the role of one factor at a time, so I shall call it the “targeted but-for test” of historical involvement.

In most cases the targeted but-for test reduces to the traditional but-for test. This is because in most cases, there is no need to remove any factors for stage I to be reached. It is simply the case that the removal of the targeted factor leaves a set that would not have produced the transition. Clearly, whether a factor is part of the history of a transition does not depend on whether a human could have foreseen its association with the outcome. It is irrelevant to the issue of historical involvement whether the outcome would have appeared, and perhaps still does appear, totally freakish. As we have seen: in history the abnormal and coincidental happen.113 Finally, as we know from the speeding car example, it is not necessary that the targeted factor (say, gravity or the speed of the car) constitute or introduce an element that increased the probability of the outcome. The outcome may be a coincidence. But to play a role in the history of an outcome there must be a perspective on the actual facts from which that element can be seen to have “made a difference” in the history of the actual outcome. The targeted but-for test is an analytical technique for searching for this perspective. Simple Applications: The Double-Hit Hunters’ Case; Gravity, etc.

In the double-hit hunters’ case the tortious conduct of hunter X satisfies the targeted but-for test of historical involvement in the victim’s death. This is because by notionally removing the tortious conduct of hunter Y the set of remaining factors is such that the victim would still have died (I), but the further removal of the tortious conduct of the targeted factor (hunter X) leaves a set where the victim would not have died as he did (II). We must also consider hunter Y: if we target that factor we find it also satisfies the test. If there is no notional sequence that shows a targeted factor making such a difference, then the factor played no role in the history of the transition. An example here would be a butterfly flapping its wings in South America at the exact instant the hunters shot the victim in Canada.

113. The character of “abnormality” relating to a factor only enters into our understanding: where there are evidentiary gaps in relation to history and its corresponding hypotheticals; or we are addressing the purposive second enquiry.

176

  Of course, gravity also satisfies the targeted-but for test. So too do the decisions of the hunters’ great-great-grandmothers to have children. But these factors did play a role in the full history of the outcome. At its most curious, science is interested in the full list of factors satisfying the criterion of historical involvement:114 until the scientist selects a purpose for his or her enquiry there is no basis on which to narrow the focus of the investigation. This helps us understand lawyer–economists. We may be surprised when they claim that the presence of the foetus in the womb was just as involved in the history of the diethylstilbestrol (DES) injuries the child suffered as was the manufacture and ingestion of the drug. But at some level this claim resonates as true, and this reflects the same point: until we choose the purpose of any more focussed enquiry there is no basis on which to emphasise one factor involved in the history of a transition over another. Pre-emption

To play a role in history of an outcome, the relevant element constituted or introduced into the scene by the targeted factor must not have been pre-empted by another factor. This is because that element must survive so that it can be a necessary member of the set of conditions that was, at least later, actually present and was sufficient to produce the outcome. In applying the targeted but-for test it is critical that only factors that did not pre-empt the relevant element of the targeted factor are notionally removed. This phenomenon of pre-emption explains why it is wrong to see history in terms of an isolated event playing on a single-set stage. History is a process where each new factor influences the range of possible futures because, inter alia, that factor may eliminate from the scene, as it stood at that point, an element necessary to a later possible fate. Let me give a simple example where an element introduced into a scene by the targeted factor is later pre-empted. Say D1 places a bomb under V’s car set to explode at noon. Judged at that point, D1 has increased the probability that V will die by bomb blast at noon. But if later D2 replaces D1’s bomb with another bomb set to explode at the same time, the conduct of D2 has completely pre-empted the effect D1’s conduct that might otherwise have been present and relevant as a NESS factor in V’s death: D1 is not historically involved in V’s subsequent death by explosion of D2’s bomb at noon. Clearly, it would not be open to a plaintiff notionally to remove D2 and thereby attempt to show that D1 played a role in the history of the death by explosion 114. See text at n. 19.

177

  at noon on the basis that D1 had increased the probability of that outcome and the outcome happened. Compare this with the case where, after D1 plants the bomb, D3 immediately shoots V dead. Here D1’s conduct did not introduce into the scene an element that was a necessary member of the set of conditions that was, at least later, actually present and was sufficient to produce the death by gunshot. Finally, contrast a case where D2 merely added his bomb to the scene. Here both D1 and D2 would have been historically involved in V’s subsequent death by explosion at noon (this being a case parallel to that of the double-hit hunters’ case). The relevant effect of D1’s conduct was not pre-empted by D2. The Desert Traveller Case

We can now turn to the desert traveller case, the facts of which are: a traveller (V) is camped in the desert; Enemy No.1 poisons V’s only water barrel with an odourless poison that would kill within seconds; later, an unassociated Enemy No.2 empties the barrel; since V has no potable water and is unable to reach help he later dies of thirst. The critical point to note about these facts is that the single piece of conduct by Enemy No.1 introduced two elements into the scene: he added poison to the water container (raising the probability of death by poison); and he left the traveller in possession of no potable water (raising the probability of death by thirst). When we target No.1’s conduct to discover if it played a role in the history of the death by thirst, we must consider (using the targeted but-for test of the NESS approach) each effect separately. We must apply the test to No.1’s conduct twice. The reason Wright and Honoré do not apply the NESS approach correctly here, and fail to recognise that No.1 played a role in the history of the death by thirst, is that they consider only one of the two effects that No.1’s conduct had on the scene: namely, increasing the probability of death by poison. Since that element was pre-empted by No.2, they dismiss the conduct of No.1 as playing no role in the history of the death by thirst. Had they also separately applied the NESS approach to the second effect of No.1’s conduct, namely leaving the traveller with no potable water, they would have found that it did satisfy the test of historical involvement. Let me explain. In an obvious sense we understand the history of the traveller’s death: we know what happened, and how and who was involved. As we will see, the story would be incomplete without both enemies. But if we try to capture this understanding we have of playing a role in the history of the death using “causal ” tests we fail. Take the traditional but-for test of necessity. After the failure of this “causal” test to capture our understanding of what happened and who was 178

  involved in the double-hit hunters’ case, it comes as no surprise that it fails as a test of historical involvement: it fails to identify the conduct of Enemy No.1 as playing any role in the history of the death. Far more surprising is that the advocates of the NESS test of “causation” disagreed on the application of that approach to this problem. Hart and Honoré originally claimed that it could not be said that No.1 caused the death, or that No.2 did, or both caused it.115 But Wright attacked this view, saying application of the NESS test gives a straightforward answer: “it is clear that the emptying of the can caused the traveller’s death and that the poisoning did not. Only those who are seduced by the butfor test would question this conclusion”.116 Tony Honoré was later persuaded by Wright and concluded that No.2 caused the death and No.1 did not.117 The NESS approach correctly captures the nature of our understanding of playing a role in history. However, if it is correctly applied using the exhaustive search engine of the “targeted but-for” test, it reveals that both enemies played a role in the history of the traveller’s death. The history of the death simply cannot be understood without reference to the fact that what No.1’s conduct achieved, apart from the intended effect118 of the introduction of poison, was the elimination of the only source of fresh water: this was a NESS factor of the death by thirst. Unless we are told about this effect, it will be incomprehensible to us how it could be that No.2’s conduct in removing poisoned water and thereby prolonging the traveller’s life could have been involved in the traveller’s death by thirst. The story is incomplete without the fact that No.1’s intervention put in place the “no potable water” feature of the scene in which No.2 intervenes. It is an error to see this feature of the scene as only part of a “setstage”. How this feature of the scene came into existence earlier is as vital to the full historical story of the death, as the pre-emption of the threat of poisoning. The Historical Involvement of Enemy No.1 in the Traveller’s Death

Let us separate and consider the two elements constituted by or introduced into the initial scene by Enemy No.1: the poisoning of the traveller’s water; 115. Causation in the Law, 2nd edn, above n. 1, at 239–41. 116. Wright, “Pruning the Bramble Bush by Clarifying the Concepts”, above n. 2, at 1024. See also Wright, Causation in Tort Law, above n. 2, at 1802; Mackie, The Cement of the Universe (Oxford: Clarendon Press, 1974) at 44–6. 117. Honoré, Responsibility and Fault, above n. 8, at 111–12. 118. Compare Honoré’s comment that that No.2’s act “adequately explains” how the death of the traveller came about: Honoré, Responsibility and Fault, above n. 8, 112. If one looked only at intended effects of conduct, then No.2 looks as if he pre-empted No.1; but then the conduct of No.1 also foiled No.2 of its intended effect, namely No.2 intended to remove the traveller’s pure water but, due to No.1, this no longer existed.

179

  and the removal of any source of potable water. The former is a NESS condition of death by that poison. The latter is a NESS condition of what actually occurred: death by thirst. Of course, judged just after No.1 acts it would have seemed that, of these two potential fates, death by thirst was less probable than death by poison. But this is irrelevant to the issue of historical involvement, as we saw earlier. The fact that No.2’s intervention was a coincidence is irrelevant to the issue of what happened and whether No.1’s conduct had any historical involvement in it. The only issue relevant to the question of No.1’s historical involvement in the death is whether No.2 pre-empted the NESS condition of death by thirst that was introduced into the scene by the conduct of No.1. This is something No.2 did not pre-empt. All No.2’s act accomplished was the preemption of the NESS condition of death by poison that was introduced by No.1. To express this in terms of the earlier NESS formulation: No.1’s conduct played a role in the history of the death by thirst because it introduced into the existing scene an element (no potable water), not later pre-empted, that was a necessary member of a set of conditions that was, at least later, actually present and was sufficient to produce the outcome. Before No.1 acted there was a fresh water supply. After he acted there was none. We can confirm No.1’s historical involvement in the death by thirst using the targeted but-for test which No.1’s conduct satisfies. If we take all factors existing at the time of the transition to death by thirst, including the factor that we are investigating (No.1’s conduct), there is a notional sequence of removing factors from that set (specifically the removal of the conduct of No.2)119 such that (I) given the remaining factors, even those that are coincidental or freakish, death by thirst might120 still have occurred, but that (II) the further removal of No.1’s conduct leaves a set that would not, in the course of things we now know happened,121 have produced the transition. The Historical Involvement of Enemy No.2 in the Traveller’s Death

No.2 is also involved in the history of the death, even though all he seemed to do was remove poisoned water and lengthen the traveller’s life. After No. 2 119. Which we can remove because it did not pre-empt the relevant effect of No.1’s conduct, namely it did not affect No.1’s elimination of potable water. 120. We must say “might” because we do not know whether, if No.2 had not acted, the poisoned water barrel would have been blown over by a gust of wind anyway. 121. Knowing what we now do about other factors, for example that a plane was not going to parachute fresh water to the traveller.

180

  acts the probability of death by No.1’s poison is nil. It is a threatened fate that has been pre-empted. No.2 has pre-empted that potential effect of No.1’s conduct. However, No.2 thereby cleared the way for the traveller to suffer the other fate that No.1’s conduct posed: death by thirst. In other words: No.2’s act played a role in the death by thirst because it introduced into the existing scene an element (the absence of the poisoned water) that, given the scene on which No.2 acted (a scene where the traveller was highly likely to have drunk the poison), seems122 necessary to the set of conditions that was, at least later, actually present and was sufficient to produce the outcome of death by dehydration. Had that poison still been present, there was a strong probability that death would have been by poison and not by thirst. We can confirm the historical involvement of No.2 in the death by thirst using the targeted but-for test which No.2’s conduct satisfies. If we take all factors existing at the time of the actual transition to death by thirst, including the factor that we are investigating (No.2’s conduct), there is a notional sequence of removing factors from that set (specifically the removal of the conduct of No.1)123 such that (I) given the remaining factors, even those that are coincidental or freakish, death by thirst would124 still have occurred, but that (II) the further removal of No.2’s conduct leaves a set that would not, in the course of things we now know happened, have produced the transition. The History of the Hibernating Hedgehog’s Death

Wright and Honoré’s application of the NESS approach to the desert traveller case was deficient for two reasons. First, they should have applied the NESS approach separately to every element constituted or introduced by No.1’s conduct. Secondly, they did not accommodate the “diverting” type of historical role illustrated by the conduct of Enemy No.2, namely conduct that diverted the fate of the victim from that which would have been the most probable to that which later actually occurred. Instead they seemed to succumb to the temptation to focus on the last NESS factor of death by thirst to fall into place, namely removal of the poison, and to treat earlier developments as part of a “given” stage-set. That is, they focus on just one 122. We see here that the historical involvement of No.2 is not certain because we do not know whether, if No.2 had not acted, the wind would have blown the barrel over before the traveller would have drunk the poison so that the traveller would have died of thirst anyway. 123. Which we can remove because it did not pre-empt the relevant effect of No.2’s conduct. 124. Because we now know there was not going to be any rescue plane, etc.

181

  moment, the moment just before No.2 acts, and so they ignore the role No.1 had in introducing into that scene an element that qualifies as a NESS factor for the death by thirst. The traveller problem is illuminating because it demonstrates that each effect of a targeted factor must be considered separately and because it shows that this single “snapshot view of history” is flawed. Both refinements to the application of the NESS approach are important to capture the nature of history as a continuous process of interrelated transitions and to identify all the factors that played a role in the history of an outcome. It is worth noting that their single snapshot view of history requires Wright and Honoré to misdescribe what No.2 actually achieved. It may be tempting to say that No.2 set up a particular NESS condition of the outcome of death by thirst: namely lack of water. It is true that lack of water was one of the NESS conditions of the traveller’s death by thirst, but in order to assert that this was set up by No.2 one has to alter a critical fact that we know actually occurred. One has to assert that what No.2 drained away was water, when in fact what he drained away was poisoned liquid. This is indeed the misdescription relied on by Wright and Honoré.125 It is true that the removal from the scene of the poisoned liquid seems to have been one of the NESS conditions of the traveller’s death and, as we have already seen, that is why No.2 seems to have played a role in the history of the death. But the lack of fresh water is also a NESS condition of the death and No.2 did not put that condition in place, No.1 did. The error in dismissing the role of Enemy No.1 in the traveller case, can perhaps be more clearly illustrated by another example. In a zoo there is a tiny airconditioned room in which a hedgehog has just gone into hibernation for the winter. The zoo’s regime is for the animal to be checked every Tuesday. Knowing this, a madman enters the room at 6am on Wednesday morning and places a bomb on the floor, which he has programmed to explode at noon that day. He does not notice that in placing the bomb on the floor it has completely obstructed the only air-conditioning vent in the tiny, otherwise sealed, room. Later at 7am on that Wednesday the zoo’s new manager is making spot checks on various facilities. He enters the room and notes that the hedgehog seems to 125. Honoré, Responsibility and Fault, above n. 8, at 112, “[No.2] introduces a condition, lack of water, that in the circumstances, including the absence of an alternative water supply, is sufficient to bring about and does bring about [the traveller’s] death by dehydration”; Wright, “Pruning the Bramble Bush by Clarifying the Concepts”, above n. 2, at 1024, “the emptying of the can was necessary for the sufficiency of a set of actual antecedent conditions that did not include the poisoning of the water, and the sufficiency of this set was not affected by the poisoning of the water”.

182

  be breathing normally.126 He sees and hears the ticking parcel. Thinking the parcel is a misplaced clock he simply turns its switch off and this deactivates the bomb. He does not realise it is blocking the air-vent. The following Monday the hedgehog dies of lack of fresh air, a fate discovered when the routine Tuesday check is made the next day. Here it is much easier to see how the first actor, the madman, played a role in the death. The death of the hedgehog is only intelligible in the light of one of the, albeit unintended, effects of his conduct. The same applies to the role of Enemy No.1 in the traveller case.

S

1. Causal terminology and usage is labile, sometimes used to express the answer to the factual question of whether a factor played a role in the history of a transition to an outcome; and sometimes to express the answer to the quite different purposive question of whether a particular factor that played such a role was important for the purpose at hand (a purpose which might be merely explanatory, or attributive or otherwise evaluative). Lawyers should unpack “causation” into these two enquiries. 2. Where the facts are agreed but the legal dispute is couched in terms of how to apply “causal” terms to the facts, the dispute is really about the second enquiry, about competing perspectives on the legal responsibility issue raised by those facts. 3. In such legal disputes, where the parties assert different causal usages, the broad patterns of causal usage identified by Hart and Honoré do not assist courts because they only mirror situations where there is a consensus about the underlying responsibility issue. 4. The identification by Hart and Honoré of two factual issues, two types of “intervention” by other factors, as central to the truncation of “causal connection” emphasises these facts at the expense of other facts that trigger diverse concerns that may be of great importance to the issue of the scope of liability for the consequences of tortious conduct, such as intention, profit-seeking, any prior undertakings between the parties and so on. 5. Though Hart and Honoré favour the traditional mixing of factual and certain evaluative issues under the legal category of “causation”, it is this 126. In other words, the blockage of the vent has not yet depleted the oxygen content of the atmosphere to such an extent that it produces laboured breathing: a useful detail for which I thank Ingrid Fuary-Wagner.

183

 

6.

7.

8.

9.

mixing that is the root of doctrinal obscurities in the area. To the extent law has in the past crudely relied on the labile and vague ways in which we use causal words outside law, it has failed to meet the minimum standards of determinacy needed for workable legal concepts. Contrary to the core thesis of Hart and Honoré, the law does need a special “legal” concept of cause. Courts should confine the legal category of “causation” to the first and the only factual enquiry here, namely whether the defendant’s tortious conduct played a role in the history of the victim’s injury. This would leave all normative issues, whether termed policy concerns or otherwise,127 to other categories such as duty or breach or the scope of liability for consequences category where they would be unpacked and fleshed out in non-causal language. Hart and Honoré correctly formulated an approach to historical involvement, the NESS approach: to have played a role in the history of the outcome, a factor must constitute or have introduced into the scene existing at that time, an element that was a necessary member of a set of conditions that was, at least later, actually present, and was sufficient to produce the outcome. This requires that there is some perspective from which that factor can be seen to have “made a difference” in the history of the outcome. We can search for this perspective using the targeted but-for test. Take all factors existing at the time of the transition to the outcome in question, including the factor that we are investigating, say a defendant’s tortious conduct. If there is a notional sequence of removing factors (except a factor that pre-empted the relevant effect of the targeted factor) from that set such that (I) a stage is reached where, given the remaining factors, even those that are coincidental or freakish, the transition to the outcome might still have occurred, but that (II) the further removal of the targeted factor leaves a set which would not, in the course of things we now know happened, have produced the transition, then the targeted factor played a role in the history of the transition to the outcome.

127. Including the normative question of when the law should require that the tortious conduct was a necessary condition for the outcome, as well as including when other normative issues such as when interventions are sufficiently relevant to truncate liability for consequences, when the intention/wrongfulness/profit-motive of conduct is relevant, and so on. See Stapleton, “Legal Cause: Cause-in-Fact and the Scope of Liability for Consequences”, above n. 28.

184

  10. The continuous interrelated nature of historical processes will not be revealed and we will fail to identify all the factors that played a role in the history of an outcome unless the targeted but-for test is applied: • separately to every element constituted or introduced by a targeted factor; and • separately to every candidate factor (the temptation to assume a single “snap-shot” view of history, with its choice of just one set-stage that forms the “static” backdrop to the enquiry, should be rejected).

185


8 PRIVATE LAW: BETWEEN VISIONARIES AND BRICOLEURS William Lucy*

The argument here differs from the others in this collection. The other essays address, in a variety of rigorous and illuminating ways, the substance of Tony Honoré’s path-breaking and challenging work on causation and the varieties of responsibility. Honoré’s arguments on these issues certainly feature here, but they are used for another purpose: to map the contours of contemporary private law scholarship. A bird’s-eye view of this terrain might be helpful for two reasons. First, it might clarify the similarities and differences between parts of the terrain and highlight some features lost to those working at ground level—labourers in the vineyard of case-law, causation, responsibility and justice. Second, it may aid in charting a path through terrain almost completely covered with thickets of ostensibly different and competing approaches; it might thus facilitate distinguishing not only wood from trees but between types of tree. The problem with such general views is that they are often incomplete: large-scale maps give the broad lie of the land but are no guide to conditions under foot. Those working at ground level, applying, developing, evaluating and criticising private law, might be tempted to dismiss them as unhelpful. And yet there might be advantage in the broader view, especially when the map provided is constructed with an eye to maps of contiguous areas. Hence, some of what is said here applies not only to private law scholarship but also to some criminal law scholarship and to mainstream legal philosophy or jurisprudence. Knowing our position in this broader intellectual milieu allows the circumvention and solution of difficulties encountered elsewhere. It also facilitates some appreciation of the problems unique to particular domains and serves to highlight recurrent snags, many of which concern assumptions about method in legal scholarship and the nature and functions of law. * Thanks are due to Simon Caney and Richard Mullender for comments on a very early draft. I am grateful to the participants at the Columbia workshop on Tony Honoré’s work, particularly John Gardner, Niki Lacey, Bernard Williams and Ben Zipursky, for their helpful suggestions. Gerry Johnstone and Neil Duxbury also offered fruitful advice and comments.

  I. H’ A

Those aspiring to the mantle of iconoclast could do no better, particularly within the confines of contemporary legal philosophy, than seek to provide a non-economic justification for strict liability. While Honoré’s vast corpus of important work does not display an obvious anxiety to be iconoclastic, that part which articulates the moral basis of strict liability1 will undoubtedly be regarded so by many. For many legal philosophers and other legal scholars, Thomas Nagel’s judgement that strict liability is “irrational as a moral position”2 and Ernest Weinrib’s view that “corrective justice . . . implies the rejection of strict liability”3 are beyond question. Honoré seeks to question such views by immediately noting the pervasiveness of strict liability. Textbook authors and philosophically inclined scholars alike are prone to assume that strict liability is relatively rare within the law. Whether or not this is so depends upon how strict liability is understood. Some thought about the logic of and thus the relations between the concepts in play, undertaken with an eye to both current usage and maximum perspicacity, suggests two points. First, our liability concepts surely occupy a continuum, at one end of which the notion of absolute liability resides and, at the other, the idea of subjective moral fault. Second, the ground between can quite plausibly and fruitfully be characterised as that of strict liability, provided that the strictness of this kind of liability is understood as being a matter of degree which varies between specific types of tort (and, perhaps, crime). This might seem more sensible once the ideas of absolute liability and subjective moral fault are clarified. The former is simply a matter of action (including refraining) or status: simply doing X, or being in situation Y, is enough to attract liability. Of course, it would rightly be claimed that even torts or crimes akin to these require more than the creation of X or being in the state of Y. The defendant must also have had the requisite capacities, and perhaps also the opportunity in which to exercise them, before he is subject to tortious or criminal liability. Hence, the defendant must be neither a minor nor insane; must lack serious volitional and cognitive impairments and must not have been compelled to act or forced into the situation. Nevertheless, a form of liability is conceivable which requires only that the defendant bring about X or be in situation Y, 1. T. Honoré, Responsibility and Fault (Oxford: Hart Publishing, 1999), ch 2. 2. Nagel, “Moral Luck” in his Mortal Questions (Cambridge: Cambridge University Press 1979), 24 at 31, cited by Honoré, above n. 1, at 23. 3. E. Weinrib, The Idea of Private Law (Cambridge Mass.: Harvard University Press, 1995), 171.

188

 :     regardless of these other requirements.4 It seems sensible to regard such liability—thankfully mainly only a product of mind and found in no or few legal systems—as absolute. At the other end of the continuum is subjective moral fault. What might this be? Understood in its starkest form it would require only being bad in mind. But moral or legal liability for nothing more than bad intentions is as rare as absolute liability; hence a more plausible account of subjective moral fault is that it requires being bad in mind and in deed. Simply bringing about nonconsensual harm to others with the sole aim of so doing must be a paradigm instance of morally faulty conduct. Here we have a combination of the defendant’s malevolent purpose and an outcome in the world that is its culmination. If content that the defendant was of capacity and not acting under compulsion, we have little compunction in judging him morally and legally liable. While the two types of liability founded upon subjective moral fault share a concern with wrongdoers’ purposes and the outcomes of their conduct, they diverge in one significant way. Assessments of moral liability based upon subjective moral fault are open, to a far greater degree than assessments of legal liability, to the particularities of the wrongdoer’s character and circumstances. Assessments of legal liability operate through doctrine which often excludes particularities (motive in the criminal law, for example) undoubtedly relevant to the assessment of a wrongdoer’s moral liability. While the latter is founded upon a fine-grained assessment of the wrongdoer’s purpose and conduct in light of his character and circumstances, the former is deliberately less sensitive and discriminating. This divergence between legal and moral assessments of liability presents a problem for those wanting to present both as consistent with subjective moral fault. It is not uncommon to blame and judge those who harm in absence of the full measure of subjective moral fault and without a complete assessment of all the particulars of their character and circumstances. For example, blame and liability often follow when the harm brought about is not the precise culmination of a wrong-doer’s malevolent purpose (the harm is greater, lesser or of a different kind), or when it results from an attitude less morally affronting (like indifference or carelessness). Furthermore, it is not uncommon to be impatient with the particulars of the wrong-doer’s character and situation and to invoke more robust behavioural and character requirements founded upon the standards of reasonable people. Hence the move is made from what the wrong-doer actually could have done to what he should have done as the basis of liability.5 4. See, for example, Winzar v. Chief Constable of Kent, reported in The Times, 28 March 1983. 5. And this move, of course, entails much about how we understand “can” and “can’t”: see the appendix to Responsibility and Fault, above n. 1.

189

  In pointing out that negligence liability is strict in so far as it holds defendants to a standard of care they could not actually (subjectively) achieve, Honoré is pointing out how far this form of liability is from the model of subjective moral fault.6 It deserves the label “strict” because there is no or little subjective moral fault. Indeed, most instances of liability in tort law in particular and law in general fall within the realm of (more or less) strict liability, since most depart to some extent from the requirements of subjective moral fault. Hence, most forms of liability do not require an exact match between the malevolent purpose of the defendant and the harm brought about; harm of some kind will usually suffice. Tort law rarely requires a specific malevolent purpose for liability. Far from focusing on the defendant’s particular purposes and aims, the law often ignores them and the particularities of his character and circumstances, over and beyond those relevant to general defences and the requirement of capacity. While most forms of liability in tort are more or less strict, they are plainly not absolute, since some defences are usually available and the normal capacity requirements for liability hold sway. Types of liability in tort are more or less strict depending upon whether they impose an absolute or a reasonable duty to avoid some harm. Even those stricter forms of liability founded upon absolute duties are not forms of absolute liability insofar as a range of defences is available and the capacity requirements must be satisfied. If negligence and most other forms of liability in tort are indeed far from subjective moral fault, is Honoré not simply playing into the hands of those that regard strict liability as morally objectionable? This might be so, but this move has additional significance. It shows Honoré taking the varieties of strict liability found in the law seriously; it consists, in effect, of taking the contours and requirements of doctrine at face value. Those, like Honoré, having a detailed grasp of doctrine and an eye to its effects on conduct must concede that tort liability is undoubtedly stricter than subjective moral fault. Those that think subjective moral fault must be part of a morally respectable and legitimate system of legal liability, are therefore committed to jettisoning much more of the existing law than the allegedly rare and eccentric segments instantiating what is often taken to be strict liability. Critics of strict liability are rarely cognisant of this. Their impatience with the very idea of strict liability bespeaks impatience with tort law doctrine itself, manifest either in the hasty assumption that doctrine embodies subjective moral fault, or the supposition that it embodies some other equally respectable conception of liability nevertheless incompatible with strict liability. 6. Responsibility and Fault, 16–23.

190

 :     This impatience and its cognate assumptions are often driven by some or other theoretical commitment. For Weinrib that commitment is to the notions of Kantian right and agency combined with an Aristotelian account of corrective justice: “the ideas of equality under corrective justice and the concept of agency under Kantian right exclude strict liability.”7 Weinrib appears to think it is an undeniable truth, given the veracity of his accounts of corrective justice and agency, that strict liability is morally and legally objectionable. However, the matter is not as obvious as he assumes, especially when his puzzling and ambiguous account of strict liability is noted. Doctrine itself has little to contribute to Weinrib’s assessment of strict liability, save to show that no doctrine supposedly embodying strict liability actually does so in the requisite sense.8 The requisite sense, for Weinrib, seems to be this: strict liability is liability without defences and without any degree of fault.9 This curious formulation, which actually looks like a characterisation of absolute liability, results from a preoccupation with Richard Epstein’s odd theory of strict liability. It is also supported by a philosophically grandiose account of the nature of fault. The principal oddity of Epstein’s account of strict liability lies in taking an incomplete account of some torts, like trespass, as a complete guide not only to the nature of strict liability but the whole of tort law. The essential limitation is that Epstein takes causation to be the principal element of these torts while actually existing instances of strict liability are much more complex, usually requiring more for liability than causation and providing a range of defences. While Epstein occasionally liberates himself from the fixation upon causation, Weinrib does not. It ensures that Weinrib makes no clear distinction between strict and absolute liability throughout his discussion of Epstein. And, although Weinrib subsequently notes that there may be a difference between the two,10 no account of what it may be is provided. The end result is this: if strict liability is as Weinrib assumes, namely, liability without defences and without fault, there is no, or vanishingly little, conceptual space left for absolute liability. The philosophical grandeur of Weinrib’s account of fault becomes clear once we pose this question: does he mean by “fault” subjective moral fault or 7. Weinrib, The Idea of Private Law, above n. 2, 177; also chs 3 and 4. 8. Ibid., 185 and 190 9. Weinrib does not give an explicit formulation of what he takes strict liability to be in ch. 7 of The Idea of Private Law, above n. 2. His view therefore has to be pieced together, as this statement is, from his discussion of Epstein. While one may disagree with Weinrib’s claim that most instances of actual strict liability are based on his conception of fault, his treatment of them rightly shows that liability depends on more that just causation. 10. Ibid., 184

191

  something less demanding? If the former, then for Weinrib most forms of liability in tort, including negligence and strict liability, must be objectionable. This is clearly not his view. Since Weinrib thinks that negligence liability is at least morally acceptable, he must regard fault in tort law as something less than subjective moral fault yet morally justifiable (which, on his view, strict liability is not). What, then, does this hybrid look like? The notion of fault receives no independent treatment in The Idea of Private Law. It is mentioned only as a feature of the correlated ideas of violation of duty and wrongdoing. Hence, “the defendant must have committed an act that violates a duty incumbent on the defendant and thus can be regarded as an act of wrongdoing. Accordingly the modern common law emphasises the importance of fault, since the defendant is duty-bound not to perform the intentional or negligent acts that constitute the faulty conduct”.11

The puzzle, then, is this: in what sense can the defendant be said to be at fault if his subjective moral fault is not in issue? Furthermore, is the sense of fault that might be in play—not the fault of the particular defendant in the sense that he breached a duty he undoubtedly could have discharged on this particular occasion, but fault understood as his failure to achieve a standard that might have been beyond his reach on this occasion—obviously insufficient to generate a justification for strict liability? Weinrib’s answer to this question is affirmative, since he can locate the fault involved in negligence in his account of Kantian right, an account inhospitable to strict liability.12 It needs be noted, though, that this is a theoretically and philosophically indulgent answer. For, before invoking Kantian right in order to understand fault in tort law, it needs be shown that no other plausible account of fault is available that (i) is less philosophically ambitious and contestable; and (ii) makes intelligible as much of the law as possible. There is at least one such account available, and it is as sensitive to the bipolar structure of private law as Weinrib’s. It is provided by Honoré in chapter 2 of Responsibility and Fault. The conditions —of philosophical or theoretical parsimony and legal intelligibility—that Weinrib’s account fails to satisfy are surely far from being either eccentric or too stringent, especially since the account purports to offer an understanding —not a reformulation, reinterpretation or deconstruction—of law.13 By contrast with some accounts of strict liability, Honoré’s treatment is not only sensitive to the varieties and range of strict liability in tort doctrine, but 11. Weinrib, The Idea of Private Law, above n. 2, 134, emphasis added. 12. Ibid., 179–83. 13. Ibid., 1.

192

 :     also avoids the rush—usually driven by some or other theoretical commitment—to normative judgement. Avoiding the rush does not, however, mean postponing judgement forever and in this respect Honoré’s treatment of strict liability differs from most textbook discussions. He accepts the necessity of both sustained, critical analysis and normative evaluation of the concept. To that sizeable group of legal scholars who craft the vast textbooks of private law, neither the nature of strict liability nor fault is particularly interesting. Standard textbook treatments of tort law, for example, usually do only two things: they offer a very brief history of the interplay between strict and faultbased liability regimes and elucidate the contours of existing strict liability and fault-based torts. A sustained, critical analysis of the nature of these concepts—not to mention many other fundamental tort law concepts such as causation, reasonableness and foreseeability—is hardly ever attempted.14 This gives rise to some problems. One is that the history story textbooks provide is critically vague insofar as two of the key concepts are left under-analysed. The standard form of the story notes the interplay between “fault” and “no-fault” types of liability. While some take account of the controversial strands in this story, others simply repeat one position in it (usually that espoused by O W Holmes on the triumph of fault) as if beyond question. But if legal history does indeed reveal a clearly discernible pattern we must, in order to see it, have some conception of what we are looking for. If the pattern is, as usually maintained, one showing the gradual rise of fault liability and the withering of strict liability— which is currently being replicated in the English law of private nuisance—it will only be evident if we have available to us clear analyses of these notions. Such analyses, which are often indiscriminately labelled, inter alia, “models”, “ideal types”, or “operational definitions” in the general historical and social scientific literature, make historical and social-scientific investigation possible. They do not, however, guarantee that the social scientist or historian will find what they seek. Textbook authors have a response to this problem, namely, that they do indeed have a conception of what fault and strict liability look like. All the textbooks invoke similar, well-worn but flimsy analyses of strict and fault liability, these having been used time and again in previous editions of the same texts, or in earlier but now defunct texts, and sometimes repeated in judicial 14. Two quite different examples of the textbook genre are nevertheless remarkably similar on this point: see R. Heuston and R. Buckley, Salmond and Heuston on the Law of Torts (21st edn, London: Sweet and Maxwell, 1996), ch. 2 and B. Markesinis and S. Deakin, Tort Law (4th edn, Oxford: Oxford University Press, 1999), 18–23 and 41–4.

193

  decisions. An important and interesting question, unlikely to be raised in the textbooks, then arises: are these analyses instantiated in the historical record or simply read back into it? Yet whether or not such analyses are so evidenced does not serve to increase their rigour and clarity. Unfortunately, most standard textbook analyses lack these qualities. There is a reason for this; the very same reason also ensures that such textbooks rarely raise the normative questions about these forms of liability that concern Honoré. One example must suffice. Sadly, it comes from one of the most challenging and interesting contemporary English textbooks. The second edition of R Dias and B Markesinis’s Tort Law located the tort of negligence within the realm of fault-based liability which, of course, was regarded as quite different from strict liability.15 This raises the problem of giving a coherent account of what it is that constitutes fault in the context of negligence and the authors were loath to involve themselves in a discussion of this type. They were, however, aware of some of the difficulties encountered in conceiving negligence liability as fault-based. Their general treatment highlighted three snags in particular: the existence of faulty conduct that does not attract negligence liability; the fact that those at fault rarely incur the full costs of their faulty conduct because they have insurance; and the especially costly nature of the negligence system. In their treatment of the contours of negligence liability itself they also highlight the snag which Honoré takes as his starting point in chapter 2 of Responsibility and Fault. This concerns those instances in which defendants who are not in any obvious sense at fault—and certainly not in the sense of subjective moral fault—are nevertheless liable in negligence. Dias and Markesinis conclude that in Nettleship v. Weston and related cases, and in liability for motor accidents in general, “negligence in fact works much more strictly . . . than in other areas, so much so that it has become artificial to continue calling it by that name”.16 This is correct only if a very close connection between negligence liability and fault is assumed. Moreover, turning this assumption into an argument demonstrating the connection requires, as its first step, a sustained analysis of the notion of fault. This is not provided. By the time of the fourth edition of the textbook, the authors had changed position on the relation between negligence liability and fault. It now occupies what seems to be a conceptual void between strict and fault liability.17 The problems previously identified with the notion of fault are still accepted and the difficulties of cases like 15. R. Dias and B. Markesinis, Tort Law (2nd edn, Oxford: Oxford University Press, 1989), at 23. 16. Ibid., 175; see also 103–4. 17. Ibid., 20.

194

 :     Nettleship are still pressing.18 But the authors are no longer sure where negligence should properly be located given the choice between fault-based and strict liability. The choice is obviously difficult since no sustained analysis of the nature of fault is offered; its difficulty is exponentially increased because no sustained analysis of the nature of strict liability is offered. The treatment the authors do provide is, however, useful. Their chronicle of the contours of strict liability torts highlights first, that these torts differ significantly from one another and, second, that they differ from supposedly fault-based torts in matters of degree only. This is important since many textbooks assume a rigorous divide between strict and fault-based liability without ever providing convincing treatments of the difference. That said, the treatment of fault and strict liability in Markesinis and Deakin is brief, lacking rigour and clarity. Their text is not however unique; all are equal in this respect—all eschew sustained analysis of the fundamental concepts of tort liability. This lacuna might seem puzzling if the primary purpose of these works is overlooked, namely, providing a chronicle of the law as it currently stands. The pressure to accommodate as much recent case-law as possible forecloses the opportunity for sustained and critical analysis of fundamental concepts. The audience for such texts might initially be thought to be law students but this is not actually the case or, more accurately, not the case once such texts make the transition from first to subsequent editions. Students do not need to know each and every recent case in order to reach an understanding of the law and, in fact, a welter of cases often impedes rather than aids understanding: the underlying concepts are lost just like wood obscured by trees. Since textbook authors regard their primary purpose as keeping on top of the welter of cases, and since this is of most value to those who must have up-to-date information, namely, practitioners, then they must be the primary audience for such works. While it is unlikely that many textbook writers would accept this, it nevertheless provides a plausible account not only of the lack of analysis of fundamental concepts in such works, but also of their lack of engagement with the philosophical and normative dimensions of those concepts. It is, for example, just as rare to find some argument as to the normative standing of strict and fault-based liability in such works as it is to find sustained analysis of the concepts themselves.19 Whereas writers like Nagel and Weinrib seem to begin their thinking about strict liability from a pre-established peak of normative judgement, textbook writers inhabit the plains of doctrine where such judgement and sustained 18. Ibid., 41–4, 155–6, 160–3. 19. Ibid., 504–8

195

  analysis of fundamentals are apparently out of place. Honoré’s approach, by contrast, seems to occupy a position between the two. Unlike the latter, Honoré regards sustained analysis of the concept and related concepts like causation and capacity as essential; he also thinks it incumbent on legal scholars to address the philosophical and normative issues such concepts raise. What might be called “justificatory ascent” is therefore almost always an option for him.20 By contrast with the former, Honoré begins by trying to make sense of legal doctrine in its own terms and is open to the possibility that doctrine might be normatively intelligible as it stands, without theoretical reformulation. Justificatory ascent is not therefore always necessary. Whether or not this apparently uncomfortable position is sustainable, displaying the advantages and eschewing the disadvantages of the alternatives, is the question the remainder of the essay attempts to answer. From the peg provided by these two different approaches to strict liability two more general and quite different pictures of the nature of private law scholarship can be hung. The two pictures are of course caricatures and, like all caricatures, are to some degree rooted in reality. They represent opposite ends of a spectrum upon which many actual instances of private law scholarship can be placed, albeit usually at intermediate points between each extreme. At one end resides visionary private law scholarship. The role of legal scholars on this view is to aid judges and the courts in the task of keeping the law on the trajectory of its normative vision. To do this effectively, legal scholars should confer with, or themselves become, moral or political philosophers, economists or rational choice theorists. This allows them to understand better the moral, political, social and economic commitments and consequences of the law. For some theorists enamoured of a visionary approach, legal scholars are philosophical counsel to courts and judges, perhaps providing a Philosophers’ Brief for every interesting legal dispute.21 Other visionaries contest the philosopher’s right to this exalted position, insisting she be replaced by, inter alios, an economist. Visionary private law scholarship assumes some body of theoretical knowledge over and beyond that immanent in private law is necessary to understand or make sense of the law. Promiscuity is evident in the range of sources mined for theoretical insight, although the dominant influences within less 20. The phrase belongs to R. Dworkin, “In Praise of Theory” Arizona State Law Review 29 (1997) 353 at 356–60. It is a matter of propositions of law and legal doctrines more generally being subject to assessment as better or worse; obviously, moral, political, economic and philosophical considerations can feed into this process. 21. See R. Dworkin (et al.), “Assisted Suicide: The Philosophers’ Brief” New York Review of Books, 27 March 1997, 41–7.

196

 :     outré visionary scholarship are fairly limited: contemporary Anglo-American analytical philosophy, particularly but not exclusively moral and political philosophy, neo-classical economics, and rational choice theory. Visionary scholarship usually consists of three steps. First, it acquaints one with the theoretical antecedents of the normative vision informing the law. The whole Anglo-American philosophical canon (membership of which is not a function of nationality) is pillaged as a source, as are other bodies of thought. While Aristotle, Immanuel Kant, Georg Hegel and John Rawls often supply the vision thing, so, too, do Jeremy Bentham, Vilfredo Pareto and John Dewey. Second, an argument is offered to illustrate the value of the selected philosopher’s or economist’s work. The argument usually consists of an attempted demonstration that, despite the selected theorists’ near complete silence on, for example, the question of whether the compensation rule in accident law should be based upon strict liability or fault, his work is nevertheless crucially relevant. Third, the vision is put to work in the following way: the law is simply read in light of it. The vision in whose light the law is read can be more or less distant from the law; the degree to which the vision is rooted in the law varies according to the theory from which it springs. Sometimes the normative vision is propped up by a mere handful of cases, sometimes allegedly supported by large chunks of doctrine and, on still other occasions, is simply held to be consistent with the law. For visionaries, theory is clearly the dominant partner in the relationship with law and doctrine. The latter are taken to be in urgent need of theoretical support. Why? To a large degree this depends upon the theory adopted, but some common concerns motivate visionaries’ recourse to theory. The first is a concern about a lack of integrity or systematicity in private law. The history of private law, like that of any persistent social practice, is a ramshackle affair that fits ill with the injunctions of any particular normative vision.22 Viewed from the perspective of a unitary and systematic vision, that history is too rich in contingency, wrong-turnings and diversity. This concern is joined by a closely related worry about the rationality of private law. That there is no single normative vision providing a coherent script for legal actors, but just a vast pile of doctrines and cases embodying a diversity of values and assumptions, is worrying because the rationality of the unfolding law is as a result doubtful. Lack of a script apparently entails the absence of systematic or coherent grounds by reference to which the rights and wrongs of particular decisions can be assessed. Latent within both concerns about 22. See, for example, D. Ibbetson, A Historical Introduction to the Law of Obligations (Oxford: Clarendon Press, 1999).

197

  coherence and rationality is one about private law’s lack of normative appeal. Once it is accepted that private law does not embody a single unified normative vision, it is assumed to have little, if any, normative appeal. Theory’s dominant position in visionary scholarship might be expected were visionaries explicitly dismissive of doctrine and the body of institutions, practices and conventions constitutive of private law. However, many contemporary private law theorists located towards the visionary end of the spectrum adopt a method for understanding and explaining almost any social practice requiring that its rules, doctrines, institutions and conventions be taken seriously, from the participants’ point of view. The puzzle here is that in the face of this consensus on method there is such divergence in the substance of visionary-inclined private law scholarship.23 The second type of private law scholarship is less self-consciously theoretical than the first. The role of the bricoleur scholar is to provide the bricoleur judge with a detailed map of legal doctrine. Such a map, which is both synchronic and diachronic, is the best aid judges can have in selecting the tools and solutions in particular cases. The bricoleur scholar strives to sharpen the available tools, to elucidate their number and how they relate to one another.24 This work is done through the journal case-note and the large, historically informed chronicle of the law. At their best, the latter set doctrines and bodies of law in relation to one another, question their compatibility and, occasionally, ask how the assumed purposes of the law could be reached more effectively. Such chronicles are rarely troubled by moral, political or sociological reflection which, given their historical antecedents, is no surprise.25 The only values that explicitly appear in such texts—over and beyond those immanent in the law itself—are generic rule of law values. Hence, it is not unusual to find bricoleurs expounding critical analyses of doctrinal developments, the critical arguments being founded upon those values—of consistency, predictability etc.,—that constitute just about any account of the nature of the rule of law. It would be a mistake to regard this work as non-theoretical. There are elements – “rule of law” values, assumptions about the impact and efficiency of law – worthy of visionary theoretical reflection in the organising framework 23. Compare R. Dworkin, Law’s Empire (London: Fontana 1986), 12–15 and ch. 8, with E. Weinrib, The Idea of Private Law, above n. 3, 11–16 and ch. 6. 24. For richly suggestive and lucid discussion, to which this essay is obviously indebted, see N. Simmonds, “Bluntness and Bricolage”, ch. 1 of H. Gross and R. Harrison (eds), Jurisprudence: Cambridge Essays (Oxford: Clarendon Press, 1992). 25. See A.W.B. Simpson, “The Rise and Fall of the Legal Treatise: Legal Principles and the Forms of Legal Literature” University of Chicago LR 48 (1981), 632–79.

198

 :     of bricolage, but those elements are usually taken for granted. The kind of theoretical activity characteristic of bricolage is therefore not the selfconscious invocation of a body of knowledge above and beyond, yet generating its own insights into, the practice, but the search for means of keeping the practice going in reasonable fettle.26 Problems explored by the bricoleur scholar are those thrown up and identified by the practice itself and the solutions seized are simply those that work. In attempting to ensure that the practice keeps going, bricolage becomes theoretical in another sense: the solutions it offers to doctrinal problems must accord with the doctrine’s and the practice’s understanding of itself. Bricoleur scholars must therefore chart the law’s self-understanding and this is surely a theoretical enterprise insofar as that understanding is itself theoretical, that is, insofar as it exceeds the immediately relevant knowledge required to move within the practice. It must not be assumed, though, that because scholarly bricolage is deeply embedded in practice, it is not creative. For example, scholarship more appropriately placed at the bricolage rather than the visionary end of the spectrum established the credentials, and thus spurred the judicial recognition, of the law of restitution in England.27 While rarely explicit about methodological matters, bricoleur scholarship is, as a detailed examination of this instance of innovation would suggest, undoubtedly founded upon the viewpoint of a few participants in the practice, namely, judges and those who argue cases before them.28 The distinction between visionaries and bricoleurs can appear both simplistic and complex. Complexity arises from the fact that the spectrum which is occupied at each end by pure forms of each approach is made up of different levels or layers in the same way as the colour spectrum consists of different colours. These levels can be labelled “theory”, “method” and “practice”. 26. Theory, for bricoleurs, is therefore akin to the conception of theory articulated by R. Dworkin in ch. 19 of M. Brint and W. Weaver (eds), Pragmatism in Law and Society (Boulder, Colorado: Westview, 1991), although this conception is not obviously present in Dworkin’s own work. Theory for visionaries, which is a little closer to Dworkin’s actual enterprise, is akin to the conception elucidated and attacked by S. Fish in, inter alia, Doing What Comes Naturally (Oxford: Oxford University Press, 1989), 14–25 and chs 14 and 17 and There’s No Such Thing as Free Speech . . . (New York: Oxford University Press, 1994), 224–30 and 302–4. 27. See Lipkin Gorman v. Karpnale [1991] AC 548. The two works that contributed most to this judicial development are Lord Goff and G. Jones, The Law of Restitution (5th edn, London: Sweet and Maxwell, 1998) and P. Birks, An Introduction to the Law of Restitution (Oxford: Clarendon Press, 1985). 28. An argument as to why the viewpoint of these particular participants is privileged is never provided; it is simply taken for granted. The possibilities that there are other participants, and that participants might have different views about the point, purpose or value of the practice or aspects of it, are not entertained.

199

  Ideal forms of visionary scholarship would rank high on the first level, would employ a method taking little account of participants’ views, and would therefore be remote from the practice in question. Ideal forms of bricolage would rank low on the first level, would adopt a method fixated upon only certain participants’ views, and would to that extent be embedded in the practice. Of course, few existing instances of private law scholarship take such pure form, ranking much higher on some dimensions than others. The distinction can appear too simple when taken to map something like a square of opposition between the two positions. Since the argument of section II carves out a via media between bricolage and visionary scholarship, it cannot be maintained that these two positions exhaust the available conceptual space.29 The opposition between them is a matter of various tensions or differences of emphasis on some issues. The first part of section II highlights some of the problems that pure forms of each approach have, thereby motivating the search for a middle way. The preferred middle way, exemplified in the work of Honoré, is explored in the third part of the section, which returns to the essay’s starting point. The second part of the section is concerned to reject two efforts at maintaining the distinction, albeit in amended form.

II. T M G 1. Problems

The problems of pure forms of bricolage and visionary scholarship are diverse and have different effects, some problems being unique only to particular instances of each approach. The cumulative effect of the problems is, however, clear: they discredit pure forms of each approach and motivate the search for alternatives. Three problems affecting visionary scholarship are examined first and three difficulties that bricolage encounters are then elucidated. Finally, a more general and altogether more tenuous difficulty raised by both approaches is noted. The first two problems that ensnare visionary scholarship are related. They are the problems of bluntness and incompleteness. Bluntness undermines visionary scholarship when it moves from founding large chunks of law upon some vision to seeking to determine particular doctrinal choices. 29. “It will not surprise the students of English law or of anything English to find that between these . . . a middle way, a compromise, has been found”: Read v. Lyons [1947] AC 156, p. 180, per Lord Simonds (quoted in Salmond and Heuston, above n. 14, 22).

200

 :     For, the better one’s grasp of existing private law doctrine is, the more likely it seems that such doctrine is neutral between the range of visions—moral, political or economic—claimed by visionaries to inform the law. The point belongs to Nigel Simmonds, who puts it thus: “The difficulty of discovering any single moral vision within the law is exhibited by a feature of legal doctrines that I shall call their ‘bluntness’. Rules and doctrines are blunt when, and to the extent that, they do not precisely embody any moral principle. Blunt laws are obliquely related to the values that they serve, and they may be so related to several distinct and perhaps incompatible values. In other cases, the failure of the blunt law to trace moral requirements more precisely is to be explained by the requirements of easy administrability”.30

In addition, doctrinal choices may often be politically and morally overdetermined. Hence, moral theories X, Y and Z, despite their many metaethical differences, might all maintain that a legal dispute must be resolved in favour of option P rather than Q, when that choice is cast in sufficiently abstract terms. If this is true across a wide range of doctrinal choices, then particular visions do not do the work that they are claimed to do. This is to be expected if “[t]he law seeks rules which will gain support from the convergence of distinct moral theories on particular issues, and which will be relatively easy to administer. Beyond the scope of such rules, it may seek to preserve its ambivalence between different moral theories by using vague formulae as a framework for more or less ad hoc decisions. The prospects of discovering an abstract and coherent moral theory in the law may look remote”.31

The problem of incompleteness undermines the effort to found doctrine upon a single, all-embracing and finely detailed vision. It is the converse claim to that about the bluntness of doctrine, since it maintains not that legal doctrine is blunt between visions, but that visions can be so incomplete as to be unable to determine doctrinal choices. Visions are incomplete insofar as they require no single legal choice in a particular situation, when they “seem inadequate to yield the necessary richness of . . . [a] legal system”, stopping “twenty feet above the ground” of doctrinal problems.32 This is a problem, since some instances of the visionary approach assume that the adopted vision is sufficiently detailed to fill whatever gaps and resolve whatever ambiguities exist in doctrine. But why should we expect a moral or political vision to be more determinate than legal doctrine itself? More 30. N. Simmonds, above n. 24, at 12; emphasis mine. 31. Ibid., at 21; emphasis mine. 32. C. Fried, “The Artificial Reason of the Law or: What Lawyers Know” Texas L R 60 (1981) 35–58, at 37 and 57.

201

  generally, what reasons do we have to expect the resources of moral deliberation to be more abundant and detailed than doctrine? It seems unlikely that there are any such reasons when we bear in mind that everyday moral deliberation usually operates at the level of rules of thumb or general intuitions; it could operate no other way if it is to be easily teachable. Yet such intuitions and rules of thumb are often unhelpful when faced with new or exceptionally difficult cases. Of course, it is at this point that professional philosophers often appear on the scene with more refined, detailed and systematic guides to moral deliberation at hand. And it is at this point that their guidance is dismissed by other philosophers as being, inter alia, too legalistic, overly obsessed with obligation and thus untrue to moral life. That the charge of legalism is aimed at such moral theories might itself be testament to a more general point, namely, that morality sometimes depends upon law for its own determinacy but should not merely mirror either the form or substance of law.33 The incompleteness problem looms even larger when specific cases are considered. Taken solely as a moral problem, is the solution to the Bland case any more obvious and determinate than the legal solution?34 Even if we set aside acutely problematic cases like these (they should probably be labelled “superhard” cases), and concentrate on “ordinary” hard cases, the question can still be asked: is there an answer in, for example, the Philosophy of Right or the Nicomachean Ethics, to McLoughlin v. O’Brian?35 Whatever power there is in this point hits at the visionary effort to solve particular cases by reference to the adopted vision. It does not impact upon the task of deriving some large area of the law from, or showing it to be compatible with, some vision or other. One form of visionary scholarship—the economic analysis of law—might be thought to have the resources to cope with the incompleteness problem. The notion of efficiency, which is the fulcrum of such work, seems admirably determinate and wide-ranging. It appears useful in generating specific answers to a vast range of doctrinal questions and can be used as a decision procedure far beyond the confines of law. The determinacy of the notion of efficiency is, however, somewhat chimerical once it is recognised that there are number of different versions of that idea that can be invoked and that each can provide different guidance in the same cases.36 Furthermore, the move from Pareto or 33. An unsurprisingly helpful discussion of the ways in which morality may be dependent upon law is found in T. Honoré, “The Dependence of Morality On Law” Oxford Journal of Legal Studies 13 (1993) 1–17. 34. Airedale NHS Trust v. Bland [1993] 2 WLR 316. 35. [1983] 1 AC 410. 36. See J. Coleman, Markets, Morals and the Law (Cambridge: Cambridge University Press, 1988), chs 3 and 4.

202

 :     Kaldor-Hicks versions of efficiency to vaguer conceptions of efficiency as wealth-maximisation not only makes the guidance generated less determinate, but also significantly undermines efficiency’s normative appeal.37 The third problem effecting visionary approaches is that of plurality. The issue is well characterised by Roberto Unger: “It would be strange if the results of a coherent, richly developed normative theory were to coincide with a major portion of any extended branch of law. The many conflicts of interest and vision that lawmaking involves, fought out by countless minds and wills working at cross-purposes, would have to be the vehicle of an immanent moral rationality whose message could be articulated by a single cohesive theory. The dominant legal theories in fact undertake this daring and implausible sanctification of the actual, and the unreflective common sense of orthodox lawyers tacitly presupposes it. Most often, the sanctification takes the form of treating the legal order as a repository of intelligible purposes, policies and principles, in abrupt contrast to the standard, disenchanted view of legislative politics”.38

This point highlights the oddity of believing, as many lawyers seem to, two claims about law. One is that a plurality of moral, political and other normative views are in play in the legislative law-making process,39 the other that the law so created is morally or politically or normatively univocal. Nevertheless, it might be maintained that once the proposed law takes legislative form questions about its normative basis are thereby fixed and univocality assured. Even if this is not so, the point may be thought irrelevant to private law since large portions of it are judge-made. The point can, however, be extended to accommodate judge-made law without too much strain. For, while what is true of the production of statute law is not true of the moment of decision in particular appellate court cases, such decisions are nevertheless to some degree wrought in the crucible of disagreement. Over and beyond the presence of dissenting judgments in these cases, there is often a background of doctrinal and lower-court controversy to them. True, this background looks nowhere near as contested or fluid—in the sense that there are fewer moral, political or other positions in play—as the political process, but it can persist for much longer than political controversy about specific legislative initiatives. This is plain once the legal and political controversy set in play by decisions such as Lochner v. New York and 37. R. Dworkin, A Matter of Principle (Oxford: Oxford University Press, 1986), chs 12 and 13. 38. R .Unger, The Critical Legal Studies Movement (Cambridge, Mass.: Harvard University Press, 1986), 9. 39. I don’t mean to imply that these are the only considerations in play; prudence and self-interest surely have a role, too.

203

  Bartonshill Coal Co v. Reid are noted.40 In light of such constestation visionaries maintain, somewhat implausibly, that the law is (at the weakest) consistent with or (at the strongest) derivable from a single normative vision. While this claim is fairly implausible when made about some moral or political vision, it is wildly implausible when the vision is question consists solely of an account of economic efficiency. Although most versions of that notion promise an admirable but ultimately chimerical determinacy, they are profoundly reductionist when taken as guide to the whole field of normative deliberation. Every value—moral, political, prudential or other—cannot without significant distortion be reduced to a cipher or echo of some or other account of efficiency. The first of the three difficulties that bricolage courts is a denial of the interpretative attitude. Conceived as bricolage, both law and legal scholarship threaten to become too unreflective, practices apparently entailing unquestioning adherence to tradition. Bricolage flourishes, perhaps is only possible, before the advent of the interpretative attitude.41 This point comes vividly to light when we realise that the bricoleur seems trapped within his context, unable to look beyond without becoming something other than a bricoleur. The idea of the bricoleur is taken from the structuralist anthropology of Claude Lévi-Strauss. He invoked it so as to explain how, while the grammar of myth, like the grammar of language, remains constant, it is nevertheless possible to develop apparently new myths and to speak in an infinite variety ways. The work of the bricoleur explains this development: the “new” myths, just like “new” instances of speech, are but re-arrangements of pre-existing components. The bricoleur cannot transcend her context, so the novelty of her myth story or speech is always limited to reorganising or redeploying what is already available: “[t]he characteristic feature of mythical thought is that it expresses itself by means of a heterogeneous repertoire which, even if extensive, is nevertheless limited. It has to use this repertoire, however, whatever the task in hand because it has nothing else at its disposal. Mythical thought is therefore a kind of intellectual ‘bricolage’. . . .”42

The bricoleur’s “universe of instruments is closed and the rules of his game are always to make do with what is at hand”.43 40. Respectively 198 US 45 (1905) and (1858) 3 Macq. 266. On Bartonshill and the common employment cases see Markesinis and Deakin, above n. 14, 693–6. M. Horwitz provides a fine discussion of Lochner in The Transformation of American Law 1870–1960 (New York: Oxford University Press, 1992), ch. 2. 41. See R. Dworkin, above n. 23, 46–9. 42. C. Lévi-Strauss, The Savage Mind (London: Weidenfeld, 1966), 16–17, emphasis mine. 43. Ibid., 17, emphasis mine.

204

 :     To elucidate the work of the bricoleur, Lévi-Strauss contrasts it with that of the engineer. “The difference is . . . less absolute than it might appear. It remains a real one, however, in that the engineer is always trying to make his way out of and go beyond the constraints imposed by a particular state of civilisation while the ‘bricoleur’ by inclination or necessity always remains within them”.44

The impossibility of going beyond these constraints and remaining a bricoleur is a function of the fact that “the elements of mythical thought . . . lie half-way between percepts and concepts. It would be impossible to separate percepts from the concrete situations in which they appeared, while recourse to concepts would require that thought could, at least provisionally, put its projects (to use Husserl’s expression) ‘in brackets’”.45

The point that thought cannot put its projects in brackets presumably means that bricoleurs cannot distance themselves from their context.46 The difficulty that bricolage so conceived presents is that we are undoubtedly children of the interpretative attitude. We think it a good thing that any living practice has a sense of itself which includes both an awareness of its point and its history. We think public practices, institutions and the agents who work within them should—perhaps even must—have a sense of their point and function and must also constantly keep this in view. Practices, institutions and actions can and should be revised in light of their point; they can and should be functionally compared to other actual or conceivable practices, institutions and actions. They can and should be transparent. Bricolage suggests more (perhaps even unwitting) deference to existing practices and institutions than the interpretative attitude could tolerate. Of course, this is not an argument about the logical or conceptual incoherence of bricolage. Rather it denies the viability of bricolage as an understanding of one of our most important public institutions at this historical juncture.47 More closely related to the kind of understanding now had of such public institutions is that they have some blueprint, some point or mission in light of which they are understood and to which they give expression. Bricolage implies that private law has no such blueprint, no coherent moral 44. Ibid., 19, emphasis mine. 45. Ibid., 18. 46. See also D. Kennedy, “Critical Theory, Structuralism and Contemporary Legal Scholarship” New England LR 21 (1985), 209 at 255–9. 47. Some of our current expectations about the rationality of public institutions are sketched by B. Williams in Ethics and Limits of Philosophy (London: Fontana, 1985), 17–18, 100–1 and 197–8. If he is right about them, then they constitute a hostile environment for bricolage.

205

  or political or normative vision or function and this is its second difficulty. This lack of a blueprint or map for private law implies that its immanent rationality is simply a matter of pick and mix, of judges patching it up—from an admittedly limited and usually ancient class of materials and tools—as they go along. Adjudication is therefore a job for tinkers rather than tailors. This fits well with the mythology of the common law—its life is allegedly ruled by experience not logic—but note that this view is both silly and disappointing. The silliness arises from the supposed separation of logic or rationality, on the one hand, and experience, on the other. It is as if the former were possible without the latter. The disappointment comes from the fact that we might expect more by way of logic and rationality in one of our most important social institutions; those educated in civilian systems, where law has been part of the university curriculum for centuries, certainly do. Why should common lawyers be satisfied with less and why think it is a virtue to be so satisfied? If law has no overarching moral claim upon us that should give us pause. Such a claim might be a condition for the law to be authoritative, although the issues here are complex and various, including the different questions about the authority of law in general, the authority of particular segments of law, and the legitimacy of a particular decision within some such segment.48 These questions become separable when it is conceded that there is no general obligation to obey the law but only obligations to obey those parts of it that are genuinely authoritative. Now, if private law articulated some morally respectable vision, the conclusion that this segment of law and, ceteris paribus, particular decisions within it, create an obligation to obey might be easier to reach than if no such vision were incorporated. The conclusion does not follow necessarily insofar as moral acceptability is not the only conceivable requirement for legitimacy. Moreover, if we somewhat artificially set aside questions of authority and focus instead solely on the other-regarding aspect of judicial decision-making—it impacts on the important interests of others —then its moral dimension is obvious. Does it not therefore require some kind of moral justification, which might be more easily provided were the doctrines that judges apply morally defensible? And, could such a justification hail from a practice conceived as bricolage conceives it, seemingly immune to demands for justification? It might be replied that private law conceived as bricolage can surely make an overarching moral claim on us and that claim could be founded upon a 48. I am taking authority and legitimacy as synonyms and thus, for present purposes, authority refers to genuine, not perceived authority. A genuinely authoritative or legitimate directive creates or states compelling reasons to act in accord with it.

206

 :     variety of moral theories. That is to say, just as the problem of bluntness elucidated the possibility of the over-determination of doctrinal choices, so it could be said here that the moral claim law makes upon us is similarly overdetermined. But it seems that this claim cannot be explicitly made from within the practice itself if it is to remain a pure form of bricolage, free from the demands of the interpretative attitude. For if the bricoleur judge is indeed trapped in context, how are we to conceive of her resort to political, moral or other arguments to support moves within that context? This appeal beyond the resources of the practice, if permitted, suggests the practice is not as impermeable as bricoleur scholars are inclined to maintain. Once regarded as permeable in this way, bricoleur scholars have the difficulty of categorising the various sources of non-doctrinal arguments and analysing the occasions of their use and their effect. Such arguments could, of course, become just another set of tools within the practice which judges use in the same way that they use other, more obviously doctrinal tools. The suspicion, though, must be that once the practice is open in this way, the interpretative attitude and all that it entails will soon take hold. Bricoleurs that look beyond the horizon of their practice are unlikely to remain bricoleurs for long. The problem that both bricolage and visionary approaches face is that they risk turning law and adjudication into something other than themselves. What, then, are law and adjudication? At least this: means of subjecting human conduct to the governance of rules, those rules having some presumptive legitimacy which extends to decisions as to what, exactly, the rules require. Law and adjudication are therefore means of legitimately resolving disputes with some degree of finality or closure. Visionary scholarship and practice can serve to turn law and adjudication into something other than this, primarily, but not exclusively, politics and morality. In so doing, law and adjudication lose the capacity for finality and closure that might mark them out as distinctive social forms and which can provide them with a rationale in circumstances of normative pluralism. The capacity is lost because law and adjudication become means to a particular end or a range of ends on the visionary view. The problem which arises is that the chaos of normative dispute in pluralistic societies testifies to the absence of a method for determining which end or ends the law embodies. In turning law and adjudication into politics and morality, visionaries ensure that both solve disputes only by the temporary fiat of voting, either in the court room or the legislature. Law and adjudication therefore only replicate, while sometimes temporarily postponing, such wider normative disputes. Bricolage does not risk turning law and adjudication into politics and morality. Rather bricolage undermines the role of law and adjudication in 207

  subjecting human conduct to the governance of rules, those rules having some presumptive legitimacy extending to decisions as to what the rules require, by turning both into simple invocations of tradition. In some societies the invocation of tradition carries substantial justificatory weight and often pre-empts questions as to the legitimacy of some decision or action. Disputes can still arise in such a context but they must be formulated in terms of what tradition actually requires. Of course, such disputes can include reference to the values constitutive of a particular tradition but these are understood as being in no sense autonomous of the tradition or practice. The problem here is that in a context in which tradition has very little presumptive justificatory weight, reliance upon it in law and adjudication undermines their legitimacy. If law and particular case decisions have no moral or political appeal beyond the fact that they might cohere with past decisions or practice, then those ensnared by the interpretive attitude are almost certain to find it unsatisfactory. What power (rational, moral, political or otherwise) does the bricoleur’s solution to a legal problem have, other than that it is one cobbled together from the ancient materials at hand? Take the image of the bricoleur as repairman seriously. He takes “whatever is available to patch up the immediate problem, while avoiding any premature judgement on what further and related problems may arise (chewing gum for the pipe and an electric fire for the bedroom: but no general replacement of old pipes, and no installation of central heating)”.49

It seems fairly obvious that now, with the expectations about the legitimacy and rationality of public institutions that dominate our culture, the bricoleur judge only brings law and adjudication into disrepute. The bricoleur repairman would, one hopes, be prevented from trading by consumer trading standards officers. It would be foolish to maintain that this argument, which aims to show that the approach of both visionary and bricoleur face a similar difficulty, was either complete or unproblematic. The minimal functions of law that both might undermine are not incontestable. Nor are they without content. They are, though, in a sense neutral, since the goal of legitimate dispute resolution has little to say about the substance of the values that inform particular segments of law save that the less controversial they are, the better. Each of these issues requires much more attention than they have received here. However, the point is clear: while visionaries run the risk of turning the law into fullblown moral-cum-political argument, bricoleurs undermine its normative 49. N. Simmonds, above n. 24, at 24.

208

 :     appeal by turning it into nothing but tradition. Both snags can be avoided if visionaries become modest and bricoleurs become ambitious. Modest visionaries and ambitious bricoleurs accept that law is founded on tradition, must have some minimal normative appeal and that it makes a claim to legitimacy founded on both. Before examining the form this via media can take in private law scholarship, two arguments for maintaining the distinction between visionaries and bricoleurs are considered.

2. Maintaining the distinction a. Horses for courses

The force of this argument is that what suits one horse will not necessarily suit another. The horses are judges and scholars and the argument holds that while visionary work should be the modus operandi of the latter, it is inappropriate for the former. Judges must be bricoleurs and scholars visionaries. Why? The argument might begin from a claim about the harmful consequences of judges directly pursuing whatever normative vision supposedly underpins the law. Judges might, for example, come to overlook the justice of particular cases before them in their haste to track and develop the underlying vision. Violence might be done to doctrine and uncertainty and incoherence result from the pursuit of some or other vision, especially so when judges are likely to interpret the underpinning vision differently. By contrast, scholarly work designed to elucidate the vision underpinning the law is unlikely to have these consequences. Further, provided this type of work is done, and provided it constitutes part of the ethos in which lawyers are educated, it might well be that judges will end up tracking the vision by something akin to an invisible hand process or an indirect consequentialist argument. The problem with this argument lies in its insistence upon a close but unintentional correlation between vision, on the one hand, and the overall outcomes of cases, on the other. While it is not in general difficult to imagine individual intentional actions unintentionally exhibiting a pattern, it is more difficult to accept that this is a satisfactory process when the pattern in question is some normative scheme. Why? Because such a scheme is often taken to have distinctive, but not always conclusive weight in our practical reasoning and leaving its achievement to chance seems unsatisfactory. If it be replied that the role of chance is minimal here, then some account is needed of the mechanism by which individual actions (judges deciding cases) are guaranteed, or very likely, to bring about the vision in question. Once it is 209

  noted that such a mechanism cannot operate via the reasons, beliefs or attitudes of the agents in question, this appears a demanding task. It will not be considered further. b. Remedying bluntness

At least two arguments fall under this heading. They aim to maintain the distinction by showing that the work of visionaries and bricoleurs is different but mutually supportive. Both arguments focus on the bluntness of law and maintain that this feature actually makes visionary work necessary. If some segment of the law is indeed blunt or neutral between one or other normative vision, surely it makes sense for scholars and judges to strive to sharpen the law so as to better embody that vision. Rather than wring hands in the face of bluntness, visionaries should get to work: their work remedies the instances of bluntness revealed by scholarly and judicial bricoleurs. However, only a very modest, perhaps even emaciated kind of visionary work will help here; as such it hardly deserves the visionary name. By contrast, standard visionary work will serve to sharpen the law’s normative edge only at the cost of vastly increasing its normative contestability. The standard visionary response to bluntness will be examined first. For the sake of simplicity, assume that the visionary work in question is moral–philosophical. The response offers the following counsel: a deeper appreciation of one or other moral theory (hereinafter “the donor theory”) allows a better understanding, and more complete elucidation, of the value or values underpinning or embodied in some area of law. Moreover, where the law seems to embody or be underpinned by more than one such value, a good donor theory will allow a ranking of such values so that conflicts can be avoided. (Note that donor theories that are economic rather than moral, or even those moral theories that can be subsumed under “utilitarian” label, can often avoid ranking, but only because they are profoundly reductionist. When it is maintained that the vast array of superficially different values are in fact simply particular instances of one meta-value, or are reducible to some conception of utility or efficiency, then ranking is otiose). The donor theory will be complete in the sense that it operates both at the value-talk and meta-ethical levels; that is, it offers prescriptions as to how to live and be, while giving some account of what it is that confers rational power or validity upon those prescriptions. Indeed, the donor theory will usually seek to derive value-talk propositions from meta-ethical propositions. The ability to generate helpful value-talk prescriptions, to offer determinate guidance as to what moral or political right and wrong entails, is usually taken as an indicator 210

 :     of the power of a meta-ethical position. Now, it might be true that some metaethical theory can sharpen the law’s moral content in the two ways noted, founding the law more precisely upon some or other value or ranking the values that provide its foundation. Insofar as the values specified and/or ranked are directly implied by a single meta-ethical position, this sharpening of the law’s moral focus will unavoidably increase the law’s moral contestability. It could only fail to do so were there only one viable meta-ethical position, only one rationally defensible donor theory, available to provide the law with sharper moral focus. And while we cannot rule out a priori the possibility of some such irrefutably correct meta-ethical position being developed, our current situation is that there is a range of such positions available, no one of which is obviously more salient than the others. This response to bluntness roots the law in a contestable metaethical theory, delivers it over to a particular meta-ethical faction, which can only lead other meta-ethical factions to dispute the law’s moral content. A bullish visionary might grasp this nettle. If giving the law a more pronounced moral focus makes it slightly more morally contestable, then so be it. That is a small price to pay for the obvious gain, which is that a better appreciation of the moral content of the law will lead to more consistent patterns of judicial decision-making. The point here is an old and plausible one: that in order to apply and interpret rules in anything but the most unproblematic situations, some account is needed of the point, purpose or value of the rule. To determine, for example, whether a memorial jeep breaches a prohibition on vehicles in the park, some account of the point of the prohibition is needed.50 The visionary offers an account of the moral value or values underpinning some area of the law and in so doing seemingly makes a high level of consistency in adjudication possible. If the law is left blunt, is taken to embody no particular value or ranking of values, then judges might well decide the same case differently because they have different views of the point, purpose or value or the law or doctrine in question. Judge X might prohibit the jeep on public safety grounds for, although it cannot move, it presents a far greater hazard to the primary users of the park, namely, children, than properly designed swings, climbing-frames, etc. Judge Y might allow it because whatever the rule in question is intended to achieve, it is surely the protection of those in the park (a recreation area) from the noise and danger of fast-moving traffic. The visionary hopes to remedy the bluntness of this rule by reference to some moral position. That position tells us which account of the point, purpose or value of the rule or doctrine in question dominates in interpretation and application. 50. L. Fuller, “Positivism and Fidelity to Law: a Reply to Professor Hart” Harvard Law Review 71 (1958), 630–72 at 663.

211

  Two points must be noted about this. First, appellate court judges do indeed often decide the same case differently and do so because they disagree about the point, purpose or value of the rule or doctrine in question. There would be a good deal less consistency in the law than our visionary desires but for one simple fact of institutional design, namely, that appellate courts are almost always constituted by an odd number of judges employing a convention of majority-rule. What consistency exists is clearly not solely a function of some or other moral theory. Second, our visionary seemingly imagines that the value or values mandated by their theory not only resolve the bluntness of legal rules and doctrines but are also not themselves blunt when particular judicial decisions have to be made. Visionaries seemingly expect not only to repair the bluntness of legal rules (e.g., those constitutive of contract law) by articulating the value or values upon which the rules are correctly founded (e.g., autonomy), but also, in so doing, expect those values to hold sway when a specific decision about the rules (e.g., can doing what you are already contractually bound to do be good consideration for a subsequent promise?) must be made. This latter expectation is problematic for the same reason that the visionary’s belief in the completeness of their donor theory is problematic: why expect a moral theory to be any more determinate than the law itself? A second response to bluntness is a variation of the first. It, too, is primarily a moral–philosophical response. Rather than remedying bluntness by means of elucidating a meta-ethical position and the value or values derived from it, this response operates solely at the level of value-talk. The visionary is still important since her job is to make plain the values the law embodies and which bricoleurs might overlook. But the values so highlighted need not be, and on this view of visionary work are not, derived from any particular meta-ethical position. Rather, the values are of such a type or nature as to be convivial to all or almost all meta-ethical theories. Unlike the first response to bluntness, this one sharpens the law’s moral focus without increasing its metaethical contestability. The values elucidated must be fairly abstract and general but also indubitably coherent with the legal rule or body of doctrine in question. It could, for example, be said that the value of bodily integrity informs large chunks of the criminal law and that such a value is more than likely to be accommodated by any meta-ethical theory and account of moral right and wrong. Similarly, it is hard to deny that the value of autonomy does indeed inform many doctrines of the law of contract and that such a value will have a place in just about any meta-ethical and moral-cum-political theory. Notice what an exceedingly modest version of visionary scholarship this is. It refines or elucidates the values underpinning the law without any meta-ethical argument as to how they should be ranked, for that path invites controversy. 212

 :     Hence, if more than one value is in play, judges can no doubt rank them differently in particular cases and thus arrive at different decisions. Nor can this response, any more than the first response, claim that knowledge of the values informing the law will determine the choices judges have to make in particular cases for, as already noted, there is no reason to think visions are any more complete than law. Undoubtedly it is important and helpful to reflect upon the values some area of law might embody. These reflections might inform practice and probably actually do so through the socialising process of education and admission into the profession. Highlighting these values, elucidating what they might demand and how they might conflict in particular circumstances, is a long way from reading the law in light of some meta-ethical theory and account of moral right and wrong. In that process the donor theory is a mould around which the law is shaped, whereas the current response accepts that there might be no single mould. Furthermore, it holds that the values informing the law must undoubtedly be embodied in it, so that the law is not just bent around preexisting autonomous values, and that those values and the law they inform might be compatible with a number of meta-ethical moulds. Although the second response to bluntness could be regarded as a very modest type of visionary scholarship, it could with equal propriety be dubbed an ambitious form of bricolage. For, in attempting to unearth and articulate the values underpinning the law while simultaneously remaining close to the law, this response avoids the snags besetting bricolage. It provides the law with a limited normative blueprint, showing that the work of adjudication is neither simple-minded tinkering nor full-blown tailoring. A pattern of values is provided which, because neither complete nor conclusive, partially constrains the nature of any future repair work. The work of modest visionaries or ambitious bricoleurs is examined further below.

3. The middle way: modest visionaries and ambitious bricoleurs

The modest visionary accepts that law has its own existence, that its doctrines and contours are not, and need not be, an exact reflection of any particular normative vision. While law is not merely a cipher of some vision, the modest visionary maintains that law is a self-reflective practice, that it is prone to contemplate its point and value. Such critical self-reflection is part of the initiation into the discipline, since one aim of legal education is that of engendering an appreciation of the history and point of various legal doctrines, as well as of law in general. Self-reflection can be triggered and informed by ideas and forces outside the law. It is not, after all, a fully 213

  autonomous segment of the social system nor is it completely porous to other aspects of the social system. Such self-reflection can be informed by a range of moral, political, economic and other arguments; it is here that the modest visionary does her work. She does not seek to reinvent some area of law in the image of an adopted vision and its supporting theory but supplements the law when and where appropriate and necessary. On the whole, appropriateness and necessity are determined by the agents embedded in the practice itself, although the effect of extra-legal influences should not be overlooked. Law has some degree of legitimacy on this view since parts of it are indeed consistent with a range of diverse normative values, as the moderate visionary is only too keen to show. Because the normative theoretical work she does here is low level, a matter of repairing gaps or supplementing specific decisions, it need not—perhaps cannot—portray the law as a reflection of a single, allpervasive moral vision. If many legal doctrines are blunt, then the gaps that are filled by recourse to some theory and its cognate vision are just as likely to be filled in the same way by a number of theories and their constitutive visions. Unsurprisingly, the ambitious bricoleur also maintains that the law has an existence independent of theories and their constitutive visions. Unlike the ordinary bricoleur, she sees that the law’s claim to legitimacy partially rests on its coherence with various normative values and beliefs. The ambitious bricoleur does not, then, simply resolve disputes by using whatever materials come to hand regardless of their normative status. If law were autonomous in that way, legitimacy would constantly be beyond it. Hence the ambitious bricoleur adopts only those solutions that add to the law’s normative veneer and thus bolster its claim to legitimacy (always assuming, as we have throughout, a connection between normative kudos and authority). This does not involve reinventing the law in light of a particular, all-embracing normative vision, since the opportunity for normative engineering arises only at the level of specific doctrinal repair. A blueprint of the moral, political or economic basis of the law, or some chunk of it, is not required from the ambitious bricoleur; what is needed is a solution to a specific legal dispute. The ambitious bricoleur, just like the modest visionary, accepts that the likelihood of such a solution being supplied by one and only one normative vision and its supporting theory is low. Moreover, the solution is likely to be normatively over-determined, in that it can be generated by a number of theories, or completely underdetermined, in the sense that no theory has an answer. Honoré’s legal philosophical work displays the hallmarks of the modest visionary or the ambitious bricoleur. From his essays on causation to those about omissions, responsibility and luck, there is a concern to articulate, 214

 :     wherever possible, the overall coherence of the legal doctrine before him. Yet his work is undoubtedly theoretical—the theory being drawn from analytical philosophy—and explores the moral, political and metaphysical premises of private law. Theory is, however, invoked sparingly and usually only as a means of elucidating doctrine or of making it intelligible where it has been shown to lack coherence. The “reading in light of ” characteristic of the visionary approach looks premature from Honoré’s perspective, since he must be persuaded that there is indeed a problem which makes some philosophical supplement necessary. The modest visionary does not assume at the outset, and under the influence of some theory or vision, that the law lacks coherence and stands in need of theoretical reformulation. If there is no snag, then why look beyond doctrine itself? When there is a perceived to be a snag, Honoré’s response takes doctrine seriously. Two examples serve to show this. The first returns us to Honoré’s discussion of strict liability. Those who read criminal and civil law in light of some donor theory which supports the intuitions, prominent in some areas of our moral thinking, that fault and actual capacity to do otherwise are preconditions of responsibility and liability, dismiss strict liability as aberrant. And so it is, when assessed against these particular intuitions. This strategy seems a little hasty, though, for strict liability is undeniably part of both areas of law and its range, as Honoré shows in his discussion of negligence, is wider than is often assumed. Strict liability is also in play in contract since the general rule is that contractual duties are strict and strict liability is, of course, the hallmark of many regimes of product liability. This should give us pause. While neither modest visionary nor ambitious bricoleur maintains that the law cannot contain theoretically or normatively dubious doctrines, both hesitate before classifying whole swathes of law as dubious. Both hold that some effort beyond convicting the doctrine by reference to some donor theory is necessary. It is just this that Honoré provides in chapter 2 of Responsibility and Fault. Having examined the difficulties of standard accounts of strict liability, Honoré arrives at the principle of outcome responsibility as both explanation and justification. Our interest is not in the plausibility of that principle—although it seems to do what other accounts of liability responsibility have failed to do, namely, make normative sense of strict liability—but in the form of the argument. The argument is not an instance of “pure” bricolage since it does not unquestioningly accept the propriety of strict liability. The bricoleur might, simply because strict liability is a well-entrenched part of the law. Indeed, the argument is quite unlike any we would expect from a bricoleur because it is a journey of philosophical and justificatory ascent, from doctrine to potentially powerful underlying account of responsibility and related notions of personhood and 215

  identity. Nor is the argument visionary in its purest sense, since, as we have noted, in this context such arguments rarely take strict liability seriously: they are quick to judge and slow to understand. It is in the course of taking the doctrine seriously that Honoré is lead to a strategy of justificatory ascent. The second example is liability for omissions. In this case the distinction between doing and refraining is well entrenched in the law and causes difficulty for visionaries but discomfits bricoleurs not a jot. The latter are, indeed, often unaware that the distinction is in need of support whereas for the former a number of substantive moral-cum-political positions maintain that the distinction is morally without foundation. Again, Honoré takes the doctrine seriously, but not unquestioningly, the principal task of chapter 3 of Responsibility and Fault being an exploration of the intelligibility of the distinction. The inclination to take our practices and institutions seriously, at least until such time as they have been shown mistaken, is part of the bricoleurs’ mentality which Honoré characterises well: “[i]f a distinction such as that between doing and not-doing is prominently marked in thought and language it is sure [likely?] to mark a difference in social evaluation too, and that difference is not likely to be arbitrary”.51

What Honoré shares with the visionary scholar, though, is sensitivity to the need for justification, but justification is tied, wherever possible, to the contours of the practice or doctrine in question. Justification works upwards from doctrine and practice rather than downwards from some donor theory or other. It is not necessary for modest visionaries that doctrine be a less abstract instantiation of the donor theory. Rather, when viewed from such a theory doctrine might well be blunt. It is once again worth repeating that neither modest visionary nor ambitious bricoleur assume that doctrine and practice must be normatively justifiable; that Whiggish assumption is not necessary for their enterprise to make sense. Honoré’s more general legal–philosophical work is also testament to the fecundity of the approach of the modest visionary or ambitious bricoleur. With Herbert Hart in Causation in the Law he defended and recommended an account of causation rooted in ordinary language and legal doctrine. The defence worked on two fronts: against those who would convert the legal conception of causation into a surrogate for policy decisions and those who would reinvent it in a philosophically pure form.52 This strikes just the right modest (or ambitious) note. Hart and Honoré’s discussion is, of course, open 51. Responsibility and Fault, above n. 1, 48. 52. See H.L.A. Hart and T. Honoré, Causation in the Law (2nd edn, Oxford: Clarendon Press, 1985), Preface.

216

 :     to philosophical reflection; they are not trapped within the horizon of the practice “law” as is a true bricoleur. Their world is not a closed one. Yet their account of causation is rooted in everyday legal and common-sense causal discourse, that discourse being held up for repair by philosophical analysis when and where necessary. That discourse is not simply read in light of some set of philosophical assumptions, being regarded as an incomplete expression or mere cipher of them. It is taken seriously in its own terms. The question articulating the investigation is: can this body of doctrine stand on its own two feet? Modest visionaries allow for a possibility that true visionaries do not, namely, that the practice, institution or discourse in question may have its own internal intelligibility, may not need to be read in the light of something else or only intermittently so. It might seem odd to say that the kind of legal–philosophical work just noted can provide the basis for a general theory of law, since the idea of a modest visionary or ambitious bricoleur formulating such a theory is curious. This, it might be thought, is a job for visionaries. However, it is not the propensity to undertake this task that distinguishes visionaries from others, but rather the substance of their account of the nature of law. And the substance of Honoré’s account is modest and his hopes for law are minimal, but nevertheless normatively significant. The general theory looks like this: law is founded upon a basic norm of co-operation, rather than something like Hans Kelsen’s formal or near content-free basic norm. The drive towards a formal foundation for law is understandable if one wants, like Kelsen, “to show that, given it is impossible to secure agreement on moral and political principle, law can be regarded as an autonomous system of social control, independent of morals and politics”.53 The argument only seems to work by employing a standard bricoleur’s trope, namely, by ignoring the claims of these other realms and discourses and the request for justification. The contemporary legal bricoleur does not arrive at a position about the autonomy of law as a result of reflection about the interplay of different realms of regulation, nor as a result of an argument about the harmful consequences of turning law into morality or politics. The bricoleur, remember, is trapped in context and so is likely simply to assume that these realms are just not relevant to his work. He needs neither philosophical seer nor prophet to go about his business. By contrast, from the perspective of the visionary, a justification for law’s normativity is essential. Honoré sees this and yet, rather than root law’s normativity in some general normative vision of how to live and how to be (a visionary trope, this), he hits upon a much more modest approach. It is not 53. T. Honoré, Making Law Bind (Oxford: Clarendon Press, 1987), 94.

217

  formal, in the way Kelsen’s grundnorm is, nor is it morally or politically “thick”: it “is the norm which prescribes that the members of a society have a duty to co-operate with one another”.54 As a vision of how to live and be, this is startlingly modest. But it is a vision attuned to the contestability of visions, its modesty a recognition that anything more may imperil the goals we need law to achieve. Law is not the kingdom ends, not a vessel loaded with our loftiest moral and political ideals, but a useful means for us to get along. That is not to claim that it is morally neutral or contains no hints of such ideals. It clearly overlaps with other normative systems and is therefore open to their influence. The crucial point here, as any modest visionary or ambitious bricoleur will recognise, is that law “overlaps with” these other systems. It does not replicate them, either in the form that citizens hold them or as unpacked and developed by moral and political philosophers, economists or social choice theorists. As we have hinted, there may be good reasons for this.

III. C

The tension between the two styles of scholarship examined here is hardly new; it most likely embodies our different, sometimes incompatible hopes for law and these are undoubtedly as old as law itself. The language used to describe the tension is not significant; the crucial points are whether or not it is indeed genuine and, if so, how it might be dealt with. Honoré’s work has been used to illustrate a way of going between and beyond the tension; his work is an exemplary instance of the kind of scholarship required if a philosophically engaged and legally meaningful understanding of law is to be achieved.

54. Honoré, Making Law Bind, above n. 53, 111.

218


9 APPRECIATIONS AND RESPONSES Tony Honoré A merit of the papers read at the Canberra and Columbia symposia and now published in a revised form is that they do not look back. They seek instead to develop, modify, or improve on ideas that have appeared in my work over the last fifty years. No scholar does more, in my view, than put items on the agenda for others to debate. Here are some comments on the way in which the contributors to this volume have carried the debate forward. I. C

Michael Smith’s and Philip Pettit’s papers make a good starting point. They relate to the idea that the capacity that morals and law require as a condition of blame or liability is a general capacity to perform successfully. It is not a capacity to have acted differently on the occasion in question. This theory I first put forward in 1964 in an attempt to explain how it can be in order to praise or blame people or hold them legally liable even when, given the circumstances, including their then state of mind and body, they could not then have done otherwise. If the “general capacity” theory is accepted we bypass the objection that determinism, if true, rules out praise, blame and legal liability. That however leaves open further questions. What precisely does the capacity for self-control consist of? How does possessing it license criticism of our failure to live up to our capacities on a given occasion? Michael Smith

Michael Smith’s essay is about self-control, a topic to which he has made a notable contribution. He starts as I did from the view that to blame someone morally or impose a legal sanction on them for what they have done requires not merely that they have done something reprehensible for which they are in a broad sense responsible, since the action is theirs. It must also be the case that that when they acted they had the capacity to reach a rational decision

  about what to do. They possessed the necessary ability to control their conduct. This capacity is a matter of degree, though the law cannot always fully take this into account. The test of whether a person is capable of doing soand-so, and more broadly of controlling his or her conduct, is a general one. Does that person generally succeed in controlling his or her conduct when they try? If so, I argue that he or she has the capacity for self-control. Apart from the last point, Smith broadly agrees with my analysis. On that point, however, to which I return, his view is that to succeed in controlling one’s conduct when one tries is evidence of capacity to control it but is not the same thing as to possess that capacity. He points out that the practice of blaming only those who have rational control over their conduct presents a number of theoretical difficulties. They concern the psychological structures needed for an agent to possess the capacity for self-control. Agents can, he argues, possess that capacity, though they intentionally act in one way when they believe they should act in another. Their desires and beliefs are in that case not sufficiently coherent and unified. (Does this reflect the fact that different parts of our brains and personalities do not always communicate efficiently?) But self-control is nevertheless possible, Smith contends, in one of two ways. Rational agents may foresee and provide in advance for the possibility of their being later out of control. They may also possess a general capacity, independent of their desires, to acquire and lose psychological states in accordance with norms of coherence. A potentially out-of-control desire may be avoided in advance or may trigger its own inhibition, as when people with an overweening desire for chocolate train themselves to think, at the moment when the desire surfaces, of the chocolate as a (repulsive) lump of fat. Smith’s account of the capacity for self-control as a capacity for co-ordination is attractive. It is a normative capacity, a capacity to order desires, beliefs and psychological states in accordance with norms of coherence. It does not follow that if an agent possesses the capacity for self-control, the agent’s desires and beliefs are ordered in a way that rests on socially acceptable norms. The agent’s conduct may be coherent and yet directed to unacceptable ends. It is this that opens the door to the possibility of criminal justice, directed at those whose behaviour is coherent enough to suggest a capacity for self-control but whose exercise of that capacity disregards the interests of others. But what makes it true that we can exercise rational self-control when we fail to do so? My suggestion was that a person can (general) do something if it is the case that when they try they generally succeed, even if it is true that, however hard they tried on the occasion in question, they could not then succeed. Applying this to a person’s general capacity for self-control, a person 220

   possesses this if, when they try to exercise self-control, they generally succeed. According to Smith, though he has some sympathy with my idea, this is to confuse the evidence for the possession of self-control with the metaphysical question of what self-control consists of. He argues that we might be satisfied that a creature, otherwise identical with a real person, possessed self-control even if the creature existed only momentarily. This argument does not convince. We might in that case think that, if the creature had lived longer and had had enough opportunities, it would have exercised (or failed to exercise) self-control and so have provided the necessary evidence. But Smith’s objection nevertheless demands an answer, for it is a standard vice, at least of lawyers, with their concern for proof, to confuse the evidence for the truth of a fact with the fact to be proved. Smith’s suggestion is that the difference between an agent who possesses the capacity for self-control and fails to exercise it and the agent who does not possess the requisite capacity consists in the difference between the possible worlds in which they fail and those in which they succeed. The world in which the agent who possesses self-control fails is closer to the world in which he or she succeeds than is that of the world in which the agent who does not possess the capacity for self-control. The clue to the distinction lies in differential similarity. This analysis has the advantage of explaining why capacity for self-control is treated as a matter of degree. My reaction to this suggestion is rather like Smith’s to mine. I sympathise with it. But what exactly does the reference to possible worlds add? Will one not be driven to inquire into the criteria of difference between actual and possible worlds? What makes one possible world more different from the actual world than another? I end, therefore, with a suggestion, not strictly a countersuggestion, of a normative sort. Whatever the capacity for co-ordination that is a condition of self-control may consist in, those who possess it have a valuable asset, as regards both their self-esteem and their ability to pursue a chosen way of life. If they misuse the capacity or fail to exercise it when that misuse or failure impinges on the interests of others they are not entitled to complain if they are exposed to blame or sanction.

Philip Pettit

Philip Pettit’s neat offering deals with a different aspect of an agent’s capacity to have done otherwise, again on the assumption that the possession of this capacity is a condition of praising or blaming the agent for doing what he or 221

  she in fact does. He contrasts an act-centred approach to the problem with an agent-centred approach. The act-centred approach (could the act have been different?) is objectionable. Either it requires that there be gaps in causal laws, which is inconsistent with determinism, or it assumes that the mere absence of compulsions of the sort that intuitively rule out free agency automatically exposes the agent to praise or blame for his or her conduct—which is not selfevident and needs to be justified. The agent-centred approach is concerned instead to ask what sort of agent the person in question is. What is the agent capable of doing in general? Like cars and horses, people sometimes perform up to standard and sometimes not. If they do not it is intuitively plausible to say that they could have done better. Sometimes, in Pettit’s phrase, they “track standards” and sometimes they do not. And if people consciously track standards this surely shows that they have a general capacity for tracking standards, for adjusting their conduct to meet whatever is required in the circumstances. This train of thought is important, I think, from a legal point of view. What lawyers need to be satisfied about in judging violations of the law is that the agent who is said to have violated it had a general capacity to conform to the standards that the law requires. Does it follow that when people do or do not come up to the standard of which they themselves are in general capable it is normatively in order to praise or blame them? Pettit, along lines pioneered by Strawson, points out that praise or blame will not be justified if it is employed merely as a technique for conditioning the agent. That would be disrespectful and a ground for resentment. But if agents hold themselves out as tracking standards, as able to meet the requirements of ordinary social intercourse, and know that others converse and deal with them on this assumption, they invite praise or blame to the extent that they meet or fail to meet the expectations they arouse. This openness to praise or blame can sometimes be rebutted, since it presupposes that the agent has acted freely in the sense of not being subject to inner compulsions or to manipulation by others. But ordinarily it can taken for granted. Pettit’s argument is persuasive, provided that the capacity he has in mind is taken to be a matter of degree. Presumably we can only know that an agent has the capacity to track standards if that is what the agent on the whole does when he or she tries. This opens Pettit’s thesis to the objection that Smith raises against mine. What exactly is success in tracking standards evidence of? At the moment I see no answer. But a person who benefits from being treated by others as a fit person with whom to converse and deal has in return to accept criticism of his or her failure to meet the standards that he or she expects others to meet. 222

   II. O

Next come four papers (Ripstein, Perry, Cane, Gardner) that are concerned with outcome responsibility and in particular its relation to the just distribution of risk in tort law. Outcome responsibility, as I conceive it, is the idea that certain outcomes of our conduct, settled according to causal criteria, are ours, even when unforeseen or unintended. We identify with them and others attribute them to us. They form a constituent of our individual character and identity, without which we should lack achievements and failures both in our own eyes and in those of others. If we possess a minimum capacity to control our conduct, we are morally responsible for these outcomes, good or bad, though we are not always, in the case of bad outcomes, morally to blame or legally liable for them. The attribution of responsibility in this way is fair, since we benefit from being regarded as rational agents. How far responsibility should carry with it moral praise or blame, or legal liability, depends on other factors, which include the virtues or vices that our conduct exhibits, the likely consequences of imposing liability and the need to distribute risk fairly.

Arthur Ripstein

Arthur Ripstein’s innovative essay “Private Law and Private Narratives” marks a step forward, it seems to me, in our grasp of the theoretical basis of tort law. He thinks that two of the ideas I have put forward—general capacity and risk-distributive justice—illuminate tort liability but that two others—outcome responsibility regarded subjectively and the argument that benefits and burdens should be in balance—do not, at any rate not without serious modification. He points out that outcome responsibility in its autobiographical aspect does not on its own provide a reason for imposing liability. If I stumble on the stairs and knock you over, having knocked you over is part of my history and being knocked over part of yours. Either of us may find the incident significant or trivial in our own lives. The law of tort is not interested in these subjective assessments. If you suffer injury from being knocked over, the question from the point of view of tort liability is on whom the risk of the injury is to be placed. Is it on me, and, if so, only if I am negligent? Or is the risk one that you must bear as an incident of your freedom to move around and interact with others? For tort liability we require that the agent had the minimum capacity needed for outcome responsibility and, in addition, violated a right to which 223

  the victim was entitled under a system of fair distribution of risk. As regards capacity Ripstein concedes that conduct is to be judged by the agent’s general capacity to control his or her conduct and, in particular, to avoid harming others. However, tort law, he points out, does not judge agents by the standard of their own capacities, but rather by that of the reasonable person of ordinary foresight and prudence. This is hard on the “shortcomers” who, though they possess the general capacity to be treated as responsible people, fall short in some respect such as speed of reaction. The justification for adopting the objective standard must rest on something more than outcome responsibility in the autobiographical sense. The something more is an aspect of what I have termed risk-distributive justice. From the point of view of tort law we bear the risk that our bodies or property or other assets will suffer injury unless that injury is one the risk of which is properly allocated to another. Under a fair system of risk distribution, the risk of injuries that others cause to us, when their conduct falls short of a prescribed standard or involves a special danger to others, is borne by them. But that system is based on reciprocity. In return for the advantage to us of others bearing these risks of injuring us, we in turn bear the corresponding risks of injury to them, from which they in turn derive benefit. The transfer of risk is a zero-sum operation and the idea of risk-distributive justice is egalitarian. All risks must fall on someone. Since it would be unfair to make me bear a greater share of the risk of injury to myself simply because you are a shortcomer, you are held to the standard of the person of reasonable foresight and prudence. This is true, if you have at least the minimum capacity to control your conduct in general, even though you are incapable of coming up to scratch in some respect such as speed of reaction. The case for an egalitarian distribution of risk in tort law is that freedom from injury (like, arguably, entitlement to medical treatment) is not something that can be unevenly distributed without injustice. In this it differs from, say, wealth, where inequality can up to a point be tolerated. I agree with this part of Ripstein’s analysis. It is true that what he calls “displacement-responsibility” (my being liable to compensate someone because the risk of my injuring that person was on me) is a narrower notion than that of outcome responsibility in the autobiographical context. It is something specifically legal, because the law alone is concerned to allocate the risk of having to pay for misfortunes. But the cases in which a person is displacementresponsible for something they have done are a subset of the cases in which that person is outcome responsible in the sense that is relevant both to that person’s conception of their own life and character and to its assessment by others. To speak of an “autobiographical” or “personal narrative” interpretation of 224

   outcome responsibility can be misleading. Though it may be, with Nietzsche, that each of us decides how to make their life into an interesting story, this is true, unless we are mentally ill, only within limits. The limits are set, in my view, by objective and interpersonal principles of attribution. These principles tell us, at least approximately, which outcomes are attributable to which acts. It seems to me that Perry is right, as against Ripstein, in regarding “third-person attribution” as essential to the concept of outcome responsibility. It may not be profitable to debate whether it, or “first-person narrative” is the more fundamental. I have reservations about Ripstein’s wish to downplay the role of my argument from the balance of benefit and burdens. Two steps are needed if in tort law an agent is to be held to the standard of conduct of the reasonable person. The first is that agents take on themselves the risk that on a particular occasion they will not come up to the standard of which they are in general capable. Why should they be judged by this standard, unless it is that they benefit from having a general capacity for self-control (Smith) or for tracking standards (Pettit)? The possession of these general capacities shows that, if not coerced or constrained by inner compulsions, they were free to act differently on the occasion on which they failed to come up to scratch. This is an additional argument against ruling out moral blame and legal liability. Both freedom on the occasion in question and overall benefit are needed if we are to justify judging people who have failed to do something on a given occasion by their general capacity to do it. Can we justify morally the rule of tort law that those shortcomers who do not possess the general capacity to do what is required of them, for example because of slow reaction time, are nevertheless to be judged by the same standard as those whose reaction time is normal? To judge them on this basis is, as Ripstein shows, necessary to an egalitarian system of risk distribution. But since this egalitarian system may operate to the disadvantage of shortcomers it needs to be justified. A promising way of doing this lies in the argument that the benefit that accrues to shortcomers from their participation in the system as a whole has to be set against the disadvantage that it imposes on them when their shortcoming lands them in liability from which they would otherwise be free. So I doubt if the allocation of risk in tort law can be justified apart from the overall benefits provided by the system to the agents on whom it places a potentially burdensome risk of liability to others. It would have been better had I not strayed, in the ostensibly respectable company of Kant, into advancing the “benefit and burden” argument in the form of a gambling analogy. Perhaps all that we need, in Ripstein’s words, is to accept that everything we do can be thought of as taking certain risks and 225

  accepting others. When we take or accept risks, in our own minds or in law, we normally have in mind potential benefits that appear to outweigh the risks. Even if faced with a choice of evils we choose on the basis that there is an advantage in avoiding the worse evil. To translate this pervasive feature of human life into a series of bets (with whom? for what stake? at what odds?) is probably to distort it. In many ways life is not a game, certainly not a game of chance. But we cannot avoid the need when we act to take and accept risks, to live with the outcomes of our acts and to take responsibility for them.

Stephen Perry

In a searching paper Stephen Perry makes the point that the notion of outcome responsibility is characterised in different ways that are not always compatible. Sometimes I understand it as relating to “personhood”, something inescapably attaching to the human condition. At others it is presented as a “social” system that could be changed, in the way that tort law could be based on fault liability or strict liability or replaced by a system of social insurance. These personhood and social understandings are, he argues, in tension. It is true that what is said about outcome responsibility in the essays collected in Responsibility and Fault is not always consistent. The essays, written over a period, were not homogenised. My present (mature?) view is that outcome responsibility, though inescapable, possesses both a subjective and a social aspect. The outcomes of what we do are attributed to us in a way that is essential to our conception of ourselves as individuals and also forms the basis on which we acquire credit or discredit in the eyes of others. Although we cannot exactly equate our successful outcomes and failures as seen by ourselves with our successes and failures as seen by others, the common-sense principles of attribution are objective enough to ensure that, except in the mentally ill, the two are not radically in conflict. It may seem pointless to ask whether outcome responsibility, so understood, is not only central to personhood (and an incentive to behave well) but also fair. Perry says that “the claim that the system is beneficial to all seems to be, empirically, simply false”. In the sense that there are people who fail more than they succeed, life’s perennial losers, this is true. Even so, we cannot pass as viable members of society without exposing our conduct and its outcome to assessment by others. In this sense outcome responsibility is a condition of the overall benefit that accrues to anyone who is treated as a rational person. Nearly everyone wants to be so treated, including the perennial losers. In this, perhaps indirect, sense I would still argue that to be 226

   treated as responsible for the outcomes of one’s actions, good or bad, is not only inescapable but a benefit. This point seems relevant to the issue that Perry raises in the second half of his paper. What is the link between the notion that in the long run outcomes make us what we are and the conclusion that, if we possess a certain general capacity for agency, we are responsible for those outcomes? Perry regards outcome responsibility as the concept that links outcomes to agents qua agents. It turns the causal link into one of authorship. Our status as moral agents requires us to acknowledge certain outcomes of our choices as ours, even if we have not chosen to act in a forbidden or blameworthy way. (It does not seem necessary to distinguish in this respect between responsibility for actions and responsibility for outcomes). But what outcomes are properly attributable to our agency? I am happy to rely for outcome responsibility in the personhood sense on ordinary ideas of attribution. According to this the outcomes for which we are responsible are cut short by certain later interventions. Parents subsidise their children’s university education. Daughter Jane works hard and gets a first class; son James takes to drugs and fails. The outcome of the parental subsidy is that each child has a university education. That is their doing and for that they are responsible. The success of Jane and the failure of James, in contrast, is their respective doing and responsibility, though the parental subsidy was a condition of both. The notion of control, on the other hand, to which Perry attaches importance in attributing outcomes to agents, seems to me to bear on risk distribution rather than outcome responsibility. Perry’s view is that an agent is outcome-responsible in the personhood sense only for those outcomes over which he or she had control, which he identifies with those that were avoidable when he or she acted. This avoidability depends, in his view, on the agent’s having a general capacity to avoid the outcome by foreseeing it and taking steps to avoid it. If this implies that the agent can always avoid the particular outcome that ensued on the occasion in question, I disagree. The possession of a general capacity for self-control, including a capacity for foresight, is compatible with responsibility for outcomes of one’s actions that were not in the circumstances foreseeable and avoidable. The idea of outcome responsibility does not dictate any particular system of risk distribution in tort law, but it does support the view that strict liability is not morally objectionable if there are pragmatic reasons for imposing it. The person held strictly liable is responsible for the outcome, even if unintended or unforeseen. To make those who harm others strictly liable for the harm done is therefore morally defensible. That is important to anyone who thinks, as I do, that it is an important aim of the law not to require what is 227

  morally indefensible. It then becomes a question of policy whether in a given case to convert outcome responsibility into legal liability, fault-based or strict. The theory of outcome responsibility is not, therefore, open to an objection that can be made to the legal libertarianism of the early Epstein. It need not lead to indeterminacy, since it rests on commonly accepted criteria of attribution. The case for translating it into tort liability depends on whether it is accompanied by other factors such as the defendant’s fault or the dangerous character of the activity in question.

Peter Cane

Peter Cane’s far-ranging discussion deals with and disputes my claim that outcome responsibility is “the basic type of responsibility in a community: more fundamental than either moral responsibility as generally understood, which requires fault, or legal responsibility, which requires either fault or special danger”. In the first part of his paper, where he gives an account of my views, one or two points need clarifying (some confusion may arise from the fact that Responsibility and Fault includes essays that, as mentioned earlier, have not been homogenised). I agree with Cane that there can be moral responsibility without moral blame. When I knock you over without being to blame for doing so I am nevertheless responsible. I must apologise and help you up. Though my obligation is not exactly to repair the harm I have done, I must acknowledge that I have knocked you over and take certain steps in consequence. I am open to moral criticism if I fail to take these steps. Outcome responsibility to that extent entails moral responsibility, at least when one’s conduct impinges on someone else; but it need not entail either moral blame or legal liability. To stress the distinction between responsibility outside the law and within it, it may help to speak of legal liability rather than legal responsibility. Of course the phrase “legal responsibility”, though elliptical, is perfectly in order. A second point is that responsibility for the outcome of one’s conduct, though in my view basic in relation to moral blame and legal liability, is not the only form of responsibility. Three species mentioned in chapter 6 of Responsibility and Fault are responsibility for one’s own conduct, the responsibilities that we undertake for other people or things and the responsibilities that are thrust upon us as members of a community. Responsibility for the outcome of one’s own conduct is a significant part of the first of these. But it is true, as Cane stresses, that in everyday life and the law, in tort law as well as 228

   criminal law, blame and liability often attach to agents for what they do irrespective of its outcome. Moreover, as he points out, legal liability can take the form of liability for someone else’s actions, as in the case of vicarious liability. In that case it is not, from the point of view of the person held liable, a form of outcome responsibility. Outcome responsibility is conceived as responsibility for an agent’s actions or their consequences. The agent is identified with his or her actions and this identification includes or extends to certain outcomes. Which outcomes? In a social context the question cannot be answered merely from the agent’s autobiographical point of view. Objective criteria are called for if moral blame and legal liability are to be fairly allocated. Hart and I contended that there are reasonably objective criteria that both courts and ordinary people resort to for this purpose (I return to this theme in connection with Stapleton’s paper). Cane, along with others, is sceptical. On our view, when a rule of law lays down that an agent is liable if he or she causes harm of a certain sort, this form of words is to be taken seriously. The rule is not to be restated in the form that an agent can be held to have caused harm if it is desirable from the point of view of morals or legal policy so to hold. It is true that a litigant is always free to put to a court the argument that a certain result would be morally unacceptable. That is a point that some legal positivists overlook. Arguments from morality and fairness can sometimes persuade a court to limit legal liability and at others to make an agent bear the risk of harm that he or she has not caused. But those arguments are what they are: arguments from morality or legal policy, not arguments that identify the agent with the outcome. So I endorse Cane’s view that the prime functions of legal and moral responsibility are different from those of outcome responsibility. But I do so on the basis that we can settle the question “which outcomes are we identified with”? independently of the question of what moral or legal consequences follow from that identification. Much the same applies to Cane’s view that the theory of outcome responsibility is not sensitive enough to the role of luck in relation to outcomes. Certainly moral and non-moral judgements about the role of skill and desert in a person’s successes or failures must take account of good or bad luck. But to determine which outcomes (achievements or botches) are ours is surely a preliminary to assessing the role of luck in bringing them about. The role of luck leads on to Cane’s views about strict liability, in the sense of liability irrespective of fault. I agree with him that luck plays no special role in regard to strict liability. If everything we do involves taking and accepting risks, luck is a pervasive element in all our achievements and botches. But there is a point to be made in this connection about fault, 229

  which Cane defines as a failure to meet a required standard of conduct. That is certainly correct for Anglo-American tort law. But to many outside the law an agent is at fault only if he or she could have done but failed to do what was required. Even if “could have” refers to the general capacity to do some specific thing, that leaves out a group of shortcomers who, though possessing the general capacity to control their conduct, lack the specific capacity, say, to react in time when driving a car. Are we not justified in treating these shortcomers, when they are unable to comply with legal requirements, as being from a moral point of view strictly liable rather than at fault? To do so is not merely a matter of terminology. Those who could but do not comply with the law exhibit a fault of character, as opposed to a constitutional defect. The shortcomers do not, as least if the latter are unaware of their shortcoming. The moral justification of strict liability for the outcome of one’s conduct seems to involve three features. One is the plain fact of attribution: the outcome is one’s own. That is not enough to justify the imposition of strict liability, but when coupled with the fact that what one has done is socially dangerous (selling adulterated milk, conducting blasting operations), the case for translating one’s responsibility into strict liability is quite strong. The argument for doing so does not reside only in the utility of minimising the incidence of harmful outcomes. The imposition of liability is fair in that the agent can recognise that he or she is being held liable for something that is his or her doing, not someone else’s. The fairness of being held strictly liable also depends, as said earlier, on the fact that the possession of the general capacity to control one’s conduct is a benefit to the holder. That benefit can properly be balanced by liability to fellow members of one’s community when their interests are seriously impaired by one’s conduct— when they drink the adulterated milk or are injured by the blasting operations. As will have appeared from these comments, I endorse a number of Cane’s views. Different considerations are involved in fixing what outcomes of a person’s conduct he or she is responsible for and in assessing that responsibility from the point of view of moral blame or legal liability. Distributive justice underlies corrective justice and the law about risk allocation seeks, as Ripstein argues, to ensure that the burden of bad luck is fairly distributed. Despite some divergences in detail, of which the most important concerns the objective character of the attribution of outcomes to people, Cane’s opinions do not seem radically inconsistent with mine.

230

   John Gardner

In a powerful paper John Gardner seeks to put the theory of outcome responsibility on a firmer basis. I had argued that it was both fair and essential to our sense of individual identity that we should be regarded as responsible for the outcome of our actions. That is not enough, he points out, to justify either strict liability or liability for causing injury by negligence as morally reputable legal institutions. For that it must be shown that an obligation not to injure someone (rather than an obligation to try not to injure them) is rationally intelligible. An obligation not to injure is an obligation to succeed in not injuring. But is the notion of an obligation to succeed defensible? Gardner begins by rebutting Brudner’s view that strict liability is ruled out because it imposes on the potential defendant an obligation of altruism and self-effacement so extreme as to be inconsistent with reciprocity between persons of worth. His counter-argument is that strict liability does not have this daunting consequence, because it does not impose on the potential defendant an obligation to take the most extreme care possible not to injure the plaintiff. It imposes no obligation to take care at all. Nor would resort to extremes of care even be a wise tactical move on the defendant’s part in an effort to avoid strict liability. It would often, like overprotective parenting, be counter-productive. Instead the important contrast, Gardner maintains, is between an obligations to try and obligations to succeed. Strict liability involves a pure success-type obligation, an obligation not to fail in avoiding injury to others, whereas the law of negligence imposes a hybrid obligation not to fail through want of trying. There can also be pure obligations to try, like the obligation in criminal law not to drive without with due care and attention. In substance this analysis seems correct but the term “try” may mislead. In the law of negligence the relation between the duty to take reasonable care and the duty not to injure is a means–end relation. French law marks this, I think wisely, by distinguishing obligations of means from obligations of result.1 The reason why this term is preferable is that, while not to try to avoid injury to another is negligent, to try to do so, however conscientious the effort, may equally be negligent if the person trying does not adopt the appropriate means. Gardner rightly points out that the obligation to take due care is not an obligation to succeed, since the notion of success is tied to ends. But it surely is an obligation to adopt appropriate means. 1. Obligation de moyen, de résultat. Most contractual obligations are obligations of result.

231

  This point is independent of the objection that to impose either sort of obligation may be unfair to defendants who lack the capacity to succeed or to adopt the appropriate means. It is true that there is no escape from moral luck. We are inevitably saddled with the risk of being held responsible for some things that we did not intend and could not control. The only dispute can be about the baseline in relation to which we can be judged lucky or unlucky. But, as will be apparent from earlier comments, I think that the particular risks that defendants have to accept in the law of negligence—the risk on a given occasion of not coming up to the standard of which they are in general capable, or of having shortcomings that reduce their capacity to do particular things such as reacting quickly—require to be justified on the basis that they are burdens balanced by corresponding benefits. It is not enough to say “robustly”, as does Gardner, that there is nothing unfair about people being held to standards that personally they are unable to meet. The main thrust of his paper, however, is to elucidate the idea of outcome responsibility as the common basis of the tort of negligence and of strict liability, to which a requirement of fault may or may not be added. He understands “outcomes” as broadly comprising both the consequences and components of action and “responsibility” as referring both to obligations to produce outcomes and to liability for failure to do so. His aim is to show that the idea of an obligation to succeed makes sense morally because the idea of a reason to succeed makes moral sense. Deciding and trying, Gardner argues, are actions of a logically parasitic type because they are decisions and attempts to do something, that is to say to succeed in doing something in particular. Hence if success is rationally insignificant so must deciding and trying be insignificant. Reasons to succeed are therefore primary in the sense that the intelligibility of reasons to try depends on the independent intelligibility of reasons to succeed. It seems to me that Gardner’s argument is correct and constitutes an advance in our understanding of moral agency. It shows how it can be possible to bridge the gap between the idea that we are identified with the outcomes of our actions, whether successes or failures, and the idea, essential if tort law is to rest on sound rational basis, that we can have obligations to succeed, and in particular to succeed in avoiding injury to others. III. C Jane Stapleton

Jane Stapleton’s ambitious paper “Unpacking ‘Causation’ ” is an item in an ongoing debate about the best way to analyse and present what are conventionally 232

   called causal issues in tort law. The debate began with her sympathetic but critical review of the second edition of Causation in the Law in 1988.2 It is not likely to end with the suggestions she makes in the present paper. Unfortunately there is space to comment only on some of its leading themes. Stapleton aims to build on and improve the ideas about tort law found in Hart and Honoré’s Causation in the Law and the later developments connected with outcome responsibility. As will become clear, I do not think that her enterprise succeeds. While not endorsing crude realism of the Leon Green variety, she would like to fix the limits of liability in tort law in two stages, roughly corresponding to the American realists’ “cause-in-fact” and “proximate cause”.3 These would be reconstructed as (i) historical involvement and (ii) the appropriate scope of liability for the consequences of tortious conduct. It seems to me that this bifurcation would be a retrograde step. It would run together factual and normative issues. Within the range of normative issues, it would confuse a number that are better kept separate. In an article on tort law in l9714 I distinguished five issues relevant to the limits of tort liability. Slightly reformulated, these perhaps cover all the points that Stapleton lists in her paper as relevant to the scope of liability for tortious conduct: i. whether the tortious conduct was a causally relevant condition of the harm; ii. whether some later human intervention or highly abnormal or coincidental event occurred that is inconsistent with the attribution of the harm to the agent; iii. whether the harm for which compensation is claimed was within the scope of the rule of law violated. This includes the question whether the rule that the agent violated was intended to make them liable for harm thereby caused or to place on them the whole risk of paying for loss that would not otherwise have occurred (questions of legislative and judicial policy); iv. whether the harm was of a highly improbable or unforeseeable type (a question of proportionality or retributive justice); and v. the moral character of the plaintiff ’s and the defendant’s conduct. Stapleton recognises the existence of these factors but proposes to run (ii) to (v) together. Even if (ii) is treated is a normative issue (and I do not think it should be) is there really a case for doing this? Admittedly, forensic necessity 2. “Law, Causation and Common Sense”, Oxford Journal of Legal Studies 8 (1988), 111. 3. The “proximate cause” test is derived from the mediaeval gloss on Justinian, Digest 4.2.9.1. 4. “Causation and Remoteness of Damage” in International Encyclopedia of Comparative Law vol. XI (ed. A Tunc, Tübingen: Mohr, 1971), ch. 7.

233

  may force a practising lawyer to lump these issues together under an impoverished label such as “proximate cause” or “remoteness of damage”. But legal scholarship should not be hamstrung by forensic constraints. What needs unpacking are the issues (ii) to (v), which are better kept separate. Is the outcome one with which the agent is identified? Do cost-benefit analysis or considerations of litigation control suggest that liability should be imposed on the agent for harm of this type? Would it be disproportionate to the agent’s conduct to impose liability for harm of a very improbable type? Would to impose liability be to reward someone who has behaved badly and has only himself or herself to blame? These are surely very different questions: questions of fact, of law, of justice, of morality—all relevant to the agent’s liability for the outcome but, in the interests of intellectual clarity, not to be fused. A brief word about (i)—whether the tortious act is a causally relevant condition of the harm for which compensation is claimed. Stapleton wishes to “crystallise” the NESS test of causal relevance (necessary element in a set of conditions together sufficient for the outcome), which Richard Wright and I favour, by adopting a “targeted but-for” test. I am not at present convinced that her suggestion improves the NESS test. A proper evaluation of it would call for an extended discussion, particularly of the role played by the condition hypothetically to be substituted for the targeted condition.5 In connection with the NESS test Stapleton’s proposes to speak of “historical involvement” rather than of a “causally relevant condition”, though she concedes that the requirement is causal. The proposed phrase hardly seems an improvement. “Involvement” is a rather loose notion. We can as a matter of history be involved in events without causing them or being necessary to their occurrence. 5. But the temptation to say a few words about the case of the desert traveller is almost impossible to resist. A poisons the keg containing the only water supply available in the desert, so that the traveller is doomed to die shortly. B then empties the keg so that the traveller in fact dies of thirst rather than poisoning. Have either or both caused the traveller’s death? The case is parallel to one discussed by the Roman lawyers (Justinian, Digest 9.2.11.3; 15.1; 51), where A inflicts a mortal wound on a slave, but before the slave dies of A’s wound B strikes a blow that ensures that he dies earlier or by a different process than he would otherwise have done. The Romans agreed that B had killed the slave but were divided as to whether A had done so. It cannot make a difference that in the desert traveller case the death ensues from deprivation rather than the infliction of wounds, nor, I believe, that in that case the act of B may have prolonged the traveller’s life as well as causing him to die of thirst rather than poisoning. The question is whether, if A dooms someone to die, because his conduct makes someone’s death certain in the near future, that counts as killing when the death ensues at a time or by a process (which can equally be a natural event such as an earthquake), other than that set in motion by A’s act. On the NESS test it does not, whereas B’s act does. Stapleton’s analysis of the desert traveller case, which holds A and B both to have been historically involved, seems to neglect the fact that B not only removed the poisoned water from the scene but also removed the water, full stop. That being so, the hypothetically substituted condition (supply of drinkable water) would anyhow have been no help to the traveller, since as it turns out B was going to neutralise it.

234

   To move on to (ii)—whether some later intervention, human or natural, is inconsistent with the attribution of the harm to the agent. Is this a normative issue, as Stapleton and others argue? Or is it an issue of fact to be settled by objective criteria, as Causation in the Law maintains and as judges in some commonwealth jurisdictions continue to assert? These criteria were independent, Hart and I thought, of the moral assessment of conduct. Moral assessment would depend on factors such as whether the outcome was good or bad, and whether the agent intended or knew that it was likely to come about. But first there was a preliminary question of fact to be settled: What did the defendant do? It was a question of fact, although the correct answer to it would quite often be a matter of degree and require the exercise of judgement. Was the outcome to be identified with the agent as part of his or her intervention in the world, or was it attributable to some later intervention, for instance to deliberate human action taken in the light of the situation brought about by the original agent? Stapleton seems to reject this analysis. The issue, she contends, is not one of fact, since ordinary usage is too fluid and variable to rely on for this purpose. But the idea of outcome responsibility, which she does appear to accept, strengthens the idea that we are identified with at least certain outcomes of our conduct in a way that commands general assent, whatever language may be used to express it. If we are identified with certain outcomes of our conduct and not with others, is not this an objective factor relevant to the extent of tort liability? If so, this supports the case for treating the attribution issue as one of fact. Hence the judges who treat the “later intervention” limit as a matter of fact seem to me correct, since that issue is not usually normative, except in the obvious sense that it depends on there being a legal norm that exempts agents from paying for harm they have not caused. (There are exceptional cases where the risk of harm is on the agent even if they have not caused it.) Stapleton and others hold that the “later intervention” criterion is not causal. It is true that it does not serve the function of explanation, scientific or historical. Its function is rather that of settling the extent of the changes attributable to human interventions in the world. That function presupposes that we each have our own achievements and failures, not usually shared with hundreds of other people, as would be the case if every outcome that would not have occurred but for an individual’s intervention was attributed, among others, to that individual. “Cut-off ” principles are basic to human individuality. We cannot dispense with them. And there is something odd about calling an inescapable feature of human society “normative”. It is true, of course, that when an outcome is attributed to us, that has normative consequences. The outcome is ours, and, if others are affected by it, we incur moral responsibility towards them. 235

  Where judges go wrong, if (as Stapleton thinks) they sometimes do, is in one of two ways. They may treat a reference to the common-sense idea of cause as enough to justify a ruling without pointing to the criteria of intervention that common-sense judgements implicitly appeal to. Or they may treat the causal issue as sufficient to dispose of the case without considering the normative issues of interpretation, justice and morality that call equally for attention—those set out in (iii) to (v) above. These judicial habits are perhaps inevitable in a system of trial by jury. But they do not invalidate, indeed they rather point to the need for the Hart and Honoré type of analysis at a scholarly level. German lawyers have an edge here, since they pay specific attention to the purpose (Normzweck) of legal rules, not only in tort law but generally. This is more difficult to do overtly if one is hamstrung by a formula such as “proximate cause” or “remoteness of damage”. But the difficulty is not insurmountable. Stapleton cites an extrajudicial statement by Lord Hoffmann that in appellate courts legal issues couched as arguments about causation are nearly always about the law. That may or may not be an exaggeration. But the insight that causal and policy issues need to be distinguished has helped the same judge on at least one occasion to deal both with a genuine causal issue, along the lines of the Hart–Honoré analysis, and with the question of legal policy involved in the case.6 Much the same must be said of Stapleton’s view of the Hart–Honoré analysis of near-causal relationships, applied in detail by Kadish to criminal law. These have some features in common with causing outcomes but differ in others. Examples are the providing or not providing reasons, opportunities, means or help on the basis of which others act for good or ill. These relationships differ from straightforwardly causing outcomes because, for example, the voluntary actions of others may intervene between the agent and the ultimate outcome. Take as an example providing means. I lend you a kitchen knife. You use it to prepare a meal, or to stab someone in the course of a quarrel. At the level of agency my relation to the outcome is the same in both cases: that of providing the means by which the outcome was achieved. It makes for clarity, surely, to distinguish this relationship both from straightforwardly causing the outcome myself and from the moral and legal aspects of the outcome of my loan. When I lend the kitchen knife, as opposed to preparing the 6. Environment Agency v. Empress Car Co. [1988] 1 All ER 481,488–9, where the issue was whether a company had caused polluting matter to enter a river. “In answering questions of causation for the purpose of holding someone responsible both law and common sense normally attach great significance to deliberate human acts and extraordinary natural events”, but “one cannot give a common sense answer to a question of causation for the purpose of attributing responsibility under some rule without knowing the purpose and scope of the rule”.

236

   meal or doing the stabbing myself, I am not automatically identified with the outcome. Whether I should be depends, among other things, on its moral aspect. This in turn depends on what I intended or knew or should have suspected when I lent you the knife. The issue of legal policy involves yet other considerations. For example, supposing that I knew you were mentally disturbed, would it be cost effective to impose liability for the stabbing? Could one draw a bright line between permissible and unlawful loans or between worthy and unworthy borrowers? Nothing is to be gained by running these issues, factual, moral and legal, together. On the contrary, the fact that my relation to, say, the stabbing is that of providing the means by which the outcome was produced rather than of straightforwardly causing it is relevant precisely because it draws attention to the need to scrutinise the moral and policy aspects of the situation before imposing legal liability. The analogy of criminal law may help to underline the point. The complex topic called “participation in crime” involves causal, moral and policy issues. Participation in outcomes that give rise to tort liability is almost equally complex and equally involves diverse issues. Stapleton’s paper makes some fascinating points but seems to suffer from strategic weaknesses. One is the running together of forensic and analytical considerations. A second, connected with the first, is the failure to distinguish between issues of fact and normative questions, and between different types of normative issue. A third is an impoverished view of the range of ways in which agents can be “historically involved” in outcomes. The possible forms of participation are too varied and intricate to be reduced to the existence or non-existence of NESS conditions.

IV. M William Lucy

A broad sweep marks William Lucy’s essay on the methodology of legal philosophy. He depicts ideal types termed “Visionaries” and “Bricoleurs”. Neither of their versions of scholarship is likely to yield a philosophically engaged and legally meaningful understanding of law, in particular private law. Visionaries do not provide this because their top-down method begins from too grand an overarching point of view. This may be, for example, utilitarian, Kantian, economic or based on a theory of corrective justice. Looking down, visionaries explain or criticise private law in terms of their chosen vision. But the visionary point of view is too blunt. It seldom yields 237

  determinate answers to legal problems. In contrast “bricoleurs” (the image comes from Lévi-Strauss) remain on the bottom, imprisoned in a set of received rules and concepts. Visionaries run the risk of turning legal debate into full-blown moral-cum-political argument; bricoleurs undermine its normative appeal by doing no more than reshuffle received ideas. There is, however, a middle way, which might be called the bottom-up method. It is practised by modest visionaries and ambitious bricoleurs. Starting from existing legal rules and institutions, we can ask, given that “we are undoubtedly children of the interpretative attitude”, whether these rules and institutions fulfil the role they claim to fulfil. Lucy thinks that my work illustrates this middle path in legal scholarship. It seems to me that the bottom-up method is not the only defensible method but that it is potentially fruitful. Legal philosophy should speak to lawyers. The middle path is sensitive to their interests in that it investigates claims made by the law itself. It is not merely that lawyers tend to think it morally and politically a good thing to have a legal system, so that there is often a good reason to obey the law. Legislators who lay down the law and judges who interpret and apply it implicitly claim that the rules they lay down and the decisions they reach are morally in order. Sometimes what the law lays down appears to diverge from what is morally acceptable, for example as regards strict liability. A legal theorist is then driven to ask whether strict liability can be morally justified and, if so, in what conditions. The inquiry can lead the theorist into controversial aspects of moral philosophy and the philosophy of action, as it does, for example, when we ask why people should be legally responsible for the harm they cause. But, as Lucy points out, given a degree of moral consensus—and to be effective law needs the backing of a certain moral consensus—it need not always do so. In principle the inquiry by a scholar who pursues the middle path is no different from the argument that an advocate puts to a court that a certain interpretation of a statute or common law rule would be morally objectionable. That argument, even if well founded, may not succeed, because the legal process has aims other than that of being morally in order. Decisions must, for example, be consistent with authority. This requirement may rule out the morally best solution to the problem before the court. But the morally best solution is always on the agenda, even though it may be impracticable for an advocate or judge to challenge such a well-established institution as strict liability in private law. One of the links between legal practice and theory is that both are concerned, though at a different level, with the moral acceptability, and hence legitimacy, of particular rules and branches of the law. Bricolage has its merits so far as the internal consistency of the legal system is concerned, for 238

   consistency is one of the values at which the law aims. But if the law is to be subject to critical appraisal in terms of the claims it asserts, bricolage is not enough. The visionary method is not similarly deficient. It is driven by the ambition to explain or reform large parts of the law on the ground that they conform to or violate economic efficiency, utility, the categorical imperative or some other general criterion of proper conduct. In principle, it seems to me, there can be no objection to this method of inquiry. But it is unlikely to attract those who share certain pluralistic or, at any rate, non-purist beliefs about law and morality. They will think, as I do, that legal systems have a number of different aims, not all of which can be pursued all the time. They will believe that there are several goods and values that people should pursue, individually and collectively, none of which can be consistently ranked above the others. These beliefs are of course controversial, but, to the extent that they are justified, they limit the fruitfulness of the visionary method.

239

INDEX absolute liability, 188–9, 191 threatened with indeterminacy, 76 see also strict liability act-centred approach, 22–5, 222 action, 7 adjudication, 207–8 job for tinkers, 206 see also judges agency, 72–4, 78, 81, 109–10, 128, 133–5, 191, 227, 236 moral agency, 232 agent-centred approach, 22, 25–34, 222 Alexander, 77 American Realism, see Realists Aristotle, 115, 191, 197 Ashworth, 93 attribution, 148–51, 153, 157, 159–60, 162–6, 171, 223, 225, 230, 235 factors affecting, 153–4 ordinary ideas of, 227–8 Augustine, 43 Austin, 153 autonomy, 73–4 avoidability: epistemic aspect of, 77–8 Ayer, 23–4 Bagshaw, 147 Becker, 145 benefit/burden principle, 55–7, 225 Bentham, 197 Bilgrami, 30 Birks, 199 Bix, 70 blame, 82–3, 222 Bratman, 2 bricoleur scholarship, 198–200, 237–9 ambitious form of, 213–18, 238 problems affecting, 204–9 Brint, 199 Brudner, 111–18, 120, 125, 133, 231 Buckley, 193 but-for test, 155, 174, 178–9 targeted, 146, 167–8, 175–85, 234 targeted test expounded, 175–6 can: particular and general, 14–15 Cane, 64, 66, 89, 100, 105, 147, 166, 223, 228–30 capacity, 1–2, 40, 83–5, 92, 101–2, 104, 107, 119, 127, 188, 219–23, 232

and intention, 2–3 degrees of 1–2, 16, 86, 220–2 general capacity, 26, 28, 39, 41–2, 78–9, 84–5, 98–9, 219–22, 224, 227, 230 misuse of, 221 moral capacity, 142 narrowing of general capacity, 49–50 normative aspect of general capacity, 79 perturbers that interfere with, 27 relation with outcome responsibility, 46–7 to act differently, 40–1 to avoid, 78 to connect, 32 to have done otherwise, 21–34 to track standards of reason, 33–4 care, 112–17, 120, 127, 129, 132 level of productive care, 116–17 objective standard of, 127, 129 reasonable care, 117 taking care, 120 see also duty of care causal language, 148, 153, 160–1 ambiguity of, 145–6, 148–9, 158, 160, 162, 167 causation, 38–9, 57–8, 69–70, 76, 89, 104, 145–85, 191, 216 analytical distribution of issues, 156–60 and fault, 107 as question of fact, 151–3, 155–6, 160, 166, 168, 173, 235 causal connection, 157–8 causally relevant condition, 157 common-sense notions of, 148, 150–2, 154, 158–60, 226, 236 constellation of casual connections, 161–6 different functions of, 149, 159 historical involvement, 146, 167–85, 233–4 in law, 173 necessary element in sufficient set notion, 146, 154, 157, 167, 174–5, 177–84, 234, 237 Realist approach to, 155–6, 158, 233 relevance to purpose, 146, 168–9, 171–3, 177, 233, 235 Stapleton’s proposed approach, 183–5 used to allocate responsibility, 89–90 cause-in-fact, 155, 160, 233 Chisholm, 24–5 choice, 24–5 civil law, 93–7, 100, 106–8 Coase, 76 coercion, 23, 34 Cohen, 105

241

 coherence: norms of, 11–13 coincidence, 169–73 Coleman, 70, 77, 135, 202 common law, 122, 125–6, 142, 192, 206 common sense, 148, 150–2, 154, 158–60, 203, 226, 236 Commonwealth jurisdictions, 155, 158–60, 165 compensation, 62, 68–9, 74–5, 94, 96–8, 129–30 consequences, 38, 47, 58–9, 130 contract, 93, 131, 215 control, 75, 76–7 conversable subjects, 31–3 criminal law, 73, 81, 86, 92–4, 96–7, 100, 103, 106–8, 231, 236–7 participation in crime, 237 criminal law scholarship, 187 Davidson, 7, 25 Deakin, 193–5, 204 deciding, 135–6 deontology, 73–4 Descartes, 23 desert traveller case, 175, 178–83, 234 desires, 4–7, 12, 25 coherent set of, 6, 11–13 non-desiderative capacity, 13 determinism, 22–3, 79, 87–8, 91, 219, 222 Dewey, 197 Dias, 194 doctrine, 147, 216 bluntness of, 201, 211–13 coherence of, 215 neutrality of regarding moral vision, 201 double hit hunters case, 174–7 duress, 23 duty of care, 122, 125, 147, 165–7 see also care; obligation Dworkin, 105, 196, 198–9, 203–4 economic analysis of law, 76, 202–3, 210 economic loss, 56, 109 economists, 149, 177, 196–7 efficiency, 202–4, 210 English anti-intellectualism, 160 English law, 193, 199 Enlightenment, 5 Epstein, 69, 191, 228 fairness, 39, 62–4, 67–8, 84–5, 92, 229 fair warning, 133 fault, 83–4, 95, 97–102, 106–8, 125, 191–5, 228–9, 232 justification for responsibility for fault, 86 subjective moral fault, 188–92, 194 Feinberg, 43 Feuerbach, 51

Finnis, 119, 121 first-person narrative, 71, 78 Fischer, 21 Fish, 199 foreseeability, 78, 169, 172 Frankfurt, 4, 21, 25 freedom: of action, 108 of action/from harm, 106–7 French law, 231 Freud, 47 Fried, 201 Fuary-Wagner, 183 Fuller, 211 fully rational state, 5–7, 11–13, 34 Gabriel, 145, 159 gambling, 53–6, 61–2, 87–8, 132, 225 individuation of bets, 55 Gardner, 93, 135, 223, 231–2 German lawyers, 236 Lord Goff, 199 Goldberg, 155 Green, 147, 155, 233 Gross, 145, 198 Hacker, 141 Harrison, 145, 198 Hart, 38, 43, 58, 69–70, 89, 95, 126, 130, 145–67, 169, 173, 179, 183–4, 216–17, 229, 233, 235–6 Hegel, 112, 197 Heidegger, 51 Heuston, 193 history: nature of, 177, 182 Hobbes, 4–5 Lord Hoffmann, 150, 160, 236 Holmes, 79, 193 Honoré, 1, 14–17, 22, 37, 39–45, 47–8, 57–9, 61–72, 74–5, 78–102, 104, 106, 109–10, 126–36, 138–40, 143, 145–67, 169–70, 173, 175, 178–9, 182–4, 187–8, 190, 192, 194, 196, 202, 214–18, 233, 236 Horder, 71, 90, 107, 147 Horwitz, 204 Howarth, 145, 148, 154, 156, 162 human condition, 143 human identity, see personal identity Husserl, 205 Ibbetson, 197 individuality, 235 intention, 2–4, 31, 75, 95–6, 107, 114, 120, 189, 192, 237 interests: balancing of, 109–10

242

 interpersonal allocative practices, 47–50 interpretative attitude, 204–5, 207–8, 238 intervention: relevance of, 164–6 irrationality, 5

role distinct from judicial, 209 via media, 209, 238 see also bricoleur scholarship; visionary scholarship Lévi-Strauss, 204–5, 238 liability concepts: continuum of, 188–9 libertarian approach, 67, 69–70, 75–6, 228 life: as a rational agent, 139, 143 Lipton, 145, 148, 161–2, 168, 170, 173 logic: and experience, 206 luck, 38, 53, 82, 88, 90–3, 98–9, 101–7, 110, 142, 229 circumstantial, 104 different types of, 104 dispositional, 104 moral, 127–8, 142, 232 sensitivity to, 104–8 Lucy, 237–9

Jackson, 23 Jones, 199 judges: aided by legal scholars, 196, 198 appellate court judges, 212 bricoleur judges, 198, 208–9 judicial decisions, 193–4 role distinct from scholars, 209 see also adjudication jury, 152, 160, 236 deciding normative issues, 160 justice, 38–9, 149, 223 corrective justice, 39, 43, 45–6, 50, 59, 64–5, 71, 105, 191, 230 distributive justice, 39, 44, 52, 65, 67, 70, 105–6, 230 risk-distributive justice, 39, 42–50, 53, 56–8, 224 Justinian, 233–4 Kadish, 236 Kaldor-Hicks, 203 Kant, 5, 53, 112, 119, 141–3, 191, 192, 197, 225 Kelley, 145, 149, 151, 162 Kelsen, 217–18 Kennedy, 205 Kennett, 1 Kenny, 130 Laslett, 44 law, 239 and morality, 121 as self-reflective practice, 213–14 authority of, 206 autonomy of, 217 finality of, 207 general theory of, 217–18 law’s self-understanding, 199 legitimacy of, 206–8, 214, 216–17 moral contestability of, 211–12 nature and functions of, 187, 207, 217 law students, 195 lawyer-economists, 149, 177 legal concerns, 147 legal doctrine, see doctrine legal philosophy, 40, 187–8, 238 legal point of view, 143, 222 legal reasoning: clarity in, 166–7 legal scholarship, 187, 196, 219, 236, 238 method in, 187

Mackie, 87, 95, 179 Macklem, 135 Mahoney, 170 mandatoriness, 140–1 Markesinis, 193–5, 204 markets, 109 McDowell, 31 McGeer, 31 method, 187, 198–9, 237–8 modal facts, 16–17 Moles, 145, 148 Moore, G, 24 Moore, M, 119 morality, 128, 143, 229, 239 depending on law for determinacy, 202 moral claims, 112 moral consensus, 238 morally acceptable, 38 morally best solution, 238 moral theory, 210 Nagel, 73–4, 91–2, 95, 104, 121, 123, 127–9, 188, 195 naturalistic constraint, 23–5, 28–9 natural science, 23 negligence, 21, 40, 42, 45, 52, 56, 58, 95, 99, 104, 107–8, 111, 117–20, 122–4, 127, 129, 139, 142, 147, 150–1, 165, 171–3, 190, 192, 194–5, 215, 223, 231–2 bare negligence, 124 contrarian interpretation of tort of, 125–7, 143 moral essence of tort of, 124–5 moral intelligiblity of, 124 New Zealand, 68

243

 Nietzsche, 47, 225 normative concerns, 173 normative constraint, 23–5, 29–33 normative vision, 197, 205, 218 basic norm of co-operation, 217–18 lack of uniform, 169, 198, 203–4, 207, 210–12 North American law schools, 123 obligation, 73–4, 111–12, 130–1, 139–40 hybrid trying-succeeding obligation, 122, 124, 142, 231 legal obligations as would-be moral obligations, 121 moral intelligiblity of obligation to succeed, 141 moral priority of obligation to succeed, 134–5, 141 not to injure, 112–13, 116, 122 primacy of obligation to succeed, 139 to succeed, 117–20, 122, 124, 127, 142, 231–2 to take all reasonable care, 117 to take extreme care, 112–17, 120 to try, 117–20, 122, 124, 127, 231 omission, 167, 216 only available means, 114 ordinary language, 227–8 advocates of linguistic analysis, 156 diversity of meanings, 145–6, 148–9, 158–9, 162 Otsuka, 21 outcome allocation, 62–4, 71 outcome responsibility, 37, 46, 61–86, 88–104, 110, 126–34, 215, 223–33 achievement conception of, 75 advertence-based conceptions of, 77 ambiguity of, 130–2 as causal concept of responsibility, 89 as concomitant of agency, 72 as form of moral responsibility, 70–1 as private narrative, 47–9, 52–3 avoidability-based conception of, 77–9 enhanced versions of, 95 justification for, 85–6, 90 legal and moral responsibility distinct from, 98 narrowing of, 49–50 non-consequentialist feature of, 74 personhood understanding of, 63–6, 68, 70–9, 132–43, 226 social/agency understanding, 40 social understanding of, 63–70, 132, 226 Owen, 38, 134 ownership of misfortune, 70 Pareto, 197, 202 Parfit, 116 Perry, 40, 63, 65–6, 69–71, 74–7, 79, 132, 223, 225–8

personhood, 63–4, 68, 71–2 personal identity, 63, 71–2, 91, 93, 109, 133–6, 147, 231 Pettit, 11, 22–3, 25, 31–2, 34, 219, 221–2, 225 philosophers, 196–7, 202, 215–17, 238 pluralism, 239 Posner, 124 possible worlds, 16–17, 221 Postema, 74 practical rationality, 113–14, 116 practical reason, 136, 209 practice, 200, 207, 213, 216–17, 238 practitioners, 195, 238 praise, 82, 222 private law: lack of unified normative vision, 198, 203–4, 210–12 rationality of, 197 private law scholarship, see legal scholarship private nuisance, 193 probability, 169–71 professional relationships, 42 proportionality, 58 proximate cause, 155, 158, 160, 233, 236 public institutions: rationality of, 205, 208 punishment, 96–8 allocation of, 94 questions of fact/value, 160 see also under causation rational choice theorists, 196–7 rational control, see capacity Rawls, 105, 197 Raz, 111, 121 Realists, 147, 155–6, 158, 233 reasonable person, 42, 45, 104, 106, 171, 189, 225 defeasible presumption of, 79 reason for action, 135–6 reasons, 139–40 reason to succeed, 134–5, 137–40, 143, 232 as obligatory reasons, 141 primacy of, 138–9 reason to try, 137–9, 143 recklessness, 95–6, 107 Lord Reid, 55 remoteness of damage, 155, 158, 236 responsibility, 15, 21, 23–4, 29, 31, 37–9, 44–5, 55, 58, 70, 87, 102, 106, 130, 150–1, 171, 215, 226, 228 action responsibility, 73, 77 and fault, 107–8 and social values, 108–10 criminal responsibility, 92 displacement responsibility, 51–3, 57, 224

244

 distinction between fault-based and strict, 100–2 functions of responsibility practices, 81–2, 94, 97–8, 103–4, 110 implication responsibility, 50–2, 57 legal and moral distinguished, 97 legal responsibility, 81–3, 90, 93, 97–8, 103–5, 108, 228–9, 238 moral responsibility, 81–3, 90, 92–3, 95–8, 103, 228–9, 235 relevant factors on attribution of, 153–4 restitution, 52, 199 see also attribution; outcome responsibility result, 130 retributive principle, 58 rights, 48–9, 56, 58, 65, 69, 100–1, 111–12, 130–1, 192 Ripstein, 70–1, 77–8, 106, 135, 223–6, 230 risk, 169–70 risk distribution, 39–40, 42–50, 56, 59, 65–70, 223–7, 230 rule of law, 43, 159–60, 198 rules, 207–8, 211–12 purpose of legal rules, 236 Runciman, 44 Salmond, 59 Scanlon, 87, 104 Scheffler, 105 self-control, 2–18, 219–21, 227 basis in intrinsic feature of agent, 16–17 diachronic self-control, 8–10, 18 synchronic self-control, 9–13, 18 self-defeating menaces, 115–16 Shapiro, 135 Simester, 107, 137 Simmonds, 198, 201, 208 Lord Simonds, 200 Simpson, 141, 198 Smith, A, 137 Smith, M, 1, 5, 11, 15, 22, 32, 219–21, 225 social insurance, 45, 57, 68 social values, 81 standards, 33–4, 59, 100–2, 222, 230 see also values Stapleton, 89–90, 109, 147, 150, 155–6, 160, 164–6, 168, 172, 175, 184, 229, 232–7 Stocker, 115 Strawson, 23, 29, 222 strict liability, 21, 45, 56, 62, 74, 81–6, 94–6, 98–102, 108, 110–43, 188, 191–3, 195, 215–16, 227–9, 231–2, 238 and balancing of interests, 109–10 and self-effacing behaviour, 111–13, 125–6, 133, 231 distinguished from absolute liability, 188–9, 191

distinguished from objective standard of care, 127, 129, 132 distinguished from subjective moral fault, 188–90 four types of, 83 institutional fairness objection, 120–3, 125–6, 131–3, 140 moral basis of, 188, 230 moral intelligibility objection, 120–5, 131–3, 140, 143 nature of, 112 relationship with negligence, 117–20, 123–5, 190 success: primacy of, 134 Swadling, 89 theory, 37, 61, 121, 123, 199–200, 210, 215, 217–18, 238 role of, 196–9 third person attribution, 225 third person narrative, 71, 78 tort law, 37–59, 65, 67, 70, 73, 78, 93, 97, 104, 108–9, 123, 126, 131, 145, 157–8, 168, 190–1, 223–4, 227, 230, 232–3, 236 American tort scholarship, 39 concepts under-analysed, 193–5 fault in, 191–2 issues relevant to liability, 233–4 justification of, 57 textbook treatments of, 193–5 tort liability, 237 tort theory, 37, 61, 121, 123 tracking standards, 30–4, 222 tradition, 204, 208–9 Tunc, 233 UK, 155 unfairness, 140 Unger, 203 US, 155, 159–60, 164 values, 81 temporally and culturally dependent, 165–6 see also standards Van Inwagen, 22 vicarious liability, 45, 82–3, 85, 97, 99–100, 229 victims, 93 visionary scholarship, 196–8, 237–9 distinction with bricoleurs, 209–13, 199–200 divergence, 198 method, 198 modest version of, 212–18, 238 problems affecting, 200–4, 207–9 von Wright, 130

245

 Waldron, 38, 58, 77 Walzer, 44 Wasik, 93 Watson, 4–5, 22 Weaver, 199 Weinrib, 53, 110–11, 121, 188, 191, 192, 195, 198

Weir, 145 Williams, 44, 53, 87, 128–9, 205 Wright, 145–6, 156–7, 174–5, 178–9, 182, 234 wrong of injuring, 125–6 Zipursky, 155

246